Loading...
HomeMy WebLinkAboutAPA4159Proceedings: Cold Regions Hydrology American Water Resources Association This document is copyrighted material. Permission for online posting was granted to Alaska Resources Library and Information Services (ARLIS) by the copyright holder. Permission to post was received via e-mail by Celia Rozen, Collection Development Coordinator on December 16, 2013, from Kenneth D. Reid, Executive Vice President, American Water Resources Association, through Christopher Estes, Chalk Board Enterprises, LLC. This symposium includes the following chapters directly relevant to the Susitna-Watana Project: The Susitna Hydroelectric Project simulation of reservoir operations by Yaohuang Wu, Joel I. Feinstein, and Eugene J. Gemperline ...................................... pages 3-11 Hydrology and hydraulic studies for licensing of the Susitna Hydroelectric Project by Eugene J. Gemperline ............................................................................................... pages 73-85 Some aspects of glacier hydrology in the upper Susitna and Maclaren River basins, Alaska by Theodore S. Clarke, Douglas Johnson, and William D. Harrison .......................... pages 329-337 Forecasting the effects of river ice due to the proposed Susitna Hydroelectric Project by Ned W. Paschke and H.W. Coleman .................................................................... pages 557-563 Freezeup processes along the Susitna River by Stephen R. Bredthauer and G. Carl Schoch .......................................................... pages 573-581 PROCEEDINGS of the Symposium: Cold Regions Hydrology UNIVERSITY OF ALASKA-FAIRBANKS, FAIRBANKS, ALASKA Edited by DOUGLASL.KANE Water Research Center Institute of Northern Engineering University of Alaska-Fairbanks Fairbanks, Alaska Co-Sponsored by UNIVERSITY OF ALASKA-FAIRBANKS FAIRBANKS, ALASKA AMERICAN SOCIETY OF CIVIL ENGINEERS fECHNICAL COUNCIL ON COLD REGIONS ENGINEERING NATIONAL SCIENCE FOUNDATION STATE OF ALASKA, ALASKA POWER AUTHORITY STATE OF ALASKA, DEPARTMENT OF NATURAL RESOURCES U.S. ARMY, COLD REGIONS RESEARCH AND ENGINEERING LABORATORY Host Section ALASKA SECTION OF THE AMERICAN WATER RESOURCES ASSOCIATION The American Water Resources Association wishes to express appreciation to the U.S. Army, Cold Regions Research and Engineering Laboratory, the Alaska Department of Natural Resources, and the Alaska Power Authority for their co-sponsorship of the publication of the proceedings. American Water Resources Association 5410 Grosvenor Lane, Suite 220 Bethesda, Maryland 20814 AMERICAN WATER RESOURCES ASSOCIATION TECHNICAL PUBLICATION SERIES TPS-86-1 LIBRARY OF CONGRESS CATALOG CARD NUMBER: 86-70416 1986 COPYRIGHT BY THE AMERICAN WATER RESOURCES ASSOCIATION All rights reserved. No part of this book may be reproduced in any form or by any mechanical means, without written permission by the publisher. These proceedings were published by the American Water Resources Association, 5410 Grosvenor Lane, Suite 220, Bethesda, Maryland 20814. The views and statements advanced in this publication are solely those of the authors and do not repre- sent official views or policies of the American Water Resources Association; the University of Alaska-Fairbanks; the American Society of Civil Engineers, Technical Council on Cold Regions Engineering; National Science Foundation; State of Alaska, Alaska Power Authority; State of Alaska, Department of Natural Resources; and the U.S. Army, Cold Regions Research and Engineering Laboratory. Communications in regard to this publication should be sent to the Circulation Department of the American Water Resources Association, 5410 Grosvenor Lane, Suite 220, Bethesda, Maryland 20814, U.S.A. PREFACE Fascination with polar regions, along with potential commerce, first attracted arctic and ant- arctic explorers. However, most adventurers and their supporters soon concluded that resource development and trading in cold regions was not very profitable. In fact, the high latitudes of North America were viewed as a physical obstacle to trade routes between southeast Asia and Europe. Substantial time and effort was spent seeking the Northwest Passage. Early commercial spirit faded with the decline of the fur trade and whaling, and with the depletion of the gold fields. The Second World War drew fresh faces to the North. Military activity and new resource development became the major motivating factors for population growth in cold regions. These activities stimulated the tremendous growth of scientific research on high-latitude phenomena over the last 40 years. The first forms of economic activity in the North, such as whaling and gold mining, were performed by temporary inhabitants who frequently retreated southward to warmer climates. Slowly this mode of operation changed; people started to make the high latitudes their year-round home. New technology made living in these cold climates more tolerable during the winter. Modern transportation and communication reduced the perceived distance between the North and the rest of the world. Since much of the increasing activity was affected to some extent by hydrologic phenomena, the need developed for both an understanding of hydrological processes dominated by snow and ice, and long-term data for hydrologically related design. Clearly, mid-latitude hydrology has attracted much rr ore attention than high-latitude phenomena during the last few decades. This is rightfully so, since most people live at the mid-latitudes. Yet the search for natural resources now extends far beyond this zone of comfortable living. The greatest strides in our understanding of hydrologic processes are now being made in both cold and tropical climates. In cold regions, science has particularly advanced our understanding of the roles of snow and ice. Also, it is imperative that we precisely understand the contribution of the polar regions and the tropics to the global climate. The objective of this symposium is to pull together researchers and practitioners in hydrology and closely related fields to discuss present hydrologic problems and interests. We are fortunate to have presentations from a large number of countries: Austria, Canada, Denmark (Greenland), England, Finland, Iceland, Japan, Norway, Sweden, USA and USSR. A review of present hydrologic data reveals that most northern countries have very sparse networks for data collection. Most data are collected around population centers that are situated at relatively low elevations. The periods of record for most hydrologic data in cold regions are quite short relative to record lengths in temperate climates. Furthermore, instru- mentation used to collect data often does not work satisfactorily in cold regions. A session on Instrumentation and Data Collection, and another on Remote Sensing will address many of these problems. iii The sessions generally follow the logical divisions of the hydrologic cycle. Watershed input is covered in a session on Precipitation -Snowpack -Soil Processes. Along these same lines, glaciers that act as storage reservoirs during the winter and provide meltwater for runoff in the summer are included in the session on Glacier Hydrology. The importance of ablation and the con- version to runoff in the hydrologic cycle of cold regions are highlighted by a session on Snowmelt Runoff. Channel processes associated with sedimentation and ice are presented in one session on Channel Hydraulics and Morphology, and two sessions on River Ice Hydraulics. Surface storage in water bodies and associated processes are discussed in a session on Reservoir and Lake Level Processes. In the field of environmental hydrology, one session is devoted to Water Quality. There are more than 20 additional papers in the poster session. These excellent papers could have fit into one of the other sessions if there had not been so many papers. The poster papers were selected by the technical chairman based solely on graphical criteria. For every session, snow, ice and frozen ground are the common threads that bind the symposium together. Despite the fact that numerous books with ominous titles (No Man's Land, Amid Snowy Waste, Ice Bound, Lost in the Arctic, Two Against the Ice, Nansen in the Frozen World, etc.) have been written about the colder regions of the world, people venture forth in ever-increasing numbers. To treat this land properly, we need to develop a clearer understanding of its natural processes. Douglas L. Kane Editor, Technical Chairman iv ACKNOWLEDGMENTS As editor of this proceedings, I would like to thank several local AWRA members who have assisted me in making this publication possible: Brent Petrie (General Chairman) of the Alaska Power Authority; Stephen Mack (Co-Chairman Local Arrangements) of the Alaska Department of Natural Resources; James Aldrich (Co-Chairman Local Arrangements) of Arctic Hydrologic Consultants; Linda Perry Dwight; Stephen Bredthauer (Finance Chair- man) and Jeffrey Coffin of R&M Consultants, Inc.; Ronald Huntsinger (Exhibits Chairman) of the U.S.D.I. Bureau of Land Management; and Charles Slaughter of the U.S.D.A. Institute of Northern Forestry. Two indi- viduals that I would like to express special thanks to are Catherine Egan of the University of Alaska-Fairbanks and Charlene Young of A WRA for helping with all the details that are required to make the proceedings a reality. Name James Aldrich Eric Anderson Gary Anderson George Ashton William Ashton Lars Bengtsson Carl S. Benson Neil Berg David Bjerklie Tim Brabets Stephen R. Bredthauer Edward J. Brown Robert Burrows Darryl Calkins Robert F. Carlson Edward F. Chacho George Clagett Jeffrey H. Coffin Samuel Colbeck Keith R. Cooley Arthur G. Crook Larry Dearborn Phil A. Emery David C. Esch James L. Foster Andrew G. Fountain John Fox Thomas George Joan P. Gosink Raoul J. Granger Dorothy K. Hall Donald R. F. Harleman Affiliation Arctic Hydrological Consultants U.S. National Weather Service U.S. Geological Survey /WRD U.S. Army CRREL R&M Consultants, Inc. Uppsala University University of Alaska-Fairbanks U.S.D.A., Forest Service· University of Alaska-Fairbanks U.S. Geological Survey /WRD R&M Consultants, Inc. University of Alaska-Fairbanks U.S. Geological Survey/WRD U.S. Army CRREL University of Alaska-Fairbanks U.S. Army CRREL U.S.D.A., Soil Conservation Service R&M Consultants, Inc. U.S. Ar~y CRREL U.S.D.A., Agricultural Research U.S.D.A., Soil Conservation Service Alaska Dept. of Natural Resources/DGGS U.S. Geological Survey/WRD Alaska Dept. of Transportation and Public Facilities National Atmospheric and Space Administration/GSFC U.S. Geological Survey University of Alaska-Fairbanks University of Alaska-Fairbanks University of Alaska-Fairbanks University of Saskatchewan National Atmospheric and Space Administration/GSFC MIT v City /State/County Fairbanks, AK Silver Springs, MD Richmond, VA Hanover, NH Anchorage, AK Uppsala, Sweden Fairbanks, AK Berkeley, CA Fairbanks, AK Anchorage, AK Anchorage, AK Fairbanks, AK Fairbanks, AK Hanover, NH Fairbanks, AK Fort Wainwright, AK Anchorage, AK Anchorage, AK Hanover, NH Boise, ID Portland, OR Eagle River, AK Anchorage, AK Fairbanks, AK Greenbelt, MD Tacoma, WA Fairbanks, AK Fairbanks, AK Fairbanks, AK Saskatoon, Canada Greenbelt, MD Cambridge, MA Name William D. Harrison Larry Hinzman Mark Inghram Peter Jordan Douglas L. Kane Bev D. Kay Esko Kuusisto Robert Lamke Jacqueline D. LaFerriere Stephen Mack David H. Male Eric A. Marchegiani Philip Marsh Mary Maurer Larry Mayo Mark Meier John M. Miller Woodruff Miller James A. Munter Gordon Nelson Jerry Nibler Mark Oswood Bruce Parks Eugene L. Peck Lawrence A. Peterson Brent Petrie William A. Petrik Steven R. Predmore Albert Rango David A. Robinson Nigel Roulet Larry Rundquist Henry S. Santeford Bernard A. Shafer Hung Tao Shen David A. Sherstone Charles W. Slaughter Heinz G. Stefan Jean Stein Robert Van Everdingen Patrick J. Webber Phyllis Weber Peter J. Williams Ming-Ko Woo Sheri Woo Yaohuang Wu John Zarling Chester Zenone Affiliation University of Alaska-Fairbanks University of Alaska-Fairbanks Alaska Dept. of Natural Resources/DGGS Consulting Services in Hydrology and Terrain Sciences University of Alaska-Fairbanks University of Guelph National Board of Waters U.S. Geological Survey /WRD University of Alaska-Fairbanks Alaska Dept. of Natural Resources/DGGS University of Saskatchewan Alaska Power Authority National Hydrology Research Institute Alaska Dept. of Natural Resources/DGGS U.S. Geological Survey/WRD University of Colorado University of Alaska-Fairbanks Brigham Young University Alaska Dept. of Natural ResourcesjDGGS U.S. Geological Survey National Weather Service University of Alaska-Fairbanks U.S. Geological Survey /WRD HYDE X L. A. Peterson and Associates Alaska Power Authority Alaska Dept. of Natural Resources/DGGS U.S. Army Corps of Engineers U.S.D.A./ ARS/BARC-WEST Columbia University York University Entrix, Inc. Michigan Technological University U.S.D.A., Soil Conservation Service Clarkson University Inuvik Scientific Resource Centre U.S.D.A., Institute of Northern Forestry University of Minnesota Laval University Environment Canada, NHRI University of Colorado Alaska Dept. of Fish and Game Carleton University McMaster University U.S.D.A. Forest Service Harza Engineering Co. University of Alaska-Fairbanks U.S. Geological Survey /WRD vi City /State/County Fairbanks, AK Fairbanks, AK Eagle River, AK Vancouver, Canada Fairbanks, AK Guelph, Canada Helsinki, Finland Anchorage, AK Fairbanks, AK Fairbanks, AK Saskatoon, Canada Anchorage, AK Ottawa, Canada Eagle River, AK Fairbanks, AK Boulder, CO Fairbanks, AK Provo, UT Eagle River, AK Anchorage, AK Anchorage, AK Fairbanks, AK Reston, VA Fairfax, VA Fairbanks, AK Anchorage, AK Eagle River, AK Buffalo, NY Beltsville, MD Palisades, NY Downsview, Canada Anchorage, AK Houghton, MI Portland, OR Potsdam, NY Inuvik, Canada Fairbanks, AK Minneapolis, MN Quebec City, Canada Calgary, Canada Boulder, CO Fairbanks, AK Ottawa, Canada Hamilton, Canada Berkeley, CA Chicago, IL Fairbanks, AK Anchorage, AK TABLE OF CONTENTS RESERVOIR AND LAKE LEVEL PROCESSES The Susitna Hydroelectric Project Simulation of Reservoir Operation -Yaohuang Wu, Joel I. Feinstein, and Eugene J. Gemperline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Reservoir Operations Planning in Snowmelt Runoff Regimes Based on Simple Rule Curves -B. A. Shafer, P. E. Farnes, K. C. Jones, J. K. Marron, and F. D. Theurer ........................ 13 Modelling Water Levels for a Lake in the Mackenzie Delta -P. Marsh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Short-Wave Heating of Lake Surface Water Under a Candled Ice Cover -J. P. Gosink and J. D. LaPerriere ....................................................... 31 Hydrothermal Modeling of Reservoirs in Cold Regions: Status and Research Needs -Donald R. F. Harleman ............................................................... 39 WATER, SNOW AND ICE MANAGEMENT Watershed Test of a Snow Fence to Increase Streamflow: Preliminary Results -Ronald D. Tabler and David L. Sturges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Survey of Experience in Operating Hydroelectric Projects in Cold Regions -Eugene J. Gemperline, DavidS. Louie, and H. Wayne Coleman ............................... 63 Hydrology and Hydraulic Studies for Licensing of the Susitna Hydroelectric Project -Eugene J. Gemperline ................................................................ 73 Ice Jam Flooding-Evolution of New York State's Involvement -Russell E. Wege ..................................................................... 87 Hydrological and Ecological Processes in a Colorado, Rocky Mountain Wetland: Case Study -Edward W. Rovey, Catherine Krcieger-Rovey, and David J. Cooper ............................. 93 Seasonal Snow and Aufeis in Alaska's Taiga -C. W. Slaughter and C. S. Benson ...................................................... 101 INSTRUMENTATION AND DATA COLLECTION Water Redistribution in Partially Frozen Soil by Thermal Neutron Radiography -Michael A. Clark, Dr. Roger J. Kettle, and Giles D'Souza ................................... 113 The Development and Use of "Hot-Wire" and Conductivity Type Ice Measurement Gauges for Determination of Ice Thickness in Arctic Rivers -David A. Sherstone, Terry D. Prowse, and Harry Gross ..................................... 121 vii Recent Developments in Hydrologic Instrumentation -Vito J. Latkovich and James C. Futrell II ............................................... 131 Problems Encountered and Methods Used in the U.S. Geological Survey for the Collection of Streamflow Data Under Ice Cover -Ernest D. Cobb and Bruce Parks ....................................................... 135 Simplified Method of Measuring Stream Slope -Jacqueline D. LaPerriere and Donald C. Martin ........................................... 143 WATER QUALITY Water Quality-Discharge Relationships in the Yukon River Basin, Canada -Paul H. Whitfield and W. G. Whitley ..................................................... 149 The Role of Snowcover on Diurnal Nitrate Concentration Patterns in Streamflow from a Forested Watershed in the Sierra Nevada, Nevada, USA -Jonathan J. Rhodes, C. M. Skau, and D. L. Greenlee ....................................... 157 Reservoir Water Quality Simulation in Cold Regions -C. Y. Wei and P. F. Hamblin .......................................................... 167 Trophic Level Responses to Glacial Meltwater Intrusion in Alaskan Lakes -J. P. Koenings, R. D. Burkett, Gary B. Kyle, Jim A. Edmundson, and John M. Edmundson ........ 179 Deep-Lying Chlorophyll Maxima at Big Lake: Implications for Trophic State Classification of Alaskan Lakes -Paul F. Woods ...................................... : .............................. 195 Factors Influencing the Quality of Snow Precipitation and Snow Throughfall at a Sierra Nevada Site -Sheri Woo and Neil Berg ............................................................. 201 POSTER SESSION Primary Production, Chlorophyll, and Nutrients in Horseshoe Lake, Point MacKenzie, Alaska -Paul F. Woods and Timothy G. Rowe ................................................... 213 Water Quality of Abandoned Mine Runoff: A Case Study of Alaskan Sites -David B. Pott, Robert E. Lindsay, and Nicholas Pansic ..................................... 221 Thawing of Ground Frost on a Drained and Undrained Boreal Wetland Site -L. E. Swanson and R. L. Rothwell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231 Probability Distributions of Rain on Seasonally Frozen Soils -John F. Zuzel ..................................................................... 237 Evidence of Groundwater Recharge Through Frozen Soils at Anchorage, Alaska -James A. Munter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 Residential Well Development of a Low Permeability Bedrock Flow System -William A. Petrik ................................................................... 253 Hydrologic Monitoring of Subsurface Flow and Groundwater Recharge in a Mountain Watershed -Michael E. Campana and Richard L. Boone Discharge Under an Ice Cover -Henry S. Santeford and George R. Alger viii 263 275 Hydrology of Two Subarctic Watersheds -Robert E. Gieck, Jr. and Douglas L. Kane ............................................... 283 The Water Balance of the Upper Kolyma Basin -V. K. Panfilova . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293 Water Balance and Runoff Analysis at a Small Watershed During the Snow-Melting Season -H. Motoyama, D. Kobayashi, and K. Kojima . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297 Estimations of Snowmelting Rate in a Small Experimental Site -N. Ishikawa, H. Motoyama, and K. Kojima .............................................. 305 A Methodology for Estimating Design Peak Flows for Yukon Territory -J. Richard Janowicz ................................................................. 313 Effects of Seasonally Frozen Ground in Snowmelt Modeling -Knut Sand and Douglas L. Kane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321 Some Aspects of Glacier Hydrology in the Upper Susitna and Maclaren River Basins, Alaska -Theodore S. Clarke, Douglas Johnson, and William D. Harrison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329 Regional Distribution of Stream Icings in Alaska -K. G. Dean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339 Estimation of Glacier Meltwater Hydrographs -David Bjerklie and Robert Carlson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345 PRECIPITATION-SNOW PACK-SOIL PROCESSES Snow Surface Strength and the Efficiency of Relocation by Wind -R. A. Schmidt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355 Water Flow Rates, Porosity, and Permeability in Snowpacks in the Central Sierra Nevada -Bruce J. McGurk and Richard C. Kattelmann ............................................. 359 In Situ Electrical Measurements of Snow Wetness in a Deep Snowpack in the Sierra Nevada Snow Zone of California -James A. Bergman .................................................................. 367 Measurements of Snow Layer Water Retention -Richard Kattelmann ................................................................. 377 Precipitation Measured by Dual Gages, Wyoming-Shielded Gages, and in a Forest Opening -David L. Sturges ................................................................... 387 The Mass Balance of Snow Cover in the Accumulation and Ablation Periods -Esko Kuusisto ..................................................................... 397 CHANNEL HYDRAULICS AND MORPHOLOGY Erosion Control for Placer Mining -Larry A. Rundquist and N. Elizabeth Bradley 407 Riverbank Erosion Processes of the Yukon River at Galena, Alaska -William S. Ashton and Stephen R. Bredthauer ............................................ 415 ix SNOWMELT RUNOFF Modelling Snowmelt Infiltration and Runoff in a Prairie Environment -D. M. Gray, R. J. Granger, and P. G. Landine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427 Using Real-Time (SNOTEL) Data in the NWSRFS Model -Keith R. Cooley . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439 Theoretical Basis and Performance Evaluation of Current Snowmelt-Runoff Simulation Models -T. W. Tesche . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449 Recent Developments in Snowmelt-Runoff Simulation -Sten Bergstrom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461 GLACIER HYDROLOGY The Role of Glacierized Basins in Alaskan Hydrology -C. Benson, W. Harrison, J. Gosink, S. Bowling, L. Mayo, and D. Trabant ....................... 471 Glacier-Climate Research for Planning Hydropower in Greenland -Roger J. Braithwaite and Ole B. Olesen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485 A Forecast Procedure for Jokulhlaups on Snow River in Southcentral Alaska -David L. Chapman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491 Suspended Sediment Budget of a Glacier-Fed Lake-Eklutna Lake, Alaska -Jeffrey H. Coffin and William S. Ashton ................................................ 501 Annual Runoff Rate from Glaciers in Alaska; A Model Using the Altitude of Glacier Mass Balance Equilibrium -Lawrence R. Mayo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509 REMOTE SENSING Seasonal and Interannual Observations and Modeling of the Snowpack on the Arctic Coastal Plain of Alaska Using Satellite Data -Dorothy K. Hall, Alfred T. C. Chang, and James L. Foster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521 Operational Demonstration of Monitoring Snowpack Conditions Utilizing Digital Geostationary Satellite Data on an Interactive Computer System -Milan W. Allen and Frederick R. Mosher . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 531 Applying a Snowmelt-Runoff Model Which Utilizes Landsat Data in Utah's Wasatch Mountains -Woodruff Miller .................................................................... 541 Initiation of Spring Snowmelt Over Arctic Lands -David A. Robinson ................................................................. 547 RIVER ICE HYDRAULICS Forecasting the Effects on River Ice Due to the Proposed Susitna Hydroelectric Project -Ned W. Paschke and H. W. Coleman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557 X A Structure to Control Ice Formation and Ice Jam Flooding on Cazenovia Creek, New York -Steven R. Predmore ................................................................. 565 Freezeup Processes Along the Susitna River, Alaska -Stephen R. Bredthauer and G. Carl Schoch .............................................. 573 Growth and Decay of River Ice Covers -Hung Tao Shen and A. M. Wasantha Lal 583 Ice Jams in Regulated Rivers in Norway, Experiences and Predictions -Randi Pytte Asvall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593 Hydrologic Aspects of Ice Jams -Darryl Calkins ..................................................................... 603 Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 611 xi RESERVOIR AND LAKE LEVEL PROCESSES JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 THE SUSITNA HYDROELECTRIC PROJECT SIMULATION OF RESERVOIR OPERATION Yaohuang Wu, Joel I. Feinstein, and Eugene J. Gemperlinel ABSTRACT: This paper presents the general concept and methodology used in the simu- lation of reservoir operation, which played an important role in the study of the Susitna Hydroelectric Project. The objective of the simulation was to find optimum operation rules which would meet projected energy requirements of the Alaska Rail belt, while at the same time satisfying flow regimes which would maintain habitat for resident and anadromous fish. Computer models were used for the simulation of reservoir operation on a monthly, weekly, and hourly basis, using streamflow records of 34 years. The results of the simulation allowed the selection of a preferred flow regime which could meet the projected energy requirements and also provide no net loss of habitat for the fish. (KEY TERMS: reservoir operation modeling; rule curve; operating guide.) INTRODUCTION The proposed Susitna Project consists of two tandem reservoirs on 'the Susitna River. The two proposed dam sites are the Watana site, a rockfill dam to be located at river mile 184 of the Susitna River, and the Devil Canyon site, a concrete arch dam 32 miles downstream from the Watana site as shown in Figure 1. The project would operate by storing the high natural flows in the summer for release during low flow periods in the winter when the energy demand is high. This alteration of the natural flow pattern would affect the quantity and availability of spawning, incubating, and rearing habitat for fish, Figure 1. Susitna Project Location Map Respectively, Principal Engineer, DeLeuw, Cather and Company, 525 Monroe Street, IL 60606 (formerly Senior Power Planning Engineer of Harza Engineering Company); Power Planning Engineer, Harza Engineering Company, 150 South Wacker Drive, Chicago, IL 60606; and Manager of Hydrologic and Hydraulic Studies, Harza-Ebasco Susitna Joint Venture, 711 H. Street, Anchorage, AK 99501 3 primarily in the middle reach of the river. The impounding of water in the reservoirs and the alteration in flow patterns would also change water quality parameters associated with mainstem flow such as water temperature, turbidity, and suspended sediment. TI1e simulation of water temperature, ice formation, and sus- pended sediment were conducted in a series of separate studies using the reservoir water levels and discharges from the reservoir operation study. To mitigate the impacts on chum salmon spawning and incubation in side sloughs and chinook salmon rearing in side channels, eight different flow regimes were developed for evaluation (Alaska Power Authority, 1985). Each of the flow regimes was designed to provide a given amount of habitat for the fish species. Each flow regime consists of a series of weekly maximum and minimum flows through a year keyed to chum and chinook salmon life cycles. The maximum or minimum flows were specified at the Gold Creek station, which is located 15 miles downstream of the Devil Canyon site. Figure 2 shows the E-VI flow regime which was selected as the preferred alternative. Minimum summer flow requirements would provide flow stability and would maintain a minimun watered area for salmon habitat. Maximum winter requirements would provide flow stability and would minimize the potential for winter water levels to overtop side slough habitats af feet ing incubating and rearing chum salmon. oL-~--~~L-~--~--------~------~~--~ FMA M J AS QND MONTHS Figure 2. Environmental Flow Requirement, Flow Regime E -VI 4 The evaluation of the flow regimes was carried out by estimating the total cost to meet the projected Railbelt energy demand. These include capital and operating costs of the Susitna Hydroelec- tric Project, other generating facilities, and any mitigation measures required to meet the objective of no net loss of habitat value. Among the mitigation measures included in the evaluation of the alternative project flow regime were hatcheries and multi-level intakes for temperature and sediment control. Since the maximum and minimum flow constraints of the alternative flow regimes would restrict the seasonal distribution of Susitna energy. production, the construc- tion and operation of additional power plants to meet the system demand were also considered. RESERVOIR OPERATION SIMULATION Reservoir operation models simulate the reservoir storage, power generation, turbine discharge, valve release, and flood release as a function of time based on reservoir and power plant characteris- tics, power demand distribution, and envi- ronmental constraints. Cone valves may operate at each dam to satisfy an instream flow requirement or to keep the water surface elevation at the normal maximum level without having to use the spillway. These simulations are normally undertaken in two parts; long-term simulation and short-term simulation (Dondi and Schaffe, 1983). The long-term simulation for the Susitna Project uses a monthly program and a weekly program for simulating the operation for 34 years of streamflow record. The monthly program was used to determine the overall trend, while the weekly program was used for refinement of operation rules and to understand the behavior of the reservoirs and flows during critical periods. The short-term simulation used an hourly program to simulate the operation over a week, using ·the output from the weekly simulations as input data. The hourly program was designed for simulation of hourly generation to meet the daily peak and off- peak loads. The monthly operation used a single rule curve as an operation guide to esti- mate the annual energy production and to satisfy the monthly instream flow require- ments. A rule curve indicates the desired reservoir water level in different months. The weekly operation program used an operation guide for seasonal adjustment of flows which produce a series of reservoir outflows with gradual changes. Operation guides consist of a series of rule curves to control the reservoir outflow. The hourly operation program used an hourly load curve as the upper limit of possible generation. The load curve was based on actual hourly load data and a load fore- cast. This program tested how well the energy obtained from the long tenn analy- ses could fit the hourly load curve, sub- ject to environmental restrictions on the daily and hourly flow changes. Power and energy production from the monthly simulation of the Susitna Project was used in the Railbelt expansion plan- ning studies, which in turn was used in both the economic and financial analyses. Although monthly simulations were suffi- cient for these studies, they were not adequate to estimate environmental effects. Therefore, a simulation with a weekly time step was needed to generate input data for subsequent computer models used in the environmental impact studies. Other environmental studies required hour- ly discharge to estimate river stage fluc- tuations. The monthly program was originally developed by Acres American for the Susitna feasibility study (Alaska Power Authority, 1982) and later improved by Harza-Ebasco Susitna Joint Venture. The weekly program was developed by Harza- Ebasco by using parts of the monthly pro- gram. The hourly operation program was developed by Harza-Ebasco. All of the programs were written in Fortran IV. The Susitna project was scheduled to be built in three stages. First an initial Watana Project would be developed, fol- lowed by construct ion of Devil Canyon downstream. Finally, Watana would be raised to its ultimate height. The full reservoir areas for the low and high dams at Watana would be 19,900 and 38,000 acres respectively, and 7, 800 acres at Devil Canyon. Because of the large reservoir surface area at Watana, release of a large quantity of water would cause a relatively small change in the project head. In contrast, Devil Canyon would have a small surface area, and \lould hence lose consid- 5 erably more head for the same volume of release. Consequently, the reservoir operation methodology attempted to keep the Devil Canyon reservoir close to its normal maximum operating level while using Watana 's storage to provide the necessary seasonal regulation. The ref ore, the modeling effort in both the single and double reservoir operation in monthly and weekly simulation was focused on the Watana operation. The operation levels of the reservoirs for the various stages are shown on Table 1. Table 1 OPERATION LEVELS Ehviron- Normal mental Nominal Min-Haxi-Sur-Plant On- imnm mum charge Capac-line Level Level Level ity Date (ft) (ft) (ft) (MW) Watana Low Dam 1850 2000 2014 440 1999 Watana High Dam 2065 2185 2193 1ll0 2005 Devil Canyon 1405 1455 1455 680 2012 MONTHLY OPERATION MODEL Monthly reservoir simulations were carried out to optimize project energy production subject to environmental flow requirements. The rule curve which would optimize energy production was determined by trial and error. The optimal energy production is a function both of the firm energy and total energy. Figure 3 shows a sample rule curve of the Watana reservoir. During the simulation in each time step, the reservoir release has to satisfy the firm energy and environmental minimum flow requirement. If the end-of-month water surface elevation is lower than the corresponding rule curve elevation, only firm energy is produced and no additional water is released. If the end-of-month water surface elevation is higher than the rule curve elevation, the water stored between rhese two elevations is released to generate secondary energy up to the system energy requirement. 2050r-----------------------------------------~ z 2000 0 f= <( > w _j ~ 1950 - u <( u.. a: ::;) (j) ~ 1900 1- <( ~ DRAWDOWN PERIOD F M A FILLING PERIOD DRAWDOWN PERIOD NORMAL MAX. El. 2000' MIN. El. 1850' M A S 0 N D MONTHS Figure 3. Example Rule Curve For Watana Operation, Year 2004 The dry season is from October to April and the wet season from Hay to September, as shown in Figure 4. The rule curve elevation at the end of September is at the nonnal maximum pool elevation because the reservoir is expected to be full at 30r-------------------------------------------, u D g 20 ::: (j) s 0 _j u.. _j <( DRY SEASON WET SEASON DRY SEASON 34-YEAR AVERAGE ,r--NATURALFLOWSAT THE WATANA DAMSITE ~10- 1- <( z F M A M MONTHS Figure 4. Natural Flows at Watana A s 0 N 0 6 the end of the wet season. In contrast, the rule curve elevation at the end of April is at its minimum because the draw- down is required for firm energy produc- tion in the dry season. The higher the minimum rule curve elevation (at the end of April), the greater the firm energy product ion. This is because the reservoir levels would be kept relatively higher and the storage available for firm energy would be more in the drought period. Alternatively, the lower the rule curve, the greater the total energy production. This is because there would be more active storage for flow regulation and less amount of spills on a long term basis. Various sets of rule curve elevations with various minimum elevations will give diff- erent values of firm energy and total energy. The acceptable m1n1muo rule curve elevation for the Susi tna Project was selected based on an operation in which the increase of total energy is about one percent when lowering the minimum eleva- tion by five feet. Once the maximum and minimum rule curve elevations were determined, the rest of the rule curve elevations were detennined by a trial and error procedure to obtain an acceptable distribution of energy through the year. The operational strate- gy was to capture additional economic benefits through adjustments of the Susitna generation by leaving the residual thermal generation during each of two periods, the summer filling period and the winter drawdown period. As stated above, the reservoir would be almost full at the end of September and would be at the low- est levels at the end of April. There- fore, Susitna energy distribution during the filling period, May to September, and during the drawdown period, October to April, could be varied as a function of reservoir water surface variation without reducing total project energy production. It was assumed that it would be more economical to provide thermal energy by running the least-cost thermal units throughout a whole period rather than running them for part of the period along with other less efficient units. Also the system was assumed to be more reliable if the thermal requirement would be about the same from month to month. This would mean that investment in additional thermal capacity could be delayed as long as pos- sible. Therefore, the energy distribution was adjusted so that the Susitna energy production would maintain constant thermal generation in both the filling period and the drawdown period as much as possible. The analysis of the project's benefits was based on its ability to meet energy requirements. These requirements were the total projected Railbelt system energy demand minus the energy product ion of existing hydroelectric facilities. Figure 5 shows the monthly distribution of energy requirements, and also indicates how the Susitna energy would be distri- buted throughout the year. 700r-----------------------------------------, 600 -500 0 ~ ~ 400 s 0 300 ~ w ~ 200 100 THERMAL GENERATION (2140 GWhl F M A ENERGY REQUIREMENT (4570 GWh) SUSITNA GENERATION (2430 GWh) M A MONTHS S 0 N D Figure 5. Monthly Energy Distribution, Watana, Year 2004 The rule curve approach is predictive because it attempts to achieve an end-of- month elevation which presumes some knowl- edge of the expected reservoir inflow during that period. The operation guide approach, which will be discussed below in conjunction with the weekly program, is non-predictive because it specifies a discharge rate through the powerhouse based on the reservoir elevation at the beginning of the period. The rule curve approach is easy to apply for simulation but can be operationally difficult to achieve, because reservoir inflows are difficult to accurately forecast. The operation guide approach is more difficult to model, but more closely approximates how the project would actually operate. The two approaches yielded similar power and energy results in many trial runs, so the monthly model (rule curve approach) was used for the economic and financial analyses in select ion of the best scheme. The rule curve approach could not be used in the weekly simulations for the environ- mental studies because the release of the 7 storage above the rule curve elevation for the secondary energy could cause an unrea- listic change of discharge between two consecutive weeks. The operation guide in the weekly simulation was designed to prevent these large changes. WEEKLY OPERATION MODEL The weekly operation model was primari- ly designed for simulation based on the operation guide. An operation guide is composed of a series of rule curves which are used as a guide for determining the turbine discharge in each week during the simulation. An example operation guide is shown on Figure 6. Each guide has two families of rule curves; increasing curves and decreasing curves. Each curve defines the reservoir level at which the power- house discharge should increase or decrease to a specified percentage rate of the expected powerhouse discharge. These specified percentage rates are in 20% intervals. The expected powerhouse dis- charges are a set of weekly discharges which would produce an expected distribu- tion of energy production over a year. The single rule curve operation is based on the target firm energy; however, the operation guide uses 60% of the expected powerhouse discharge as the minimum dis- charge and 140% as the maximum normal discharge. fz~~~---t---r--i---t-~---t~~~~~~~---1 z 0 ~ < ~ 1~o~~~~~---r--,_--~~77~~4r--~~r-~~~ ...I w w ~ ~ 1~~~~~~~~~~~7Y~-+---r--+---~~---i ::> "' a: w ~ 1~~~---+~-F~~~~~---+---r--+---~~---i 3: F M A M A S 0 N 0 MONTHS Figure 6. Operation Guide For Watana Operation, Year 2004 At the beginning of each simulation week, these curves are used to determine whether the current discharge rate is kept the same, increased to the next higher rate, or decreased to the next lower rate. The water surface elevation at the begin- ning of the week is put on the operation guide and compared with the elevations of increasing and decreasing curves. If the water surface elevation is higher than the elevation of the increasing curve of the next higher rate, the discharge is in- creased to that rate. If the water sur- face elevation is lower than the elevation of the decreasing curve of the next lower rate, the discharge is decreased to that rate. Otherwise, the discharge is kept the same. The change of rates in two consecutive weeks is limited to 20%. For example, if the discharge rate is at 100% of the expected powerhouse discharge in the preceding week, the rate may be changed to either 80% or 120% or stay at 100%. Because of the difference in elevation between the increasing curve of the next higher rate and the decreasing curve of the next lower rate, the rate will generally be kept the same for a few weeks. In contrast with the simulation using a single rule curve, an operation guide will give an out flow hydrograph with relatively gradual changes in the discharges. A smooth curve giving the energy requirements throughout the year is shown on Figure 7. The average energy produc- tion of the Susitna Project in the draw- down and filling periods was determined from the monthly simulation. Similar to the monthly modeling, efforts were made to capture the additional economic benefits by leaving the thermal generation constant in both the drawdown period (October to middle May) and the filling period (middle Hay to September). In weekly runs the reservoir level is the lowest in the middle of l'1ay and full in most years at the end of September. The energy pro- duction in these two periods are inex- changeable, but redistributing energy pro- duction within either period does not cause additional valve and spillway releases or flow deficits. Adjustments of energy production within either period by leaving thermal requirements constant was assumed to increase the economic value of the project. Gradual changes of thermal energy requirements at the boundaries of 8 the filling and drawdown periods were considered for a smooth transition of the operation. The resulting weekly distribu- tion of energy production over a year was used for computation of the weekly ex- pected powerhouse discharge. 160~----------------------------------------. 140 ;2120 ENERGY REQUIREMENT (4570 GWh) LJ.J LJ.J ~ 100 _c :s: s 80 >- <.:J a:: 60 ~ ~----------~ LJ.J 40 20 THERMAL GENERATION (2140 GWhl M A M SUSITNA GENERATION (2430 GWhl A S 0 N MONTHS Figure 7. Weekly Energy Distribution, Watana, Year 2004 D Development of an operation guide is an iterative process. An assumed set of rule curves for the ope ration guide were put into simulation initially to find flow deficits resulting from not satisfying the minimum flow constraints or from discharg- ing less than the minimum powerhouse dis- charge ( 60% of expected discharge in the example). The curves were then gradually improved by satisfying these two require- ments through the whole simulation period. The curves were again adjusted to maximize the average energy product ion and improve the energy distribution through the year. A good operation guide should provide: 1) turbine discharges close to the expected powerhouse discharge, 2) discharge rates generally constant for a period of at least several weeks, and 3) average energy production maximized. Figure 8 shows the historical inflow hydrograph at Watana. Figure 9 shows the simulated outflow hydrograph of Watana operation for the load year 2004. The comparison of duration curves between the pre-project and post-project flows at Gold Creek Station is shown on Figure 10. Note that, through flow regulation by Watana, the high flows in summer would be substan- tially reduced and the low flows in the winter would be increased for power generation. The simulated out flows from Watana were, in general, consistent with the expected discharges and rapid changes of discharge during floods. there were no except those 50r-----.-----.-~-,.-----~----~----~----. J----j~-----jJ--j--f1pM!!!AX 0•§6J""'53"".0'-'C""-F,S_ +----+------1 ~40+-----+-----+--+-i+-_,--+-.---+-----+-----~ 0 0 g 53ot--t--+rrt1-~.t~t-~,-+-++1-~-+--+-+---4 ~ ~ z ~ 1- --1-· ~20+H~+;HH~~~~~~~~+r~~~~-14~~~ ~ w ~ w ~ o~~~~--~~--~~~~~--~~~--+-~--4 1950 1955 1960 1965 1970 1975 1980 1985 40 : 0 g 30 0 ;: 0 ~ ~ ... 6 20 ~ 0 > ~ w "' w ~ 10 Figure 8. Watana Inflows YEAR II II II II n1 Ill ill I 1111 I~ l~ ~I JIAIIIIWI IIIII 'II 0 1950 1955 .1\1 ~ I I 1960 I' 'U II 111\UIU I I I I 1965 1970 1975 YEAR Figure 9. Watana Outflows, Stage I, Load Year 2004 HOURLY OPERATION HODEL I 1980 1-- 1985 The hourly program modeled a reservoir operation over a week, using an hourly load curve (for the demand distribution within a week) as the upper limit of possible generation and discharges from 9 40 \ ~ • " g 30 0 = ~ w w a: u :bo 0 <:> 1- <( "' s: 0 -' u. 10 LEGEND: -•--•-PRE·PROJECT FLOWS AT GOLD CREEK STATION -X---X-POST-PROJECT FLOWS AT GOLD CREEK STATION PERCENT, TIME Figure 10. Duration Curves of Flows at Gold Creek the results of weekly simulation as input data. Outflow fluctuations were restrict- ed by maximum hourly and daily variation constraints. The model tests how energy obtained from the long term analyses can fit the hourly variation of demand \vi th environmental constraints. The output was used in river stage fluctuation studies. An example of the output is plot ted as shown in Figures 11 and 12 to aid in understanding the results. Figure 11 contains a plot of generation showing the system demand, the existing hydro genera- tion, and Susitna Project generation. Figure 12 shows the reservoir discharge through the powerhouse and the valves, and the minimum flow required to meet the environmental requirements at Gold Creek. The reservoir outflow constraints limit how the plant can operate in the sys tern. For example, if the daily variation con- straint is very SP.lall, then the plant is essentially base loaded, but if both hour- ly and daily variation constraints are large, the hydro plant can load follow. Load following operation means that the plant can increase or decrease its genera- tion by following the hourly fluctuation of the system demand, and will leave the 1200.-----------,------,----~,-----,,-----,-----, ~ 800 :::; z 0 f- <( a:: 600 w z w l9 w a:: l9 ~ 400 1------+-<( 0 I Z "-f- Ui ::::J I EXISTING HYDRO PEAKING I NON·SUSITNA GENERATION + . ! t -4 ~ -+ ---1. ---------------EXISTING HYDRO ON BASE SUN MON TUE Figure 11. Hourly Power Generation SAT 20.-----~-----.-----,-----,------r-----,-----~ u 0 ~ 12~---1~--~~----~~--~~--~~--~44--~~ w l9 a:: <( I u ~ 8 0 POWERHOUSE DISCHARGE__/ I MINIMUM ENVIRONMENTAL RELEASE i ------·---- ---+----. J_ _____ ( __ ---+ ----- 0~--~----~--~-----L----L---~--~J~ SUN MON TUE WED THU FRI SAT Figure 12. Hourly Reservoir Discharge energy from other sources at constant capacity. Base loaded operation means that the plant generation is constant in principle, but a certain percentage fluctuation may be allowed. Load following operation would provide more 10 capacity value for the project than base loaded operation. Base loaded operation would provide stable flows in the downstream channel. Project operation is currently constrained to be base loaded with allowable variations of 20% in total project discharge within a week, therefore, providing stable flows and minim1z1ng impacts to salmon. Chum and sockeye salmon spawn in sloughs along the river. These sloughs collect sediment and organic matter throughout the normal course of the year, which make it difficult for the fish to spawn and may reduce egg survival. Naturally occurring floods clean out the sloughs and thus provide better habitat. The dams would tend to reduce these naturally occurring floods which may decrease the natural fish habitat. Some of the flow regimes incorporate spikes of flow to create artificial floods to clean the sloughs. The hourly program can also model these spikes as instrearn flow requirements at Gold Creek. Input to the hourly model consists of the initial storage at the beginning of the week and the amount of water to be released during the week (obtained from the results of the weekly simulation), The program generates a curve called a template for use as a guide in simulation of hourly power generation. The first template is equivalent to the hourly sys- tem load minus the existing hydro produc- tion. Turbine release in each time step is determined from the energy requirement of the template. The turbine release is then checked with the flow constraints and the total release is adjusted to satisfy the constraints if there is any violation. After the first iteration with the tem- plate, the template is adjusted according to the ratio of the amount of water to be released and the total outflow in the previous i te rat ion. The simulation is iterated with new templates until the out- flow for the week is equal to the amount of water to be released for the week. The hourly program modeled a single reservoir. When the Devil Canyon reser- voir is in operation, the Watana plant will load-follow and the Devil Canyon plant will be base loaded. Therefore, the release from Devil Canyon would be stable and the variation of discharge on the downstream channel could be easily con- trolled. Because of low flow from Watana in off-peak hours and high flow in peak hours, Devil Canyon will draw down in off-peak hours and fill in peak hours. The maximum drawdown for daily fluctuation at Devil Canyon was estimated at one-half foot. The hourly program as well as the week- ly program have prov1s1ons for flood operation. During a large flood when the reservoir is full, the reservoir inflow could be greater than the sum of turbine and valve capacities. If the spillway is used, nitrogen would be entrained in the water and there would be the potential for nitrogen con centra t ion to exceed tolerable levels. In order to minimize use of the spillway, the reservoir is allowed to surcharge above the normal maximum level up to an environr.1ental sur- charge level. The environmental surcharge for Watana low dam would be 14 ft and that for Watana high dam would be 8 ft. These levels were determined on the basis of avoiding the use of the spillway in a flood of less than a 50-year return peri- ad. The spillway would not be open unless the water surface elevation reaches the environmental surcharge level. In non-flood operation the valves would not release water unless it is necessary for the instream flow requirements. When the water surface elevation is at or above the normal maximum level, the excess water would be released from the valves. As the water starts to surcharge above the normal maximum level in a flood, the total out- flow could be increased hourly at a special flood rate, designed to minimize i1npacts on the fishery from changes in flow and temperature, until the valves are fully open. However, the outflow would never be allowed to be greater than the peak discharge of inflow. As stated pre- viously, if the water surface .elevation reaches the environmental surcharge level, the spillway would be opened for release so that the outflow would be equal to inflow. The falling limb of the outflow hydrograph would also be cons trained by an hourly decreasing rate for flood opera- tion. CONCLUSION The monthly simulation with rule curve operation is simpler and less expensive than the others. It was ef feet ively used ~ the economic analysis of the project. 11 The weekly simulation \·lith the opera- tion guide more closely simulates the dis- charge variations for the studies of envi- ronmental impacts. The operation guide restricts the discharge variation in a specified limit to secure the protection of fishery habitat. The simulation with the weekly model was successfully used for the evaluation of the flow regimes. The hourly simulation was used to test how the energy obtained from the weekly analysis could fit the hourly load curve. It was also used for the study of peaking capacity with respect to the allowable fluctuation of discharge in the downstream channel. Honthly. weekly, and hourly operation models are all indispensable in the study of the Susitna Hydroelectric Project. REF EREJIIC ES Alaska Power Authority, 1985. Alternative Flow Regime, Hydroelectric Project, Volume Report, Document No. 2600. Case E-VI Susitna 1 -Main P.H. Dondi and G. Schaffe, 1983. Simula- tion and Optimization of a Series of Hydro Stations, Water Power and Dam Construction, Nov. 1983. Alaska Power Authority, 1982. Feasibility Report, Susi tna Hydroelectric Project, Volume 1, Ehgineering and Economic Aspects. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 RESERVGIR OPERATIONS PLAKNING IN SNOWMELT RUNOFF REGIMES BASED ON SIMPLE RULE CURVES B.A. Siafer, P.E. Farnes, K.C. Jones, J.K. Marron, and F.D. Theurer] ABSTRACT: Selecting appropriate storage and release rates for reservoirs 10 snowmelt runoff envi- ronments is a prerequisite to sound water rr.anagement. A significant null'her of small impoundments, oper- atf;ci tor single or multiple purpose use 1n the Western U.S., lack adequate management tools to guide tl:is process each year. A metho- dology is presented to use seasonal streamflow volume forecasts issued by the u.s. Soil Conservation Service and N::Jtional Weather Se;vice to improve rr.anagement c;;pabi 1 i ty at many of these reser- vons. The technique involves rcntrating a family of simple rule rurves for each forecast period. Lese curves permit operators to use predicted inflow volume to set t<:get outflow rates that will Pnable them to reach a full reser- voir after passage of the seasonal ~eak. Forecasts at three probabil- i t y 1 eve 1 s h e l p e s t <t b 1 i s h t h e r a 11 g e of 1:ikE'ly seasonal ruGoff events. Tr:e derivation of the curvt-:s 1s presented along with a mathematical algorithm to produce them objec- tively from historical inflow records. The rule curves· provide ill' operational tool useful for developing effective water manage- ment plans /for reservoirs where forecast information is available. Seven reservoir operating plans have been developed and implemented using this procedure in Hontaoa and Oregon. (KEY TEKhS: reservoir management, rule curves, streamflow forecasts, snowmelt runoff.) INTRODUCTION Annual water supply in the Western U.S. is highly variable, often fluctuating between extr~ffies. This inherent natura 1 varia ;J i 1 it) imposed by climate and topograph:; makes it difficult for water managers to plan their operation to 0ptimally use available runoff. Tbe seasonal nature of runoii Bls~) complicates management because usually the bulk of flow occurs in. only a four-month period (April~ July typically) in response to r::elting mountain snowpack.s. This srowmelt component produces from SO to 85 percent of the region's an~ua1 runoff and is therefore the foundation of most water management decision-making. Many water users have found iL necessary to construct storage re- servoirs that enable them to caJ,;ture seasonal flow volumes <:.r;c.i regulate releases consistent v.:itb an,ual demar,ds. There are thou- sacds of these reservoirs of all sizes in the West ranging from only lr:espective1y, .b.A. Shafer, K.C. Jones, and J.K. Marron, Soil Conser- vation Service, )11 J.I.JW Broadway, Portland, Oregon 97209; P.E. Farnes, boil Cor. sr: r v a t i o n S e r v i c e , F e c e r a 1 n u i 1 d i n g , B o z em a n , Non t a n a 5 9 7 1 5 ; an d f' • D • Theurer, Soil Conservation Service, 301 South Ho"es, Ft. Collins, Colorado 80 52 2. 13 a few acre-feet up to millions of acre-feet in storage capacity. Most of the water held in storage in these impoundments is utilized for irrigation, hydropower product- ion, municipal water supply, flood control, fisheries, recreation, or some combination of these uses. Often there is competition among uses for water held in storage, making it important for managers to apply the best analytical tools available to fill and empty reser- voirs in a manner which maximizes beneficial water use and minimizes negative impacts. Inordinately high or low flows downstream from reservoirs caused by uninformed management are to be avoided if possible. Reservoir capacities on snow- melt dominated streams are fre- quently smaller than average annual runoff. As a result, reservoir operators must balance their abil- ity to store and eventually release runoff volumes with expected sea- sonal inflows on a recurring basis. A substantial number of irrigation reservoirs and multi- purpose structures currently lack a formal management plan to guide this process each year. Instead a philosophy of "fill and spill" is adopted by default; this approach sometimes has disastrous conse- quences and does not take advantage of information on current hydro- logic conditions that is readily available. Water supply forecasts based on snowpack, precipitation, temperature, and streamflow are made routinely by the Soil Conser- vation Service (SCS) and National Weather Service to aid reservoir managers in their decision-making. These predictions have not been effectively utilized in many cases. A major factor contributing to this situation was an inability to integrate the forecast informa- tion into a practical and easily applied scheme that could be under- stood by relatively unsophisticated operators. A procedure has been developed to address this deficiency. It provides a means for reservoir operators who depend on snowmelt 14 runoff for all or a significant portion of their inflow to set outflow rates that enable them to meet storage goals based on stream- flow forecasts. The technique is principally oriented toward improv- ing management capability on small reservoirs whose primary function is to supply storage for agricul- tural interests. Reservoir management plans have been developed and implemented on six impoundments in Montana and one 10 Oregon using this procedure. One of these plans on Hebgen Lake 10 southwestern Montana has been in use for.over ten years. It can also be applied to evaluate the feasi- bility of planned storage projects. The methodology involves generating a family of simple rule curves which permit an operator to relate forecasted inflow volumes at three probability levels to target outflow rates. The target outflows are designed to help the operator reach a defined storage level (usually a full reservoir) after passage of the seasonal peak flow. Principles used to construct the curves are presented. A mathema- tical algorithm is described to produce the curves objectively from historical inflow records. A computer program that incorporates the rule curve algorithm, data input, data analysis, and curve plotting is explained. RULE CURVE PRINCIPLES Typically, snowmelt runoff begins on most streams from late March to early May, with the re- cession continuing into July or early August. Peak inflows usually occur between mid-May and early June. Although this flow pattern 1s annually repetitive, there is substantial variability 1n total volume and timing of runoff, making it necessary for reservo1r opera- tors to tailor their operations to prevailing conditions. Examination of runoff hydro- graphs on snowmelt dominated streams revealed that on an indivi- dual watershed there is a discern- ible consistency in hydrograph shape and time distribution for similar seasonal flow volumes. This observation implied that it should be possible to construct a stable relationship between seasonal volume and reservoir outflow settings to achieve a specific storage goal. In this context, seasonal volume serves as an index to hydrograph shape. To accommodate the need to reflect current conditions, relationships can be developed for several flow periods for which forecasts are routinely made; e.g., April-July, May -J u 1 y , J u n e -J u 1 y • Specifically, a relationship is desired for each forecast period to make it possible for an operator to select an outflow setting that would produce a specified storage level if the forecasted runoff actually occurred. Figure 1 illus- ~ 0 .J lL ~ _J 0 > _J <t z 0 (/) <t w (/) RMX -..., __ OUTFLOW (CFS) Figure 1. Conceptualized represen- tation of how seasonal snowmelt runoff at three probability levels--RMX, MP, RMN--can be used with reservoir rule curves to arrive at reservoir outflow rates that meet storage goals. sl,s2··· Ss are storage increments that an operator wishes to achieve during spring runoff. Note ss>s4>s3, etc. 15 trates conceptually how an operator would use the forecast with a set of rule curves derived from an analysis of historical reservoir in f 1 ow s • In a c t u a 1 p r a c t ice , mo s t probable (MP), reasonable minimum (RMN), and reasonable maximum (RMX) forecasts are made corresponding to 50, 90, and 10 percent exceedance probabilities. Entering the graph with these forecast values and moving horizontally to the volume of storage left to fill would produce a range of outflow rates. The rate chosen would be based on an assessment of local operating constraints and magnitude of runoff expected. This concept allows the reservoir manager to fill the reservoir while maintaining fairly constant release rates. Besides determining outflow rates during the main runoff period, early season forecasts and reservoir operating curves can be used to determine desirable storage levels in the reservo1r. When the present storage and forecasted run- off indicate an outflow less than that needed for downstream uses, it would be desirable to 1ncrease storage prior to spring runoff so these needs could be met. The largest storage level that would be desirable could be determined by reading the storage at the inter- section of the forecast and upper range of desirable outflow. RULE CURVE GENERATION Development of operating rule curves depicted 1n figure 1 is unique for each reservoir. The first step to develop rule curves is to obtain a minimum of 10 years of daily reservoir inflow data that includes both high and low runoff seasons. Daily streamflow observa- tions for only the snowmelt period satisfy the minimum data require- ments. However, complete annual records are desirable to detect unusual runoff sequences that affect reservoir operation but are unrelated to snowmelt runoff. For each year, individual storage versus outflow curves are constructed by selecting a series of outflow rates and tabulating the corresponding storage that would result given the seasonal inflow hydrograph. Storage volumes are obtained as the sum of all daily flows greater than or equal to the selected outflow rates. This stipu- lation prevents a drop in stored contents and implies outflow is set equal to inflow when the target re- lease rate is higher than inflow. The storage-outflow data pairs are then plotted and a smooth line fitted to the points either by eye or a least square analytical method. Figure 2 illustrates this procedure for a single year's hydrograph. Setting outflow rates I Ool w ' f-I <( a: 3: 0 _J u. TIME I W I STORAGE-OUTFLOW CURVE (.!) I <t Ss --,--a: I 0 I I ~ Sc --~--:----1 I I o So --;---~ - --:-- --- o ' I OA Os Oc Oo OUTFLOW RATE Figure 2. Storage vs. outflow curve constructed for each year of record by determining how much water could be stored at various release rates (OA,Os,Oc, on). 16 of OA,Os,Oc,On results in storing volumes SA,Ss,Sc,Sn (cross hatched area). In actual practice, many more points than four are generated to fit the storage-outflow curve. Figure 3 shows examples of s1x storage-out- flow curves for a hypothetical stream based on April-July daily inflow data. High, low, and intermediate seasonal volumes are represented. From these relation- ships, operating curves can be created for specific storage levels; i.e., S1,S2,S3,S4. Th~s task is accomplished by generating a series of seasonal volume-outflow rate data pairs for each incremental storage level;, e.g., S1 in figure 3, and fitting a: smooth curve to the points. The process of generating the pairs is to select a desired storage level, and for each year, read the outflow setting required to produce it from the curves of figure 3. Doing so for a storage increment of sl results in outflow rates 01,02···06 corresponding to years 1971-76, respectively. Data pairs consisting of these seasonal volumes and associated outflow rates are plotted and a line fitted to the points either by eye or using a curvilinear least square regression method. Figure 4 shows an operatinl curve for the S1 to store increment derived from the graphs of figure: 3. Successive repetitions of this procedure for other storage increments produce a family of operating curves like those of figure 1. Families of curves for any desired seasonal flow period can be developed by the method outlined. RULE CURVE ALGORITHM Building on this conceptual understanding of the principles ot operating curve development, an; objective, automated technique to: produce the family of curves is: explained. The rule curve' s4 1971 s4 1972 s4-1973 wS3 V=30,000 A-F s3 V= 40,000 A-F s3 V=20,000 A-F (!) <( ~s2 s2 1- CJ) sl---: s, --~---s, So So o, s4 1974 s4 1975 s4r 1976 ~s3 V=25,000 A-F s3 V=37,500 A-F I :r V=15,000 A-F <( 0::: ~s2 s2 CJ) s,k s, I s, • + So I So I 04 05 So Os OUTFLOW OUTFLOW OUTFLOW Figure 3. Six storage vs. outflow curves for a hypothetical stream illustrate how seasonal volume (V) influences shape. To produce the same storage level (Sl) each year, outflow rates are set at Ol,Oz ... 06. algorithm to generate the curves is constrained to be an analytical least square fit of a three- dimensional data set consisting of seasonal inflow volume, outflow rate, and storage level. The regression model relating these variables is: where T = V-S; X = 0; y (0)0.5 Z = (O)(S); (1) V = total seasonal volume inflow to reservoir; S = desired storage level; 0 outflow setting required to obtain desired storage leve 1; and 17 al,az,a3 are regression coefficients. The transformed variables (X,Y,Z) were chosen to meet certain rational and/or observed behavior. The dependent variable, T = V-S, was chosen so that when the outflo~ setting was at or near zero, the predicted storage level would be at or near the inflow volume; i.e. S = V when 0 = 0. The first indepen- dent variable, X = 0, was set as a standard linear regression term. The second independent variable, Y = (0)0. 5, was set to reflect the observed curvature behavior between inflow volume (V) and outfloli setting (0) for a given storage (S). The third independent variable, Z = (S)(O), was chosen to reflect the skewness between the various storage level curves because they are not necessarily parallel. ........ 1-w w lL I w a:: (.) <( 3: 0 ...J LL .....J ::> -, I a:: a.. <( 50....---------------------., 40 30 20 10 06 03 04 01 05 02 OUTFLOW (CFS) Figure 4. A reservoir rule curve for the sl storage level is generated from the graphs of figure 3 by plotting seasonal volume vs. release rate for each year of record. A necessary step in obtaining the rule curve equation is to pro- duce a three-dimensional matrix of seasonal volumes, storage, and out- flow rates. These values are derived from the storage-outflow relationships for individual years (figure 3). A fifth degree least square polynomial regression pro- cedure is used to objectively fit a curve to the storage-outflow data pairs for each year. It takes the form: s = b 1 o + b 2 o2 + b 3 o3 ( 2 ) + b4o4 + b5o5 + b6 where bl, b2, b3, b4, b5, b6 are regression coefficients and S and 0 are as previously defined. It is now possible to employ a Newton iteration technique (Carnahan et 18 al., 1969) to find outflow rates at previously selected storage volumes for each year. This process yields the requisite three-dimensional matrix of data elements that is input to the rule curve algorithm. The coefficients a1, a2, and a3 are found by solving the normal equations dictated by the form of the model in equation 1 (McCuen, 1985). The normal equations are: a1 EX2 + a2 EXY + a3 EXZ = EXT (3) a1 EXY + a2 EY2 + a3 EYZ = EYT (4) a1 EXZ + a2 EYZ + a3 Ez2 = EZT (5) The s u mro at ion ( E ) 1. s c a r r i e d out for p elements. where p 1 (m) n (6) 1 = number of years analyzed; m number of outflow rates chosen; n = number of storage levels chosen. The solution to the system ofl three simultaneous equations is given by finding the values of a1, a2, a3 in the following matrix! representation of the normal equa- tions: Ex2 EXY EXZ EXY Ey2 EYZ al a2 a3 = EXT ( 7) E YT E ZT Gaussian elimination or matrix inversion can be used to solve for a1, a2, and a3 (Carnahan et al., 1969). Having determined the regres sion coefficients, it is now possi- ble to solve either directly or iteratively for any one of the original variables given the other two. In particular, we desire to produce the operating curves for· specific storage levels. This re- quirement can be satisfied by fix- ing the storage level, incrementing through a range of seasonal volumes, and solving for outflow settings at each step using Newton's iteration method. The rule curve algorithm has been subjected to verification tests to insure its mathematical integrity. In addition, validatio~ analyses have been conducted with actual data from various locations throughout the West. ROMP PROGRAM To facilitate SCS field personnel's ability to effectively utilize this procedure, a reservoir operation and management planning (ROMP) computer program was written. The program was designed to integrate data entry, streamflow screening, flow analysis, curve fitting, plotting, and error analysis. The initial vers1.on of ROMP was written in BASIC for Tektronix 4050 series graphics systems because they support high resolution screen and plotter graphical displays and were avail- able in each of the SCS state offices 1.n the Western U.S. A second version has been adapted to run on a Data General MV 8000 mini- computer with graphical output directed to a Tektronix 4105 color terminal. ROMP is fully menu driven making it easy for field personnel to use without extensive training. ROMP's architecture is comprised of eight modules that are normally executed sequentially to produce a set of reservo1.r operating curves. Following l.S a list of the modules 1.n the ROMP program: 1. Data Input 2. Hydrograph Plotting 3. Reservoir Inflow Correction 4. Flow Analysis 5. Storage vs. Discharge 6. Rule Curve Equation 7. Rule Curve Error Analysis 8. Rule Curve Plotting RULE CURVE APPLICATION It is instructive to go through an abbreviated example of developing and applying the results of the ROMP program. Middle Creek Reservoir 1.n southwestern Montana 1s used to show how reservo1.r 19 operating curves are generated and how they guide the management decision making process. This reservoir is managed primarily for irrigation and municipal water storage but potential impacts on other interests affect how the project is operated. Middle Creek Reservoir l.S on Hyalite Creek in the Gallatin River Basin. It has a drainage area of 27.4 square miles. The streamflow regime l.S dominated by snowmelt runoff. Reservoir capacity is 8,261 acre-feet. Daily inflow records are only available April- July for a 17-year period (1966-83, 1970 missing). The outlet tunnel will pass 800 cfs when the reser- voir is full. However, considerable downstream erosion occurs at flows 1.n excess of 400 cfs. A flow of 125 cfs l.S necessary to satisfy decreed water rights and irrigation demands. A minimum outflow of 25 cfs l.S required to support downstream fish populations. These operating constraints dictate that the desirable operating range for outflow be between 125 and 400 cfs during the irrigation season. Streamflow forecasts are issued for reservo1.r inflow monthly January through June. Daily reservoir inflow data for the April-July period were entered from the keyboard using the ROMP Data Input module. The Flow Analysis, Storage vs. Discharge, and Rule Curve Equation modules were next executed sequentially to create the following rule curve regression equation for the April- July period: S = [V -7.4247 * 0 -1537 * (8) o0.5]/(l + .0025 * 0) The rule curves shown in figure 5 were generated using equation 8 and the Rule Curve Plotting module. A similar procedure was followed to produce the May-July rule curves shown 1.n figure 6. These two charts used in conjunction with streamflow fore- casts provide the means to make outflow adjustments commensurate with anticipated runoff. Illustra- tions of how the operating curves might be used 1n a high snowpack year and low snowpack year are given for comparison. Low Runoff Year On March 1, assume the reser- V01r has 3,261 acre-feet of water in storage and the most probable (MP) April-July inflow forecast 1s 22,000 acre-feet or 76 percent of average. Average April-July inflow is 29,000 acre-feet. The reason- able minimum forecast (RMN) is 17,500 acre-feet; the reasonable maximum forecast (RMX) is 27,500 acre-feet. Subtracting current storage from the reservoir's capacity leaves 5,000 acre-feet required to fill the reservoir. Entering the April-July operating curves of figure 5 with RMN, MP, and RMX and moving horizontally to the 5,000 acre-feet to store curve yields release rates of 60, 110, -,_ Ill Ill II. I Ill ex: u ~ ~ ::;) ..J 0 ,.. ... ..J ::;) .., ..J 40000 35000 30000 25000 20000 15000 10000 APRIL -JULY RESERVOIR OPERATING CURVES "IDDLE CREEK RESERVOIR a: 5000 A.. < oL-~----------~----~--~~~_.--~ 0 50 1 00 1 50 200 250 300 350 400 450 500 OUTFLOW (CfS) Figure 5. Reservoir rule curves for Middle Creek Reservoir, Montana, applicable for April-July runoff period. Curves were developed from 17 years data. 20 and 170 cfs, respectively. Based on these figures, a logical decision would be to store as much water as possible and to only release enough water to support a viable fishery. By April 1, reservoir storage was increased by 200 acre-feet, thus requiring 4,800 acre-feet to fill the reservoir. The April-July forecast has been raised slightly to 18,000 acre-feet for RMN, 22,500 for MP, and 2 7, 000 for RMX. Entering the April-July operating curves with these values indicates outflow settings of 67, 112, and 167 cfs, respectively. These figures show that only with above normal subsequent precipitation will the reservo1.r fill and meet withdrawal demands. Both outflow settings found with RMN and MP inflows are well below the minimum desirable outflow of 125 cfs necessary to satisfy downstream water rights. At this point, the decision would probably be made to continue possible received. 40000 35000 storing until May 1 as much forecasts ~y -JULY RESERVOIR OPERATING CURVES "IDDLE CREEK RESERVOIR as are ;:: Ill 30000 Ill II. I Ill ex: 25000 u ~ !I! 20000 ::;) ..J 0 ,.. ... ..J ~ ... 5000 < ll: 0 0 50 1 00 1 50 200 250 300 350 400 450 500 OUTFLOW (CFSl Figure 6. Reservoir rule curves for Middle Creek Reservoir, Montana, applicable for May-July runoff period. During April, runoff was sufficient to provide an additional 500 acre-feet of storage leaving 4,300 acre-feet necessary to fill the reservoir. The May-July MP has been raised to 80 percent of normal or 21,500 acre-feet. The May-July average is 2 7, 000 acre-feet. RMN is 18,000 acre-feet, and RMX is 25,000 acre-feet. Using May-July reservo1r operating curves of figure 6, the outflows for RMN, MP, and RMX would be 75, 110, and 158 cfs, respectively. The decision would probably be made to maintain releases at the minimum levels dictated by fishery considerations and continue storing as much water as possible until irrigation and municipal withdrawal demands exceed inflow. With the expected low runoff, the probability of generating channel damaging flows in excess of 400 cfs is remote. In such a low runoff year, when reservoir storage is low, demand for irrigation water may be greater than the outflow rates which would permit filling the reservo1r. Water users must then decide whether they would rather use the water for early irrigation or to put it into storage for later delivery when crop consumptive demands are highest. HIGH RUNOFF YEAR On March 1, assume 3, 000 acre-feet are needed to fill the reservoir and the April-July MP is 35,000 acre-feet or 121 percent of average. RMN and RMX are predicted to be 30,000 and 40,000 acre-feet, respectively. Entering figure 5 with these predictions yields out- flows of 257, 300, and 410 cfs. These outflow rates are in the range that raises concern about the possibility of being forced into r e 1 e as e s h i g he r t h an a r e de s i r a b 1 e. later 1n the season. They also support the contention that there is likely to be plenty of water to fill the reservo1r and satisfy all downstream requirements. The decision will probably be made to 21 hold the reservoir at the same level or decrease it a little by setting outflow equal to, or slightly greater than, inflow and wait until the April 1 forecasts are received. When April 1 comes, the reser- voir storage has dropped to a level requiring 3,300 acre-feet of water to fill. Streamflow projections of April-July runoff are for RMN, MP, and RMX values of 31,500, 35,000, and 39,500 acre-feet, respective- ly. Entering figure 5 with these numbers translates into potential release rates of 275, 335, and 395 c f s corresponding to RM N , M P , and RMX. The dec is ion wou 1 d 1 i kely be made to set outflow rates 10 to 15 cfs above inflow rates to begin reducing reservoir storage before runoff begins. Any significant increase in storage or leaving the reservo1r at its present level could create a potentially hazard- ous situation if abnormally high precipitation and temperatures are experienced 1n the next several months. On May 1, the MP 1s for a May-July flow of 130 percent of average or 35,000 acre-feet. RMN and RMX are for flows of 32,000 and 38,000 acre-feet, respectively. Reservoir storage has been reduced and there are 4,000 acre-feet of available storage. Using these figures in the May-July reservoir operating curves (figure 6) gives release rates ranging from 260 cfs for RMN to 351 cfs for RMX with an intermediate value of 305 cfs for MP. These numbers continue to indicate a heavy runoff but with a lessening probability that release rates will have to approach or exceed 400 cfs, the threshold at which damage occurs at downstream reaches in the channel. The deci- sion would probably be made to set outflow rates about 300 cfs and continue to carefully monitor weather temperatures during the month for unusually warm conditions or heavy precipitation. If either of these events occurred, it would be appropriate to raise outflows to levels between 350 and 400 cfs. DISCUSSION Melting winter snowpack and spring and early summer ra1.n are the primary sources of water in many locations in the West. Flows from these sources into a reservoir vary significantly from year to year and day to day within a given year depending on a number of factors. These include volume of water accumulated in the winter's snowpack, basin soil wetness, areal extent of snow cover, temperature conditions during the main snowmelt period, and the amount and rate of spring and early summer precipita- tion. Reservoir managers must assess these factors and the uncertainty they impose in terms of risks associated with storing too much or too little water. To the degree that operators can reduce uncertainty about future runoff, they incrementally reduce their exposure to risk. An analytical tool to help them define the magnitude and probability of runoff events several months in advance and the consequent implications for project regulation is highly desirable. Ideally, reservoir operation would regulate outflow to minimize spilling excess water, satisfy senior downstream water rights, m1.n1.m1.ze eros1.on and downstream flooding, provide sufficient water for recreation, fisheries, and wildlife, and enable the reservoir to be full near the end of the high water period. Sometimes heavy snowfall or ra1.n occur 1.n late spring and may prevent achieving ideal outflow conditions each year to satify all these requirements. However, most of the time it l.S possible to successfully base each year's reservo1.r management on expected runoff conditions by using water supply predictions. During heavy snowpack years, downstream flooding can be reduced; in low snowpack years, effects of low runoff can be somewhat moderated. 22 SUMMARY A procedure has been developed and 1.s being used operationally to assist reservoir operators to manage their facilities with sea- sonal forecasts of snowmelt run- off. Informed decisions can be made based on the probability of occurrence of seasonal flow volumes. The methodology· employs reservoir operating rule curves that are generated from historical inflow data. The derivation of the operating curves has been explained and a mathematical algorithm to calculate · them presented. An integrated, menu-driven computer program called ROMP has been developed to aid technical specialists in the development of these curves. Examples of how the curves might be used as guides 1.n high and low runoff years are offered. Given reasonably accurate streamflow forecasts and knowing the amount of available storage space, a reservoir operator can use the family of operating curves to analyze available options objec- tively to help reduce uncertainty and manage exposure to risk. Literature Cited Carnahan, B., H.A. Luther, and J.O. Wilkes, 1969. Applied Numerical Methods. John Wiley & Sons, New York, N.Y. McCuen, Richard H., 1985. Statistical Methods for Engineers. Englewood, Prentice-Hall, N.J. pp. 221. Inc., JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 MODELLING WATER LEVELS FOR A LAKE IN THE MACKENZIE DELTA P. lfarshl ABSTRACT: A detailed hydrologic study of a perched lake in the Mackenzie Delta was carr\ed out during the sl.llTIIer of 1985. The hydrologic regime of thls lake may be divided into three d\stinct periods: flooding, discharge, and evaporation. Over the sll11ller of 1985 the 1 ake ex per\ enced a sma 11 negative water balance. However, 1f the lake was not f1 ooded, 1t would have experienced a severe negative ba 1 ance s i nee evaporation exceeded predpitation. Th\s type of lake is probably dependent on flooding to keep water levels at the present level. A si100lation model, whlch accurately predicted water level over the Sl.llTIIer period, could be used to predict the flooding frequency required to maintain lake levels. (KEY TERMS: lake hydrology; water balance; Mackenzie Delta.) INTRODUCTION The Mackenzie Delta is dominated by the myriad of 1 akes which occur throughout 1t s entire 65 by 180 ~ extent. These lakes play a significant role in the delta ecosystem. They affect the distribution of permafrost (Smlth, 1976), support large populations of flsh, mcmnals, and waterfowl (Gill, 1973; Peterson, Allison, and Kabzems, 1981), and provide storage for water, sediment, and pollutants. In spite of the il!llortance of these lakes to the hydrology, geomorphology, and wildlife of the Mackenzie Delta, few hydrologic studies have been conducted to date. Mackay (1974) discussed the origin of Mackenzie Delta lakes and their general hydrologic characteristics, and Bigras (1985) has considered 1 ake 1 eve 1 s in the eastern sector of the Mackenzie Delta. Changes in the natural environment of these delta 1 akes, though not a prob 1 em at present, may become an issue as development proceeds in the Mackenzie Valley. A major concern is that pollutants introduced into the Mackenzie River upstream of or within the delta, will be distributed into Mackenzie Delta lakes by the vast distr\butary channel network. A second concern is that further flow regulation in the Mackenzie River system may have an effect on lake levels, as has already occurred in the Peace-Athabasca Delta (Peace-Athabasca Delta Group, 1972) with the resulting consequences to permafrost and wildlife habitat (Gill, 1973). It is the purpose of this paper to describe the hydrological processes operating in a typical perched lake in the Mackenzie Delta. In addition a si100lation model for predicting lake level will be presented. MACKENZIE DELTA LAKES Lakes dominate the physical landscape of the Mackenzie Delta both in sheer number and area covered. An analysis of aer\al photographs for a typical 70 km2 area southwest of Inuvik, N.W. T., showed that 282 1 akes covered 50'1. of the surveyed area. Large river channels covered an additional ~ of the area, and land the remaining 47\. Two conmon types of lakes, each with different hydrologic conditions, are found in this area. The primary factor controlling the lake type is the sill elevation between the lake and main channel . Even though only two lake types are described here, there is probably a continuum of lake types between the two. The first lake type has a well defined, water filled channel connecting it to a main channel. As a result, the water level in the lake and main channel are similar. Due to small changes in the 1 National Hydrology Research Institute, Environment Canada, Ottawa, Ontario, Canada, KlA OE7. 23 relative 1 eve l of the channel and lake they are constantly interchanging water. These connecting channels may became dry when the main channels reach low water levels in the fall. There are 61 connected lakes covering 39\ of the surveyed area. The second lake type is perched above the surrounding main channel and connected lake sys tern . These perched lakes have small connecting channels which carry water into the lake for a one or two week period, but may discharge water to the main channel for an additional two to five weeks. Flooding is normally an annual event, but during years with low Mackenzie water levels and for lakes w lth the highest closure levels, overtopping and subsequent flooding may not occur every year. In the study area there are a total of 207 perched lakes, covering 8'1. of the area. Bigras (1985) ca ~led these perched lakes either low or high closure lakes depending on their sill elevation. The 1 and area contributing water to these 1 akes is generally small. In most cases the actual contributing area is hard to define since the land is very flat and the poorly defined drainage system is covered by dense vegetation. In the study area there were also 14 lakes (4\ of the area) which from the aerial photographs, could not be classified as either connected or perched. STUDY AREA AND METHODOlOGY Study area Fl e l d work was carried out from early June to early September 1984 and 1985 at a typical perched lake (unofficial name NRC lake) approximately 5 km southwest of Inuvik (Figure 1). This lake is about 2 m above the low water level of the surrounding connected 1 akes and channels. During the spring flood, the lake level may be over 2.5 m higher than in mi d-slJ111le r. The small channel connecting NRC lake to a nearby connected lake is active during and inmediately after the flood period. For the rest of the year the channel is dry. During mid-summer NRC lake is approximately 300 m in length, 220 m in width, has a mean depth of 0.88 m and max iiTllJTI depth of l . 6 m, and is 0. 069 kJn2 in area . The basin surrounding the lake has an area of 0. 43 kJn2 and is covered by an open, mature spruce forest. The area surrounding the lake is underlain by permafrost in excess of 80 m in thickness (Johnston and Brown, 1964), and the active layer is up to 0.5 m deep by late summer. The zone beneath the 1 ake is COfll>OSed of unfrozen 24 Figure l. location of the study site near Inuvik, N.W.T. silts and clays (Johnston and Brown, 1964, 1965), with bedrock occurring at a depth of 80 m (Johnston and Brown, 1964). In the Inuvik area, air tefl1)erature rises above 0°C in mid May (AES, 1982). Snowmelt and the first deterioration of ice on the Mackenzie River East Channel are initiated at this time and on average the river is clear of ice by June 5 (Allen, 1977). As wanner floodwater enters the lakes during spring breakup, the lake ice cover melts rapidly and lakes such as NRC lake are usually ice free by early to mid-June. Air tefl1)erature falls below 0°C by late September (AES, 1982) and the lakes freeze shortly afterwards. The first permanent ice forms on the Mackenzie River East Channel by October 11 on average, and it is completely frozen by October 19 (Allen, 1977). Field Methods and Instrumentation Hicrometeorological and hydrologic measurements were made from an instrument platfonn near the centre of NRC lake and a 15.9 m lower at a forest site. A Campbell CR21 data logger recorded air temperature, relative humidity, wind speed, net radiation, solar radiation, precipitation, and water or soil temperature at both sites. Hourly averages of these parameters were obtained from measurements taken at 60 second intervals. A Stevens Type F water 1 eve 1 recorder was used to record lake water level. Manual measurements were made of water and bed temperature from the 1 ake surface to 4 m below the lake bed, soil moisture, supra-permafrost groundwater levels at sites around the perimeter of the lake, frost table, and surface flow in a small rill entering the lake. The 1 ake discharge was measured a number of times in order to obtain a rating curve of discharge versus 1 ake stage. In addlt ion the 1 ake out 1 et channe 1 was surveyed to detenni ne the sill elevation controlling inflow and outflow. Water 1 eve 1 measurements for the Hackenz i e River East Channel were obtained from Water Survey Canada. Lake Water Balance The lake water balance is given by (1) P + E + Oo + Oi n + Osp + Osb + Os + e -= ds/dt where P is precipitation on the lake surface, E is evaporation from the lake surface, Q0 is outflow discharge, Oin is inflow discharge from the Mackenzie River, Qsp is supra-permafrost groundwater flow to the lake from the surrounding active layer, Osb is sub-permafrost groundwater flow through the tal ik beneath the lake, Os is surface flow from the surrounding basin, e is an error tenn, and ds/dt is the change in 1 ake storage. In a 11 cases the tenns have positive ~lues when water is added to the lake and negative when removed from the lake. The P, Oin• Q0 , and ds/dt terms were measured on an hourly basis. Methods to calculate E, Osp• and Os are given below. Since all tenns were measured or calculated, the error tenn (e) could be calculated asaresidual. Water balance and groundwater studies at NRC lake have suggested that sub-permafrost groundwater flow occurs. However, the volume is small because the lake bed is composed of fine grained material (Johnston and Brown, 1965) with a low penmeabillty, and permafrost surrounding the lake is over 80 m thick (Johnston and Brown, 1964). Because of its 25 small magnitude, sub-permafrost groundwater will not be considered in this paper. Evaporation Evaporation was computed using the Priestley and Taylor 0972) approach as applled to northern forests by Rouse, et al. (1977), to shallow northern lakes by Stewart and Rouse (1976), and to arctic tundra by Harsh, et al. (1981). In this approach evaporation is calculated by (2) E =a' [ s/(s +g)] (Q*-Og) I lvP where E is evaporation, s is the slope of the temperature -saturated vapor pressure curve, g is the psychrometric constant, Q* is net radiation, Og is the ground heat flux, lv is the latent heat of vaporization, pis water density, and a' is an empirical constant which relates actual to equilibrium evaporation (Harsh, et al., 1981). Various studies (Priestley and Taylor, 1972; Rouse, et al., 1977) have found that for saturated surfaces a' averages 1.26. The slope of the temperature -saturated vapor pressure curve (s) may be calculated as a function of air temperature (Dilley, 1968). . For the lake, Og was estimated from the change 1n water and bed temperature Zo (3) Og = f c dT/dt dz z-=-0 where z is the depth be 1 ow the water surface and z0 is the depth at which annual bed temperature amplitude equals zero, cis the bed or water heat capacity, and T is the bed or water temperature. Observations showed that over the summer period z0 is approximately 4 m. Hourly averages were used for air temperature and net radiation in equation 2, and for water and bed temperature in equation 3. The ground heat flux tenn varied greatly on a daily basis and must be included if hourly or daily evaporation is required. Over the summer period however, Og was sma 11 , accounting for only 3'1. of the available energy. Surface and Supra-permafrost Groundwater Flow Surface and supra-permafrost groundwater flow entered the 1 ake in 9 sma 11 ri 11 s spaced around the lake. Total surface flow 1nto the lake was obta1ned by (4) Qs(T) = Qs(m) (A I A(m)) where Qs(T) 1s the total surface flow enter1ng the lake, Qs (m) 1s the measured surface flow, A 1s the total area contr1but1ng water to the lake (.43 l<lrf>, and A(m) 1s the area contr1but1ng water to the measured r1ll (.021 l<lrf>. Supra-penmafrost groundwater flow (Qsp> 1n a s1ngle r111 was calculated from (5) Qsp = K dh/dl W D where K 1s the hydraul1c conduct1v1ty, dh/dl the hydraul1c grad1ent, W 1s the w1dth of the r11l, and D 1s the th1ckness of the saturated port1on of the act1ve layer. Total flow 1nto NRC Lake <Qsp<D was then estimated by where N 1s the number of r1lls. NRC LAKE HYDROLOGY The hydrolog1c reg1me of NRC lake dur1ng the per1od Hay to Septermer may be d1v1ded 1nto three d1st1nct per1ods. Dur1ng each per1od one process usually dom1 nates the 1 ake water ba 1 ance. These per1ods are: (1) flood1ng reg1me, (2) d1scharge reg1me, and (3) evaporat1on reg1me. Flooding Regime Dur1ng 1985 NRC lake was flooded on Hay 21 (Flgure 2), when the ma1n channel water level rose above 3.813 m. After th1s date, water levels 1n NRC rose rap1dly, reach1ng a peak of 6.363 m on June 2. levels then decl1ned quickly, reach1ng 4.25 m by June 9. After th1s date lake discharge was controlled by the outlet channel geometry, not by the Hackenz1e R1ver water level. Throughout th1s per1od, NRC lake water level was s1mllar to that measured by Water Survey of Canada on East Channel 5 km to the northeast (F1gure 2). Pr1or to the remova 1 of 1 ce on East Channe 1 (June 2), NRC lake water level averaged .246 m h1gher than East Channel. From June ? to June 9, the d1fference was only .136 m. For a 6.34 km channel length, the water slope was .00004 and .00002 respect1vely dur1ng th1s per1od. The r1ver slope dur1ng the ice 26 covered per1od was cons1derably h1gher than the open water slope of .00002 measured 1n late summer. ---------~~l -NRC Lake Julian day (1985) + WSC East Channel Figure 2. NRC lake level and Hackenz1e R1ver East Channel level, Hay to September 1985. Jullan Day 130 1s Hay 10, 1985. The lake water balance dur1ng th1s per1od (Table l) 1s dom1nated by the 1nflux of Mackenzie TABLE 1. Water Balance, NRC Lake 1985. --Inputs------Outputs--Error Date P Qsp Qs Q1n Qsp Q0 E ds/dt e Hay 21 -June 9 5 0 0 2578 0 -2113 0 465 -5 June 9 -Aug 4 22 9 239 0 0 -579 -185 -437 56 Aug 4 -Sept 1 4 0 0 0 0 0 -60 -62 -6 Total 31 9 239 2578 0 -2692 -245 -34 45 floodwater (2578 mm) and subsequent outflow (-2113 mm) as the Mackenzie level dec11ned. Prec1p1tat1on was only 5 mm and because of the lake 1ce cover, evaporat1on was close to zero. The overall result was to 1ncrease the lake storage by 465 mm. In add1t1on the flood water saturated the land surround1ng the lake. Th1s flood water was the pr1mary source of water enter1ng the lake as surface and groundwater flow follow1ng the flood period. The water balance error term was only -5 mm. Discharge Regime After June 9, discharge from NRC lake was controlled by the outlet channel geometry. The peak discharge on this date was .20 m3/s and the flow decllned gradually as the lake level dropped. Discharge ceased on August 4 when the lake level reached the sill elevation (Figure 2). Channel discharge of -5 79 nm was the dominant component of the lake water balance over this period (Table l). After the lake ice cover was removed in early June evaporation became an important component of the lake water balance, with a total of -185 nm (Table l). When flood waters declined, water drained from the land into the lake. The surface flow totalled 239 mm from June 9 until June 20, while from June 9 to August l supra-permafrost groundwater flow contributed 9 nm to the lake. Ora i nage and evaporation from the surrounding forest was greater than prec1p1tation and the active layer dr1ed significantly over the summer. During June and July, precipitation was only 22 mm. The result was that lake storage dec rea sed by -437 nm over the per1od June 9 to August 4. The water balance error was 56 nm during this per1od. ~ll errors in detenmining surface inflow and channel discharge dur1ng the rapid decllne in lake level could account for this error. Since the error is only 12'1. of the change in storage, it is acceptable. Evaporation Regime After August 4, groundwater inflow, surface inflow, and channel outflow had ceased. The only source of water was prec1pltation directly onto the lake surface and the only removal of water was by evaporation from the lake surface. Predicted evaporation from August 4 to Septeni>er l was -60 nm, prec1pltation was only 4 nrn, and the change in lake storage was -62 nm. The residual error tenm was -6 nm (Table 2). Since the lake acted as a large evaporation pan, this period provided an excellent test of the Pr1estly- Tayl or evaporatlon approach (equation 1). The predicted and measured evaporation were wlthin 6 nm or 10\ of each other. This is an acceptable result and justifies the use of this technique to calculate evaporation from lakes in the Mackenzie Delta. 27 Summer Period Even though 1t received a large volume of floodwater in 1985, NRC lake experienced a decline in storage of -34 nm from Hay 21 to September l. The water balance components responsible for this are listed in Table 1 and are described below. The Mackenzie River added a total of 2578 nm to the 1 ake as flood water. Surface and groundwater flow, of which most was Mackenzie flood water draining from the surrounding land, added an additional 248 nm of water. Total water input from flooding was therefore 2826 nm. The only other source of water, precipitation on the lake surface, totalled 31 nm from Hay 21 to September 1. This was less than normal. At the AES Inuvik weather station for example, the June, July, and August precipitation was only 31 nm compared to the 30 year normal of 101 nm. Host of the 2927 nm of water added to the lake, was discharged as channel flow which totalled -2692 nm. Evaporation was the second largest tenm, equalling -245 nm. Total calculated output was -2937 nm, and the water balance error tenm, calculated as a residual, was 45 nm. This is an acceptable error since 1t is such a small percentage of elther the total input or output of the lake. However, 1t is a significant proportion of the measured change in lake storage. The lake is in fact in a very fine balance, with large inputs and outputs of water, but very small changes in storage from year to year. PREDICTED NRC LAKE LEVEL Simulation Model Changes in lake level over the flooding, discharge and evaporation periods were predicted using: (l) adjusted East Channel water levels during the flooding period and (2) hourly water balance calculations for the ren~inder of the summer. The model is initiated wlth a measured pre·~lt lake level. In 1985 this was 3. 785 m. Lake level remains constant until the channel water level rises above the NRC sill elevation (3.813 m) and the lake is flooded with Hackenz i e River water. The 1 ake 1 eve l then equa 1 s the channe 1 1 eve 1 , which is estimated from East Channel level adjusted for channe 1 s 1 ope. A s 1 ope of . 00002, detenmi ned from summer observations, was used. Once the lake level falls below the elevation where NRC Lake discharge is controlled by the outlet channel geometry L B 0 "' (4.25 m), the lake level is controlled by the lake water balance. The hourly lake levels were then calculated as (10) LS(t) ; LS(t-1) + P + Osp + Os + E + Oo where lS is the 1 ake stage at time t and (t-1). The other parameters, as defined earlier, are measured or calculated over the time period (t-1) to (t). Prec1p1tation was measured at the lake, and groundwater, surface flow, and evaporation were calculated using equations 6, 4, and 2 respectively. Discharge was calculated from a relationship between stage and discharge. Observations showed that discharge ceased when lake level dropped below 3.813 m. Results Predicted NRC lake level is similar to that observed over the entire summer period (Figure 3). ,_, __ , _____________ , 5.6 -j 4 t 5.4 ~ J \ +J ', 5.2 J +; * 5 "~ +I I 4.8-1 l * 4.6 ~ I I 4.4 ~ I I I I 4.2 J ,I 4 J! I I ::::J 3,8 ~~ 3.6 -J----,----r ---,--.-----r---.-----r----,----,- 140 160 180 200 220 Julion doy (1985) --PREDICTED + OBSERVED Figure 3. Observed and predicted NRC Lake level, May to September 1985. Julian day 140 is May 20, 1985. 240 The maximum differences between predicted and observed mean daily lake level during the flooding, discharge, and evaporation periods were only .176, . 028, and . 023 m. By the end of the s imul at ion period the difference between observed and predicted was only .016 m. The largest error occurred in the flooding period when lake level was calculated using WSC East Channel level adjusted using an open water slope. During this period however, water slope varies as the channel changes from ice covered to open water. These changes in slope were not accounted for in the present model. However, considering the large changes in level 28 during this period, the error involved is relatively small. DISCUSSION The water ba 1 ance 11 sted in Tab 1 e 1 shows that flooding contributes a large volume of water to NRC Lake. Low closure lakes like NRC Lake probably flood annually, however certain high closure lakes are flooded infrequently. The frequency of flooding, controlled by the lake sill elevation and the Mackenzie River flood level, could change H flow regul~tion occurred in the Mackenzie River Basin. The water balance of a perched lake during a non-flooding year can be estimated from the NRC Lake data. In 1985, evaporation from NRC Lake was -245 mm and prec1pltation was 31 mm. Since in the absence of flooding the only significant input and output is precipitation and evaporation, the resulting change in storage at NRC Lake would have been -214 mm. Even in a year wlth normal precipitation, evaporation would probably exceed prec1pltation. In 1984 for example, a year with near normal precipitation, NRC Lake level decreased over the summer period. This implies that evaporation was greater than prec1p1tation. Therefore, without flooding, water level in NRC lake would be expected to decline significantly within a few years. This is probably true for most perched lakes. This conclusion is substantiated by data from a perched lake 50 km southwest of Inuvik. This high closure lake was last flooded in June 1982. The 1 ake does not have a surface out 1 et, but the 1 ake level declined 1.43 m between June 3, 1982 and August 27, 1985 (Bigras, personal communication). The average summer decline of .36m, is larger than the difference between precipitation and evaporation experienced at NRC Lake in 1985. This is probably explained by a higher evaporation rate at the southern site due to warmer air temperatures. Further work is required to substantiate this explanation. The important point from this example is not the rate of decllne, but that without flooding it will probably be dry within a few years. One aspect of the lake hydrological regime not inc 1 uded here is the sno..ne lt period. It is not known how much of the winter snowfall of approximately 130 mm water equivalent runs off into the 1 ake during the spring melt. However the dry soil, low relief, and poorly defined drainage system must llmlt the runoff. Future work will consider this aspect of the lake hydrological regime. Perched lakes certainly do not require annual flooding, but they probably require flooding on a frequent basis to maintain water levels. The frequency of flooding required could be detennined by running a version of the simulation model described earlier. CONCLUSIONS The primary conclusions of this paper are: (1) the lake sill elevation is a major factor controlling the hydrology of Mackenzie Delta lakes, detennining the duration and magnitude of flooding (2) permafrost 1 imits groundwater movement from NRC lake, allowing it to remain perched above channel level for most of the year (3) because evaporation from the lake surface is greater than precipitation, the lake experiences a negative water balance. Without replenishment of water from the Mackenzie River during the spring flood, the lake level would probably decrease significantly within a few years (4) as a result of evaporation from the land surrounding the 1 ake, surface and sub-surface drainage to the lake was small for much of the SlJlTiler. Most of the drainage to the 1 ake was a result of flooding of the land by the Mackenzie River (5) A simulation model was able to accurately predict lake level throughout the spring flood and slJlTiler period. This model, with a few additions, could be used to detennine the frequency of flooding needed to sustain perched lakes in the Mackenzie Delta (6) perched lakes are susceptible to pollutants introduced from the Mackenzie River since these lakes are only flushed once per year Acknowledgements: The generous 1 og is t i ca 1 support of Polar Continental Shelf Project, Department of Energy, Hines, and Resources, and the Inuvik Scientific Resource Centre, Department of Indian and Northern Affairs is greatfully acknowledged. I would like to thank R. Smith, M. Suzor, and R. Storey for their assistance in the field. 29 REFERENCES Atmospheric Environment Service, 1982. Canadian climate normals. Volume 2, Temperature, 1951 to 1980. Environment Canada. 306 p. Allen, G.D., 1977. Freeze-up, break-up and ice thickness in Canada. Atmospheric Environment, Publication CLI-1-77, 185 p. Bigras, S.C., 1985. lake regimes, Mackenzie Delta, NWT, 1981 . Nation a 1 Hydro 1 ogy Research Institute, Internal Report, Environment Canada, 39 p. Dilley, A.C., 1968. On the computer calculation of vapor pressure and specific humidity gradients from psychrometric data. Jour. of Applied Meteorology, 7, 717-719. Gill, D., 1973. Modification of northern alluvial habitats by river development. Canadian Geographer, 17, 138-153. Johnston, G.H., and R.J.E. Brown. 1964. Some observations on permafrost distribution at a lake in the Mackenzie Delta, N.W.T., Canada, Arctic, 17, 163-175. Johnston, G.H., and R.J.E. Brown. 1965. Stratigraphy of the Mackenzie River Delta, Northwest Territories, Canada. Geol. Society of Amer1ca Bull., 76, 103-112. Mackay, J.R., 1974. The Mackenzie Delta area, N.W.T., Geol. Survey of Canada Misc. Report 23, 202 p. Harsh, P., W.R. Rouse, and M.K. Woo., 1981. Evaporation at a High Arctic site. Jour. of Applied Meteorology, 20, 713-716. Peterson, E.B., l.H. Allison and R.D. Kabzerns., 1981. Alluvial ecosyterns. In: Mackenzie River Basin Committee, Mackenzie River Basin Study Report Supplement 2, 129 p. Peace-Athabasca Delta Group. 1972. The Peace-Athabasca Delta, A Canadian Resource. 144 p. Priestley, C.H.B. and R.J. Taylor. 1972. On the assessment of surface heat flux and evaporation using large-scale parameters. Monthly Weather Review, 100, 81-92. Rouse, W.R., P.F. Mills, and R.B. Stewart. 1977. Evaporation in high latitudes. Water Resources Research, 13, 909-914. Smith, M.W., 1976. Permafrost in the Mackenzie Delta, Northwest Territories. Geol. Survey of Canada Paper 75-28, 34 p. Stewart, R.B. and W.R. Rouse, 1976. A simple method for detennining the evaporation from shallow lakes and ponds. Water Resources Research, 12, 623-628. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 SHORT-WAVE HEATING OF LAKE SURFACE WATER UNDER A CANDLED ICE COVER J, P. Gosink and J, D. LaPerriere* ABSTRACT: During spring, as the snow covering the lakes and rtvers begins to melt from a positive surface heat balance, meltwater percolates through microcracks in the ice, initiating vertical channels through the ice. The iet> develops into a porous fabric, known as candled ice, characterized by closely packed vertica1 "candles" of ice interspersed with chan- nels of meltwater. Candled ice is quite transparent to short wave radiHijon, with optical extinction coefficients approach- ing those found in the lake water. This implies that the incident solar radiation can penetrate through the ice cover, warming the water immediately below the melting ice. La Perriere (1981) reported lake temperatures in excess of 6°C one to two meters belo,;r the hase of the candled ice during field studies in Harding Lake, Alaska. This laye·c of water, although heated above the temperature of maximum density, appeared to be stable due to the slight density difference between the meltwater with low specific cond11ctance, and the more saline lake water. The si:ahility of the u•lJ,,r·-icP water column is critical for lake overturn and subsequent rf~oxygenation of the lake. In a typical dlmictic lake, overturn occurs during strong winds at or soon following complete disappearance of the ice. The overturn res•tlts ln reoxygenation of the lake to near saturation levels. However, when ner1.·r surface warming of the lake occurs t:hront;h the candled ice column, the potentlal exists for initiation of summer stratification conditions below the ice cover, with a stable density strut·.tqre ln the water co11Wlil• If the wind stress over the lake is low as the final ice melts, it is possible that little or no overturn and rexoygenation of the lake water will take place. This situation occurred in Harding Lake in May 1975. This paper presents the results of a one-dimensional heat balance study of the near surface water below candled ice. Short wave penetration of heat through the candled lee and into the underlying water column, latent heat effects, and thermal conduction and convection are considered. The model predicts the oc- currence of a subsurface water tempera- ture maximum close to 6°C at a depth one to two meters below the base of the candled ice. The analysis includes a discussion of convective stability in the meltwater layer below the meltin~ ice cover. Velocities in the meltwater are shown to be between 0.1% and 10% of the corresponding Monin-Oboukhov velocity at a free surface. (KEY TERMS: lake ice; candled ice cover; penetra- tive convection, surface mixing.) INTRODUCTION Thermal energy transfer at a snow or ice surface on a lake or river consists of heat flux due to sensible heat trans- fer, conduction through the snow and ice, latent heat exchange due to sublimatl{)n and melting or frc,.ezing, long-wave radia- tive exchange and short-wave radiative transfer. The short-wave transfer differs from the other modes of heat exchange in that attenuated short-wave radiation *J, P. Gosink, Geophysical Institute, University of Alaska, Fairbanks, Alaska 99775-0800; J, D. La PP.rriere, Alaska Cooperative Fishery Research Unit, Arctic Health Research Building, University of Alaska, Fairbanks, Alaska 99775-0810. 31 penetrates throneh i:he surface, into the snow and ice, and ultimately into the underlying water column. If the snow layer is thick, then most of the avail- able short-wave energy is expended warming or melting the snow layer. In spring, as the snow layer melts, an increasing amount of shortwave radiation is absorbed within the ice layer ancl in the underly- ing water column. The radiation absorbed in the ice raises the ice temperature to 0°C, and excess radiation results in melting at gn1i r1 houndaries (Knir;ht, 1962; Lyons and Stoiher, 1959). The effect of this melting process is to cause deteri.•JI'>tl. in11 of the ice fabric (A>1hton, 1985; Bulatov, 1970). The deteriorated ice becomes porous and assumes the appearance of closely packed "candles". Candled ice is substantially weaker than an intact ice layer. La Perriere (l9Rl) observed candled ice on Harding Lake during late spring, and measured water temperatures below the bottom of the j C'~ layer. A somewhat surprising result was the discovery of a stable layer of water heated above the temperature of maximum density which had not mixed downward. It was hypothesized that the heat gain in the water was due to the transparency of the candled ice to short-wave radiation. It was further assumed that melting of the ice produced a shallow and relatively buoyant fresh IJater lens below the ice surface. Verti- cal water velocities were not measured; however, it was suggested that convective velocities due to buoyancy differences were rnLnimal. '!'his article evaluates the mixing potential of the surface layer water. This is accomplished through an analytic model of heat transport in the water column with three modes of heat trans- port: conduction, convection and radia- tion. This model, in the form of an analytic solution to the governing equa- tion, is previously unpublished. Scaling :-1rguments are used to prescribe maximum and minimum limits for the convective velocity. Short-wave radiation and light extincUon coefficients are approximated from typical ~easured values at the HarJing Lake study site. 32 T~lSORY Short-wave radiation penetrating into a \vater column is attr""""t'~'l VJith depth according to Bouger's relation (Fischer et al., 1979). It is known that the light extinction is dependent upon the wave length (Hutchinson, 1957); however, for studies involving small changes in depth, a bulk extinction formula is appropriate. The usual form of the governing equation for short-wave radiation Ln water is: S(y) = Sw exp(-ay) (1) where Sw is· the radiation at the top of the water column, y is depth and a is the extinction coefficient. Measurements were 1nade of the short->vave radiatio11 in. Harding Lake during summer 1976 (La Perriere et al., 197R). The extinc- tion coefficients for various wavelengths ranged between 0.4 and 1.0 m-1. For the present study, a value of 0.8 m-1 was assumed for all calculations. The short-wave radiation at the top of the water column, s,,, is the radiative flux at the botto~ of the ice layer. Sw is proportional to the radiative flux arriving at the top of the ice. Bulatov (1974) determined that the radiation attenuation in ice is ~ependent upon the radiative wave length. Several investiga- tors (I.Jilrrr~n, 1982; Grenfell and Maykut, 1977) measured radiative extinction in fresh and sea-water j r~r~, and established empirical relationships defining the extinction as a function of depth in these media. The study by Grenfell and Haykut (1977) suggested that the bulk short-wave radiation attenuation in ice is defineci by the formula: (2) where S0 is the short wave radiation at the top of the ice. This implies that, for ice thicknesses less than 0.5 m, more than 40% of the short wave radiation at the ice surface penetrates into the water column and is absorbed there as heat. Furthermore, the short-wave radiation at the top of the water column, Sw "' S0 exp (-1.36 d•5) where d is the ice thickness. The short wave radiation flux given hy equation (1) may be considered a distri- buted heat source in the water Utellor, 1964; Tien, 1960). The equation for heat transport in the water, including both conduction of heat and the short-wave radiation acting as a distributed source, may be written as follows: where K is the thermal conductivity of the water. This equation assumes that a steady state condition exists and that no convective heat transport is consiJered. The solution to the equation is straight- forward (Carslaw and Jaeger, 1971): T(y) T 0 + y (T1-T0 )/b + Sw [1-exp(-ay) -y(1-exp(-ab))/b]/aK (4) where the boundary conditions T(o) = T0 and T(b) = T1 have been applied, y = o is the position of the \vater-ice inter- face, and y = b is a given water depth. The model defines a modified linear temperature distribution, sneh that the curvature in the profile is proportional to the short-wave radiative heating. Two limitations of the model are the assumptions of steiidy-state condit tons and that of no vertical convection. Clearly, the pos Lt ion of the water-ice interface changes with time as the ice melts, but the temperature at this posi- tion remains at the freezing point. This suggests that a solution to e~uation (3) in a coordinate system which moves upward at the rate th;ti· 1l:e j c:.e is 1nelting \vould ]J<· lHlr"t~ rlppropriate. If the ice melting rate is assumed to he constant and equal to v, then a transformation of coordinates for the transient form of equation (3) results in the following quasi-steady equation: pc v dT/dy 33 '"here the coordinate y is measured from the moving ice-water interface, p is the water density, and c is the specific heat. Note that v ~ Q/L where Q is the residual heat available for melting (\.J m-2), and L is the volumetric latent heat of jce (.J m-3). Q can be clet'-'rni:1ed as the melt rate of the ice by a heat balance approach. However the meteorolo- gical data required for a complete surface heat balance was not available for the Harding Lake study site. Since heat con- duet i 011 t11rOtigh an isothermal ice cover is zero, a reasonable approximation of Q is Sw, the short wave energy flux trans- mitted thro~~h the ice. For e~~mple, ~l Sw ~ 15 W m -, then v = 5 • 10 ' m sec or 0.5 em day-1. This is a very small melting rate which possibly repn:•s:·rii::> early springtime wan:J.:i "E in central Alaska. Another interpretation oi ·~'lUA.tion (5) may be inferred when v is assumed to represent a vertical convective velocity in the water. With this interpretation, there are three modes of heat transfer: conduction, convection and short-wave radiation. If v is a gravitational velocity due to density differ~nces, solutions of equation (5) would repre- sent the complete convective-diffusive model with a distributed heat source. Therefore, the same equation, ·~'l!lati.on (5), is a model for the quasi-steady Ch•' \ti t1: n •·1oving phase front (when v = Sw/L) and for the convective case (when v > Sw/L). The solution to equation (5) when v is constant is: T(y) = T0 + (T1-T0 ) [1-exy(vy/k)] 1-exp(v1J/k) +Tsw [1-exp(-ay)] -Ts,v [0-exp(-a b)) (1-ex_p_(vy/k))] (6) 1-exp(vb/k) where k = K/ pc Equation (6) then represents a f[llasi- steady solution to the equation for he<J t transport in the water when all t 11rPP me chard srts of he.qt transfer are included: conduction, convection due to density differences and short-wave heating. The lower limit for v is the melting rate of the ice. An upper liflit for v may be found from Monin-Oboukhov scaling for velocity due to gravitational differences (Tennekes and Lumley, 1972; Fi.scher et al., 1979): u = <s B g b)1 13 f ~\1~---'.L_;::_ (7) pc where B is the thermal expansion coeffi- cient of water and g is the gravitational acceleration. Uf ls often called the p!_'ll<~t rative convective velocity and is a measure of the free-fall velocity due to density changes at the water surface. Monin-Oboukhov scaltnc; has been used widely in geophysical applications (Turner, 1973; Fischer et al., 1979; Niiler, 1975) to estimate the mixing or penetration velocity in mixed layer dynamics. It may he considered an upper limit for the convective velocity under the ice layer. It is an upper limit since it does not include the effects of the surface drag which would be present at the water-ice interface. Penetrative convection may be ex- pected if the density differences associ- ated with temperature change are suffi- ciently large to overcome viscous forces. The critical parameter for the initiation of convection when a temperature gradient equal to hT/b is imposed is the Raleigh number (Raleieh, 1916; Bear, 1972; Turner, 1973): Rat= B ~T~3 kv (8) where v is the kinematic viscosity. Two other forms of the Raleigh number are pertinent in the present analysis. The first is a sali.ni.ty Raleigh number associated with the melt water -lake water salinity difference (Turner, 1973): (9) 34 where ~P is the density difference due to salinity variation between the melted ice and the lake water. The other pertinent Raleigh number is associated with an internal heat source as the cause of the gravitational convection (Turcotte and Schubert, 1982): aSHBgb 5 Kkv (10) Each of these three Raleigh numbers have somewhat different critical values de- pendint; upon the assumptions made con- cerning boundary conditions. However, in all cases the critical Raleigh number is less than 5000. Estimates of these Raleigh numbers approp~iate for Harding Lake are; Rat = 5 • 10', Ra£; = 0.0, and ~ai _,. 10 • These estimates were obtained by assuming a temperature difference of 2°C across a 0.5 m water depth, a thermal expansion coefficient equal to 0.3 • 1o-5oc-1, an electrical conductivity difference of 60 ~ mhos cm-1, short-wave radiation equal to 10 W m-2, and an ex- tinction coefficient of 0.8 m-1. Accord- ing to the formulae given by Bennett (1976), this electrical conductivity difference implies a negligible salinity difference. These values correspond approximately to tlv~ w~asured quantities determined at Harding Lake by La Perriere et al. (1978). The estimated Raleigh numbers suggest that penetrative convec- tion due to thermal gradients and radia- tive heating was indeed occurring, but that measured salinity differences were inadequate to account for gravitational convection. RESULTS AND DISCUSSION Upper and lower limits for the velocity v in equation (5) and (6) have been previously defined. The lower l ir.li t is the ice melting rate and the upper limit is the Monin-Oboukhov (M-0) convec- tive velocity. Both these limjts are dependent upon the heat exchange term Sw• In Table 1, a few values of melting rate and M-0 velocity are presented as a function of Sw• Table 1. Hinimum and maximum velocities -2 Helting_yate H-0 velo~fty ~Vlm ) (m sec ) (m sec ) --,1 II 10-8 5 II 2 • II 2 • 10-4 15 5 • 10-8 3 • 10-4 10-8 50 ~ 16 • ----- L~-4 Equation (6), which defines the temperature distribution with convection, conduction and heat source, is depicted in Figure 1 for the two limiting values of v, i.e., -8 -1 vmin = 3 • 10 m sec and 0 • 0 4 TEMPERATURE (() 2 3 4 5 • Jl 6 ? • • • D 0 8 E p ll T ll H • M 1 2 • 1 6 2 0 Figure 1. Heasured (*) and calculated (-) temperature distributions under a candled ice cover in Hay, 1975. Depth is measured from the bottom of the ice, which was about 0.5 m thick. Both calculated temperature distribu- tions are defined by equation (6). The assumed values for v (v = Vmin and v = Vmax) are defined on the curves. 35 vmax = 2 • 10-4 m sec-1 • The measured temperature distribution at the Harding Lake study site is also shown on the fig- ure. The calculated temperature distri- bution for the lower limit velocity, Vmin• is effectively identical to the zero convection solution, equation (4). This might have been anticipated through a consideration of the Peclet number, Pemin = Vmin b/k ~ 0.3, since Pe can be considered the ratio of convection to diffusion of heat. The calculated tem- perature distribution for the upper limit velocity, Vmax> indicates an intensely mixed surface layer in which latent heat losses at the water-ice interface are rapidly transported downward. TEMPERATURE (() 0 2 3 4 5 6 ? 0 4 D 0 8 E p ll T ll H • M 1 2 • " • 1 6 • • 2 0 Figure 2. Measured (*) and calculated (-) temperature distributions under a candled ice cover in Hay, 1975. Depth is measured from the bottom of the ice, which was about 0.5 m thick. The calculated temperature distribution is defined by equation (6), with values of v = 10-5 m sec-1 in the upper 0.2 m and v = 10-7 m sec-1 in the lower 1.8 m. The measured temperature distribu- tion suggests that an intermediate value for convective velocity may more accurate- ly define the heat transfer mechanisms. Furthermore, it appears that a larger velocity is indicated in the upper portion of the profile, and a somewhat smaller velocity below. In Figure 2, the calcu- lated temperatures have been graphed with a convective velocity equal 10-5 m sec-1 in the upper 0.2 m and equal to 10-7 m sec-1 in the lower 1.8 m. These veloci- ties for upper and lower layers are about 10% and 0.1% respectively of the associ- ated Honin-Oboukhov velocity at a free surface. The agreement between the two layer convective model and the measured temperature distribution is excellent. The reasons for a greater convective ve- locity in the upper portion of the pro- file relative to the lower portion are not apparent. However, possible expla- nations may be related to deceleration due to viscosity or to the smaller heat- ing in the lower portion associated with attenuated radiation penetration. The upper 0.2 m may be interpreted to be a region in which the convection of cool water from the ice surface is suffi- ciently high to overcome the warming by short wave penetration. In effect, the time scale for removal of heat by convec- tion is smaller than the time scale for heat generation by short wave penetration. This downward velocity is necessary to explain the positive curvature of the temperature profile in this region. Note that the application of equation (4), the zero velocity solution, to the upper layer would imply a negative curvature, contradicting the measured values of temperature. In the lower layer, a smaller value of the convective velocity permits radiative heating and conduction to warm the water above the temperature of maximum density. The convection time scale in the lower region is smaller than the time scale for short wave heat generation. 36 SUHMARY AND CONCLUSIONS Short-wave radiation penetration through a candled ice cover has the potential to raise the near ice water temperature above the temperature of maximum density. This occurred during a spring 1974 investigation of Harding Lake (La Perriere, 1981). The resul- tant density excess induced an insta- bility with downward gravitational convective velocity. Density differ- ences due to the measured salinity difference between lake water and melt water were negligible. However, the calculated Raleigh numbers for both conductive.heat transfer and internal radiative heating were sufficiently high to ensure convective instability of the near-ice water. The appropriate convective velocity below the ice cover appears to be a small fraction of the Monin-Oboukhov velocity at a free sur- face. Apparently the ice-water inter- face acts to decrease the expected mo- tion. A new and previously unpublished solution of the heat transport equation with a heat source has been presented. Heasured temperatures at Harding Lake under the ice cover show excellent agreement with the calculated tempera- tures, verifying the assumptions of weak convective instability under the ice cover and the penetration of radiative heating below the ice. SYHBOL DEFINITIONS b water depth [m] c = specific heat of water [J kg-1 oc-1] d ice thickness [m] g gravitational constant [m sec-2~ k thermal diffusivity of water [m sec-1] K thermal conductivity of water [Wm-1 oc-1] L volumetric latent heat of ice [J m-3] Q available heat for melting ice [W m-2 ] Rag s v y (l B p ~p v Raleigh source, Raleigh 9 [ number for equ. 10 [ number for ] internal heat salinity, equ. Raleigh number for temperature gradient, equ. 8 [ ] Short wave radiation in water, equ. 1 [W m-2 ] Short wave radiation in ice, equ. 2 [W m-2 ] Short wave radiation at ice upper surface [\v m-2] Short wave radiation at water sur- face or water-ice interface [W m-2] water teBperature [°C] water temperature at y = o [°C] water temperature at y = b [°C] Monin-Oboukhov velocity, equ. 7 [m sec-1 ] velocity defined in text [m sec-1] downward distance [m] short wave extinction coefficient in water [m-1 ] thermal exyansion coefficient of water [°C-] average water density [kg m-3] densit~ difference by salinity [kg m-] kinematic viscosity of water [m 2 sec-1] Acknowledgements: Partial support for this work was provided by the Department of the Army under contract DAAG29-85-K- 0260 and by State of Alaska funds through the Geophysical Institute, University of Alaska. REFERENCES ~hton, G. D., 1985. Deterioration of Floating Ice Covers. J. Energy Res. Tech. 107:177-182. Bear, J., 1972. Dynamics of Fluids in Porous Media. American Elsevier Publ. Co., Inc., N.Y., 754 P• Bennett, A. s., 1976. Conversion of in-situ measurements of conducti- vity to salinity, Deep Sea Res. 23:157-165. 37 Bulatov, s. N., 1970. Calculating the Strength of Thawing Ice Cover and the Beginning of Wind-Activated Ice Drifts (in Russian). Trudy Vypysk 74, Gidrometeorologicheskoe Izdated'stvo, Leningrad, p. 120. Carslaw, H. s. and J. c. Jaeger, 1971. Conduction of heat in solids, Oxford at the Clarendon Press, London. Fischer, H. B., E. J. List, R. Koh, J. Imberger and N. H. Brooks, 1979. Hixing in Inland and Coastal Waters. Academy Press, N.Y. Grenfell, T. C. and G. A. Haykut, 1977. The Optical Properties of Ice and Snow in the Arctic Basin. J. Glac. 18(80): 445-463. Hutchinson, G. E., 1957. on Limnology. Vol. 1. Physics and Chemistry, 1015 P• A Treatise Geography, Wiley, N.Y. , Knight, c. A., 1962. Studies of Arctic Lake Ice. J. Glac. 4(3):319-355. La Perriere, J.D., 1981. Vernal Over- turn and Stratification of a Deep Lake in the High Subarctic Under Ice. In: Proceedings of the Inter- national Association for Theoretical and Applied Limnology, Congress in Japan, 1980, pp. 288-292. La Perriere, J. D., T. Tilsworth and L. A. Casper, 1978. Nutrient Chemi- stry of a Large, Deep Lake in Sub- arctic Alaska. Ecological Research Series Report EPA-600/3-78-088. Environmental Protection Agency, Environmental Research Laboratory, Corvallis, OR 97330. Lyons, J. B. and R. E. Stoiber, 1959. The Absorptivity of Ice: A Critical Review: Scientific Report No. 3, Dartmouth College, Hanover, N.H., October 31, 1959; also, Air Force Cambridge Research Center Technical Note 59-656. Hellor, M., 1964. Properties of Snow. USA CRREL Report II-A1, Hanover, N.H. Niiler, P. P., 1975. Wind-Mixed Layer. 33(3):405-422. Deepening of the J. Marine Res. Raleigh, L., 1916. On Convection Cur- rents in a Horizontal Layer of Fluid \vhen the Higher Temperature is on the Under Side. Phil. Mag. 6(32):529-546. Tennekes, H. and J. L. Lumley, 1972. A First Course in Turbulence. The MIT Press, Cambridge, Mass. Tien, c., 1960. Temperature Distribution of Snow with Gamma Ray Radiation, USA CRREL Research Report 67, Hanover, NH. Turcotte, D. L. and G. Schubert, 1982. Geodynamics. John Wiley and Sons, N.Y. Turner, J. s., 1973. Buoyancy Effects in Fluids. Cambridge at the Univer- sity Press, Cambridge, England. Warren, s. G., 1982. Optical Properties of Snow. Rev. of Geophys. and Space Phys. Vol. 20(1):67-89. 38 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 HYDROTHERMAL MODELING OF RESERVOIRS IN COLD REGIONS: STATUS AND RESEARCH NEEDS Donald R. F. Harleman 1 ABSTRACT: A review of mathematical models for the prediction of the thermal struc- ture and water quality of reservoirs in cold climates is presented. Recent research and research needs are discussed in the following subject areas: dynamics and thermodynamics of ice cover formation and decay, wind stress with weak stratification and wind stress attenuation due to ice formation, suspended sediment effects and local mixing and water quality effects. (KEY TERMS: reservoir thermal stra tifica- tion; reservoir water quality; cold regir- impoundmen ts, ice cover formation; ice cover decay; heat transfer in ice.) INTRODUCTION The development of the cold regions of the earth for human habitation and for energy resources is proceeding at a rapid pace. Thus the potential for reservoir construction for hydroelectric development and water supply in high latitudes is greater than in the more developed temperate and tropical zones. There exists, therefore, a need for deterministic models to predict the physical, thermal and associated water quality aspects of impoundments in cold regions. The obvious starting point is the considerable body of knowledge that has been accumulated over the past 30 years in the mathematical modeling of lakes and reservoirs located in mid-latitudes. The objective of this paper is to focus on the features of the modeling problem that are characteristic of, and in many cases, unique to cold climates. The first section will review a number of impoundment models that have been used in temperate zones with an emphasis on attempts that have been made to adapt them to cold region conditions. Subsequent sections will discuss some of the unique modeling problems such as the dynamics and thermodynamics of ice cover formation and decay; wind stress on weakly stratified lakes and attenuation due to ice growth; effects of suspended sediment inflows due to glacier melt; and localized mixing and overall water quality effects. In each case a brief literature review will be followed by a discussion of future research needs. REVIEW OF ONE-DIMENSIONAL RESERVOIR MODELS Diffusion Models Thorough documentation of the early history of hydrothermal modeling of impoundments has been given by Harleman (1982) and Orlob (lq83). The first attempts at mathematically modeling 1Ford Professor of Engineering, R. M. Parsons Laboratory, Department of Civil Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139. 39 temperature distributions in deep, stratified impoundments were carried out in the late 1960's and early 1970's independently by Orlob and co-workers at Ua ter Resources Engineers (\-IRE) and by Harleman and co-workers at MIT. The Orlob (1983) review documents one, two and three-dimensional modeling efforts and gives criteria for use of the one-dimensional approach in which the water body is assumed to be stratified in horizontal planes. In this paper attention will be restricted to one-dimensional models. The classical advection-diffusion equation approach is illustrated by Figure 1 showing a one-dimensional schematization for a variable area reservoir with an inflow and one or more outflows. 4Z T FIG. 1 -Schematization and control volun• for mathematical model for reservoir temperature distribution. The thermal energy equation for an internal element is given by Huber, Harleman and Ryan (1972) as oT 1 o 1 o oT ~ + A ~ ( Qv T) = A ~ ( AE z ~ ) B(uiT. -u T) + 1 0 A ( 1) 40 where T is temperature within the reservoir (Ti is inflow temperature), z is measured from a bottom datum (the surface elevation zs is a variable), A is the horizontal area of the reservoir, Ez is the eddy diffusivity of heat and B is the width. Horizontal velocities due to inflow ui and outflow u0 are computed by assuming Gaussian velocity profiles. The height of the Gaussian outflow profile is related to the vertical density gradient by selective withdrawal theory. The river inflow is mixed with water from the surface element in a specified ratio, changing both the volume and temperature of the inflow. The mixed inflow enters the water column centered at the elevation at which the reservoir density is equal to the mixed inflow density. Vertical velocities are computed from the continuity equation for each element, thus the vertical flow rate Qv is given by 0 ( z 't) 'v B Jz u.(z,t)dz 0 1 f z - B u (z,t)dz 0 0 ( 2) The last term in equation (1) represents the internal heat source due to the absorption of shortwave solar radiation. The terms p and c are the density and heat capacity of water, ~s is the heat flux per unit area due to short wave radiation reaching the water surface, ~ is the fraction of that radiation absorbed in the surface (approximately 0.5) and ~(m-1) is the extinction coefficient which depends on water clarity. The surface boundary condition, assuming that the water surface remains free of ice is given by oT pcEz ~ = -~~s -~a+ ~b + cp + ~ at z = z ( 3) e c s where ~Pa is the incoming atmospheric, or long wave, radiation which is completely absorbed at the surface; tPb is the flux due to back radiation from the water surface; ~Pe and ~c are heat fluxes from the water surface due to evaporation and convection. The latter three components are functions of the water surface temperature, wind speed and vapor pressure gradient. The bottom boundary condition, usually assumed to represent zero heat flux (oT/oz=O), and a specified vertical temperature distribution initial condition completes the mathematical specification. Various empirical representations have been used for the vertical eddy diffusivity Ez· These range from assuming it to be a constant in both time and depth to assuming a functional re1a tionship be tween Ez and the densimetric water column stability (1/p) op/oz. Harleman (1982) pointed out that in reservoirs having near surface inflows and relatively deep outlets the vertical heat convection term (the term to the left of the equal sign in equation 1) dominates ilie vertical diffusion term. The result is that the vertical temperature profiles are relatively insensitive to the vertir 11 diffusivity. This is in contrast to la'k<>s ~ which there is generally a high degrPP of sensitivity. Emperical evidence suggests that turbulent mixing induced by wind shear at the surface is highest in the near surface or epilimnion region. In addition, the epilimnion is frequently found to be well-mixed with little or no vertical temperature gradient at the surface. These observations, coupled with ilie lack of a generally satisfactory resolution of the problem of specifying Ez=f ( z, t) , led to the development of mixed layer models in the latter half of the 1970's. Mixed Layer Models The surface mixed layer may be thought of as encompassing three· zones: (1) the near surface region in which turbulent kinetic energy (TKE) is produced by wind shear as the stirring agent. The TU is exported to the fluid below; (2) a central zone in which the TKE exported from above is used to homogenize the fluid; and (3) a thin frontal zone separating the turbulent interior from the quiescent fluid beneath the upper mixed layer. In this zone the remainder of the TKE exported from the surface, plus any that may be locally produced by the shear of the advec ting mixed layer, less that 41 which is locally dissipated or radiated downward by internal waves, is used to entrain quiescent fluid into the mixed layer. Density instabilities, resulting from surface cooling, can also cause mixing through the production of turbulence associated with vertical buoyancy fluxes. This is known as penetrative convection mixing. The entrainment velocity ue, defined as the time rate of increase of thickness, h, of the upper mixed layer, is ue = dh/dt and the problem is reduced to finding an expression for ue• Stefan and Ford (1975) at University of Minnesota developed a mixed layer model (MLTM) for application to small lakes. Independently, Harleman and Hurley-Octavio (1977), Bloss and Harleman (1979, 1980) and Imberger, et al., (1978) developed similar models at MIT (MITEHP) and University of California, Berkeley (DYRESH). All of these models consider the turbulent kinetic energy budget integrated over the thickness of the mixed layer. Parameterizations of the various processes leading to the production and decay of TKE result in expressions for the entrainment velocity Ue that are functions of the local and bulk Richardson numbers. One of the differences between DYRESM and the previous mixed layer models was the change from a fixed or Eulerian layer system to a variable or Lagrangian grid in which each layer can expand or contract to balance the movement of water in and out of that layer. The advantages of the Lagrangian approach are twofold: (1) conservation of mass is improved because the need for addition or deletion of layers at the reservoir surface is removed, (2) numeri~al dispersion is reduced because no vertical flows between layers are required. The mixed layer models have worked quite well for lakes and reservoirs in temperate climates. Admitedly there are a fair number of "constants" in the TKE budgets employed in the various models and these have yet to be assigned "universal" values. The following section will discuss attempts to modify existing models for cold region applications. Modification of Existing Models for Cold Regions In the mid 1970's the Orlob-WRE model was modified to include a freeze-thaw cycle for applications to two impoundments near the US-Canada border (Chen and Orlob, 1973) (Norton and King, 1975). An additional term for the heat exchange accompanying the change of state of water from liquid to solid was added to the heat budget equation (1). When freezing temperatures are reached, further loss of heat proportional to the latent heat of fusion, results in ice formation. Other surface heat transfer processes (e.g., evaporation) were modified to account for the ice sheet. Findikakis, Locher and Ryan (1980) modified the MIT mixed layer model to include ice cover formation and melting. They noted that if wind mixing in the surface layer is underestimated, the simulated water surface temperature decreases at a faster rate than the observed, resulting in an early formation of ice in the reservoir. Their model and field data comparisons were made for Spada Lake, a reservoir on the Sultan river in Washington. Additional modifications to existing models include a proposal by Ashton at CRREL (Ashton, 1982) to include an ice and snow cover in CE-QUAL-Rl, the water temperature and quality model currently being used by the Waterways Experiment Station (Environmental Laboratory, 1981). The hydrothermal component of this model is a modified version of the Orlob-WRE model. The modifications to the original WQRRS model to form CE-OUAL-Rl include the change to a Lagrangian grid and the replacement of the eddy diffusivity approach by a mixed layer algorithm. (Ford, et al. 1980). Ashton's cold region modifications included the following "threshold conditions" to be met before an intact ice cover is established: volume averaged water temperature < 2°C average wind speed < 5 m/s average daily air temperature < -5°C Ashton states that the choice of water temperature less than 2°C requires that the water mass be cooled below the 4°C stability point due to surface cooling. 42 This criterion therefore implicitly considers the effect of wind mixing in retarding the initial formation of the ice cover. The requirement that the wind velocity be less than 5 m/s is based on his experience that strong winds prevent ice formation. The threshold air temperature of less than -5°C is a subjective one that allows the ice sheet to attain a thickness sufficient to resist break up. As soon as an intact ice cover is formed the wind stress is set equal to zero. There has been an extension of DYRESM to include an ice cover by Patterson and Hamblin (1983). They allow the ice cover to gradually decouple the wind stress from the surface layer dynamics. A further modification to DYRESM to include suspended sediment as well as ice has been described by Wei and Hamblin (1986). Independently, Gosink (1986) formulated a snow and ice cover algorithm for DYRESM. These will be discussed in the following sections. DYNAMICS AND THERMODYNAMICS OF ICE COVER FORMATION AND DECAY Ragotzkie (1978) and Ashton (1980) have summarized the essential features of the heat budget of lakes including lakes with ice and snow covers. In the latter case the relevant heat flows are shown in Figure 2. Once an ice cover has developed the thermal response of a lake or reservoir to climate undergoes a drastic change. Evaporation, a major component of the heat budget in an unfrozen lake, ceases and radiation becomes the primary mechanism of heat exchange. Hhen snow is added to the ice cover the heat budget becomes much more complex. Sensible heat is stored in water, ice and snow and latent heat is stored in the ice and snow. Defining heat content per unit area on a base of 0°C, the "negative" sensible heat in ice is given by the product of its temperature belo~ °C, its thickness, density (0.92 g/cm ) and specific heat (0.5 cal/°C-g). Latent heat is equal to the mass of ice or snow per unit area times the latent heat of fusion for water (80 cal/g). Heat loss by outgoing longwave radiation from the .... ong WOllE Nad10I1on Reflected Solar Sensible Heat Snowfall * Evaporation * • * * * Fir.. 2 -Heat Flows in a frozeq lake ice or snow surface resu 1 ts in the growth of the ice layer. Some heat enters the water by short wave solar rarliation which is able to penetrate an ice cover. However, if the snow is present on top of the ice the change in the albedo effectively eliminates short wave penetration. One portion of this enterinrs heat is stored in the water while another portion is lost by conduction back through the ice layer. Sensible and latent heat exchanges occur between the atmosphere aTJd the snow or ice surface and snow accumulation represents a negative latent heat input that will have to be balanced by heat addition before the lake loses itS ice cover in the spring. Adams and Lasenby (1978) also discuss the · calculation of latent heat terms for ice and snow energy budgets. Some of the earliest observations of the abi 1i ty of short wave radiation to penetrate an ice cover were made in permanently ice covered Antarctic lakes "Jy Shirtcliffe and Benseman (1964). Their measurements in Lake Bonney with a snow-free ice cover of approximately 3.5 I"' thickness, shown in Figure 3, indicate a maximum temperature of 7.5°C at a depth of 13 m. The density profile is stable 43 because the salt content of the lake increases with depth such that the density in contact with the ice is 1.0 g/ml, whereas at the lake bottom (30 m) the density is 1.2 g/ml. The absorption length for solar radiation (reciprocal of the extinction coefficient ~ in eq. 1) is 8.2 meters. The lake is therefore a natural "solar pond". Stewart (1972) measured isotherms under ice covers in a number of small non-saline frozen lakes in New York and Wisconsin. These lakes show the typical temperature gradient from 0°C at the bottom of the ice cover to 4°C at the lake bottom. Stewart (1973) also measured isotherms under Lake Erie ice and found that in contrast to findings from smaller ice-covered lakes, Lake Erie has almost no winter vertical stratification. For example, in early March most of the lake water was less than O.l°C. Stewart attributed this to the fact that even though the lake is covered extensively with ice there are irregular openings and cracks;and heat exchange and wind mixing are effective in preventing a stable winter stratification. \~ake and Rumer (1979 a) developed a mathematical model for the ice regime of Lake Erie that assumed a vertically mixed water column during the ice season (in accordance with Stewart's observations). They neglected super-cooling effects, snow accumulation on the ice cover and assumed temperature gradients in the ice cover and at the ice-water interface, when they exist, to be linear. They further assumed the ice cover to be static at all times and well drained during the ice dissipation period. The neglect of the effect of accumulated water on the dissipation rate was justified in an earlier study (Wake and Rumer, 1979 b). The model was two-dimensional in the plane of the lake surface and the outputs of the numerical model are water temperature and ice thickness during a 20 day ice growth period from January 1 and a 25 day period in the spring ice dissipation season. Wake and Rumer (1983) extended the model to account for ice transport and deformation due to wind and water stresses. They adopted a macroscopic continuum hypothesis for the fragmented ice field and coupled the thermodynamic model of the previous study with an ice East Distance. m West FIG. 3 -Vertical section of Lake Bonney showing isothermal profiles. transport model. The hydrodynamic component produces two-dimensional ice drift velocity fields under time varying wind inputs. Additional outputs are areal ice concentration fields and areal internal ice pressure distributions. In a later paper Chieh, Wake and Rumer (1983) presented the calibration of the model using observed data from specific Lake Erie ice transport events during the winter of 1979. The simulation for the freezing period agrees quite well. During the melting period the simulation agrees well with the data when wind-driven surface currents and ice melt at the ice-water interface are included. Very few model-prototype comparisons specifically directed at high latitude reservoirs with extensive periods of ice cover have been presented in the literature. Exceptions are the recent studies by Gosink (1986) and Wei and Hamblin (1986) using a data set for Eklutna Reservoir in south central Alaska. The measured temperature profiles were obtained by R & M consultants (1982) for the open water season and by Osterkamp and Gosink (1982) (unpublished field data) for the ice-covered season. Gosink (1986) developed a numerical model (DYRSMICE), a modifies tion of DYRES~1, in \-Jhich the main modification is the incorporation of ice growth and dissipation. This requires a solution of the heat transport equations in the sno\oJ and ice layers as schema tized in Figure 2. A quasi-steady assumption is used for the surface heat flux and the snow and ice temperatures. The heat 44 transfers are assumed to be in equilibrium recognizing that meteorological observations may be available on at most a daily basis at remote locations. For example, the governing heat transfer equation for temperature in the snow layer is k o 2T/oz 2 + ~ ~ exp(-~ z) = 0 (4) s s s s where k 8 is the conductivity of snow, ~s is the short wave extinction coefficient in snow, z is the vertical coordinate, positive downward from the snow surface, and ~s is the short wave radiation. The surface and bottom boundary conditions are T(z=O) = T8 and T(z=h ) = Ti where h 8 is the snow dept~ and Ti is the temperature at the snow-ice interface. Because of the steady state assumption an analytical solution to equation (4) can be obtained. Since T8 and Ti are not known a priori, they must be determined by simultaneous consideration of the surface heat fluxes and the temperature distribution in the ice layer. The latter requires solution of a heat transport equation similar to equation (4) for the lower ice layer. Wei and Hamblin (1986) modified DYRESM to include a snow and ice cover in a similar manner. Their model includes turbulent transport of heat from \-Ja ter to ice due to the under-ice flow field generated by inflows and outflows and the formation of frazil ice. They compared, predicted and measured ice cover thickness as well as reservoir isotherms. This section, on dynamics and thermodynamics of ice cover formation and ~cay, will conclude with a brief assessment of some future research needs in this area. Prior to the formation of an ice cover, evaporation is one of the dominant terms in the heat budget. Fog may form in ilie surface layer above a lake or reservoir when the air becomes saturated due to the temperature having fallen to the dew point. Thus fog may occur with a higher frequency in cold climates with resu 1 ting condensation or negative evaporation. Resea reb is needed on techniques for factoring in the occurrence of fog on surface heat exchange. Some research in this area has been done in connection with pre dieting the formation of fog induced by cooling ponds (Tsai, 1980). Reservoirs used for pumped storage power production experience large daily fluctuations in water level. Very little is known about the effect of large water elevation changes on ice formation and decay. Karnovich et al. (1983) presented some operating experience on ice effects on pumped storage reservoirs in the USSR. There is a need for additional field data and analysis on this problem. The prediction of the decay and deterioration of ice covers during spring melt appears to be a somewhat more difficult problem than the formation and growth of ice. Ashton ( 1983, 1985) has analyzed the deterioration process by applying an energy budget to determine when and where the ice temperature is belov1 freezing and where it is at freezing but sufficient energy has been absorbed to melt a portion of the ice. The resulting ice porosity is related to failure stress and this may be used to predict the load-carrying ability of the ice· sheet. Snow cover generally prevents solar radiation from penetrating to the ice cover; however, when the snow cover is gone conditions are set for deterioration to begin. It is unf or tuna te that data does not exist for the testing of these hypotheses. 45 WIND STRESS WITH WEAK STRATIFICATION AND ATTENUATION DUE TO ICE FORMATION In most aspects the application of mixed layer models, that have been developed and tested on temperate climate lakes and reservoirs, to cold region water bodies is straightforward. Two exceptions are (1) the effect of surface wind stress induced mixing under open water conditions with relatively weak stratification and (2) the attenuation of surface wind stress induced mixing due to the formation and growth of an ice cover. Mixing with Weak Stratification Lakes and reservoirs in temperate climates tend to stratify rapidly and strongly during the spring v1arming period. In summer, surface to bottom temperature differences of the order of 20°C are common. The resulting strong stratification effectively isolates the hypolimnion in regard to mixing. In fact, if a late spring temperature profile is used as an initial condition, it is usually found that heat transfer in the hypolimnion throughout the summer is adequately represented by molecular diffusivity. However, it has also been observed, during the short period of weak stratification following the spring overturn, that episodic mixing events are necessary to explain the slight warming of the hypolimnion. Because the period of weak stratification is a matter of a few weeks, data is scant and the question of hypolimnetic mixing has not received the attention it deserves. By an interesting coincidence, related to the non-linear temperature-density relationship for water, stability conditions in tropical water bodies are quite similar to those of cold regions. For example, a tropical lake having a maximum surface-bottom temperature range from 28° to 26°C has the same density stability as a northern lake with a temperature range from 13° to 4°C. Harleman, Adams, Aldama and Bowen (1986) used Lewis' (1983) data from Lake Valencia, Venezuela (which has seasonal temperatures in the 28°-26°C range) for determining an improved hypolimnetic diffusion component for the MIT mixed layer model. The expression for Ez in equation ( 1) is E = z ( 5) where CU is a dimensionless (non-universal) parameter that probably depends on water body area and shape; u* is the friction velocity, u* = <~s/p)l/2, where ~s is the water surface wind shear stress; As is the surface area; h, the total lake depth and Ps the potential energy of the stratification, defined by Jh -P = [p-p(z)] g Azdz s 0 (6) where p is the volumetric mean density of the water body. This parameterization of the hypolimnetic diffusivity which is capable of capturing episodic wind mixing events is similar to that proposed by Imberger and Patterson (1981) except that it was found to be unnecessary to include their local stratification effect. It is necessary to use an upper bound for Ez during isothermal conditions when Ps approaches zero. The value of Ez = l.lo-4 m2 /s (about 700 times the molecular diffusivity of heat) recommended by Imberger and Patterson (1981) was used. It is suggested that the diffusivity relation represented by equations (5) and (6) could be used for open water conditions in cold climates. Farmer (19f.n) discusses some special features of mixed layer dynamics near the temperature of maximum density. In particular, he points out that as the temperature cools through 4°C,· the surface buoyancy flux changes sign and the wind works to drive the heat deeper. Gosink (1986) in her discussion of modifications to DYRESM refers to the parameterization 46 of the surface stress term in the TKE budget and indicates that in high latitudes it may be necessary to introduce a Coriolis "cut-off" time scale to limit mixed layer deepening. Wind Stress Attenuation Due to Ice Formation Ashton's (1982) modification to CE-QUAL-Rl to induce ice formation hased on "threshold conditions" has already been mentioned. Based on experience on Eklutna Reservoir, Gosink's (1986) modification to DYRESM differs in that she elects to modify the wind stress by applying a damping co~fficient that decreases linearly from 1 to 0 as the calculated ice thickness increases from 0 to 10 em. This allows for partial ice cover development in the early stages of formation. It would be interesting to see how well the modified model predicted the formation and growth of the Eklutna ice cover. SUSPENDED SEDIMENT EFFECTS Gosink (1986) noticed difficulties in calculating the heat balance in Eklutna Reservoir in the late summer which she attributed to an order of magnitude change in the short wave extinction coefficient caused by the inflow of high turbidity glacier melt water. Incorporation of this effect into a predictive impoundment model requires inclusion of a mass conservation equation for suspended sediment and modification of the equation of state to incorporate sediment as well as temperature effects on density. Findikakis, Locher and Ryan (1980) added a suspended sediment mass balance equation in their modification of the MITEMP model for Spada Reservoir. They expressed particle concentration in turbidity units by assuming a linear relationship. The sediment inflows in this reservoir consisted of very fine clay material and they assumed that settling was negligible. In this aspect the situation may be similar so that encountered with "glacial flour" inflows. Dhamotharan and Stefan (1980) modified MLTM to include the simulation of nservoir turbidity. The resulting model, designated RESQUAL, has been applied to Lake Chocot, Arkansas. The Corps of Engineers CE-QUAL-Rl model, previously discussed also contains suspended sediment as a state variable. Wei and Hamblin (1986) extended DYRESM to include the concentration of suspended sediment. Density inversions resulting from suspended sediments are checked and mixed to establish stability. Settling velocities appropriate to the size range of the inflowing sediments are input to the model. Additional research is needed into ilie settling characteristics of fine sediments characteristic of glacier fed reservoirs. Floculation and coagulation effects must be recognized since much of the material of interest is outside of the discrete particle settling range. LOCAL MIXING AND WATER QUALITY EFFECTS The interaction between seasonal ice cover and hypolimnetic dissolved oxygen depletion in lakes and reservoirs is beginning to receive attention. Thomas and Orn ( 1982) report on DO conditions in ilie Greifensee, a Swiss Alpine lake. Disastrously low oxygen conditions result if low mixing during autur.m cooling is followed by an early ice cover of long furation, succeeded by rapid stratification after the ice melt. Babin and Prepas ( 1985) report on efforts to Mdel winter oxygen depletion rates in ice covered lakes in Canada. Jain (1980) analyzed the plunging phenomena at the en trace to reservoirs using gradually varied two-layer flow theory. A limitation of this type of analysis is that it does not account for the three-dimensional lateral spreading r.haracteristic of actual reservoirs. Fischer and Jensen ( 1983) made ~servations of reservoir inflow mixing in Lake Mead. A tributary inflow was dyed continuously for nine days and permitted computation of the entrainment into the inflow at the plunge point. They found that the entrainment flow was about equa 1 to the inflow. However, a clear correlation between inflow slope, densimetric Froude number and other hctors remains to be established. 47 Coleman and Armstrong (1983) measured horizontal diffusivities in a small, ice-covered lake by injecting a tracer as a point source under the ice cover. A number of studies have been carried out to investigate the effects of heated effluents or waste discharges into ice-covered water bodies. Ashton (1979) prepared a comprehensive report on the suppression of ice by thermal effluents. In the first part he considers the effluent to be fully mixed across the flow section while in the second part he assumed a side discharge. A similar study of mixing in the near bank zone is reported by Gerard, Putz and Smith (1985). Significant differences were found between open water and ice-covered conditions. Belore and McBean (1983) carried out laboratory experiments on entrainnent of heated discharges at the ice-water interface. In contrast to open water conditions the dilution was founri t-> have relatively little dependence on the discharge densimetric FroutiP. number. Wells and Gordon (19e()) discussad :_·,'"' need for multi-dimensional reservoir models for water quality management using field data from two TVA reservoirs. They concluded that there was no justification for 3-D models, that under certain circumstances 2-D (longitudinal and vertical) models might be needed but that for the most part 1-D models were adequate. CONCLUSIONS A number of one-dimensional reservoir models have been modified to include cold region effects. In most instances the applications have been limited to temperate region lakes and reservoirs in which ice cover is a secondary effect of relatively short duration. It is clear that much additional research is needed for the prediction of temperature and water quality effects in high latitude regions in which ice cover is the dominant effect. In this respect the most pressing need is for field data. Theoretical advances necessary to improve mathematical modeling will continue to be limited by the sparcity of cold region lake and reservoir data. A significant data collection resource that does not appear to have been adequately utilized is that of satellite observations. Stewart (1985) discusses a number of remote-sensing techniques that are directly applicable to ice observations. These have been widely used in oceanographic studies of the polar ice regions. The techniques include Landsat visible light scanners (VISSR) and radio-frequency radiation signals. The latter provide open water surface temperatures to an accuracy of l°C, areal ice-open water concentrations and distinction between first year and multi-year ice. Synthetic-aperture radars accurately map reflectivity changes over areas 50 to 100 km on a side with resol tion of 10-40 m Simultaneous modeling and data collection efforts are vitally needed. It is unfortunate that research in lake and reservoir managment has not had access to the data collection resources that have routinely been associated with ocean0?raphic research in polar regions. References Adams, W.P. and n.c. Lasenby, "The role of ice and snow in lake heat budgets," Limnol. Oceanogr. 23(5), 1978. Ashton, G.D., "Suppression of river ice by thermal effluents," CRREL Report 79-30, u.s. Army Corps of Engrs, Cold Regions Research and Eng. Lab., 1979. Ashton, G. D., "Freshwater, ice growth, motion and decay," Ch. 5, Dynamics of Snow and Ice Masses. Academic Press, 1980. Ashton, G.D., Details of an ice cover algorithm for the CE-QUAL-Rl reservoir water quality model, unpublished note, 1982. Ashton, G.D., "Lake ice decay," Cold Regions Sci. and Tech., 8, 1983. Ashton, G.D., "Deterioration of floating ice covers," Trans. ASME, Jour. of Energy Res.Tech., Vol. 107, June 1985. Babin, J. and E.E. Prepas, "Modeling winter oxygen depletion rates in ice-covered temperate zone lakes in Canada," Canadian Jour. of Fish. and Aq. Sci., Vol. 42, No. 2, Feb. 1985. 48 Belore, R.C. and E.A. McBean, "Physical modelling of dilution entrainment of heated discharges under ice," 20th IAHR Congress, Uoscow, Vol. II, 1983. Bloss, S. and D.R.F. Harleman, "Effect of wind-mixing on the thermocline formation in lakes and reservoirs," Tech. Rep. No. 249, R.M. Parsons Lab., MIT, Nov. 1979. Bloss, S. and D.R.F. Harleman, "Effect of wind-induced mixing on the seasonal thermocline in lakes and reservoirs," 2nd Intern. Symp. on Stratified Flows, Trondheim, Norway, 1980. Chen, C.W. and G.T. Orlob, "Ecologic study of Lake Koocanusa," Rep. to US Army Corps of Engineers, Dist. of Seattle by Water Res. Eng., Inc., 1973. Chieh, S. H. , 1\. Wake, and R. Rumer, "Ic- forecasting model for Lake Erie " ' ~roc. ASCE, Vol. 109, Jour. of Waterway, Port, Coastal and Ocean Eng., No. 4, Nov. 1983. Coleman, J.~. and D.E. Armstrong, "Horizontal diffusivi ty in a small, ice-covered lake," Limnol. Oceanogr. 28(5), 1983. Dhamotharan, S. and H. Stefan, "Mathematical model for temperature and turbidity stratification dynamics in shallow reservoirs, ASCE Symp. on Surface Water Impound., Vol. I, Univ. of Minn., 1980. Environmental Laboratory, CE-QUAL-Rl: "A numerical one-dimensional model of reservoir water quality; User's Hanual," Tech. Rep. E-18-8, U.S. Army Engineer Haterways Exp. Station, Vicksburg, Miss., 1981. Farmer, D.M., "Mixed layer dynamics in a lake near the temperature of maximun density," 2nd Int. Symp. on Stratified Flows, Trondheim Norway, 1980. Findikakis, A.N., F.A. Locher and P.J. Ryan, ASCE, Symp. on Surface Water Impound., Vol. I, Univ. of Minn., 1980. Fischer, H.~. and A.R. Jensen, "Observations of reservoir inflow mixing," 20th IAHR Congress, Vol. IV Moscow, 1983. Ford, D.E. K.W. Thornton, A.S. Lessem, and J.L. Norton, ASCE, Symp. on Surface Water Impound., Vol. I, Univ. of ~inn., 1980. ------------------~----~-~----~-- Gerard, R., G. Putz and D.W. Smith, "Mixing in the near-bank zone of a large northern river," 21st IAHR Congress, Melbourne, Australia, 1985. Gosink, J.P., "Northern lake and reservoir modeling," submitted to Cold Regions Science and Technology, 1986. Harleman, D.R.F. and K.A. Hurley-Octavio, "Heat Transport Mechanisms in Lakes and Reservoirs," 17th IAHR Congress, Baden Baden, August 1977. Harleman, D.R.F., "Hydrothermal analysis of lakes and reservoirs," ASCE Proc. Vol. 10~, No. HY3, 1982. ~rleman, D.R.F., E.E. Adams, A. Aldama and J. Bowen, "Hypolimnetic mixing in a weakly stratified lake," R.M. Parsons Laboratory, MIT, (in preparation), 1986. Huber, W.C., D.R.F. Harleman and P.J. Ryan, "Temperature prediction in stratified reservoirs," Proc. ASC~, Vol. 98, No. HY4, 1972. Irnberger, J., J. Patterson, B. Hebber'., and I. Loh, "Dynamics of Reservo: r of Medium Size," Vol. 104, No. HY5, Proc. ASCE, 1978. Imberger, J. and J.C. Patterson, "A dynamic reservoir simulation model - DYRESM:5." Transport Models for Inland and Coastal Waters, H.B. Fisher (Ed.) Academic Press, New York, p. 542, 1981. Jain, S.C., "Plunging phenomena in reservoirs," ASCE, Symp. on Surface Water Impound. Vol. IV, Univ. of Minn. 1980. Karnovich, V.N., "Winter operation of pumped storage power plant basins and canals," 20th IAHR Congress, Moscow, Vol. II, 1983. Lewis, W.M., "Temperature, heat and mixing in Lake Valencia, Limnol. Oceanogr. ( 2 8) , No. 2, 19 8 3. Norton, W.R. and I.P. King, "Mathematical simulation of water temperature to determine the impact of raising an existing dam," Prog. in Astro. and Aero. (36), 1975. Orlob, G.T.,(Ed.), Mathematical Modeling of Water Quality: Streams, Lakes and Reservoirs. (Ch. 7, "one-dimensional Models" by G.T. Orlob; Ch. 8, Two and Three-Dimensional Models by M. Watanabe, D.R.F. Harleman and O.F. Vasiliev) Wiley, 1983. 49 Patterson, J.C. and P.J. Hamblin, "Simulation of a lake with winter ice cover," Env. Dyn. Rep. ED-83-043, Univ. of Western Australia, 1983. R & M Consultants, Inc., "Alaska Power Authority, Susitna Hydroelectric Project, Glacial Lake Studies," Report prepared for Acres American Inc., Buffalo, N.Y., 1982. Ragotzkie, R.A., Heat budgets of lakes, Ch. 1, in Lakes, Chemistry, Geology, Physics. A. Lerman (Ed.) Springer-Verlag, 1978. Shirtcliffe, T.G.L., and R.F. Benseman, "A sun-heated antarctic lake," Jour. of Geophy. Res., (69) No. 16, 1964. Stefan, H. and D.E. Ford, "Temperature Dynamics in Dimictic Lakes," Proc. ASCE, Vol. 101, HYl, 1975. Stewart, K.M., "Isotherms under ice," Verh. Internat. Verein. Limnol. 18, 1972. Stewart, K.M., "Winter conditions in Lake Erie with reference to ice and thermal structure and comparison to lakes Winnebago (Wisconsin) and Mille Lacs (Minnesota)," Proc. 16th Conf. Great Lakes Res., 1973. Stewart, R.H., Methods of Satellite Oceanography, Univ. of Calif. Press, 1985. Thomas, E.A. and C.G. Orn, "Ice cover and hypolimnetic reoxygenation in the Greifensee from 1950 to 1980," Schweizerische Zeitschrift fur Hydrologie, Vol. 44, No. 1, 1982. Tsai, Y.J., "Predictive Model of Fog Induced by Cooling Pond," ASCE Symp. on Surface Water Impound., Vol. II, Univ. of Min •• 1980. Wake, A. and R.R. Rumer, "Effect of surface meltwater accumulation on the dissipation of lake ice", l\la ter Resour. Res., (15) No. 2, 1979. Wake, A. and R.R. Rumer, "Modeling ice regime of Lake Erie," Proc. ASCE, Vol. 105, No. HY7, July 1979. Wake, A. and R.R. Rumer, "Great Lakes ice dynamics simulation," Proc. ASCE, Vol. 109, Jour. of Waterway, Port, Coastal and Ocean Eng., No. 1, Feb. 1983. Wei, C.Y. and P.F. Hamblin, "Reservoir water quality simulation in cold regions," Proc. Cold Regions Hydrology Symp., Am. Water Res. Assoc., Fairbanks, Alaska 1986. \.Jells, S.A. and J.A. Gordon, "A three-dimensional field evaluation and analysis of water quality - management and modeling implications of the third-dimension," ASCE, Symp. on Surface Water Impound., Vol. I, Univ. of Minn. 1980. 50 WATER, SNOW AND ICE MANAGEMENT 51 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 WATERSHED TEST OF A SNOW FENCE TO INCREASE STREAMFLOW: PRELIMINARY RESULTS 1 Ronald ~ Tabler and David L. Sturges ABSTRACT: Although snow fences have long been proposed as a method to increase water yield, the feasibility of this prac- tice has not been established. This paper reports hydrologic changes the first 2 years after construction of an 800-m-long snow fence, 3.78 m tall, on a 307-ha paired watershed having 10 years of pretreatment data. Although final assessment will require more years of study, initial treatment effects are so apparent that preliminary results are meaningful. Snow accumulation increased about 58%, and streamflow averaged 237% of that predicted by the pretreatment calibration with the control watershed. Flow duration on the ephemeral stream increased about 18 days. Overall water- yield efficiency of the snow fence drift averaged about 43%; however, the yield efficiency of the added snow was about 85%. To amortize the fence construction cost over a 25-year physical project life, the value of the new water would have to be about $0.06721 per cubic meter. (KEY TERMS: snow fences; snow management; water yield improvement; blowing snow.) INTRODUCTION Although snow fences can augment water supplies by altering the distribution of snow, they also make "new" water available by reducing the evaporative losses from blowing snow particles (Tabler, 1973). In Wyoming, for example, more than half of the blowing snow evaporates over a transport distance of 3000 m (Tabler 1975). It should also be possible to pro- long snowmelt runoff by using snow fences to increase snow depth. Indeed, snow fen- ces may have been invented as a method to augment water supplies because, according to Brown (1983), the first known reference to snow fences was in an 1852 Norwegian publication describing snow fences to pro- vide water for livestock. Another early reference was an article that appeared in an Alaskan newspaper (Seward Weekly Gateway, 1909) describing how miners used snow fences to augment water supplies for placer mining. Since then there has been considerable speculation on the possibi- lity of using snow fences to increase water yield or local water supplies (Lull and Orr, 1950; Martinelli, 1959; Berndt, 1964; Costin, 1968; Tabler, 1968; Swank and Booth, 1970; Tabler, 1971; Tabler and Johnson, 1971; Martinelli, 1973; Rechard, 1973; Saulmon, 1973; Tabler, 1973; Cooley et al., 1981; Sturges and Tabler, 1981). But~espite these numerous references, no studies have shown an effect of snow fen- ces on streamflow. The snow fence treat- ment on the only watershed-based study (Cooley et al., 1981) was too small to have a significant hydrologic effect, leaving in question the feasibility of this form of snowpack management. This paper reports hydrologic changes during the first 2 years after construc- tion of a large snow fence on a "calibrated" paired watershed having 10 years of pretreatment streamflow data. Although final assessment will require more years of study, initial treatment effects are so apparent that preliminary results are meaningful. 1Hydrologist and Research Forester, respectively, Rocky Mountain Forest & Range Experiment Station, Forest Service, USDA; 222 South 22nd Street, Laramie, Wyoming 82070. 53 STUDY AREA AND METHODS The study watersheds are part of the Stratton Sagebrush Hydrology Study Area (41° 26'N, lOJC 07'W), located about 32 km west of Saratoga in south-central Wyoming. Elevation of the gently rolling terrain ranges from 2320 m at the lowest stream gage, to 2440 m at the upper watershed divide. Annual precipitation averages 53 em, about half of ~.Thich falls as snow from November through March. Precipita- tion data are from a well-sited recording gage located in a forest grove about 4.5 km from the streamgage on the fenced watershed. Mean monthly values of preci- pitation, air temperature, and wind speed are shown in Figure 1. Because snow is blown off of wind,.Tard slopes and ridges, and redeposited on leeward slopes and in topographic depressions, snow depths prior to melt range from a few centimeters on windward slopes to 6 m or more in topographically controlled deposition zones. Q; Hl ' ~ 8 E3 6 w fli 4 0 2 z ; 111 ""' u 0 '-' 2111 lJ.J c:t: ~ 1111 < c:t: lJ.J ~ 111 lJ.J I- c:t: -1111 ...... < 'E _:; 6 z 0 ...... 4 I-< I-...... 2 a_ ...... u lJ.J 111 c:t: a_ J F M A M J J A S 0 N D Figure 1. He~m monthly wind speed, air temperature, and precipitation at the Stratton Study Area, 1967-1981. 54 Soils developed in place from sand- stone of the Brm.rn' s Park Formation, and have a loam or sandy loam texture in the A ann B horizons. Both infiltration and soil stability are good. On moist sites, ve.~etation is dominated by mountain big sagebrush (Artemisia tridentata ssp. vaseyana) having a typical height of 60 em. tolyoming big sagebrush (A. t. ssp. wyomingensis ), typically about 20 em tall, occupies upland areas where effective pre- cipitation is less because of snow removal by wind. Hydrologic data have been collected on the Stratton Study Area since 1967 when streamgages were installed on the two perennial ·streams, Loco Creek and Sane Creek, to determine the hydrologic effects of sagebrush eradication. Although a final publication on this aspect of the research is sti 11 in pre par at ion, hydrolo- gic characteristics of the area have been reported bv Sturges (1975a, 1975b, 1979). Loco Creek (664 ha) has remained untreated, and serves as the control watershed for the study reported here. The 307-ha snow-fenced watershed (North Draw) is drained by an ephemeral stream charged primarily by snowmelt, with very little runoff from summer storms. This watershed has been continuously gaged since 1974, and remained untreated until the summer of 1983 when the snow fence was built. The primary streamgages are 120-degree V-notch weirs having cutoff walls extending about 3 m below the surface and equipped with water-stage recorders. The entire weir pond, measurement section, and discharge apron of each streamgage are housed within a steel multiplate pipe arch (Figure 2) for protection against the deep snowdrifts in the stream channels (Johnson and Tabler, 1973). A Parshall flume was installed immediately above the snow- fenced channel reach in 1983. Permanent snow courses have been measured since 1968 on Loco and Sane Creek watersheds. N.easurements on two permanent snow courses crossing the lower North Draw channel were hegun in 1969 and measure- ments began in 1979 on seven additional snow courses to better quantify channel snow storage in anticipation of the snow fence treatment. Although snow densities are measured each year along selected transects using a Federal snm.T sampler, Figure 2 . Construction of the 3 .78-m-tall snow fence on ~orth Draw watershed in 1983, with the streamgage shown in foreground . these values are not representative of true snow densities in the deeper drifts because of measurement and sampling errors . ~ater -equivalents reported here are therefore calculated using the rela- tionship between snow density and snow depth reported by Tabler (1985). D~SIGNING THE SNQ{-l FENCE TREATt1ENT The most logical location for a snow fence was such as to augment snow deposi- tion along the lowest 815 m of the main stream channel. This section, having a mean azimuth of 292°, was closest to being perpendicular to the prevailing southwest winds, and natural snow storage capacity was inadequate to store all transported snow in winters with average precipita- tion. In addition, concentrating the snow in the primary channe 1 would minimize snowmelt conveyance losses . This section of channel is the principal snow accumula- tion area on the North Draw watershed becaus e it is incised about o m below the surroundi ng t e rrain . Snow depths elsewher e are controlled prir:~.ari ly by sagebrush height, and range from only a few centimeters to a meter or so at peak accumulation. It is estimated that prior to fencin~, the drift th at formed in the incised chan- nel contributed 60% of total discharge, even though 79 % of the total watershed area lies above thi s section . Snow 55 3 storage capacity averaged abont 7H m of water-equivalent r>e r meter of channe 1 3 length, for a total of about 64, 000 m (52 acre -ft). Although the efficiency with which blowing snow was deposited in this topog raphic feature was unknown, a SPJ.ooth snow surface extended across the channel in most years, implying a relatively low trapping efficiency by the end of the accumulation seaso n and suggesting that a snow fence would increase deposition. Above this incised section, the channel branches into shallow tributaries that are nearly parallel to the wind and thus not as well suited to fencing . The length of the snow fence (80 0 m) was therefore dic- tated by the length of the incised chan- nel . Although it would be possible to construct additional snow fence on the upper part of the watershed, we would expect the yield efficiency to be less than for fencing along the incised channel section . The type of fence selected was the standard wood snow fence used by the \lyoming Highway Department since 1971 (Figure 2). This structure consists of horizontal hoards 15 em wide separated by 15-cm spaces, a bottom gap of approxima - tely one-tenth of the fence height, and a 15-degree inclination downwind (Tabler, 1974). Net porosity (open area) of the structure, excluding the bott om gap , is about 48%. This type of fence was selected because it is known to be effec- tive, economical t o builrl and :naintain, and has a known capacity and drift geometry (Tabler, 1980). At the time the snow fence treatl'l.ent was designed, plans were available for heights of 1.>53, 2 .44, 3 .17, and 3 .78 m. The 3 .7H-m hei ght was chosen as that reqni red to store the maxi - mum snow transport anticipated ove r the 25-year physical life of the snow fence , using the equation developed by Tabler (1975): Q = 1515 Pr [ l-(O.l4)R/3 0JO ] (l) where Q is snow tr ansport in cubic meters of water-equivalent per meter of width across the wind , Pr is relocat ed water- equivalent precipitation in mete r~, and R is the fetch or "contributing" distance in meters . This equation has been used to desig n successful snow fe :1.ce systems for highway protection (Tahler and Furnish, 1982) and water supply augmentation (Sturges and Tabler, 1981 ), and to design snow fences on Alaska's North Slope, including the system constructed at the village of Wainwright in 1982. For the North Draw watershed, mean winter precipi- tation (November 1 through March 31) is estimated to be 0.24 m. Of this amount, about 16 em would be retained by the vege- tation, which averages about 45 em tall. The maximum transport was estimated using a value for winter precipitation having an e~ceedance probability of 4% (return period of 1 year in 25), as computed from the frequency distribution reported by Tabler (1982). This value was 1.5 times the mean, or 0.36 m. Pr was then computed to be 0.20 m by subtracting the anticipa- ted snow retention by vegetation (0.16 m). The fetch distance, R, was taken as the average distance (2000 m) between the North Draw channel and the nearest snow accumulation feature upwind. Although some blowing snow is expected to escape these upwind deposition areas, which are similar to the North Draw channel, using this fetch distance assured a conservative (low) estimate of snow transport. Maximum snow transport was thus esti- mated from Equation (1) to be about 190 m3 water-equivalent per meter of channel length. A 1·78-m-tall fence will store about 172 m on level terrain if both windward and leeward drifts are considered (Tabler, 1980). Total storage 3 capacity would therefore be about 250 m per m~ter of fence length after adding the 78 m storage in the channel. The extra storage capacity provided by using a 3.78-m-tall fence assured that most of the blowing snow would be deposited, because snow trapping efficiency does not decrease significantly until a fence is filled to about RO% of its capacity (Tabler, 1974). The fence line was curved to parallel the stream channel at an average distance of 38 m (ten times the fence height) upwind (Figure 3). This placement was chosen to minimize the distance snowmelt runoff had to travel to the channel, while being far enough upwind that the drift would not be expected to block the stream channel in a winter with average (or below) precipitation. However, because it is necessary to gage the streamflow accurately for research purposes, per- forated plastic drain pipe was installed 56 in the stream channel to provide for the possibility of flow obstruction by the drift. In addition to the main 800-m fence, a separate snow fence, 112 m long and 3.8 m tall, was installed 200 m downwind of the North Draw channel to provide a measure of the trapping efficiency of the main fence (Figure 3). This fence was located near the watershed boundary, and because of the small size of the drift in relation to total channel snow storage (and its distance from the channel) it would not be expected to have a discernible effect on streamflow. The fences were built during the 1983 summer after a calibration neriod sufficient to detect a 15% change in streamflow from snowmelt. HYDROLOGIC EFFECTS OF THE SNOW FENCE Nearly identical results were obtained the first 2 years after the fence was built, even though the 36 em of precipita- tion the first winter (1983-84) was the highest in 14 years of record on the study area, and the 22 em recorded the second winter was about average. A comparison of snow accumulation on a single snow course in the incised channel section before and after fencing is shown in Figure 4. Based on the pretreatment relationship between maximum snow accumu- lation on two index transects on the control and treated watersheds (Figure 5), snow accumulation was 54% and 62% greater the first and second year after fe~cing, respectively, than would have occurred without the fence (Figure 6). Pretreat- ment data in Figure 'i are restricted to values at peak accumulation; after fencing, however, the snow courses have been measured on several dates each winter to provide an estimate of the future effect of treatment on snow retention. The ratio of the slopes of the snow accu- mulation relationships in Figure 5 suggests that before fencing, efficiency of the stream channel in trapping blowing snow averaged about 56% of that with the fence in place. Peak snow water-equivalent volume stored by the downwind fence was less than the precipitation relocated between the fences in both posttreatment years, -- Figure 3 . Aerial view of No r th Draw watershed on 25 1arch 1985 , the date of peak accumulation . The wind is from the upper left, and the streamgage inst r ument shelter LS visible n e ar the bottom of the photo . "' E ~ v IS WINO -> ~ ~ ':1------ iil 0 0 10 20 30 40 50 60 70 DISTANCE Cm> Figure 4 . Snow profiles on a North Draw snow course before and after fencin g , in years wit~ similar winter precipita- tion . 57 STRATTON STUOY AREA WATERSHEDS 300 " I .,.,)1~ PEAl< N ~ I AFTER <•> SNOW FENCING• ~I 0 Y •1.7SX-36 , X>S0 LIJ I (r -0 . 99) I (f) 0::: 9 5 PE~ I LIJ ..... I* < 200 I 31< 0 I LIJ ':'·· u z LIJ I IJ... z 0 z 0 10 0 Col SNOW FENCING• .... Y•0.96X+3o (.--0. 96) ..... < ..J ::l ::£ . · ... '\..._ 0. 95 CONF IOENCE INTERVAL ::l u u < 31< 0 z 0 (f) 0 100 200 300 SNOW ACCUMULATION ON UNFENCED WATERSHEDS (m 2) Fi~ure 5 . Snow accumulation on North Draw wa te rshed (average cross-sectional area of two transects), versus the average of t wo transects on untreated wa tersheds . Dat a before fencin g are at peak accumulation only . Posttreatment values include measurements taken through the accu~1latton pe ri ods as well as at peak accumulation. Dotted lines indicate the 95 % confiden ce in t erval for the pretre~tment regression . ---.1. Figure 6 . Snowdrift on the lee side of the fence at time of peak accu mulation in 1985 . The top of the streamgag e ins t rument shelter is visible in t he background . suggesting that no significant transport escaped the main snow fence. Snowmelt discharge on the snow-fenced watershed was computed as total annual streamflow minus any rainstorn runoff occurring after the first cessation of flow. On the control watershed, snowmelt discharge was calculated as total streamflow from the first to the last day of sno\Ymelt surface runoff, minus mean base flow for the interval. Snowmelt discharge was 232% of that predicted by the pretreatment regression the first year, and 242% the second year (Figure 7), both exceeding the 99% con- fidence limits for the pretreatment rela- tionship. The fact that the 137% average increase in streamflow has been greater than the percentage increase (58%) in snow water-equivalent volume suggests a higher efficiency of water yield from the addi- tional snow caught by the fence. The ratio of the increase in streamflow to the ,.., E 4 STRATTON STUDY AREA WATERSHEDS 0 '-" CD BEFORE SNOW FENCEo Y • B.23BX-8.87 o (r • 111.96) LLo 3 LL.w * WITH SNOW FENCE CJ:r ZUl ::::la: * a:w ...... 2 _J< IJJ:Jt ::Eo :Jtw ou Zz Ulw LL. z Ill CJ Ill 1 2 3 4 5 6 7 SNOWMELT RUNOFF ON UNFENCED WATERSHED (em) Figure 7. Annual snowmelt discharge on the snow-fenced watershed in relation to that on the control (Loco Creek). The pretreatment regression is shown as a solid line, with dotted curves indi- cating the 95% confidence interval. increase in water-equivalent snow storage was 0.78 in 1984, and 0.92 in 1985, suggesting an average yield efficiency of 85%. This high efficiency is attrihuted to the fact that soil water recharRe requirements are ~atisfied independently of the additional snow, and that evapora- tion losses are Lelatively small during the extended melt period. It is possible 58 to estimate overall yield efficiency of the fence drift hy subtracting flow through the flume above the snow-fenced channel section from total watershed dischan~e, and comparing this volume with water-equivalent snow storage. In 1984 the efficiency calculated in this manner was 44%, and in 1985 it was 42%. These values are consistent with the evap•)rat ion losses from snowdrifts reported by Rechard (1975), suggesting that other conveyance losses were small. If streamflow continues to average 237% of pre-fencing levels, annual water yield would be increased by an average of 28,200"rn 3 (22.9 acre-ft). Thefence construct~on cost (including site preoara- tion) was $59.25 per lineal meter, for a total of $47,390. In order to amortize the initial construction cost of the fence over the anticipated life of 25 years, the value of the incr~ased water would have to be $0.06721 per rn ($83 per acre-ft). Flow duration on the fenced watershed was extended 19 days in 1984, and 18 days in 1985, based on the pretreatment rela- tionship with the control watershed (Figure 8). Observed values exceeded the 95% confidence limits for the pretreatment relationship. CONTRAST WITH AN EARLIER SNOW FENCE TEST The dramatic streamflow effects of the North Draw snow fence contrast markedly with those from an earlier paired watershed test of snow fences conducted on Pole Hountain about 30 km east of Laramie, \\Tyoming. Located at 41° 15'N, 105° 21'W, at an elevation of about 2450 rn, weather conditions were similar to those on the Stratton Stndy Area, except winter preci- pitation averaged about 15 ern. Hore importantly, however, the Pole Mountain soils are coarse and extremely permeable, being derived fro~ Sherman granite. High infiltration rates and a deep zone of fractured rock result in very little sur- face runoff from snowmelt. After two small watersheds had been calibrated for 7 years (1963-1969), a 3.8-m-tall snow fence, 396 m long, was built on a 45-ha watershed in 1970. The second watershed, 36 ha in size, remained untreated. The design of the snow fence treatment (Tabler, 1971) was based on the same Cl LJJ o BEFORE FENCING, Y • 1. 27X -43. 4 , Cr • Ill. 99) • liFTER FENCING ill 2111111 Q: LJJ f-< )1: .. i·/ .· .····· Cl LJJ u ~ 15111 lL. z 0 lL. lL. 0 z ~ 1111111 .··· .. ·· .. ···· .. ·· ., .. ······m CD __ ••• ••• .......... CD ~-········ .. ··· .. ··· f- ...J LJJ l: )1: 0 z (() ... /~Ill. 95 CONFIDENCE INTERVAL lL. 5111 0 >-< 0 f- ((J < ...J 5111 1111111 15111 2111111 LAST DAY OF SNOWMELT RUNOFF ON LOCO CREEK Figure 8. Last day of snowmelt runoff on the snow-fenced watershed (North Draw), versus that on the control (Loco Creek). The pretreatment regression is shown as a solid line, with dotted cur- ves indicating the 95% confidence interval. reasoning used and eventually Equation (1). for the North Draw fence, led to the development of This snow fence also was planned to augment snow deposition in the stream channel immediately above the streamgage, and was located about 20 m upwind from the channel centerline to minimize snowmelt conveyance losses. Snow accumulation and streamflow were monitored for 4 years after treatment and then the study was discontinued in 1974. Although snow accumulation on the watershed approximately doubled .after the fence was built (Figure 9), there was no significant increase in snowmelt discharge because all melt water percolated to such depth that it was not intercepted by the streamgage (Figure 10). Although the increased snow accumulation must have had some effect on groundwater, the treatment failed to increase local surface discharge. This comparison demonstrates that edaphic and geologic characteristics of the watershed must be considered when designing snow fence treatments to increase surface runoff. 59 ~ 12 <~ a:: E ~ .!!11!1 :.C oLII z:E: B UIUI It: ,_LII zl-6 LJI< ...J;»r <c >LJI -u 4 ;;)z CJLJI 'jlLL a::z ~0 2 < :. 1!1 1!1 POLE MOUNTAIN WATERSHEDS o BEFORE FENCING. Y • I. 17X • I. 5t1 , Cr • I. 75> • AFTEII FENCING. Y • I. 31X + I. 38 , Cr • I. Ill> -----..·--..... ~·············· .. -···················- ····································································································-·--···········- 2 4 6 B 11!1 12 14 16 lB WIITER-EQU1VALENT SNOW STORAGE ON UNFENCED WATERSHED Ccm> Figure 9. Water-equivalent sn01.r pack on the Pole Mountain snow-fenced watershed versus that on the control. 'Tl-te pretreatment regression is shown as a solid line, with dotted curves indi- cating the 95% confidence interval. 1111 'E lJ.. .8 8 ~fa ?5 ffi 6 C::c:: wW ?5~ ,_,Jt ±fa 4 uu ~ m 2 :::Eu. z 0 POLE MOUNTAIN WATERSHEDS e BEFORE SNOW FENCE, Y • e.32X + e.92, Cr • e.49> * WITH SNOW FENCE .. ·· .. ·· 111 2 4 6 8 1111 12 14 MARCH-JUNE RUNOFF ON UNFENCED WATERSHED (em) Figure 10. March-June (inclusive) streamflow on the Pole Mountain snow- fenced watershed, versus that on the control. The pretreatment regression is shown as a solid line, with dotted curves indicating the 95% confidence interval. CONCLUSIONS Although final conclusions must await additional years of study, it seems cer- tain that the snow fence has significantly increased both quantity and duration of streamflow on North Draw watershed. Results the first 2 years after fencing indicate that the method and criteria used to design the treatment were entirely satisfactory. Although the geology of North Draw watershed is probably ideal for realizing maximum benefits from this form of snowpack management, even larger increases in water yield are possible in areas having more snow transport, less natural snow storage capacity, or where longer fences can be built. Properly designed snow fence treatments appear to be a viable means of augmenting water supplies on windswept lands. ACKNOWLEDGEMENTS Research reported here was supported in part by the Bureau of Land Management, u.s. Department of the Interior. The Rocky Mountain Forest and Range Experiment Station laboratory at Laramie is main- tained in cooperation with the University of Wyoming. Station headquarters is at Fort Collins, in cooperation with Colorado State University. LITERATURE CITED Berndt, H. w., 1964. Inducing Snow Accumulation on Mountain Grassland Watersheds. J. Soil and Water Conserv. 19(5):196-198. Brown, R. H., 1983. Snow Fences: Then and Now. J. Cultural Geog. 4(1):87-98. Cooley, K. R., A. L. Huber, 0. C. Robertson, and J. F. Zuzel, 1981. Effects of Snowdrift Management on Rangeland Runoff. In: Proc. 49th Western Snow Conf., P• 55-64. Costin, A. B., 1968. Management for Improved Water Yield from Victorian High Mountain Catchments. Victoria's Resources, June-Aug. 1968:4-8. Johnson, K. L. and R. 0. Tabler, 1973. An Enclosed Weir for Small Streams in Snow Country. DSDA For. Serv. Res. Note RM-238. 8 P• Lull, 1-I. w. and H. K. Orr, 1950. Induced Snow Drifting for Water Storage. J. Forestry 48:179-181. Martinelli, M., Jr., 1959. Some Hydrolo- gic Aspects of Alpine Snowfields Under Summer Conditions. J. Geophys. Res. 64(4):451-455. 60 Martinelli, M., Jr., 1973. Snow Fences for Influencing Snow Accumulation. In: The Role of Snow and Ice in Hydrology. Proc. Symp. on Measurement and Forecasting, WMO, Sept. 1972. Banff, Alberta. p. 1394-1398. Rechard, P. A., 1973. Opportunities for Watershed Management in Wyoming. Proc. Irrig. and Drainage Div. Specialty Conf., April 22-24, 1973. Colorado State TJniv., Fort Collins. Amer. Soc. Civil Eng. p. 423-448. Rechard, P. A., 1975. A Study of Evapora- tion from a Snowdrift. In: Proc., Snow Management on the Great Plains Symposium, July 29, 1975. Bismarck, N. Dak. G~eat Plains Agric. Counc. Publ, 73, P• 65-84. Saulmon, R. w., 1973. Snowdrift Manage- ment Can Increase Water-Harvesting Yields. J. Soil and Water Conserv. 28(3): 118-121. Seward Weekly Gateway, 1909. Alaska Miners Store Snow. Seward Weekly Gateway, April 10, 1909. p. 2C2. Sturges, D. L., 1975a. Sediment Transport from Big Sagebrush Watersheds. In: Proc. ASCE Irrig. and Drainage Div., Watershed Manage. Symp., August 11-13, 1975. Logan, Utah. p. 728-738. Sturges, D. L., 1975b. Oversnow Runoff Events Affect Streamflow and l.J'ater Quality. In: Proc., Snow Management on the Great Plains Symp. July 25, 1975. Bismarck, N. Dak. Great Plains Agric. Counc. Publ. 73, p. 105-117 Sturges, D. L., 1979. Hydrologic Rela- tions of Sagebrush Lands. Proc. Symp. on the Sagebrush Ecosystem, April 27-28, 1978. Logan, Utah. p. 86-100. Sturges, D. L. and R. D. Tabler, 1981. Management of Blowing Snow on Sagebrus Rangelands. J. Soil and Water Conserv 36(5):287-292. Swank, G. W. and R. W. Booth, 1970. Snc Fencing to Redistribute Snow Accumula· tion. J. Soil and Water Conserv. 25(5):197-198. Tabler, R. D., 1968. Physical and Econ~ mic Design Criteria for Induced Snow Accumulation Projects. Water Resourct Res. 4(3):513-519. Tabler, R. D., 1971. Design of a Watershed Snow 'Fence System and First Year Snow Accumulation. In: Proc. 39 Western Snow Conf., p. 50-55. Tabler, R. D., 1973. Evaporation Losses of Windblown Snow, and the Potential for Recovery. In: Proc. 41st Western Snow Conf., p. 75-79. Tabler, R. D., 1974. New Engineering Cri- teria for Snow Fence Systems. Transp. Res. Rec. 506:65-78. Tabler, R. D., 1975. Estimating the Transport and Evaporation of Blowing Snow. In: Proc., Snow Management on the Great Plains Symp., July 25, 1975. Bismarck, N. Dak. Great Plains Agric. Counc. Publ. 73, p. 85-104. Tabler, R. D., 1980. Geometry and Density of Drifts Formed by Snow Fences. J. Glaciology 26(94):405-419. Tabler, R. D., 1982. Frequency Distribu- tion of Annual Peak Water-Equivalent on Wyoming Snow Courses. In: Proc. 50th Western Snow Conf., p. 139-148. Tabler, R. D., 1985. Ablation Rates of Snow Fence Drifts at 2300 Meters Eleva- tion in Wyoming. In: Proc. 53rd Western Snow Conf., p. 1-12. Tabler, R. D. and R. P. Furnish, 1982. Benefits and Costs of Snow Fences on Wyoming Interstate-SO. Transp. Res. Rec. 860:13-20. Tabler, R. D. and K. L. Johnson, 1971. Snow Fences for Watershed Management. In: Proc., Snow and Ice in Relation to Wildlife and Recreation Symp., February 11-12, 1971. Ames, Iowa. p. 116-121. 61 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 SURVEY OF EIPEI.IEIICE D OPDA'I'IIIG BYDROKLKCTR.IC PR.O.O:CTS D COLD U:GIOWS Eugene J. Gemperline, DavidS. Louie, H. Wayne Coleman! ABSTRACT: In response to operational and environmental concerns, a literature review, mailed survey and site visit have been made to evaluate the procedures adopted by operators of hydroelectric projects in cold regions similar to the Susitna Hydroelectric Project site. Four specific areas were addressed and are discussed in this paper. 0 0 Reservoir procedures flooding, and powerhouse operating to mitigate ice jam related The effects of reservoir ice cover and bank ice on animal crossing, Management of reservoir ice control cracking and the danger to animals, and cover to associated Bank erosion resulting from reservoir and river ice movement. The effect of this movement on suspended sediment and turbidity levels. Ice-related flooding is a major concern in cold regions and several comprehensive manuals have been published by concerned organizations. Summaries of the subjects covered in these documents are presented. In order to obtain more specific information on operating hydroelectric projects, a mail survey was conducted. Questionnaires were sent to hydropower utilities, water supply utilities," federal and state agencies for the environment, universities, research organizations, and other oganizations located in Canada, northern states in the United States, Europe, and Asia. The questions asked concerned ice management policies in use or being adopted, for operations or environ- mental purposes. The results of the survey are presented and discussed. Additionally, a site visit was made and discussions were held with operators of an existing hydro- power utility in Canada and an agency concerned with the operation. A summary of the results of those discussions is included. (Key Terms: Cold Regions Hydrology, Hydroelectric Projects, Experience Survey, River Ice, Environmental Impacts, Reservoir Ice Cover.) INTRODUCTION The Susitna Hydroelectric Project 1s being proposed by the Alaska Power Authority to meet projected electrical energy demands in the Alaska Railbelt in the 21st century. The project would con- sist of two dams, powerhouses and appur- tenant facilities, to be built in three stages on the Susitna River about midway between the principal load centers of Anchorage and Fairbanks. The upstream dam (Watana) would be an earth and rockfill structure about 270 m high having a s1x unit powerhouse capable of generating 1,110 MW at a flow of 650 m3/s. The downstream project would be located 52 km below Watana at the Devil Canyon site and would be a thin concrete arch having a height of 197 m and capability of 680 MW 1Respectively, Manager Hydrologic and Hydraulic Studies, Harza-Ebasco Susitna Joint Venture, 711 H St. Anchorage, Alaska 99501; Chief Hydraulic Engineer, Harza Engineering Co.; Section Head, Hydraulic Design and Analysis, Harza Engineering Co., 150 S. Wacker Dr., Chicago, Illinois 60606. 63 at a flow of 430m 3 /s, through four turbine/generator units. The Watana dam would be built in two stages. In the first stage the dam would be raised to 56m below its final height and four of the six units installed in the powerhouse. This would become operational in 1999. The second stage of the project would be the construction of the Devil Canyon dam and powerhouse. This would become operational in 2005. In the final stage, the Watana dam would be raised to its full height and the powerhouse completed. The average annual generation from the final project would be 6,900 gwh. The project is located in the cold region of south-central Alaska. Air temperatures in the region are below freezing for seven months (October-April) and the river is subject to 1ce cover during this entire period. River flows during this period are normally low, but proposed project operation would result in much higher winter flows. The river ice regime would be altered due to both flow and temperature effects induced by the reservo1r. Additionally, a large reach of the river to be impounded by the reservoirs (approximately 130 km) would be covered by a stable but large ice sheet as compared to the normal river ice cover composed of frazil pans and frozen slush. As a result of these effects there was concern that wildlife inhabiting the basin area and anadromous fish (salmon) in the river downstream of Devil Canyon would be affected. Therefore, a series of hydrolo- gic and hydraulic studies were undertaken to estimate the effects of project construction and operation on the river to aid biologists in their impact assess- ments. These included studies of project operation (Wu, et .al., 1986), studies of reservoir water quality including tempera- ture, 1ce cover and suspended sediment (Wei and Hamblin, 1986), river tempera- tures for open water reaches of the river downstream of the dams (AEIDC 1983) and river ice processes (Paschke and Coleman 1986). A more complete description of the project, basin, river and the hydrologic studies is contained in a paper by Gemperline (1986). In addition, there were several aspects of the proposed project's effects, beyond the scope of model studies, which were 64 addressed by a survey of experience in operating hydroelectric projects in cold regions. These concerns dealt with the effect of the reservoir ice cover on animals crossing the reservoir; the effect of potential powerhouse operation on the stability of the downstream ice cover, and the effect of ice cover on reservoir and river bank erosion. The experience survey was carried out in three parts. In the first part a literature review was made of several ice management and ice engineering manuals compiled by U.S. and Canadian agencies. In the second part, a concise set of questions was developed and mailed to hydroelectdc project operators in cold regions, environmental agencies, water supply utilities, univers1t1es, research organizations and others who might have information. The replies were campi led and evaluated. Follow-up phone calls and disussions were held which yielded additional information~ In the third part of the study a visit was made to an operating hydroelectric project in a cold region. The results of the study were documented in a report (Harza-Ebasco 1985) and integrated with the modeling studies to determine the potential project effects during the winter and to evaluate proposed operating constraints. LITERATURE REVIEW Several organizations have published general information and guidelines for ice considerations in the design and operation of hydroelectric and other projects. These sources were reviewed: 1. Evaluation of Ice Problems Associated with Hydroelectric Power Generation in Alaska, Fina 1 Report to the State of Alaska Department of Commerce and Economic Development by J.P. Go sink and T.E. Osterkamp, of the University of Alaska Geophysical Institute. The problems dealt with in this study pertain more to energy generation than to environmental concerns. There is discus- sion of problems related to hanging dams and ice jams which is of interest. Various methods for determining the open water reach downstream of a reservoir are discussed. A survey of hydropower plants was conducted to determine potential ice- related problems and possible solutions. 2. Course Notes for Ice Engineering on Rivers and Lakes by the University of Wisconsin, Madison in cooperation with the U.S. Army Corps of Engineers, Cold Regions Research and Engineering Laboratory and the University of Wisconsin Sea Grant Institute. The course notes include articles by leading authorities in the field of ice engineering, dealing with: a) formation and breakup of a river ice cover and methods for analyz- ing and solving associated problems, b) ice problems at hydroelectric structures, and c) mechanical properties of ice and ice forces on structures. The notes provide a good compendium of potential ice problems and engineering solutions. However, they do not deal with environmental effects other than effects of flooding on human habitation. 3. Design and Operation of Shallow River Diversions in Cold Regions by the u.s. Department of the Interior, Bureau of Reclamation REC-ERC-74-19. This report contains information on potential ice problems and design guide- lines. Although the report is written for shallow river diversions, many of the design guidelines are applicable to hydro- electric projects as well. The report does not deal with environmental problems. 4. Winter Ice Jams on the Gunnison River, by the U.s. Department of the Interior, Bureau of Reclamation R EC-ERC-7 9-4. The report details ice jam flooding pro- blems associated with operating projects and methods used in an attempt to allevi- ate the problem. The flooding affected residents along the reach of the Gunnison River between Blue Mesa and Taylor Park 65 reservoirs. Relationships were developed between ice jam location, weather condi- tions and level of Blue Mesa Reservoir water surface. 5. Ice Management Manual, by Ontario Ministry of Natural Resources. This report includes guidelines for dealing with chronic ice problems including procedures for monitoring, predicting and acting on freeze-up and break-up ice jamming related flooding. It includes information on conditions causing ice jam~ing, explains causes and predictive methods for break-up, lists the data which should be collected in a monitoring program, and assesses the success rate of various remedial measures. 6. Ice Engineering by the u.s. Army Corps of Engineering, EM 1110-2-1612. This is a very comprehensive report summa- rizing potential problems at all types of civil works structures including hydroelectric projects. The report provides guidance for the planning, design, construction, operation and maintenance of ice control and ice suppression measures and is used by the Corps of Engineers for their projects. The manual discusses ice formation processes, physical properties and potential solutions. 7. Behavior of Ice Covers Subject to Large Daily Flow and Level Fluctuations by Acres Consulting Services, Ltd. for the Canadian Electrical Association. This report contains much valuable information on the types of problems encountered relative to river ice covers downstream of hydroelectric projects. An attempt was made to establish the state-of-the-art in predicting the stability of a river ice cover subject to flow and level fluctuations. Theoretical computations were made to establish stability criteria for the ice cover and to provide a means for developing guidelines for flow and level fluctuations to prevent ice cover break-up. The study concludes that: " ... generalized criteria do not exist at present, and designs cannot be prepared for many cases of ice structure, or shoreline, interaction." The report concludes that extensive labo- ratory and field studies are necessary before a generally applicable model can be formulated and that the guidelines pre- sented in the manual can be used to establish that field program. Site speci- fic studies would also be required. 8. Reservoir Bank Erosion Caused and Affected by Ice Cover by Lawrence Gatto for the U.S. Army Corps of Engineers Cold Regions Research and Engineering Laboratory (Special Report 82-31). This report describes a survey of reser- voir bank erosion problems at various places throughout the world and provides a reference list. Many photographs of existing installations are presented. Criteria for various types of erosion are presented. 9. Proceedings of the Ice Management Seminar, January 30, 1980, London, Ontario by the Ministry of Natural Resources, Southwestern Region, London, Ontario. This report provides the experience of many Canadian and U.S. experts in the field of ice management and control. Its scope is similar to Ice Engineering by the U.S. Army Corp of Engineers (6. above) and course notes for Ice Engineering on Rivers and Lakes (2. above). Additionally, many articles available in the general literature were consulted. Some of these were provided by partici- pants in the mail survey described later. MAIL SURVEY A mail survey was conducted to determine the experience of operators of hydroelectric projects in cold regions. The survey asked for information on the following four subjects: 1. Procedures or operating policies used in the control of ice levels in rivers 66 downstream and upstream of dams and hydropower plants caused by en vi ron- mental water releases and power generating flow fluctuations in order to minimize the formation of ice jams and more importantly to minimize the associated flooding. 2. Environmental impact on terrestrial animals such as caribou, elk, bear, moose, etc., due to the formation of ice on wet reservoir banks exposed by reservoir drawdown or due to reservoir surface ice which has broken up at the banks. This ice may cause the animals to lose their footing and slip into the reservoir, resulting in injuries or drownings. What procedures, if any, have been taken to minimize this hazard? 3. The method of reservoir fluctuation management or precautions used in order to control the width of opening and pattern of crack development in the ice sheet such that after snowfall with cracks covered, the traversing animals would not fall into and be trapped in the cracks. 4. Problems of bank erosion caused by break-up and movement of ice resulting in increase of sediment in the reser- voir and in the r1ver downstream. What 1s the permissible degree of turbidity in parts per million or its equivalent that is acceptable for aquatic life such as salmon, trout, etc.? The questionnaires were sent to hydro- power and water supply utilities; federal and state agencies for the environment, fish, wildlife, natural resources, energy, and in 1 and w ate rw a y s ; u n i v e r s i t i e s; research organizations; and engineering companies located in each of the Canadian provinces and northern states in the United States, Europe and Asia, involved in ice engineering, and also selected concerned citizen groups. After a reasonable time lapse, follow-up letters were sent to some non-respondents. During the process of compiling the replies, telephone contacts were made in an attempt to clarify certain points. Over 160 letters and telexes or cables were sent to various throughout the world, of replies were received - considered a good response. organizations which over 80 50%. This is The replies were carefully reviewed and those that addressed the question(s) were tabulated; the rest of the respondents normally stated that nothing is known about the queried subject. Technical information in the replies was quite sparse although many respondents sent papers and manuals. Many respondents suggested names of persons and organizations to contact. In general, these suggestions were followed through. The following is a summary of the replies, organized by question. Question No 1. -Reservoir Operatin~ Procedures to Mitigate Ice Jam Related Flooding: 1. During freeze-up it is important that powerhouse discharges remain relative- ly high until a stable ice cover is formed, in the downstream river, at a high enough stage and of sufficient thickness and strength to allow full flexibility of discharge throughout the winter. Thereafter, the outflow may be reduced as required. This action permits the water to flow free- ly under the ice cover. When short term increased discharge is necessary, it should not exceed the discharge at ice cover formation until the ice cover has had a chance to strengthen as a result of heat loss and consoli- dation of ice blocks forming the initial cover. The consequences of increasing discharge over that at cover formation are the lifting of the ice cover, and a tendency to cause ice build-up. The ice build-up or hanging ice could result in increased back- water and ice jams during the break-up period. 2. During ice cover formation, the rate of freezing is monitored, and daf\y discharge is kept as constant as pos- sible to reduce ice shoves at the leading edge of the ice cover and minimize flooding. If shoving should occur and water stage should increase, the discharge is moderated to reduce the hazards. 67 3. British Columbia Hydro attempts to coordinate ice break-up of the Peace River with the various tributaries on its river system. However, the timing and rate of break-up depend primarily on prevailing weather conditions and spring freshet flood peaks from the tributaries, and cannot be controlled at the dam. Therefore, extensive field observation posts at various stations have been established to monitor ice conditions. Where necessary and fea- sible, operations were modified in order to minimize hazards. 4. Each plant in the Manitoba Hydro system is associated with a unique set of operating policies. These policies are usually established out of concern for the environment, but also with recognition of a preferred mode of operation for power production pur- poses. Attempts to mitigate effects of flooding, etc. are made. If damage should occur, compensation procedures are adopted. 5. Ontario Hydro states that operational procedures at dams and hydroplants are still primarily based on operator's experience because the necessary understanding of ice jams is still not available. 6. Most respondents state that no attempts have been made to control ice levels to affect ice jamming or flooding. 7. Some respondents state that they have no written operating policy. Water levels are not regulated with effects on wildlife in mind, but only with the intent of providing required power generation or adequate water supplies for the users. Potential downstream effects at many of these sites are generally negligible because of sparse population or wildlife. 8. In other cases, operational con- straints are employed to prevent the formation of hanging dams downstream of a hydro project or to reduce water levels upstream of the hanging dam after it forms. Hanging dams may result in high water levels which can reduce the plant generating capacity, endanger the powerhouse, or result in flooding of areas adjacent to the river. These types of constraints include: o Inducing an early ice cover on the river upstream of known sites of hanging dams, by artificial means such as ice booms or other obstruc- tions. When an ice cover forms, frazil ice production stops and the hanging dams, which result from frazil accumulation, are minimized. o Inducing an early ice cover on the river by keeping powerhouse dis- charges low while the ice cover forms. This may result in more rapid ice cover advance, preventing further frazil production. After the ice cover is formed, powerhouse discharges can be increased. o Preventing ice cover formation in sensitive areas by fluctuating discharges, continually breaking up the ice cover and keeping it downstream. This may result in higher water levels further downstream, but lower water levels in sensitive areas. o Reducing discharges after a hanging dam forms in order to reduce water levels upstream of the hanging dam. 9. The Canadian Electrical Association and many plant operators indicated that powerhouse operations during the winter to maintain a stable cover would be site specific and require operating experience over a number of years. Reservoir discharge, climate conditions, channel morphology, and water temperature are all variables which must be considered. Question No. 2. -Ice on Reservoir Banks: In general, all organizations take no specific actions on their reservoirs to alter the state of ice on reservoir banks for wildlife safety reasons. Two organi- zations indicated sporadic cases of deer drowning within ice covered drawdown zones 68 but have no quantitative or documented information. Others reported no known problems with animal injuries or drowning as a result of reservoir drawdown. Vattenfall (The Swedish State Power Board) reports some potential problems related to the need for reindeer to pass regulated rivers have been discussed when planning for new hydro power stations and in some cases the Power Board has con- structed special reindeer bridges where "natural crossings" cannot be used anymore. The Union Water Power Company reports, "The nature of reservoir freezing during drawdown does not allow wet reservoir banks to ·exist. The drawdown is gradual thus allowing solid freezing of the water. There are no exposed areas where an animal would become entrapped in a combination of wet mire and reservoir ice. At the time of freezing, the reservoir ice has formed sufficiently to support the weight of animals." The United States Fish and Wildlife Service reports "There is the potential, if reservoirs freeze, for terrestial animals to become stranded on ice and become easy prey to predators. Animal loss can be prevented by predator control, fencing of reservoirs and providing access to winter feeding areas away from iced surfaces." question No. 3. -Reservoir Management for Ice Crack Control: All respondents except one state that no procedures are used to control cracks in reservoir ice that might be a hazard to animals. Also, most stated that no known problems with animals falling in cracks or openings along reservoirs have been docu- mented. Many routes for migratory species do not cross existing reservoirs. Appar- ently, this is coincidental and not by choice of design because many of the reservoirs were in place prior to public awareness of environmental problems. Kennebec Water Power Co., Maine states "In Maine, most large animals stay off the ice as they are unable to maintain mobil- ity -especially the hooved animals." At the Lucky Peak Dam, an irrigation and flood control structure in Idaho, many deer drownings occurred between its construction in 1956 and the institution of measures to minimize the problem. The reservoir is reportedly in a major migra- tion path and, up unti 1 about 10 ago, as many as 150-175 deer per year would drown in the reservoir. This was apparently caused by the animals crossing the reser- voir when the 1ce cover was still thin. Pockets of unsupported 1ce or cracks apparently formed in the ice cover as the reservoir was being drawn down. Deer stepping on these areas would fall through the cover. Later, when the 1ce cover thickened, the problem ceased. The reser- voir is now maintained at a stable level during cover formation to prevent the formation of these cracks or pockets. Deer drownings have reportedly been reduced to 5-10 animals per year. At the Blue Mesa Reservoir in Colorado there has been one major incident of elk drowning during recent years. The exact cause is not known. However it appears, from the location where the elk were found, that they may have fallen through thin ice at the edge of the reservoir when the cover was first forming. It also appears that the elk do not normally travel on the reservoir and were there because of any of a number of reasons including a harsh winter and poaching hunters. The elk had apparently travelled at least a mile on the ice. The Blue Mesa Reservoir normally draws down continually through the winter by 40 to 100 feet. No measures have been instituted to control ice cover formation. Isolated instances of animals being trapped on the 1ce do occur and rescues have been made. At the Revelstoke Hydrolectric Project in British Columbia, moose cross the im- poundment ice with no apparent reluctance or difficulty when it is stable and gener- ally avoid crossing when it is not. Many observations at this reservoir confirm that in almost all cases moose. can climb out of the reservoir onto the ice after falling through weak spots. No ice relat- ed caribou mortalities have been noted at the Revelstoke Project. Woodland caribou readily cross the reservo1r during the winter when ice conditions permit, indi- vidually or in groups of up to 20 animals. Question No. 4 -Bank Erosion Due to Ice Movement: The answers to this question dealt mostly 69 with the question of turbidity and not bank erosion. The report by Gatto (1982) was found to be the best source of information on bank erosion. Regarding turbidity, permissible levels are diffi- cult to define and vary from province to province in Canada and from state to state in the U.S. Usually the levels are set for drinking water standards or human recreation standards and seldom for aqua- tic life. Yne province of Ontario does not permit Secchi disc readings to change by more than 10%. Alberta's objectives suggest changes be less than 25 JTU's over seasonal natural background levels. These are drinking water standards. The state of Michigan replied with information on maximum water surface fluctuations in impoundments for: o cold water rivers (salmon, char, trout, etc.) at 8"-10" per day o warm water rivers (bass, walleyes, etc.) at 12"-18" per day. These criteria are based on exper1ence 1n the state. Many agencies stated that project oper- ation results in turbidity 1ncreases over natural conditions during spring floods and ice movements but this is not within design control. The increase in sediment gives an apparent large increase in turbi- dity. However, turbidity changes due to project construction such as high velocity sediment sluicing operations, are more critical to fish. The agencies generally concluded that operational experiences gained each season, monitored by special- ists, should be used to guide future operations. Large amounts of bank erosion were ex- perienced at Southern Indian Lake (Hecky, and McCullough 1984) as a result of wave action on an exposed permafrost shoreline. The Southern Indian Lake water level is nearly constant all year and the shoreline at one level is continually exposed to erosive forces from waves. The soils at Southern Indian Lake are predominately silty clay, with widespread permafrost at a depth of up to 10 m. The exposure of the permafrost to the warm water and waves melts the ice and creates "thermal niches" which contribute to erosion. Soils eroded from the banks of Southern Indian Lake tend to stay in suspension and contribute to high turbidity levels because of their small grain size and low settling velocity. Additionally, the ratio of shoreline length to volume at Southern Indian Lake is very high compared to narrow deep reservoirs such as are proposed for Watana and Devil Canyon. Shoreline erosion which occurred at Southern Indian Lake would be expected to have a larger effect on suspended sediment concentrations than at narrow, deep reservoirs. VISIT TO BRITISH COLUMBIA HYDRO AND PEACE RIVER TOWN Officials of British Columbia Hydro in Vancouver and Alberta Environment in Peace River Town were visited. Information was obtained on operating policies of the W.A.C. Bennett and Peace Canyon Project and records of the flooding of Peace River Town which was related to wintertime oper- ations of B.C. Hydro's upstream projects. Conclusions regarding the effect of the Portage 1-iountain Development on Peace River ice conditions, are as follows: 1. Freeze-up staging on the order of several meters can result from consolidation of an ice front following severe flow fluctuations from a load following power plant. 2. 3. 4. 5. This consolidation and staging can extend over 100-150 km. associated a range of Such consolidations occur naturally to some extent, but can be more frequent and of greater magnitude with the higher winter power flows and if flow is fluctuated. An important aspect of the freeze-up staging is flow surge from water released from storage under a backwater profile following consolida- tion of an ice front, resulting in unsteady flows which may be 1.5-2.0 times the steady flow. The generally accepted procedure for operation in the vicinity of a sensitive area, is to maintain steady, 70 high power discharge while the ice front is passing through the area. Once the front is well upstream, and a competent cover has developed, which period may be 1-2 weeks depending on the air temperatures, load following operations can resume. The ice front is always subject to consolidation, but the sensitive area will be safe if the front is far enough upstream. 6. Break-up consolidation and jamming is much less controllable. Factors other than power releases can be more important, such as development of intervening flow from snowmelt, effect.s of tributaries, and rate of warming of air temperatures. 7. On the Peace River, the procedure on break-up seems to be to provide high, fluctuating flows as far as possible in non-sensitive areas. When approaching a sensitive area, it is desirable to reduce flow and hold steady until the front is downstream of the sensitive area. SUMMARY AND CONCLUSIONS The conclusions of this study are: 1. Reservoir operating procedures which are in use at other projects to mitigate downstream ice jam related flooding include: a) b) Establishment of a stable ice cover on the downstream river early in the season during freeze-up. The ice cover should be high enough and strong enough to allow full flexibility of discharges throughout winter. Operational procedures flay also oo employed to prevent hanging dams. These include inducing an early ice cover on the river upstream of known sites by artificial means, or by keeping powerhouse dis- charges low while the cover forms. Hanging dams may also be prevented in sensitive areas by fluctuating discharge to keep the ice cover broken up and downstream of the area. c) Measures to prevent release of ice from the reservoir to the river downstream. This ice, if released, could contribute to jamming downstream. The Canadian Electrical Association and many plant operators indicated that powerhouse operations during the winter to maintain a stable cover would be site specific and require operating experience over a number of years. Reservoir discharge, climatic conditions, channel morphology, and water temperature are all variables which must be considered. 2. All respondents to the mail survey state that their organizations take no specific actions on their reservoirs to alter the state of ice on reservoir banks for wildlife safety reasons. The Power Board in Sweden has provided reindeer bridges where "natural crossings" cannot be used. Two organizations indi- cated sporadic cases of deer drowning within ice-covered drawdown zones but have no quantitative or documented information. Usually, in Canadian provinces, animals have returned and crossed the reservoir prior to ice break-up. Others reported no problem of a similar nature. It was also noted that the freeze-up and break-up periods are dangerous times to cross the natural river. 3. All respondents except one state that their organizations take no actions to control cracks in reservoir ice cover that might be a hazard to animals. A policy has been instituted at Lucky Peak Dam to minimize drownings of deer. This includes keeping the reservoir water level stable during the initial ·ice cover formation period, to prevent cracks or pockets of unsupported ice. Reservoir drawdown to required spring levels for flood control is accomplished either before initial ice cover formation, or after the cover has thickened sufficiently that the unsupported areas would be less of a hazard. 4. The report by L. Gatto ( 1982) indi- cates that local bank erosion can be expected in reservoirs which drawdown 71 continually in the winter. This is generally limited to the period during ice cover melt out when winds may blow the ice cover against the exposed banks. This would result in local increases in suspended sediments but would not af feet the overall sediment load of the outflow significantly. Some erosion of bank material and vegetation removal also occurs in reservoirs with ice covers continually at one level because of thermal expansion and wind force induced "ice push" of the stable ice cover. ACKNOWLEDGEMENT This study was funded by the Alaska Power Authority and was carried out by the Harza-Ebasco Susitna Joint Venture. Particular thanks are given to all those respondents to the mail survey. The cooperation of British Columbia Hydro, and Alberta Fnvironment is deeply appreciated, especially the help of c. V. Kartha and Les Parmly of the Hydrology Section of B. c. Hydro and Gordon Fonstad of Alberta l'hvironment. REFERENCES Alaska Power Authority, 1985, Before the Federal Energy Regulatory Commission, Application For Major License, Project No. 7114, The Susitna Hydroelectric Project, (Amended Draft) FXhibi t E Chapter 2 Water Use and Quality, Section 3. Prepared by the Harza-Ebasco Susitna Joint Venture. Arctic Environmental Information and Data Center (AEIDC), 1983, Susitna Hydroelectric Project, Stream Flow and Temperature Modeling in the Susitna Basin, Alaska, for the Harza-Ebasco Susitna Joint Venture for the Alaska Power Authority. Gemperline, Eugene J., 1986, Hydrology and Hydraulic Studies For Licensing of the Susitna Hydroelectric Project, Proceedings of the Cold Regions Hydrology Symposium, American Water Resources Association, Fairbanks. Harza-Ebasco Susitna Joint Venture, 1985, Susitna Hydroelectric Project Survey of EXperience in Operating Hydroelectric Projects in Cold Regions, for the Alaska Power Authority. Hecky, R.E., and G.K. McCullough, 1984, Effect of Impoundment and Diversion on the Sediment Budget and Nearshore Sedimentation of Southern Indian Lake, Canadian Journal of Fisheries and Aquatic Science, Vol.41. Paschke, Ned W., and H. Wayne Coleman, 1986, Forecasting the Effects of River Ice Due to The Proposed Susitna Hydroelectric Project, Proceedings of the Cold Regions Hydrology Symposium, American Water Resources Association, Fairbanks. Wei, C.Y., and P.F. Hamblin, 1986, Reser- voir Water Quality Simulation in Cold Regions, Proceedings of the Cold Regions Hydrology Symposium, American Water Resources Association, Fairbanks. Wu, Yaohuang, Joel I. Feinstein, and Eugene J. Gemperline, 1986, The Susitna Hydroelectric Project, Simulation of Reservoir Operation, Proceedings of the Cold Regions Hydrology Symposium, American Water Resources Association, Fairbanks. 72 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 HYDROLOGY Aim HYDRAULIC STUDIKS FOR LICERSIIIG OF TBK SUSITRA HYDROELECTRIC PR.O.JECT Eugene J. Gemperlinel ABSTRACT: The planning for and licensing of a major hydroelectric project require many hydrologic and hydraulic studies. These range from observations of existing conditions in the watershed, to estimates of project related effects on water use, water quality and impacts on the eco- system. The number and breadth of these studies for a project located in a cold region is discussed. Examples of analyses used to predict changes to plants and animals resulting from the construction and operation of this major hydroelectric facility are presented. Hydrologic con- siderations in the design and operation of such a facility which are additional to considerations in a more temperate zone are included. For example, the effects of glaciers on streamflow and on sediment and the effects of ice on river stage and reservoir heat transfer are topics which are not addressed in temperate reg1on hydro -projects. Eva 1 uat ion of sue h a development in a cold region, therefore, requires the coordinated efforts of hydrologists, hydraulic engineers, fishery, wildlife and plant biologists. (Key Terms: Cold Regions Hydrology, Hydro- electric Projects, Licensing, Environmental Impacts, Alaska Rail belt.) INTRODUCTION Project Description The Susitna Hydroelectric Project has been proposed by the Alaska Power Authority to provide for the projected electrical energy needs of the Railbelt region 1n the 21st century. The Railbelt region is the area of southcentral Alaska extending from Homer at the southern tip of the Kenai Peninsula to Fairbanks and including the large metropolitan area of Anchorage. The region is so-named because its principal cities are linked by the Alaska Railroad (Figure 1). The project would consist of two dams, powerhouses and appurtenant facilities, to be located on the Susitna River about midway between Anchorage and Fairbanks. The upstream development at the Watana site is located 296 km (184 miles) upstream of the river's mouth at Cook Inlet. This dam would be an earth and rockfill structure and would be built in two stages. In the initial stage the dam height would be raised approximately 214 m (702 ft.) above its foundation to El. 617.2 m (2,025 ft. msl). A powerhouse with four turbine/ generator units (units) having a total average capability of 440 MW at a discharge of approximately 340 m3fs (12,000 cfs) would become operational in 1999. This dam would be raised to El. 672.1 m (2,205 ft. msl) in the third stage of the project, following completion of the downstream dam. Two additional units would be added to the powerhouse increasing the total average generating capability of the Watana development to 1,110 MW at a discharge of approximately 650 m3Js (23,000 cfs). The two additional units would become operational 1n 2012. The downstream development at the Devil Canyon site is located 245 km ( 152 miles) lManager, Hydrologic and Hydraulic Studies, Harza-Ebasco Susitna Joint Venture, 711 H St. Anchorage, Alaska, 99501 now at Stetson-Harza, 185 Genesee St., Utica, New York, 13501. 73 upstream of Cook Inlet. The dam at this site would be a thin concrete arch structure with a crest at El. 446 m (1463 f t • m s 1) 19 7 m ( 6 4 6 f t • ) above its foundation. The downstream impoundment would extend to the upstream dam. The powerhouse at Devil Canyon would contain four units and have a total ave rage generating capability of 680 MW at a flow of 430 m3/s (15,200 cfs). These units would become operational in 2005. The Watana dam site is located in a broad U-shaped canyon and the Devil Canyon dam site is located in a narrow, steeply incised canyon. The Watana reservoir would provide the flow regulation for its O\m and the Devil Canyon powerhouses. The Devil Canyon dam would provide little flow regulation but would develop additional head. The Watana reservoir would impound 5.3xl09m3 (4.3xl06 ac-ft) of water in Stage I and 11.7xl09 m3 (9.5xl06 ac-ft) of water when it is raised in Stage III. The Devil Canyon dam would impound 1.4xl09 m3 (l.lxl06 ac-ft) of water (APA 1985). History of Project The proposed project is a result of a series of reconnaissance, prefeasibility and feasibility studies performed by various agencies of the Federal Government and the State of Alaska (Acres 1981). The initial reconnaissance level work by the U.S. Bureau of Reclamation (USBR) identified five damsites from a list of 25 as being most appropriate for further investigation. These sites were all located in the river reach upstream of the major confluences with the Chulitna and Talkeetna Rivers. These areas were considered appropriate because the site characteristics generally allow for high heads to be developed and substantial flow regulation to be achieved with dams located in relatively narrow canyons. Additionally, dams located in this reach would have less effect on the river's large anadromous fishery than dams at downstream sites. Later studies by the USBR, Alaska Power Administration and H. J. Kaiser Co. for the State of Alaska built upon the original USBR study with some slight refinements to the site locations. All proposed the Devil Canyon site as the initial damsite with upstream sites to be developed in the future. The 74 U. S. Army Corps of Engineers (COE) prepared comprehensive basin studies in 1975 and 1979 and proposed the damsites at Watana and Devil Canyon as the most appropriate. Following the COE's 1979 study the State of Alaska formed the Alaska Power Authority (APA) for the purpose of planning for the power needs of Alaska and developing the projects to meet the needs. The APA reassessed the previous studies and confirmed the conclu- sions of the COE. The initial License Application before the Federal Energy Regulatory Commission (FERC) was filed by the APA in 1983 (APA 1983). This applica- tion was amended to include refinements and staging the Watana dam (APA 1985), The latest application has recently been withdrawn in favor of a study of alternative energy sources for t~ region. The Basin The drainage basin upstream of the Devil Canyon site is located approximately between latitude 62°05' and 63°40' North and between longitude 146 ° 10' and 149° 30' West in south central Alaska, approxi- mately 225 km (140 miles) north-northeast of Anchorage and 177 km (110 miles) south-southwest of Fairbanks (Figure 1), The drainage areas upstream from the Devil Canyon and Watana damsites are about 15,050 and 13,400 square kilometers, (5,810 sq. mi. and 5,180 sq. mi) respec- tively. The basin is geographically bounded by the Alaska Range to the north and west, and the Talkeetna Mountains to the south and east. The topography is varied and includes rugged mountainous terrain, plateaus, broad river valleys and lakes. Mount McKinley (El. 6,194 m) is located on the northwest divide of the basin. Elevations within the basin upstream of the Devil Canyon site range from approxi- mately 260 meters above mean sea level (850 ft, msl) at Devil Canyon site to over 2,100 meters, msl (7,000 ft. msl) near the head reach of the Susitna River. Approximately 5% of this basin is covered by glaciers. Three major glaciers -West Fork Susitna, East Fork Susitna and Maclaren, exist in the basin. The landscape consists of barren bedrock mountains, glacial till-covered plains and --- 6UL \OF A ASKA ~ \ !t !t 51 • 0 on o -~ !: \ \ \ \ '\... \ ~ ~SHINE .... \ !/"':!--.. J'-..., l U"' I; 't'l \ I . SUSITNA I. _ \ g_ATION \ I I \ I I \ I \ l \ \ II ~ LEGEND • DAMSITE • STREAM GAGING STATION ALASKA RAILROAD FIGURE 1 LOCATION MAP exposed bedrock cliffs in canyons and along streams. Soils are typical of those formed in cold, wet climates and have developed from glacial till and out-wash. They include the acidic, saturated, peaty soils of poorly drained areas, the acidic relatively infertile soils of the forest and gravels and sands along the river. The basin is generally underlain by discontinuous permafrost. The River The Susitna River originates in the East Fork and West Fork Susitna Glaciers at an altitude of approximately 2,380 m (7,800 ft. msl) and travels a distance of about 512 km (318 miles) before dis- charging into Cook Inlet. The head waters of the Susitna River and the major upper basin tributaries are characterized by broad, braided, gravel flood plains below the glaciers. Several glacierized streams exit from beneath the glaciers before they combine further downstream. Below the confluence with the West Fork Susitna River, the river develops a split-channel configuration with numerous islands and is generally constrained by low bluffs for about 89 km (55 miles). The Maclaren River, draining the Maclaren Glacier and a few small lakes, and the non-glacial Tyone River draining Lake Louise and swampy lowlands of the south-eastern part of the basin, join the main river downstream of Denali. Below this confluence, the river flows west for about 155 km (96 miles) through steep-walled canyons before reaching the mouth of Devil Canyon. River gradients average about 0.3 percent in a 87 km (54-mile) reach upstream of Watana, about 0.2 percent from Watana to the entrance of Devil Canyon and about 0.6 percent in a 19 km (12-mile) reach between Devil Creek and the outlet of Devil Canyon. The Susitna River is typical of glacial rivers with high turbid summer flow and low, clear winter flow. The discharge generally starts increasing during early May. The base flows during July through September are due to groundwater, glacial melt and melt of long term snowpack. Peak flows during this period are associated with general frontal type of thunderstorm activities. The river flow rapidly decreases in October and November as the 76 river freezes. The break-up generally occurs in early May. The May through June flows are caused by snowmelt combined with rainfall. Melting of snow, firn and ice from the glaciers has accounted for about 13% of the annual streamflow at Devil Canyon. The average summer and winter flows at a few selected stream gaging stations are given in Table 1. Figure 1 shows the locations of the stream gaging stations. Project Operation The project will operate by storing the high summer flows in Watana Reservoir to provide a· dependable source of power in the winter for the Rail belt. The reser- voirs will generally be full in late August or September and the Watana Reservoir will be drawn down throughout the winter. It will reach its lowest level in early May and begin to fill as river flows increase from snowmelt and rainfall. Filling will continue throughout summer unt i1 the water level reaches its normal maximUill level. This can occur as early as late June in a wet year or as late as early September in a dry year. When the reservoir is full, inflow in excess of power and environmental flow requirements must be released. High inflows in July and August may often exceed these requirements resulting in the need to release flows through outlet works to prevent the reservoir water level from encroaching on dam safety requirements. Table 1 compares natural and with-project flows for the Susitna River at Gold Creek for summer (May -September) and winter (October April) periods based on 34 years of record and simulations of project operation (Wu et. al. 1986). Gold Creek is 26 km (16 miles) downstream of the Devil Canyon site and is the location at which environmental flow requirements will be gaged. There are no major tributaries between the damsites and Gold Creek. Average monthly flows and floods during Stage I, II, and early Stage III would be similar. Energy demands are projected to increase in late Stage III and the summer flows would decrease accordingly. Flood peak discharges would also be reduced due to the storage capacity of the Watana Reservoir as shown by Table 2. Summer (May -Sept) Winter (Oct -Apr) Susitna River Drainage Stages Stage Stages Stage Gaging Station Area Natural I, II III Natural I, II III (Sq. km) Near Denali 2,460 179 179 179 11.7 11.7 ll. 7 Near Cantwell 10,700 365 365 365 37.8 37.8 37.8 At Gold Creek 16,000 572 374 285 64.1 207 271 At Sunshine 28,700 1,380 1,180 1,090 153 296 360 At Susitna Station 50,200 2,680 2,480 2,390 354 497 561 Table 1. Average Summer and Winter Flows (m3/s) at Selected Stream Gaging Stations for Natural and With-Project Conditions Return Natural1l Gold Creek Sunshine Period Gold (Years) Creek Sunshine Stages I, nJJ Stage nil/ Stages I, II Stage III 2 1,360 4,050 1,030 626 3,650 2,970 5 1,790 4,700 1,220 844 4,190 3,430 10 2,090 5,180 1,250 968 4,560 3, 770 25 2,470 5,670 1,270 1,080 4,930 4,160 50 2, 770 6,060 1,320 1, 210 5,270 4,500 Table 2. Natural and With-Project Floods Susitna River (m3/s) ~ Annual series, occurs in May -June at Gold Creek and July -September at Sunshine. J./ July -September series. Under natural conditions the highest peak floods occur in June as a result of snowmelt and precipitation runoff. Regulation of floods by the reservoir will delay the highest floods until the July -September period except in late Stage III. In late Stage III regulation by the project will be so large that July -September floods will be less than those in June. Overview of Hydrologic Studies The planning for and licensing of a major hydroelectric project require many hydrologic and hydraulic studies. The initial requirement, during the reconnaissance level studies, is for a reasonable estimation of streamflow quantity, time distribution and reliability. As the need for the project increases and the proposed sites must be screened to develop plans worthy of more detailed and costly investigation, the scope of the hydrologic studies must also increase. More accurate knowledge of flows is required in these prefeasibili ty level studies and potential project effects on the ecosystem must be more 77 accurately evaluated. For the feasibility and licensing level of work, the selected development will be compared to other projects on the bases of economic and engineering feasibility and environmental impacts. For a large, capital intensive project located in an ecologically sensi- tive area to survive comparison against smaller, less capital intensive projects with less visible environmental impacts requires accurate determination of the hydrologic resource available to produce energy and comprehensive studies of how project operation will affect the environ- ment. During feasibility and licensing of the project, hydrologic studies are carried out for three purposes: one, to develop information on flows required to judge the project economics; two, to develop information necessary for the planning and preliminary design of project structures; and, three, to estimate potential project effects on the water resource and resulting impacts to humans, animals and plants which use the water. From an engineering or project design standpoint there are many hydrologic considerations. The most important is the time distribution and reliability of river inflow and how this affects the need for active storage capacity in the reservoir. This was a factor in the select ion of possible dam sites and in the scheduling of Watana dam construction ahead of Devil Canyon. Other hydrologic considerations in design were the potential for glacial outbreak floods and the influence of mass glacial wasting on streamflows. The location of the project in a cold region with its great variation in summer and winter streamflows, the importance of snowmelt and glacier melt and the presence of glaciers which could surge or cause jokulhlaups has resulted in studies which would not be carried out in a more temperate climatic region. The proposed project is located in a wilderness like area on a stream which supports a diverse anadromous fishery in a basin which contains much wildlife. The potential for affecting this ecosystem is an important issue and is addressed primarily by hydrologic and hydraulic studies coordinated with biologic studies. Such factors as the project influence on downstream flows, water temperature, sediment concentration, river ice regime, and dissolved gas concentration have been evaluated in great detail with hydrologic and hydraulic studies and have influenced the proposed project design and operation. Again, the breadth of these studies is larger in a cold region than in a more southerly area because of the occurrence of ice on the river and proposed reservoir, and its affect on water levels, river and reservoir temperatures. Hydrologic studies will not end with project licensing. In fact, they will likely increase as project operators and fish and wildlife agencies seek to use the water resource to greater advantage. Efforts will be made to forecast reservoir 78 inflows (Hydex, 1985). Project effects on temperature, ice, sediment, etc. will be monitored and predictions made during licensing will be refined. Effects on fish and wildlife will be observed. Ehergy demand growth, now just a predic- tion, will occur. Project operation will need to be modified to meet the need for energy and to preserve and enhance the environment. HYDROLOGIC STUDIES FOR PROJ FCT H:ONOMICS The hydrologic studies required to evaluate project economics center on three subjects: one, the quantity of flow in the river; two, the distribution of this flow throughout the year, and three, the reliability of this flow from year to year. These three factors along with the topographic features of a reservoir site (depth, volume, surface area) determine the average energy which can be generated, the reliable or firm energy, the amount of storage which must be provided in the reservoir and the manner of reservoir operation. The location of the Susitna Project in a cold region influences the three parameters. The first parameter, average quantity of flow, is a function of precipitation, evaporation and transpiration since, over the long term, runoff must equal precipi- tation minus the other losses. This is largely controlled by the basins' geo- graphic location, topography and large scale weather patterns. The main influ- ences on the quantity of flow due to the cold climate, which are different than in a more temperate climate, would be the effects on evaporation and transpiration losses. For the Susitna Project the estimation of streamflow quantities was relatively simple. The U.S. Geological Survey has collected streamflow information at a site near the proposed project since the potential project was first considered. Thus, thirty-four years of flow data are available (USGS, 1949-1984). These values were transposed to the project site using multi-site regression analyses (Harza- Ebasco 1985a). While its location in a cold region may not af feet the quantity of flow, the location does affect the distribution of flow within the year and the reliability of flow from year to year. The location of the energy demand centers in a cold region also affects the demand for the power over a year and thus affects the project operation. In a warmer climate, such as in some areas of the 48 contiguous U.S. states, summer temperatures are typically hot enough to require air conditioning. These areas may experience their highest electrical energy demands in the summer. In contrast, the Alaska Railbelt has mild summers not requiring air conditioning. Winters are cold, long, and relatively dark resulting in highest electrical energy demands in December and January. This pattern of energy consump- tion is expected to continue in the future and contrasts with the pattern of stream- flows. The long period of subfreezing air temperature (October -April) results in extreme differences between summer (May - September) and winter streamflows. Average summer streamflows are 470 m3 Is cfs compared to average winter flows of approximately 53 m3/s. Therefore, the Watana Reservoir must provide an active storage equal to 0. 6 of the ave rage annual inflow in order to provide a dependable capacity equivalent to 211 m3/s in the winter of a very dry year. While the extreme seasonal distribution of inflow results in the requirement of a large storage capacity, other factors offset this. These are the minimal net evaporative loss from the reservoir surface and the presence of glaciers and occurrence of long term snow pack. In effect, the river streamflow is regulated by the glaciers and snowpack. Studies were undertaken to estimate the net difference between evaporation from the reservoir surface and evapotranspiration from the same area under natural conditions (Harza-Ebasco 1985a)'. These established that net loss of water would be less than 0.1% of the annual inflow. Thus, this was not a factor in sizing the reservoir as in warmer climates. Studies were also made to determine how the glaciers act to regulate streamflow (R&M 1981, 1982, Clarke et.al. 1985, Clarke, 1986). Although they cover only 5% of the basin they have a significant regulating effect. In wet years they tend to accumulate snowfall and in dry years they tend to waste. A study of the mass 79 balance of the glaciers was undertaken to determine whether there were any discerni- ble trends in the glacier's behavior to indicate whether the streamflow estimates during the 34 years of record were influ- enced by any gain or loss of glacier mass. These studies were, by necessity, carried out on a reconnaissance level since the only aerial photos of the glaciers in 1949 were uncontrolled, and the only controlled photographs of the glaciers in 1980 comprised less than 5% of the glaciated area. Additionally, a reconnaissence level study of the glacier surface eleva- tions was undertaken. These studies tended to confirm that the streamflow measurements were probably not unduly influenced by changes in the glacier mass (APA 1985). Studies were also made to determine the influence on project eco- nomics if the glacier melting were to diminish (Harza-Ebasco 1985b). These confirmed the project's viability even if the glaciers' mass balance were to change. HYDROLOGIC STUDIES FOR PROJOCT DESIGN Basin hydrology affects the design of major project features in addition to reservoir size. The most prominent hydraulic structure in a major hydroelectric project is the spillway or outlet works which must pass flood flows through the project without endangering the dam. In the Susitna Project there are two means for passing non-power releases. Outlet works controlled by fixed cone valves are planned at both dams to release all floods up to the 50-year event. Less frequent floods would be released through gated overflow spillways. The outlet works are provided so that the more frequent floods can be discharged to the river through the cone valves which disperse the flow over a large area and minimize the potential for elevated gas concentrations in the river downstream. High gas concentrations can be deleterious to the fish. Hydrologic studies included development of the 50-year flood hydrograph for annual, spring and fall series and routing of these floods through the project reservoirs. These studies established the necessary outlet works and flood storage capacities (Harza-Ebasco 1985c). Project spillways were designed to pass the Probable Maximum Flood (PMF) without endangering the dam as set out in guidelines of the CO E and the U.S. Committee on Large Dams (COE 1965, USCOLD 1970). Hydrologic studies included estimation of the PMF hydrograph (Acres, undated) and routing of the PMF through the projects to establish required spillway capacities and surcharge levels (APA 1985). An important factor in the PMF determi- nation was the estimation of snowpack and the manner of snowmelt since the PMF would occur during the May-June period (Acres, undated). A probability approach was adopted to estimate the total snowpack during the event and snowmelt was assumed to occur in a manner to maximize runoff. The PMF was estimated by assuming the maximum possible precipitation concurrent with a 1000-year snowpack and various antecedent conditions and the runoff routed through the basin. This is a standard, accepted method. However, in a glaciated basin, there is always the potential for a jokulhlaup or flood caused by the break-out of a glacially dammed lake. Discharges from such occurrences can be very high, potentially exceeding a PMF. Therefore, a survey was made to determine the potential for glacial dammed lakes which might af feet the project ( R&M 1981). The study indicated little likelihood of this. Almost all large reservoirs are subject to some degree of sedimentation and the Susitna Reservoirs would be no exception. Hydrologic studies were made to estimate the suspended and bed load in the river (Knott and Lipscomb 1983, 1985) and to determine the effects on reservoir life (Harza-Ebasco 1984a, 1985d). The average annual sediment load of approximately 6.0 x 109 kg. (6.5 million tons) would require 1 ,400 years to fill Watana dead storage and 2,300 years to fill the Devil Canyon dead storage. The average suspended sediment concentration in the inflow is 800 rng/1 and is comprised of a high percentage of very fine rock flour (27% less than 10 microns). This is the result of glacial weathering of underlain rock. This material has a very slow settling velocity (10-6 -w-5 m/sec) and much is expected to remain in suspension in the reservoir. The trap efficiency of the 80 reservoir is expected to be about 80% - 90% (APA 1985) as contrasted to reservoirs of similar characteristics in areas with coarser sediment which have trap efficiencies near unity (USBR 1977). A mathematical model was developed, and is described below, to more accurately estimate the potential sediment concen- trations downstream of the project, for estimating impacts to fish. Another important project feature is the means of handling water during project construct ion. The divers ion facilities will consist of tunnels to pass normal river flows around the construction areas and cofferdams at the upstream and down- stream ends of the areas. These facili- ties will be sized using risk/ cost analyses to minimize their cost and the potential losses resulting from failure, This means that cofferdam heights and tunnel sizes will be determined for various frequency floods to assign probabilities to the risk of failure. Another hydrologic consideration in diversion tunnel design is it's elevation relative to the streambed and t~ potential for bed load material to become trapped in the tunnel, if it is set too low, thus reducing its capac! ty and affecting the hydraulics at the tunnel outlet. The Susitna tunnels have been located to prevent this (Wang, et. al 1986). The diversion facilities design must also consider the need to pass ice and the potential for ice jam floods. The diversion tunnel intakes at both Watana and Devil Canyon would be located on the outsides of bends for reasons of economy in tunnel construct ion. They are thus well located for passing incoming frazil ice in October and November and broken ice sheets in April and May (USBR, 1974). The tunnel sizes are believed wide enough ( 11 m.) to handle ice sheets during break- up. Nevertheless, careful consideration will be given to the intake design, to minimize potential jamming in this area. Breakup jamming is also a potential problem downstream of the diversion tunnel. A bend in the river downstream of the tunnel outlet may provide a site for jamming of broken ice passed through the tunnel. Therefore, consideration was given to this and the downstream cofferdam crest elevation was set to prevent over- topping and flooding of the construction site by water backed up behind the potential jam. Other hydraulic considerations due to the project's location in a cold region are also primarily the result of ice. The design of the power intake towers includes heated floating ice booms to prevent ice forces on the trashracks and gates. The potential for entrainment of frazil and broken ice in the flow through the intake may dictate the submergence of the operating intake below the water surface at some times. However, as the intake has openings at several levels this will not preclude safe operation of the powerhouse. HYDROLOGIC AND HYDRAULIC STUDIES FOR ENVIRONMENTAL IMPACT ANALYSIS The primary environmental concern is the potential effect of the project on the downstream fishery. Other concerns include the project's potential effect on terrestrial wildlife and riparian vege- tation. The mechanisms responsible for the potential impacts are the proposed project's effects on the quantity and quality of water in the Susitna River. The primary concerns relate to the potential impacts on river flows, floods, water temperature, river ice conditions, suspended sediments, turbidity, and river morphology. Salmon utilize the peripheral areas of the river (such as sloughs which have favorable velocities, depths, tempera- tures, turbidities and substrates) for spawning, rearing and incubation. The amount of area available for fish use is related to the magnitude and stability of river flow. In conjunction with fisheries experts, who developed models of fishery habitat versus flow, the amount of habitat for all stages of project operation was estimated by simulating flows with project operation from initial construction to full use of project capacity, approxi- mately 30 years (Trihey, et. al. 1985). Flow constraints were developed to provide fishery habitat of equal or greater value than natural conditions. The quality of the water can also affect the fishery. For example, tempera- ture can be lethal in the extremes or can affect fish growth. Suspended sediment can af feet fish gills. Settling of 81 sediment in spawning beds can af feet intergravel flow through these areas. Turbidity can provide protect ion from predators and can retard production of waterborne insects which provide food for the fish. The hydrologic and hydraulic evaluation of the effects of the Susitna Project on water quality were evaluated with a system of three models: a reser- voir water quality model, a river tempera- ture model and a river ice model. Reservoir water quality was evaluated using the Dynamic Reservoir Simulation Model (DYRESM) (Imberger and Patterson, 1981). Modifications were made to the model to handle cold regions conditions and features of the Susitna Project (Harza-Ebasco 1984b Wei and Hamblin 1986). The model was modified to include: o Formation of an ice cover on the reservoir and winter stratification, o Outflow from the reservoir through multiple level offtakes, and o Simulation of suspended sediment including settling and the effect of sediment on density and thus, reservoir stratification. This latter modification was necessary because of the small size of inflowing sediment and the need to estimate the downstream sediment concentration. A program of collection of hydrological and meteorological data was undertaken at Eklutna Lake (R&M 1985b) a small, glacially fed, lake-tap hydroelectric project near Anchorage to provide the data needed for development and testing of the modifications. Upon completion of testing, the model was applied to the proposed sites using hydrologic and meteorologic data collected for the purpose at the sites (R&M 1985a). EXten- sive studies were made, at the request of regulatory agencies, to provide informa- tion for evaluating impacts and to determine the most favorable method for operating the multi-level offtake. Temperatures in the river downstream of the reservoirs were evaluated using the Stream Network Temperature Model (Theurer, et. al. 1984), driven by output from DYRESM. The modeled reach extended from the Watana and Devil Canyon dam faces to Sunshine, 23 km (14 miles) downstream of the confluence with the Chulitna River a distance of about 160 km (100 miles). The potential for lethal temperatures to occur was found to not be a problem and the modeling effort focused on the potential for effects on growth. While the DYRESM model provided outlet temperatures on a daily basis, the SNTEMP model was used on an average weekly basis. Several refine- ments were made to the SNTJMP model as well (AEIDC 1983). These include: o Estimation of solar radiation from radiation incident at the edge of the atmosphere corrected for atmospheric and topographic effects, o Inclusion of frictional heating, o Inclusion of tributary temperature effects on mainstem temperature and regression modeling of tributary temperatures, and o Inclusion of air temperature lapse rates between the site of the temperature recorder and the upstream end of the study reach. River temperatures were measured both in the mainstem and tributaries to allow calibration and verification of the model. The SNTEMP model was used to estimate river temperatures downstream of the reservoir throughout the year for all DYRESM simulations. In the summer the downstream end of the study reach was at Sunshine. Modeling of temperatures was not considered necessary downstream of that point because with-project tempera- tures were generally found to be within 1° C of natural. In the winter the downstream end of the SNTEMP modeled reach was the location of 0°C. Modeling of winter river conditions, with ice, was done using a model developed for the project (!CECAL) (Harza-Ebasco 1984c). This model computes the amount of ice produced, hydraulic conditions in the channel, development of border ice, formation of an ice cover from frazil ice and staging of water levels due to the ice cover. The model was used primarily to determine how peripheral habitat areas 82 would be affected by the increase in winter flows (from 60 m3/s -250 m3/s) coupled with the change in the extent of ice cover. There was concern that increased water levels in the river area affected by ice would overtop peripheral habitat areas and introduce cold water into the sloughs thus stressing the salmonids. The results of the modeling allowed prediction of the impact, and development of mitigation measures. During development and testing of the model an extensive program of field observations was carried out (R&M 1981-85) to develop information for verifying the model and to better understand the basic ice proceSses in the river. Several other hydrologic studies were undertaken in conjunction with the evalua- tion of biologic impacts. A mailed survey was undertaken and a site visit was made to determine the experiences of other hydroelectric project operators in cold regions (Gemperline et. al. 1986). River- bed stability was evaluated to estimate potential aggradation and degradation (Harza-Ebasco 1984a, 1985e). This involved determination of bed load, bed material sizes and bed material transport equations. Impacts evaluated included the potential for aggradation near tributary mouths possibly affecting fish access and degradation in the mainstem possibly affecting peripheral habitat. Potential effects of project operation on riparian vegetation were evaluated using notes on vegetation types observed during river surveys. The observed elevations of various types of vegetation were correlated to river flows and floods. Based on a model of vegetation succession and predicted with project flood flows, the vegetation encroachment on the river was, to some degree, quantified. CONCLUSION This paper presents some of the more important hydrologic and hydraulic studies which have been made for the licensing of the Susitna Hydroelectric Project, in- cluding considerations because of the project's location in a cold region. For the purpose of the paper the studies were separated into those required for eco- nomic, analyses, engineering design and environmental lmpact analyses. However, in reality, the studies were not separated. For example: the evaluation of fishery habitat and the establishment of minimum flows affected estimated project energy production; the design of power offtakes and release facilities affected estimated downstream water quality. Coordination was required between all participants to develop the information necessary for licensing of the project. ACKNOWLEDGEMENT The studies described in this paper were funded by the Alaska Power Authority. The studies were carried out by the Harza- Ebasco Susitna Joint Venture, R&M Con- sultants, Inc., the Arctic Environmental Information and Data Center, the u.s. Geological Survey and the University of Alaska Geophysical Institute. The support of all these organizations is deeply appreciated. REFERENCES Acres American, Inc., 1981. Susitna Hydroelectric Project, Development Selection Report, for the Alaska Power Authority. Acres Arne rican Inc, undated. Susi tna Hydroelectric Project, Feasibility Report, Hydrological Studies, Final Draft, Volume 4, Appendix A, for the Alaska Power Authority. Alaska Power Authority, 1983. Before the Federal Energy Regulatory Commission, Application for Major License, Project No. 7114, The Susitna Hydroelectric Project, prepared by Acres American Inc. Alaska Power Author! ty, 1985. Before the Federal Energy Regulatory Commission, Application for Major License, Project No. 7114, The Susitna Hydroelectric Project (Amended Draft), prepared by Harza-Ebasco Susitna Joint Venture. Arctic Environmental Information and Data Center (AEIDC), 1983. Susitna Hydroelectric Project, Stream Flow and Temperature Modeling in the Susitna Basin, Alaska for the Harza-Ebasco 83 Susitna Joint Venture for the Alaska Power Authority. Clarke, T. s., D. Johnson, and w. D. Harrison, 1985. Glacier Mass Balances and Runoff In the Upper Susitna and Maclaren River Basins, 1981-1983, for Harza-Ebasco Susitna Joint Venture for the Alaska Power Authority. Clarke, Theodore s., 1986. Glacier Runoff, Balance and Dynamics in the Upper Susitna River Basin, Alaska, Thesis presented in partial fulfillment of the requirements for the degree of Master of Science, University of Alaska, Fairbanks. E. w. Trihey and Associates, 1985. Response of Juvenile Chinook Habitat to Discharge in the Talkeetna to Devil Canyon Segment of the Susitna River, for Harza-Ebasco Susitna Joint Venture for the Alaska Power Authority. Gemperline, E. J., D. S. Louie, and H. W. Coleman, 1986. Survey of Experience in Operating Hydroelectric Projects in Cold Regions, Proceedings of the Cold Regions Hydrology Symposium, American Water Resources Association, Fairbanks. Harza-Ebasco Susitna Joint Venture, 1984a. Susitna Hydroelectric Project Reservoir and River Sedimentation, for the Alaska Power Authority. Harza-Ebasco Susitna Joint Venture, 1984b. Susitna Hydroelectric Project, Eklutna Lake Temperature and Ice Study, with Six Months Simulation for Watana Reservoir, prepared for the Alaska Power Authority. Harza-Ebasco Susitna Joint Venture, 1984c. Susitna Hydroelectric Project Instream Ice, Calibration of Computer Model, for the Alaska Power Authority. Harza-Ebasco Susitna Joint Venture, 1985a. Susitna Hydroelectric Project, Case E-VI Alternative Flow Regime, Appendix D, Stream Flow Time Series, for the Alaska Power Authority. Harza-Ebasco Susitna Joint Venture, 1985b. Letter of June 4, 1985 to the Alaska Power Authority. Harza-Ebasco Susitna Joint Venture, 1985c. Susitna Hydroelectric Project, Flood Frequency Analyses for Natural and With Project Conditions, for the Alaska Power Authority. Harza-Ebasco Susitna Joint Venture, 1985d. Susitna Hydroelectric Project, Effects of the Proposed Project on Suspended Sediment Concentration, for the Alaska Power Authority. Harza-Ebasco Susitna Joint Venture, 1985e. Susitna Hydrolectric Project, Middle Susitna River Sedimentation Study, Stream Channel Stability Analysis of Selected Sloughs, Side Channels and Main Channel Locations, for the Alaska Power Aut ho ri ty. Hydex Corporation, 1985. Streamflow Fore- casting Feasibility Study, for Harza-Ebasco Susitna Joint Venture for the Alaska Power Authority. Imberger, J. and J. C. Patterson, 1981. A Dynamic Reservoir Simulation Model - DYRESM: 5, in Transport I1odels for Inland and Coastal Waters, Chapter 9, Academic Press. Knott, James, M. and Stephen W. Lipscomb, 1983. Sediment Discharge Data for Se- lected Sites in the Susitna River Basin, Alaska 1981-1982, U. S. Geological Survey Open File Report 83-870 prepared in cooperation with the Alaska Power Authority. Knott, James, M. and Stephen w. Lipscomb, 1985. Sediment Discharge Data for Se- lected Sites in the Susitna River Basin, Alaska, October 1982 to February 1984 U. S. Geological Survey Open File Report 85-157, prepared in cooperation with the Alaska Power Authority. R & M Consultants, Inc, and W. D. Harrison, 1981. Susitna Hydroelectric Project Task 3 -Hydrology; Glacier Studies, for Acres American Inc., for the Alaska Power Authority. R & M Consultants, Inc. 1981-1985. Susitna Hydroelectric Project, Annual reports of Ice Conditions during Freezeup and Breakup, for Acres American, Inc. and 84 Harza-Ebasco Susitna Joint Venture for the Alaska Power Authority. R & M Consultants, Inc, and W. D. Harrison, 1982. Susitna Hydroelectric Project; Task 3 -Hydrology; Glacier Studies, for Acres American Inc. for the Alaska Power Authority. R & M Consultants, Inc., 1985a. Susitna Hydroelectric Project, Processed Climatic Data, October 1983 -December 1984, for Harza-Ebasco Susitna Joint Venture, for the Alaska Power Authority. R & M Gonsultants, Inc. 1985b. Susitna Hydroelectric Project Glacial Lake Physical Limnology Studies, (Draft) for Harza-Ebasco Susitna Joint Venture, for the Alaska Power Authority. Theurer, F. D., K. Voos, and w. J. Miller, 1984. Ins tream Water Temperature Model, Instream Flow Information Paper 16, U. S. Fish and Wildlife Service (Draft). U. s. Committee on Large Dams, 1970. u. Criteria and Practices Utilized in Determining the Required Capacity of Spillways. S. Department of the Army, Corps of Ehgineers, 1965. Standard Project Flooo Determinations, Ehgineering Manual No. 1110-2-1411. U. S. Department of the Interior, Bureau of Reclamation (USBR), 1974. Design and Operation of Shallow River Diversions in Cold Regions, REC-ERC-74-19. U. s. Department of the Interior, Bureau of Reclamation, (USBR) 1977. Design of Small Dams, (2nd ed.). U. s. Department of the Interior Geologi- cal Survey (USGS), 1949-1984. (Annual Reports) Water Resources Data for Alaska. Wang , B. H. , S. R. Bred t haue r, and E. Marchegianni, 1986. Design Problems in Gravel Bed Rivers, Alaska, Proceedings of the International Workshop on Pro- blems of Sediment Transport in Gravel Bed Rivers (In Press) 12-17 August 1985, Colorado State University. Wei, C. Y., and P. F. Hamblin, 1986. Reservoir Water Quality Simulation in Cold Regions, Proceedings of the Cold Regions Hydrology Symposium, American Water Resources Association, Fairbanks. Wu, Y., J. I. Feinstein, and E. J. Gemperline, 1986. The Susitna Hydroelectric Project, Simulation of Reservoir Operation, Proceedings of the Cold Regions Hydrology Symposium, American Water Resources Association, Fairbanks. 85 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 ICE JAM FLOODING-EVOLUTION OF NEW YORK STATE'S INVOLVEMENT Russell E. Wege 1 ABSTRACT: This paper outlines the development of New York State's involvement in assisting flood plain communities impacted by ice jam flooding. While this paper is neither a technical nor historical disserta- tion the discussion illustrates the state's process by which state-of-the- art ;echnology has been translated and communicated to village, town and city officials. This training effort has encouraged local government to help themselves and has substantially reduced the number of re- quests for federal and state assistance. (KEY TERMS: involvement; ice jams; training.) INTRODUCTION -BACKGROUND The many large rivers in the northeastern United States encouraged colonial settlement in the floodplains. Nineteenth century industrialization and development brought a constant flow of immigration and the expansion of villages and cities in the floodplains. The transformation of forest land into agriculture and settlements modified both fluvial flow and ice related problems. Records of fluvial flooding go back to in the colonial period. In New York State, ice jam flooding began to be recorded in the latter half of the 19th century. After the U.S. Department of Interior's stream gaging program was expanded, early in the 20th century, ice jam flooding was more frequently reported. Continued development in the floodplains increased the frequency of ice jam flooding and rapidly expanded the record of ice jam problems. Economic development in New York State included several large public works projects that affected ice jam and fluvial flood events. Ashokan, the largest of several New York City water supply reservoirs, was completed in 1915. The 500- mile barge canal system, with its summit level storage reser- voirs and control structures, was completed in 1918. The 42-square mile Great Sacandaga Reservoir, which regulates flow in the upper Hudson River, was completed in 1930. In addition, many private and utility-owned hydro projects were constructed throughout the state during the first quarter of the 20th century. With the exception of the Great Sacandaga Reservoir, none of these projects was constructed specifically for flood control purposes, although all had an important and mitigating im- pact on ice jam problems. These large pools of water stopped the moving ice from causing downstream jams and reduced frazil ice production by regulating stream flow. In addition, rivers channalized for navigation eliminated obstacles that previously had triggered ice jams. New York winters may be described as ranging from moderate to severe. The up-state area, away from the coastal influence, annually receives 60-180 inches of snow. Winter months commonly record sub-zero temperatures. It is not uncommon for night temperatures to drop to -20°F and, on occasion, to -40°F in the mountain valleys. Almost annually, New York will experience a mid-winter thaw during the 4th week of January. Temperatures will often rise into the 40's and even the SO's and sometimes last for several days. The mid-winter thaw can cause streams and rivers to rise, breaking up their ice cover. The ice begins to run, only to jam, flooding roads and, on occasion, forcing evacuations. HISTORY OF STATE INVOLVEMENT Until the mid 1930's, ice jam and fluvial flooding were considered a local problem and not a specific state issue. However, the great floods of 1935 and 1936 overwhelmed countless river communities and cities in several northeastern states. As a result, federal and New York State legislation in 1936 provided authority to develop local flood protection projects in severely damaged communities. Flooding prob- lems were longer ignored by federal and state government. The success of this legislation is reflected in the statistics. Over 80 flood control projects involving construction of 142 miles of improved channels and 105 miles of levees and walls have been constructed in New York over the last 45 years. However, New York's 1936 legislation limited state involve- ment to a cooperative partnership with the U.S. Army Corps of Engineers. The state does not have standing authority to construct flood control projects unless they are approved 1New York State Department of Enviromnental Conservation, 50 Wolf Road, Albany, NY 12233-0001. 87 federal projects. There are 1480 communities that have been identified as being flood prone along the thousands of miles of rivers and tributaries in New York State. However, almost all of these communities will never qualify for a federal flood protection project and therefore, remain subject to ice jam and fluvial flooding events. Without a legislative mandate for involvement, ice jam problems from the late 1930's to the late 1960's were viewed by the state as little more than winter statistics. This attitude was soon to change. A reorganization of state government in 1967 led to the development of a flood control bureau in the State Conservation Department. The flew bureau had a small professional staff and was responsible for flood response train- ing, operation and maintenance of flood control projects and advocating new federal projects on behalf of flood-damaged communities. The first flood control project to address ice jam flooding was in the rural community of East Branch in the Delaware River Basin. During the winter of 1971 an ice jam so threatened the community that it made international news. The crisis was met jointly when the Army Corps of ~ngineers built an emergency levee on the right-of-way expeditiously obtained by the state. The quick solution to the East Branch ice jam problem was unusual. More commonly, when communities made requests to the state for ice jam removal, the requests were simply forwarded to the Corps of Engineers. It was hoped that the federal agency would solve the problem. Rarely did that happen. However, the Corps usually did send a representa- tive to look at the problem. The Corps soon realized that many of the requests were for minor ice jam problems that were not a serious threat to the health and well being of the community, so it began to develop criteria for federal involve- ment. Eventually, the criteria authorized federal involvement in ice jam breaching only as a supplement to state, county and local efforts. This required local governments "to exhaust" local resources before federal involvement was authorized. However, the state did not have legislative authority, aside from state emergency powers, to spend money on ice jam problems and county governments were under similar restraints. Since adoption of this federal policy several years ago, only two ice jam problems in New Yotk State have qualified for federal intervention. Both cases took several weeks to satisfy federal requirements. The Corps policy has essentially eliminated federal involvement in ice jam mitigation work and has forced the state to be more responsive. The Flood Control Bureau in the newly organized State Department of Environmental Conservation began to respond to community requests for assistance by sending a flood con- trol engineer to look at the problem and offer technical assistance. By December of 197 6, it had become official state policy to provide technical assistance to communities having ice jam problems. It was during the early 1970's that an understanding of the causes of ice jams and the techniques to alleviate ice jam 88 flooding developed. This learning process was greatly en- hanced by the cooperation of personnel from the Corps of Engineers Cold Regions Research and Engineering Laboratory in Hanover, New Hampshire. By the mid 1970's, community requests for state assistance had become overwhelming. There were simply not enough experienced people to meet the need. This led to the develop- ment of a state training program, which has as its goal the development of field personnel trained in ice jam problems and mitigation measures. With more people trained, the state could respond more promptly with sound technical assistance. Over the years, the training sessions have increased in frequency and improved in content. State field personnel invited local emergency management and community officials to attend the training sessions, which are held throughout the state. Class si~e is usually 20 to 30 people, but as many as 100, have attended these sessions. A brief description of the material covered in the training sessions is as follows. TYPES OF ICE JAMS Ice problems are dependent upon several factors: frost depth, amount of snow in the basin, amount and intensity of a rain storm, severity of winter temperatures, and rate of temperature rise. These factors play an interrelated role. Annual variances of any one of these conditions can spell the difference between major ice jam problems and no problem at all. Ice problems fit into two general categories: A. The warm weather breakup. Runoff from a thaw or rain storm feeds water into the tributaries and rivers. The ice/ground contact weakens and begins to fail. Continued inflow begins to raise the ice sheet or inun- date it, introducing upward forces. The ice sheet be- comes unstable and begins to move and break up. The breakup develops into an ice run and will begin to move downstream until it meets an obstruction. There, the run will stop, back up water and usually breach. This is a true ice jam, where a jam forms and breaches, only to jam again downstream. This happens countless times each winter as small streams discharge into major river systems. In general, high gradient streams will loose their ice cover sooner than low gradient streams. B. Cold weather blockage. The most common and troublesome form of ice is called frazil. Research (Ice Engineering, 1982) has determined that ice nuclei crystals develop in waters moving 2 FPS or greater in an atmospheric environment of 20°F or less. These crystals are very cohesive and join together to form frazil. Frazil ice will stick to most objects. It forms collars around rocks and piers and sticks to the stream beds and banks. It adheres to broken pool ice and often forms masses greater than 10 feet thick in our larger rivers. Hydro electric generation, with daily releases in cold weather, is a magnificent frazil ice generating machine. As an example, hydro generation on the· West Canada Creek in Central New York annually produces frazil ice that fills a one-mile reach of a 300-foot wide channel to a depth of8-15 feet. The resultant high water surface increases local ground- water depth and causes wet basements at Herkimer, a village some 25-miles downstream from the power dam. Frazil ice can also develop into a hanging dam. This phenomenon usually occurs in a pool at the end of a rapids section in larger rivers. Frazil ice develops in the higher energy upstream reach and moves into the low velocity pool where it begins to float to the surface. The ice covered pool begins to collect frazil under the sheet ice. The hanging mass of ice severely restricts river flow, causes bottom scour, and often produces upstream backup and localized flooding. Hanging dams are usually massive. One can assume that such ice blockages are the last to shear during the mid-winter or spring breakup. This stability often creates an obstruction that can quickly plug the restricted channel and produce a rapid backup of the river. CAUSES OF ICE JAMS A. Changes in stream slope. New York is a headwater state. Many rivers originate in its mountain areas. Steep sloped streams produce both frazil and pool ice. The mid-winter or spring thaw can produce a runoff that rapidly breaks up the ice cover on these upland streams. Broken sheet ice and large masses of frazil ice move down the river system in a series of jams and breaches until the ice reaches the floodplain of a large river. Often the slope of the tributary flattens out at this point and the stream velocity de- creases. The drop in velocity causes the larger pieces of ice to ground out which commonly triggers a jam in the tributary. B. Bridge piers. New York State has many old bridges with massive stone or concrete piers that are spaced closely together. These piers frequently catch large pieces of solid sheet ice and trigger jams behind the structure. Replacement structures now'include more widely-spaced, piers which offer a higher probability of breaking or crushing of large pieces of ice, allowing them to continue down river. C. Sheet ice. New York State is blessed with over 7,000 lakes and ponds. In addition, hydropower development has created many run-of-river pools. Each pool, either natural or constructed stops ice. A jam will often occur at the point of entry. Usually, the backup is minor because the water will circumvent the jam and find entry into the pool at a nearby location. Lakes, reservoirs and ponds are especially beneficial in miti- gating ice damage because they prevent moving ice from traveling further downstream. 89 D. Sharp bends. The geometry of the stream channel is another factor causing ice jams. A river channel that suddenly losses one fourth of its width at a sharp bend has a high probability of stopping an ice run and plugging the channel. Flooding is the inevitable result. E. Man-made obstructions. Aside from bridge piers many New York rivers have other man-made obstructions. The 19th century timber industry drove countless wooden piles into river beds to anchor log booms. Thousands of these piles remain today and are capable of catching ice. In addition, stone fllled wooden cribs and derelict dams dot the river beds in mountain rivers and streams. These also play a part in triggering ice jams. F. Modified channeL This type of problem could be classified as a man-made obstruction, but it is better discussed as a separate cause of ice jamming. It is usually associated with a new road bridge. Design engineers, intent on accommodating a high rate of flow, will widen the channel immediately above and below the structure. Such a change in stream bed geometry can have embarrassing and sometimes devasting results. The widened channel will often reduce stream velocity to the degree where the ice run may ground out and result in a jam at that point. Several years ago the writer observed an ice jam in the Vil- lage of Mohawk that was almost 20 feet high and did enor- mous damage to many private homes. It was caused by doubling the width of a high gradient stream under a replace- ment bridge. The jam never touched the bridge but forced ice well above the elevation of the bridge railing. This type of problem can be mitigated by including a low flow channel through the modified channel section. The low flow channel concentrates stream flow offering a good prospect for passing an ice run. G. Combinations. Rarely is there a single cause for an ice jam. It is not uncommon that three or even four causes are present. In addition, the triggering mech- anisms can be spread over an extensive distance in large rivers. These possibilities need to be considered carefully prior to initiating any action. WHAT GOES INTO A DECISION TO "DO SOMETHING"? A very important part of ice jam training is to convey to local governments that federal and state agencies will not respond by breaching ice jams. Therefore, instead of focusing on big government resources, which are not there, local governments must turn to and focus on their own resources. Repeated communication of this message is changing the thinking of local interests and reduces the number of requests for ice jam assistance. The federal and state government can be relied upon only for technical assistance. Secondly, it is important to evaluate the damage potential of the ice jam. For example, does the jam threaten the pro- perty and well-being of a substantial number of people? A judgment concerning property damages must be made. It is one thing to worry about a few wet basements and it's another to evacuate a subdivision, cut off a hospital or shut down a major employer. Finally, a judgment concerning public utilities and facilities such as roads must be made. The flooding of a low usage, rural town road is nowhere near as disruptive as flooding a primary road system. State and federal technical assistance can assist a com- munity in evaluating damages or potential damages. However, the decision to take action remains with local officials. The training sessions emphasize that point. WHAT CAN BE DONE TO ALLEVIATE ICE JAM FLOODING? The evaluation of what caused the jam and the significance of damages leads to the question, "what can be done to alle- viate ice jam flooding?" A decision to "do something" costs money, requires time and planning, and incurs a certain degree of risk. Often the practicality of "doing something" is ~mited. The community which sends its public works person- nel to the river's edge to throw dynamite onto the ice has no chance for success, although it may reap certain public relations benefits. Likewise, a community that faces a several mile long jam in a major river has little hope of success if it tries to breach the jam. Let's look at the more realistic alternatives for alleviating damages. A. Raise or remove damage-prone materials and equip- ment. If something can be moved out of harms way the recommendation is simple. As example, the writer advised the management of a paper company, which had stored $1,000,000 worth of paper inventory on the riverbank, to relocate the inventory when an ice jam backed water to within two feet of the material. B. Emergency measure at the riverbank. Often a jam will ftll or even overfill the river channel. The top of the ice may be above the riverbank. Its massive extent precludes attempts to breach it. A temporary levee separating the river from the community will reduce damages and is a logical mitigation measure, provided the length of levee needed is not prohibitive. Planning time is required for this alternative, and it is particu- larly appropriate after a mid-winter breakup when there is a heavy snow pack which increases the danger of flooding in a &pring runoff. gn C. Evacuation/Flood insurance. All riverside communi· ties should have emergency management plans. Such plans should include river stage elevations and indicate when evacuations are to begin. Evacuation is often the only feasible alternative to minimize the impact from ice jam flooding. Public education and availability of flood insurance can minimize personal property losses. In New York State approximately 1460 of the 1480 flood-prone communities are participating in the National Flood Insurance Program. Eligibility for this program allows individuals to purchase insurance protection against flood ·damages. D. Ice jam ·breaching. Our experience indicates that the best chance for successfully breaching a jam is to attack the jam immediately after it has formed. Methods of breaching a jam are outlined below. These methods are most applicable to small and moderate width rivers. 1. Mechanical. Many New York tributaries and rivers are shallow and have a stony bottom. We have found that the most effective way of breaching an ice jam in these shallow, hard bottom rivers is with bulldoz~rs, beginning downstream and progressing upstream through the jam. It is not necessary to clear the entire river channel of ice. Construction of a pilot channel through the blockage is sufficient to breach the jam. Clamshell buckets have also been used success· fully in the area near bridges or along short stream reaches. The Town of West Seneca near Buffalo, New York, contracts for a small crane with a clam· shell bucket during the winter months. As the ice begins to run, the contractor responds on short notice to the known ice jam location behind a shopping center. The equipment operator simply keeps the ice moving where it tends to ground out on a bar. 2. Explosives. Through the years, there has been much misuse of explosives by many communities attempting to breach ice jams. Research (Ice Engineering, 1982) by the Army Corps of Engineers has developed exten· sive information on the best way to breach ice jams with explosives. Briefly, the following principles should be followed when this alternative is considered. a. Explosives must be placed under the ice. Explo· sives break ice by forming a gas bubble under the ice. The gas bubble lifts the ice, causing it to fail in shear. b. Blasting operations must be under the direct supervision of a licensed blastor. c. Blasting operations must begin at open water downstream of the problem area. This will allow broken ice to float away and results in a clean channel through the jam. d. Streamflow must be sufficient to carry away ice debris. e. Blasting should not be considered near structures. Several years ago the writer was informed of an incident that resulted in litigation after a village begin ice blasting operations near a greenhouse. f. Moderate temperatures are desirable. Cold weather will cause rapid refreezing of ice debris. g. Downstream effects need to be considered. Blasting may instantly release a large volume of water. Usually, such releases become insignifi- cant a short distance downstream, but, if a down- stream community experiences flooding from a high river stage coincidental to up-river blasting, litigation may result. The burden of proof will rest on the community doing the blasting. 3. Dusting. Dark material spread thinly over sheet ice will successfully weaken the ice, causing it to fail at the beginning of breakup, thus reducing the chances of its obstructing an ice run. However, dusting is not effective in breaking a mass of frazil or a true ice jam. The melting of a few inches of frazil or the top of a jam several feet thick has no effect on breaching the blockage. 4. "Ducks. " The City of Buffalo has an: amphibious vehicle. It has been used successfully to break up ice in the Niagara River. It has also been used to breakup competent sheet ice on low-gradient small streams. Success depends upon the availability of downstream open water and sufficient depth and flow to float the broken ice away. ·Without suffi- cient depth and flow, the machine will simply break the ice and push it into the mud, possibly increasing the ice jam potenti;u. 5. Imaginative solutions using reservoirs and hydro plants. a. Reservoir releases for flushing out river ice have been suggested for years. Usually, these sug- gestions are made by laymen when considering a course of action to breach a troublesome ice jam. Only recently has the suggestion been developed into an apparent workable plan of 91 action for ice control. CRREL (Ferrick, 1985) is studying this technique in the Lake Luzerne/ Corinth reach of the upper Hudson River in eastern New York. Releases originate from the Great Sacandaga Reservoir through a cooperative effort with the owner of the reservoir, the Hud- son River Black River Regulating District and the Niagara Mohawk Corporation, the regional electric power supplier. The purpose of the re- lease is to break up and flush out early winter sheet ice in order to prevent development of thick sheet ice that may stop an ice run. b. Reservoir releases may also be employed to breach an ice jam. Water in deep reservoirs will be warmer than river water. A reservoir release introduces heat into the river system which widen the passage ways through the ice jam. Eventually, the roof of the ice jam will collapse thus, breach- ing the obstruction. c. New York's larger mountain rivers have numer- ous low-head hydro plants. On one occasion, near Corinth in eastern New York, a breach was accelerated when the pool containing an ice jam was lowered several feet. The drawdown intro- duced additional stress through the blockage and accelerated its breaching. E. Overflow channel. Ice jams sometimes can be bypassed as a mitigation measure. Bypassing eliminates the expense of attempting to breach the blockage. Main- tenance costs of a bypass channel may be minor com- pared to the cost of efforts to breach recurring jams. A bypass channel involves the removal of trees and woody growth and the grading of a high flow channel downstream of a damage-prone area. A small bypass channel has been constructed by the City of Norwich, in central New York, as an outcome of the state ice jam training session. In addition, the Philadelphia District of the Army Corps of Engineers is planning a major bypass channel to protect the twin communities of Port Jervis, New York, andMatamoras,Pennsylvania, on the Delaware River. Three other bypass channels have been recommended by New York State engineers but have not been constructed as of this date. Design criterions for ice jam bypass is in its in- fancy. The following design guidelines are suggested. The bypass channel should have a width between one- half to full channel width for rivers up to 200 feet wide. The width of the bypass channel can be sub- stantially smaller than the natural channel for larger rivers. The entrance into the bypass channel should be wider than the bypass channel in order to insure bypass if jamming occurs at that location. The invert into the bypass channel should be of sufficient elevation to exclude normal flow rates. Ideally, this elevation should be high enough to prevent masses of frazil ice from entering the bypass and low enough to accommo- date a high water flow rate without damaging the upstream community. F. Ice booms. Aside from the large booms at the en- trance of the Niagara River and in the St. Lawrence River, only one ice boom has been installed in New York State for ice control. Unfortunately, that boom was severely damaged by a late fall flood and is not operational at the present time. The boom is located in the upper Hudson River at Hadley and was con- structed by state and local governments under the technical guidance of CRREL. Ice booms are inexpensive structures that encourage development of upstream sheet ice. Sheet ice deters frazil development and is a barrier to an ice run. Favor- able ice boom locations are at the downstream ends of pools, upstream from damage-prone areas. SUMMARY The success of ice jam training cannot be accurately deter- mined. However, as a result of such training, many com- munities are recognizing state and federal limitations and, based on a better understanding of the causes of ice jams, are taking action to mitigate their ice jam problems. Local communities have removed channel obstructions and, in one case, have even constructed a bypass channel. In another community an ice boom was installed. In addition, approxi- mately 1,460 flood-prone communities participate in the National Flood Insurance Program. State training has con- tributed significantly to an understanding and awareness of ice jam flooding. As a result, reports on ice jam problems and requests for state and federal assistance have steadily de- clined. REFERENCES Ferrick, M., L. Lemieux, L. Gatto, and Mulherin, 1985. Hudson River Ice Management. 42nd Eastern Snow Conference, Montreal, Canada. Ice Engineering, Engineering and Design, 1982. Department of the Army Corps of Engineers Office of the Chief of Engineers, Ice Formation and Characteristics, pg. 2-2, October 15. Ice Engineering, Engineering and Desij:n, 1982. Department of the Army Corps of Engineers Office of the Chief of Engineers, Ice Jams, pp. 3-3 to 3-5, October 15. 92 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 HYDROLOGICAL AND ECOLOGICAL PROCESSES IN A COLORADO, ROCKY liDJNTAlN WETLAND: CASE STUDY Edward W. Rovey, Catherine Kraeger-Rovey, David J. Cooper * ABSTRACT: Wetlands of Cross Creek exist in the vicinity of and are controlled by flows in the main channel. Water levels in the creek have a seasonal cycle dominated by snowmelt runoff. Groundwater levels in the wetland rise and falL~ in response to water levels in the main channel and levels in the tributaries affected by backwater from Cross Creek. Vegetation conmun it i es have deve 1 oped at locations in the wetlands related to depths of standing water or depths to ground water. Stream meandering a 1 so affects the wetlands. Large diversions of surrmer runoff proposed by the Homes take I I wi 11 1 ower water 1 eve 1 s in Cross Creek. Distribution of wetland plant species will ~ange in response to the diminished creek water levels. (KEY TERMS: wet 1 ands, groundwater, snowmelt, vegetation conmunities, water levels.) INTRODUCTION Wetlands of the Cross Creek Valley exist high in the Colorado Rocky ~untains where the controversial ~mestake II water diversion project is proposed (see Figure 1). The project is planned to annually divert 25 million cubic meters (20,000 acre-feet) of water from the Holy Cross Wilderness Area. Eleven miles of tunnels are planned beneath 4,270 meter (14,005 feet) high Mount of the Holy Cross to carry snowmelt runoff to an adjacent watershed where existing storage and tunnel facilities will be used in the transbasin diversion of water to the cities of Colorado Springs and Aurora. A general definition of wetlands is an area where soils are saturated for a duration and frequency that allows the establishment of vegetation adapted to saturated soil conditions. Federal law and Executive Orders are intended to regulate filling operations within wetlands. This paper is written by the professional consultants retained by the Holy Cross Wilderness Defense Fund, a nonprofit organization, to evaluate the potential impacts of the project on wetlands. While the project proponents claim there will be no impact to wetlands below diversion points, it is our opinion, based upon research and evaluation completed to date, that removal of up to ninety percent of the natur~l streamflows at diversions will have significant, adverse impacts to wetlands downstream. METHODS Hydrologic and hydraulic calculations have been made of pre-and post-diversion flows (Sundeen and Fifer, 1981) from historical records at the creek mouthand projections into the upper basin. Stream * Rovey and Kraeger-Rovey, hydrologists, Terra Therma Inc., 2927 W. 36th Ave Denver, CO 80211; Cooper, ecologist, 3047 Redstone Ln., Boulder, CO 80303 93 hydraulics were calculated using direct step backwater and uniform flow methods, channel slope, typical channel cross sections and reasonable flow resistance parameters. Groundwater hydraulic calculations were made with parameters obtained from a brief field investigation and seven laboratory tests of permeability from one wetland (Ward, 1983). Water table responses were analyzed using a standard drain algorithm. ---~ESfD 0 r,w:g~~ .......... ~~~i[i~ • f:~1~~ COLORADO PROJECTeDENVER KILOMETERS 0 1) l • 0 1 2 MILES Figure 1. Wetlands and Vicinity Map Twenty four homogenous stands of vegetation were selected, sampled and classified using standard Braun-Blanquet methods {Westhoff and van der Maarel, 1978). Depth to water table was measured in each stand. Gradient analysis, using the methods of Dix and Smiens {1967), was used to analyze the behavior of species along the water table/drainage gradient regime. Plant species nomenclature follows Weber {1976). 94 AREA DESCRIPTION Snow and snowmelt control the hydrologic regime of Cross Creek wetlands and surrounding areas. Eighty percent of the annua 1 runoff occurs from 1 ate May through July (USGS,1984). Though the region is cold due to the high elevation, there is no permafrost in the area. In fact, the wetland surface may not freeze due to the early,insulating covering of a snowpack • Wetland characteristics and those of surrounding areas are listed in Table 1. Montane wetlands, as described by Kilburn (1983} and considered in this report, ·occur at locations where geologic outcrops have created erosion-resistant features that obstruct the gradient of the stream. Soils in the wetlands are TABLE 1. Description of the Area Wetland.Elevation Range Mean Annual Precipitation Cross Creek Drainage Area Annual Runoff Volume 2,900 to 3,070 m. (9,500 to 10,060 ft) 50 to 125 em (20 to 50 inches) 8,675 hectares (33.5 square miles) 71 million cu. m. (58,000 acre feet) primarily composed of silts with little clay-sized material (Ward, 1983}. Main stream channels have meandered across the wetlands leaving buried remnants of material varying from large boulders and gravels to some lenses of clay material. Slopes across lateral sections of the wetlands are nearly flat. Close inspection of the wetland surface reveals the existence of numerous tributary channels extending from the main channel of Cross Creek to the valley walls. Six major communities comprise the bulk of the wetland vegetation. These are: (1} Ranunculus tricophyllus (water crowfoot) -Sparganium sp. community in shallow water averaging 10 em in depth, but intermittently exposed during summer and fall; (2) Carex utriculata (bladder sedge) community is one of the most abundant communities in the wetlands. It occurs in both marshes with mineral soils and fens with largely organic soil. Depth to water table averaged 16 em; (3) Carex aguatilis (water sedge) community occupies a variety of habitats and was observed at an avera9e depth to water table of 31 em; {4) Calamagrostis canadensis (Canada reedgrass) community occupies sites with an average water table depth of 56 em; (5) Calamagrostis canadensis -Mertensia ciliata (bluebell) community occupies the tops of levees and is characterized by coarse herbaceous dicots. Average water table depth was observed at 65 em; (6) Salix planifolia (plane-leaf willow) community occupies the most well drained sites in the wetlands with an average water table depth of 70 em. RESULTS WATER SOURCES AND TIMING Snowmelt is the primary source of water to the wetlands. Snowmelt can be subdivided into components dependent on location and timing of the melt. Primary components are: onsite snowmelt, sideslope runoff and main channel flows. The other component contributing to wetland water sources is the growing season precipitation, i.e., rainfall, falling directly on wetlands. Snowmelt on the wetland surfaces at 3,050 meters (10,000 ft.) elevation commences durin9 early May (Kraeger-Rovey and Rovey, 1984). Melt calculations and observations indicate the 43 em (17 in.) average water equivalent snowpack on the wetland surface can melt in two to three weeks. Using 7 deg. C (45 deg. F) as an indicator of the commencement of growth (SCS, 1970), the onsite ~now melt is completed on the wetlands pr1or .to wetland plant growth. This early snowmelt replenishes soil moisture deficien~ies that are carried over from the prev1ous growing season and is a adequate quantity to raise the shallow water table to the surface of the wetland. However, an adequate quantity of water is necessary but not the only condition sufficient to provide saturated soil conditions in the wetlands. Mechanisms of groundwater flow must be considered in determing how water drains or is retained within wetland 95 soils. Groundwater is discussed in a subsequent section. Melt from sideslope areas above the wetlands contributes water to the wetlands during May through early fall. This sideslope water continues beyond the duration of active snowmelt due to the timing delay of melt water flowing through the porous medium of talus and rock slopes. Estimates of these ~ideslope contributions have been made us1ng data generated form a water budget procedure (Enartech, 1983) and reported (Kraeger-Rovey and Rovey, 1984) as 13 em (5 in.), 23 em (9 in.), ~n~ 13 em ~5 in.) for typical wetland cond1t1ons dur1ng May, June and July, respectively. An evaluation of the mechanisms and factors controlling the sideslope flows was made. Numerous small tributaries were located leading from the valley walls to the main channel. The frequency of these wetland channels has been measured from air photos as one tributary channel per 60 m (200 ft) of wetland perimeter (Kraeger-Rovey, 1984) for those tributaries larger than 1.0 m (3 ft) wide and 0.3 to 1.0 m (1 to 3 ft) deep. Smaller rivulets are more abundant yet often not observed until walking across the wetland surface. A hydraulic analysis of these wetland channels determined their capacity to average four times the estimated sideslope contribution for the May to July period. Thus, the wetland channels could efficiently convey the sideslope contributions through the wetlands to the main channel absent high water in the main channel which back water up causing it to spread across the entire wetland surface. Onsite, summer precipitation on wetland surfaces provides only a portion of the moisture requirement of wetland vegetation (Sundeen, 1983). Consumptive water requirements from the wetland surfaces were estimated to be 42 em (16.5 in.) for the May through September period while the effective precipitation was 14 em (5.5 in.) during that interval. The net water requirement of 28 em (11 in.) must come from other sources. This water deficit applies to the plants and does not include the substantial amount of water to keep from draining the wetland soils. The main channel of Cross Creek conveys streamflow through the wetlands and provides the two main mechanisms that maintain saturated soil conditions in the wetlands. Main channel flows overtop the natural streambank during the peak runoff season and during heavy rainfall periods as occurred in late July, 1985. These overbank flows saturate the wetlands during that time. The mechanism that provides the longest duration of wetland saturation is when the stage (stream level) is at or near bank full. High water level in the stream may not directly provide water to the wetlands but creates a backwater extending up the tributaries that flow across the wetlands, and also causes the groundwater table throughout the wetland to be maintained at a high elevation. These mechanisms are discussed in later sections of this paper. WETLAND VEGETATION AND WATER LEVEL RELATIONSHIPS Main Channel Flows and Backwater Characteristics. By early June the onsite wetland snowpack has melted and the remaining dependable sources of wetland water supply are sideslope runoff and high stream flows in the main channel of Cross Creek. Flow estimates at Reeds Meadow, the first major wetland below two of the Homestake II diversions in the upper portion of the watershed (see Figure 1) have been made. Table 2 shows the average June and July Cross Creek discharges, for natural conditions and with the depletions caused by Homestake II diversions. TABLE 2. Cross Creek Flows at Reeds Meadow Month Natural Post-Diversion Flow Flow June 3.40 cu.m/s 0.99 cu.m/s (120 cfs) ( 35 cfs) July 2.12 cu.m/s 0.71 cu.m/s ( 75 cfs) ( 25 cfs) 96 Water level variations due to diversions were evaluated. An analysis was made in Reeds Meadow where stream flow estimates are available and some surveyed information on the slope of the channel water surface profile was collected (Ward, 1983). Streamflow from Reeds Meadow discharges over a resistent rock outcrop at the lower end of the meadow. The outcrop appears to function as a weir with a steep increase in downstream gradient where the flow goes through critical depth. Using an approximate rock outcrop elevation of 3,060.20 meters (10,040 feet), a water surface slope of .05 percent (.0005), and constant channel cross sectional shape, backwater calculations were made to calculate the water surface elevations for the flow conditions shown in Table 2 for natural and post-diversion flows. The results are shown in Table 3 for a location 700 meters (2,300 feet) above the mouth of Reeds Meadow. This is a location where a transect of groundwater piezometers was installed (Ward, 1983). TABLE 3. Water Surface Elevations in Reeds Meadow Month Natural Post-diversion Water surf. Water surf. June 3,061.50 m 3,061.01 m (10,044.3 ft) (10,042.7 ft) July 3,061.29 m 3,060.86 m (10,043.6 ft) (10,042.2 ft) The decrease in water surface elevation in Cross Creek channel is 0.49 meters (1.6 feet) for June and 0.43 meters (1.4 feet) for July. The average summer season decrease in channel water surface is 0.46 meters (1.5 feet). Backwater effects, extending into the wetlands from the full-flowing main channel, provide the essential mechanism giving the wetlands access to sideslope runoff and controlling groundwater drainage from the wetlands. With the exception of a few isolated areas, the Cross Creek wetlands are traversed by a dense network of small to medium sized tributaries, ranging from 0.3 to 2.0 meters (1 to 6 feet) across and from .15 (0.5 feet) to 1.5 meters (5 feet) deep. In every wetland, the aggregate capacity of these channels is many times greater than the total available sideslope runoff (Kraeger-Rovey, 1984). The high water level in Cross Creek 11 backs up 11 outflow from these side tributaries, prevents it from running downstream, and causes the sideslope runoff to spread over the surface of the wetlands during the growing season. If this backwater effect is eliminated, the network of tributary channels will very efficiently drain the sideslope runoff through wetlands into .cross Cre~k, leaving wetland surfaces l1terally h1gh and dry. The proposed diversions of Homestake II would substantially reduce peak runoff flows in Cross Creek to levels now typical in late summer and fall. The average reduction in water level in Cross Creek that is expected to result from the diversion has been estimated at about 65 em (2.2 feet) (Ward, 1983). As a result, backwater restrictions into side tributary channels will be substantially reduced or, for the shallower tributaries, eliminated. The tributaries will be able to convey a greater portion of the sideslope runoff through wetland channels, substantially reducing or eliminating the portion that overflows onto wetland surfaces. Groundwater Responses. Groundwater seepage provides the link between water levels in Cross Creek and its tributaries, and wetland soil saturation. The water-holding capacity of soils and alluvial material in the wetlands provides a time-de 1 ay between the drop . in creek levels in mid-summer, and drainage of groundwater from the wet 1 ands. If that time delay is sufficiently long, wetland soils may remain saturated for . a considerable period of time after snowmelt has ceased. Prolonged moisture retention in the soils would reduce the wetland's dependancy on high water levels in the main creek channel. A decrease in water level in a stream channel induces drainage from the adjacent streambank. With time, the subsurface gradient propagat~~ away from the 97 streambank, eventually inducing drainage throughout the wetland area. As drainage occurs, the water table drops. Drainage occurs most quickly in areas close to a streambank, where the soil is permeable. Both these conditions predominate in the Cross Creek wetlands. Numerous observations have shown that in most areas, wetland groundwater levels respond very rapidly to changes in Cross Creek water levels, indicating that the hydrology is governed by permeable materials. Given the dense network of tributary channels in most wetland areas, no point is very far from a surface stream channel. A range of permeability values for the Cross Creek wetlands was obtained from field testing and laboratory analy~is of soil samples (Ward, 1983). Us1n~ . a geometric mean of these permeab1l1ty values 0.0065 em/sec, the time delay of drainage was computed using a standard drain formula (Todd, 1967). The estimated time delay between a drop in channel water level of 70 em (2ft), and a decline in wetlands water table of 15 em (0.5 ft) was computed for a range of distances from the nearest drainage channel. For a distance of 2.4 m (8 ft) from the channel, the water table would decline 15 em (6 in) in about two hours. At a distance of 12 m (40 ft), the 15 em water table decline would take about five days. Few wetland areas are farther than forty feet from at least one minor drainage channel. The conclusion drawn from these calculations is that wetlands are strongly dependent on high creek levels for maintaining a high water table necessary for keeping wetland soils saturated. Except for a few isolated locations, the groundwater reservoir drains too quickly to provide any useful function in independently maintaining the wetlands. Vegetation Resronse. The gradient analysis shown in igure 2 plots the percent coverage of 6 key indicator species an~ community dominants along the water table/drainage regime gradient. This ordination gives a model of species behavior along the moisture gradient and outlines the niche of each species. It also gives a predictive model of change in species abundance if the water Calamagroatls canadensis Carex aquatllls 0:: w > 0 () 1-z w () 0:: w CL Carex utriculata so\ I 60 40 Sparganlum sp. Ranunc!l!us trlcophyllus 20 1 wettest 2 Salix planlfolla 3 4 5 driest WATER TABLE-DRAINAGE REGIME GRADIENT Figure 2. Gradient Analysis of Indicator Plant Species for Cross Creek Wetlands table/drainage regime at any site were to change. Figure 3 plots the average depth to water table or the depth of standing water for the 6 communities described here. The average difference between communities in depth to water table is 13.3 em (5.2 in) • Thus, small differences in water table relations make large differences in composition of the stands. It is hypothesized that any permanent change in summer water table depth greater than 13.3 em will initiate secondary succession. The pathway of secondary succession and the floristic change to be expected can be predicted using Figures 2 and 3. In general, succession would proceed from one species and community to the next up the water table/drainage gradient. For example, if the water table in Carex utriculata communities is dropped an average of 50 em (20 in.), conditions similar to what presently support the 98 Calamagrostis canadensis-Mertensia ciliata community wi 11 deve 1 op on that site. If the water table dropped 15 em, conditions similar to the Carex aguatilis would occur. Whether or not tree species would invade the wetlands is more difficult to predict since forests do not currently occur in the Cross Creek wetlands. OTHER PROCESSES AFFECTING WETLANDS High flows associated with spring runoff perform two functions essential to the maintenance of fen and marsh wetlands: (1) stream meandering and (2) overbank flooding in the meadows. Sustained high volume, high velocity flows, such as occur during spring runoff, provide the energy which allows streams to meander. Only under the influence of high flows are major sections of streambank dis 1 odged, and sediment transported to form point bars. AVERAGE THICKNESS OF SOil ABOVE WATER TABLE (em) 70 SaNa planlfolla community C•l•magroatla c•nadenala-llertenale clll•t• community 60 Celamagroatla canedenala community 50 40 Carex equatllla community 30 20 Carex utrlculate community 10 WATER TABLE LEVEL----0------------------------ AVERAGE DEPTH OF STANDING WATER (em) 10 Aanunculua tr~ophyllua-Sperganlum ap. co,..munlty 20 Figure 3. Average Water Table Depth or Standing Water Depth for Major Communities, late July-early August, 1985 Meandering reworks the floodplain sediments. This serves two critical ecological functions in the Cross Creek meadows. (1) By eroding high terrain, such as beaver dams and natural levees, the stream 1 imits floodplain aggradation that, left unchecked, would eventually convert the floodplain into a river terrace. ( 2) Water provides the nutrients necessary to sustain the growth of the existing, fen-type and marsh-type vegetation. Absent this infusion of nutrients, the soils would become depleted, and the vegetation would be replaced with plant species that flourish in nutrient-poor soils. Overbank flooding assures saturated soil conditions by maintaining high water levels during the growing season. Saturated soil conditions also prevent the invasion of upland plant species into wetland areas. Overbank flooding during the six-week, high streamflow period from late May to early July assures that wetland water levels are near the ground surface. 99 The proposed Homestake II project will capture and divert the high flows associated with spring and early summer runoff. This will drastically reduce meandering and overbank flooding in the wetlands. (1) (2) (3) (4) CONCLUSIONS Main channel water levels are predicted to decrease an average 0.46 meters (1.5 feet) during the growing season as a result of the proposed Homestake II diversions, Groundwater levels will decline in response to declining stream levels during the growing season, The distribution of wetland plant species will change in response to declining groundwater levels during the growing season and some communities would probably die out, Therefore, Homestake II will have an adverse effect on wetlands along Cross Creek. REFERENCES Aqua Resources, Inc., 1984. Wetland Impact Evaluation for Homestake Phase II, prepared for Sacramento District, Corps of Enginners. Burke, R., H. Hemond, K. Stolzenbach, 1980. An Infiltrometer to Measure Seepage in Salt Marsh Soils. In: Estuarine and Wetland Processes, P. Hamilton and K. McDonald (Editors). Plenum Publishing Corp., New York, New York. pp. 413-423. Burton, T., 1985. The Effects of Water Level Fluctuations on Great Lake Coastal Marshes. In: Coastal Wetlands, H. Prince and F. D'Itri (Editors). Lewis Publishers, Inc., East Lansing, Michigan. Dix,R. L. andF.E. Smiens, 1967. The praire meadow and marsh vegetation of Nelson County, North Dakota. Canadian Journal of Botany 45:21-58. Enartech, Inc., 1983. Stream and Wetland Hydrology -Homestake Phase II Project. Kilburn, P., 1983. Preliminary Ecological Investigation of Wetlands for the Homestake II Project, Eagle County, Colorado. International Environmental Consultants, Inc., Denver, Colorado. Kraeger-Rovey, C., E. W. Rovey, 1984. Microscale Engineering Analysis of the Holy Cross Wetlands. Rovey, E., C. Kraeger-Rovey, 1984. Comments of Draft Homestake II Wetland Report. Sundeen, K., R. S. Fifer, 1981. Homestake Phase II Fishery, Water, and Wetland Evaluation for Cross and Fall Creeks. USDA White River National Forest, Rocky Mountain Region. Sundeen, K., 1983. Analysis of Wetland Conditions -Fall and Cross Creeks. USDA White River National Forest, Rocky Mountain Region. Todd, D.K., 1967. Ground Water Hydrology. John Wiley & Sons, New York, New York. U.S. Department of Agriculture, Forest Service, White River National Forest, 1983. Final Environmental Impact Statement on Application for Land Use, Homestake Phase II Project. U.S. Department of Agriculture, Soil Conservation Service, Irrigation Water Requirements, Technical Release No. 21, April, 1967. U.S. Geological Survey, 1984. Flow Frequency Analysis of Discharge Records at Cross Creek near Miniturn, Colorado. WATSTORE Data Retrieval. Ward, J.R., 1983. Hydrological Investigation of Wetlands, Homestake I and II Projects, Eagle County, Colorado. Weber, W.A., 1976. Rocky Mountain Flora. University of Colorado Press, Boulder, Colorado. Westhoff, V., E. van der Maarel, 1978. The Braun-Blanquet Approach. In: Classification and Ordination of Communities. The Handbook of Vegetation Science, R. H. Whittaker (Editor). Junk, the Hague, pp. 617-726. 100 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 SEASONAL SNOW AND AUFEIS IN ALASKA 1 S TAIGA C.W. Slaughter and C.S. Bensonl ABSTRACT: The unglaciated taiga of central Alaska is subjected to seasonal snow and ice for 6 to 8 months of every year. Snow and ice thus play a major role in hydrologic regime. A typical taiga snowpack is less than 100 em in depth, has a mean density at deposition of 0.05 to 0.10 g cm-3, and mean 11 ripe 11 density at time of spring snowmelt of less than 0.30 g cm-3. Low snowpack density (the result of intensive depth hoar formation in response to very steep vapor pressure gradients from base to surface of pack during the entire winter) contrasts with high-density (0.40 g cm-3) tundra snow at wind-affected taiga sites and in the high Arctic. Aufeis can occupy major sectors of stream channels and flood plains, and modifies hydrologic regime by temporary storage of groundwater (winter baseflow) and release of that water to streamflow after the snowmelt season. (KEY TERMS: snow; ice; aufeis; subarctic; taiga; hydrology.) INTRODUCTION Alaska has over 1.5 million km2 of land, and extends through 20 degrees of latitude, from 5l 0 N lat. at Amatsignak Island in the Aleutians to 7l 0 N lat. at Barrow, and through 58 degrees of long- itude from Hyder, on the Alaska-British Columbia border, to Attu at the tip of the Aleutian chain. The westernmost point on the Seward Peninsula is only 88 km from Asia; Little Diomede Island (U.S.) is only 3.7 km from Big Diomede Island (USSR). Mean annual runoff for the State is estimated at 9.8 x lOll m3, or about one-third of the total estimated water yield of the entire United States. South-central and southeastern Alaska account for some 5.7 x lOll m3 of water, while the Yukon and Kuskokwim River systems contribute about 2.5 x 1011 m3 (Hartman and Johnson, 1978). Four major physiographic regions are recognized in Alaska (Wahrhaftig, 1965): the Arctic Coastal Plain, the Rocky Mountain System, the Intermontane Plateaus, and the Pacific Mountain System. Alaska may be divided into four climatic zones that roughly parallel the physiographic regions (Hartman and Johnson, 1978). The Arctic zone has a mean annual temperature of -12 to -l5°C, generally low precipitation, and is subject to strong winds. The Interior zone is characterized by a continental climate; summers are short and warm, with air temperature occasionally exceeding 30°C, and winters are commonly long and cold with -40°C commonly and -sooc occasionally observed. Mean annual temper- ature is -10 to -5°C; precipitation varies from 25 to over 90 em annually. The Transitional zone has less yearly temper- ature variation, higher precipitation and more cloudiness than the continental Interior, and is intermediate between the Interior and the Maritime zone of south- central and southeastern Alaska. The Maritime zone is characterized by relatively low variation in annual lRespectively, USDA Forest Service, Pacific Northwest Research Station, Institute of Northern Forestry, 308 Tanana Drive, Fairbanks, AK 99775-5500, and University of Alaska, Geophysical Institute, Fairbanks, AK 99775. 101 Alaska SCALE: 102 lf)O Miles ~00 Krlometers GENERAL AREA OF THE AlASKAN TAIGA Figure 1. Distribution of taiga (after Viereck, 1973) and climatic zones (after Hartman and Johnson, 1978). A: Arctic; C: Continental; T: Transition; M: Maritime. temperature, a great deal of cloudiness, high precipitation (to over 500 em annuallt), and mean annual air temperature above 0 C. The taiga (Figure 1) generally coincides with the Interior climatic zone and with Wahrhaftig's (1965) Intermontane Plateaus physiographic region. According to Viereck (1973), "In Alaska the taiga extends from the south slope of the Brooks Range southward to its border with the coastal forests, eastward to the border with Canada, and westward to a maritime tree line very close to the Bering and Chukchi Seas. Within this area of 138 x 106 hectares, approximately 32% ..• is forested .•.. The unforested land consists of extensive bogs, brush thickets, grasslands, sedge meadows, and some alpine tundra." This region is composed of rolling, occasionally rugged uplands and extensive low-lying wetlands. The region was unglaciated in the Wisconsin Glaciation save for localized activity at high elevations. The taiga generally corresponds to the zone of discontinuous permafrost--perennially frozen ground--and periglacial geomorphic processes are active in the entire region. Permafrost is conmon on north-facing slopes and in poorly-drained lowlands but is generally 1 acki ng on southern exposures. The 102 proportion of frozen ground increases with latitude; only sporadic to discontinuous permafrost is found in the south, but permafrost is essentially continuous from the crest of the Brooks Range north to the Arctic Ocean. TAIGA SNOW Snow is a dominant feature of the taiga and mantles the entire landscape for 6 to 8 months of every year, depending on specific location and year. Surface energy relationships which affect biological and physical processes, to say nothing of.human activities, are markedly responsive to this seasonal snowpack. Two primary northern snow t¥pes and regions are distinguished by Pruitt (1970): "taiga snow" and "tundra snow." Hare (1971) showed that densit¥ of taiga sno~acks in Canada reaches 0.25 to 0.30 g cm-3 in late winter, while density of tundra snow is consistently greater than 0.35 g cm-3. The seasonal snowpack of interior Alaska may be considered typical taiga snow: densities at initial deposition are low, 0.05 0.10 g cm-3, movement and reworking by wind is uncommon, and snowpack metamorphism is primarily through vapor transfer within the pack in response to strong vapor pressure gradients. Relatively warm soil (rarely below -5°C) at the base of the snowpack and cold ambient air produce steep vapor pressure gradients, with upward water vapor flux within the snowpack averages 0.025 g cm-3 day-1 throughout the winter (Trabant and Benson, 1972; Trabant, et al., 1970). Final "ripe" snO\'Ipack density is commonly 0.20 to 0.28 g cm-3. In contrast to taiga snow, "tundra" snow is conmonly deposited during \~indy conditions, which lead to errors in measurement. Long-term climatic records have underestimated snowfall on Alaska's Arctic slope by a factor of three; the annual flux of windblown snow is on the order of 50 ± 20 t/m (Benson, 1982). A significant, but unknown, amount of snow is lost by sublimation during wind transport; in Wyoming, Tabler (1975) showed that about a third of windblown snow sublimated during transport. The t¥pical structure of tundra snow includes "Tundra" Snow • above treeline, or otherwise exposed • extensive re-working by wind • higher density on windslab layer at top of pack • relatively low density at base· of pack Typical Profiles Temperature, 'C Stratigraphy Density, gem"' -20 -10 0 0.10 0.20 0.30 0.40 0.50 8or-L-~~r-~~~~========~=.~ ~ ~ 20 E 3: ~ ~ 40 =-! ~ 60 68 Water Equivalent =17.5 em 5 March 1975 Caribou Peak 0 0.10 0.20 0.30 0.40 0.50 ~ • f Water Equivalent= 14.0 em Figure 2. Typical taiga and tundra snowpack stratigraphy. a hard wind-packed layer with density 0.35 to 0.50 g cm-3 overlying a low-density depth hoar layer. Depth-hoar may be very thin while snow beneath the wind-packed layer is soft with grain size of about 1 mm. 11 Tundra 11 snow is very irregularly distributed. Stream valleys may be comoletely filled with dense (0.35 0.40 g cm-3), hard snow while adjacent slopes are snow-free (Benson, 1967; Liston, 1986; McKay and Findlay, 1 971 ). Tundra snow prevails on the Arctic coastal plain of Alaska, and is also found in wind-affected settings in the taiga--areas above treeline (Benson, 1967) and in valleys subject to frequent cold-air drainage 103 "Taiga" Snow • generally below treeline • low snow pack density • little to no re-working by wind • subject to strong air temperature inversions Typical Profiles Temperature, •c Stratigraphy Density, gem"' -20 -10 0 0.10 0.20 0.30 0.40 0.50 1\1\ ,...,.. ""'"" ,. ,. """""" ~ ",.",., Water Equivalent = 13.2 em Stratigraphic Symbols + + + + + New snow, some original crystal forms still recognizable . . . .. -· ,· .·. ··-·· Fine or very fine grained (snow < 1 mm). Same as above, but very hard--a result of wind packing --often referred to as a windslab. <:?:\!\'.;:;i, Medium grained snow 1 to 2 mm. 1\ 1\ A 1\ Depth hoar, coarse loosely-bonded crystals often with well /\ 1\ A " developed crystal faces. Larger chevron patterns indicate A A /\ larger crystal size. winds, such as at Delta Junction, Alaska (Benson, 1972; Bilello, et al, 1970). A 11 transitional 11 snow zone has been delineated by Benson ( 1982). This zone is climatically intermediate between the cold continental climate of the central Alaska taiga and the cold maritime climate of the western Alaska coast and Bering Sea. West of Koyukuk (158° W long.) in the Yukon- Kuskokwim delta, the temperature is higher and winds are stronger than they are farther east, and the climate becomes more maritime than continental. Many storms produce mixed rain and snow, and the· snow cover is characterized by significant amounts of ice crusting, though depth hoar is still developing at the ground surface {Benson, 1982). Typical profiles of taiga and tundra snowpacks from interior Alaska sites are depicted in Figure 2. Tundra and taiga snowpack types may be closely inter- mingled. At many tree-line or otherwise wind-affected sites in the taiga, deep, high-density wind-drifted tundra snow grades rapidly to shallow, low-density taiga snow over very short distances {Benson, 1978). SEASONAL ICE Rivers and lakes in the taiga can have an ice cover for 5 to 7 months of the year. Floating ice, frazil ice, anchor ice, ice breakup events, and the hydraulic and thermal conditions attendant to these seasonal ice forms have received detailed attention esl ewhere {Benson, et al, in preparation; Calkins, 1979; Michel, 1971; Osterkamp, 1975, 1978) and are not considered in the present discussion. Another important seasonal ice phenomenon forms when water in or adjacent to a stream channel rises above the surface of an existing ice cover. Such an accumulation of ice, superimposed on the frozen surface of a stream, river, or sector of landscape is generally termed 11 aufei S 11 {Grey and !~acKay, 1979) or 11 icing11 {Carey, 1973); the Russian term 11 nal ed11 is preferred by some authors {Akerman, 1982). Aufei s is commonly associated with permafrost-affected terrain. Comprehensive reviews of aufeis are provided by Carey {1973) and Grey and MacKay {1979). Kane et al. {1973) and Kane and Carlson {1977) summarize knowledge of mechanisms of aufeis formation. Aufeis is a circumpolar phenomenon, and has been studied or documented in locales ranging from Yukon {van Everdingen, 1982) to Spitsbergen {Akerman, 1982) to Mongolia {Froehlich and Slupik 1982). Aufeis deposits are multilayered and may be several meters thick in streams that normally have water depths of 50 em or less. In Alaska•s taiga, aufeis is most commonly found in stream valleys and river floodplains. Extent and thickness of aufeis deposits in the taiga are generally less than for those on Alaska• s 104 Arctic Coastal Plain {Dean, 1984) or in major north-flowing river systems of northeastern Siberia {Kane, personal communication). HYDROLOGIC RELATIONSHIPS Seasonal snow and seasonal ice constitute forms of detention storage for precipitation. The snowpack binds virtually all precipitation falling during the period from October through t1arch in a veneer overlying most of the taiga 1 andscape. This water may be redistributed vertically over short distances by snowpack metamorphism and depth hoar formation, and may be redistributed laterally by wind action. In either case, the redistribution is essentially local, and incoming precipitation, at least below treeline, is largely retained on the watershed of original deposition. The hydrologic relationships of seasonal snow have been studied over many years in many settings. An extensive body of literature is available concerning snowmelt physical processes, snow/terrain/ vegetation relationships, meltwater routing, co nc:eptua 1 and quantitative modeling, and engi nee ring/resource management applications of this knowledge {e.g., Colbeck and Ray, 1979; Glen, 1982; Meiman, 1969; National Academy of Sciences, 1974; U.S. Army Corps of Engineers, 1956). The seasonal snowpack is subject to the same processes and physical laws in the taiga as in other regions. Major differences in snow hydrology, or in application of snow hydrology knowledge developed in more temperate settings, derive from several factors: the prolonged and continuous nature of the snow-cover season, extreme seasonal variability in energy available at the earth surface at high latitudes, the existence of cold soil {either seasonally or perennially frozen) at the base of the snowpack during most of the winter and during spring snowmelt, pronounced development of depth hoar in the snowpack, occurrence of aufeis in streams and valleys, and a relative paucity of quantitative data concerning prec ipi tati on and streamflow. Depending upon the specific setting and year, snow covers the taiga for 6 to 8 months of each year. Often the ambient air temperature does not exceed 0°C from October through March. Average snow cover duration in the central Alaska 1 owlands (Fairbanks) is 214 days (Haugen, et al., 1982). Snowpack persistence is a direct response to avai labi 1 i ty of energy for warming and melting; the extreme seasonal variability of incoming short-wave radiation at high latitudes is well documented (Baker and Haines, 1969; Seifert, 1981). Daily short-wave radiation available at a horizontal surface at Fairbanks (Figure 3) varies from less than 200 W-hr m-2 in December to over 5000 W-hr m-2 in June (Wendler, 1981). The winter c 1 ima te of the 1 owl a nds of the taiga is typified by cold air and 1 ittl e wind. The snow-covered surface favors development of strong surface inversions which trap the calm, cold air in a surface layer between SO and 1 00+ m thick. The snowpack lying within the altitude range of the inversion layer is subject to negligible winds, often through the entire winter. The sno\\1)ack is subjected to very 1 ow ambient air temperature at the air/snow interface--from -30 to -45°C is normal. The ground beneath the sno\\1)ack rarely becomes colder than -6 to -l0°C, and strong temperature gradients preva i1 in the snow; this leads to vertical water vapor flux and extensive depth hoar development, with average density generally less than 0.20 g cm-2. The steep temperature gradient within the sno\'4)ack is accompanied by a steep vapor pressure gradient, leading to vertical flux of water vapor of up to 0. 025 g cm-2 day-1. This vertical flux of water leads to dessication of the underlying soil (Kane, et al., 1978; Trabant and Benson, 1972). Santeford (1979) reports that about 50% of water initially available in an organic soil-surface layer overlying permafrost mgrates upward into the snowpack in the course of a winter. Such soil desiccation strongly influences the disposition of initiru snowmelt water (Slaughter and Kane, 197 9) • Spring, the period of rapid increase in solar insolation and day length, is accompanied by rising solar angle so that an increasing proportion of north-aspect 105 a "'""' 200 N I s 1-< 0 ..c: I 2 4 6 8 10 12 14 16 18 20 22 ;3: Hour of Day ......., 1:: 0 .... b +J ell •.-l "0 1,..---... \ Max ell ~ I \ I \ I \ I \ ' ' \ \ ,, \ I \ \ I \ ,~in \ I \ I ......... \ I .... \ I ..... , ' ' \ ......... .... ..... 0 .... A M J J A s 0 N D J F M Month of Year rigure 3. (a) Hourly mean values of global radiation (horizontal surface) in December, March, June, and September, Fairbanks (65°49 1 N). (b) Daily sums (maximum, mean, minimum) of global radiation on a horizontal surface, Fairbanks (after Wendler 1981 ) • terrain is subject to the increasing energy load; this coincides with rapid decrease in surface albedo as the snow surface ages and settles, and twigs and branches become exposed and act as energy absorbing and reradiation media. The result is a relatively brief, intense snowmelt runoff season. The complete snowmelt runoff period can be as little as 7 to 10 days (Kane and Stein, 1984). Disposition of meltwater is affected by presence of seasonally or perennially frozen ground. Kane and Stein (1983, 1984) demonstrated that antecedent (fall) soil moisture conditions affect volume of snowmelt runoff the following spring on small plots and on a large (5,132 km2) watershed in the discontinuous-permafrost zone of Alaska. The proportion of snowmelt water that infiltrated into the soil rna ntl e on a penna frost-free site was sharply reduced, with a concomitant increase in runoff, by increased soil moisture levels in the fall. The presence of permafrost similarly influences disposition of meltwater (and summer precipitation). Permafrost essentially precludes infiltration beneath the seasonally thawed .. active layer, .. thereby restricting meltwater detention and transmission to the soil mantle (largely organic soils of low bulk density and high tra nsmi ss i vi ty) (Kane, et al. , 1981; Slaughter and Kane, 1979). This effectively increases the proportional yield of meltwater in spring, relative to nonpermafrost catchments. In an upland watershed north of Fairbanks, a permafrost-free 5.2-km2 basin yielded 1 e ss than 50% as much tot a 1 snowme 1 t runoff volume as did an equal-sized (5.7 km2) basin which is dominated by pennafrost (Slaughter, 1981); the spring period of high flow lasted more than twice as long in the permafrost-dominated basin, 8 days versus 3 days. Aufeis derives from groundwater sources (van Everdingen, 1982). Aufeis immobilizes water that is being yielded to the stream channel or to the ground surface during the winter and which would otherwise contribute to current streamflow. A consequence of aufeis formation in a catchment is solid-state storage of baseflow and diminution of streamflow during the winter and spring, with subsequent augmentation of streamflow when aufeis melts during spring and early s ul1lller. The occurrence of aufei s has consequences for both hydrologic regime and geomorphic processes. Akerman (1982) showed that aufeis has both direct and indirect geomorphologic effects. The physical presence of ice in stream channels and on floodplains during spring runoff may protect stream banks and adjacent ground surfaces from fluvial action, and may divert flow to landscape sectors not normally abraded by water---such as at the lateral margins of ice, perhaps far removed from the normal (ice-free) stream channel (Froehlich and Slupik, 1982; Kane and Slaughter, 1972). Such protection during major runoff events 106 may influence either local bed profile or local drainage patterns (Akerman, 1982). Presence of aufeis in cor!bination with an ice cover on streams can produce marked effects on streamflow. Kane, et al. ( 1973) monitored pore water pressure in two stream valleys in discontinuous permafrost, and demonstrated a very dynamic system during the 11 frozen 11 period of aufei s format ion and continuous ice cover on streams. Hydrostatic head variations as great as 2.1 m within a single week were recorded, reflecting changes in groundwater flow and subsurface conditions for flow in the aquifer/stream/aufeis system. Peaks in hydrostatic head were generally accompanied by an increase in formation or thickness of aufeis (Kane, et al., 1973). The water mass held as aufeis can be appreciable; in a central Alaska watershed of 104 km2, aufeis covered 1% of the drainage area (Benson, 1978), comprised approximately 4% of total annual streamflow and 40% of expected winter streamflow (Kane and Slaughter, 1972). This ice melts more slowly than the adjacent snowpack, and thus contributes meltwater (release from storage) to the stream system over a period of weeks or even of several months into summer. Although aufeis cannot be reliably predicted in terms of intensity and magnitude, Sloan et al. (1975) monitored a transect across Alaska and demonstrated that aufei s has a tendeocy to recur at the same 1 ocati ons year after year; however, not all 1 ocat ions s hawed aufeis activity in every year of their study. A similar pattern of recurrence was documented by Slaughter (1982) in a central Alaska watershed. SUMt·1ARY Seasonal snow and ice are dominant components in the annual hydrologic regime of Alaska's taiga. A veneer of snow covers the landscape for 6 to 8 months of the year and strongly affects energy balance at the earth surface. At high latitudes, rapidly increasing day length and solar angle combine with decreasing albedo to produce short-1 ived spring snowmelt runoff periods; the flood of record may result from snowmelt water encountering aufeis-blocked stream channels. Presence of permafrost and of associated organic soil layers modifies translation of meltwater from snowpack to stream. Aufeis comprises another form of detention storage; the aufeis affects fluvial geomorphic processes, modifies winter base flow in streams, and provides meltwater well into summer months in some locations. LITERATURE CITED Akerman, J., 1982. Studies on naledi (icings) in West Spitsbergen. pp. 189-202 In: Proceedings, 4th Canadian Permafrost Conference. The Roger J.E. Brown f~emori al Volume. National Research Council of Canada, Ottawa. Baker, D.G. and D.A. Haines, 1969. Solar radiation and sunshine duration relationships in the north-central region and Alaska. Minnesota Agricultural Experiment Station Technical Bulletin 262. Benson, C. S., 1967. A reconnaissance snow survey of interior Alaska. UAGR-190. Geophysical Institute, Universi~ of Alaska, College, Alaska. 71 pp. Benson, C.S., 1972. Physical properties of the snow cover in the Ft. Greely Area, Alaska. SR-178, U.S. Army Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire 24 pp. Benson, C.S., 1978. Studies on the seasonal snow cover of the Caribou-Poker Creeks Research Watershed during the winters of 1974-75 and 1975-76. Final Report. Geophysical Institute, Universi~ of Alaska, Fairbanks, Alaska. 38 pp. Benson, C.S., 1982. Reassessment of winter precipitation on Alaska's Arctic slope and measurements on the flux of wind blown snow. UAG R-288, Geophysical Institute, University of Alaska, Fairbanks, Alaska. 26 pp. Benson, C.S., W. Harrison, J. Gosink, S. Bowling, L. Mayo, and D. Trabant, in prep. Workshop on Alaskan hydrology: problems related to glacierized basins. 107 Geophysical Institute, University of Alaska, Fairbanks. Bilello, t4.A., R.E. Bates, and J. Riley, 1970. Physical characteristics of the snow cover, Fort Greely, Alaska, 1966-67. Technical Report 230, USA Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire. 33 pp. Calkins, D.J., 1979. Accelerated ice growth on rivers. CR 79-14, U.S. Army Cold Regions Research and Engineering Laboratory, Hanover, N.H. 5 pp. Carey, K.L., 1973. Icings developed from surface water and ground water. Cold Regions Science and Engineering Monograph III-03. USA Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire. 65 pp. Colbeck, S.C. and M. Ray (editors), 1979. Proceedings, Modeling of Snow Cover Runoff. USA Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire. 432 pp. Dean, K., 1984. Stream-icing zones in Alaska. Report of Investigations 84-16. Alaska Department of Natural Resources, Division of Geological and Geophysical Surveys, Fairbanks. 20 pp. Froehlich, W. and J. Slupik, 1982. River icings and fluvial activity in extreme continental climate: Khangai Mountains, t~ongo 1 i a. pp. 203-211 In: Proceedings, 4th Canadian Permafrosy-conference. The Roger J.E. Brown Memorial Volume. National Research Council of Canada, Ottawa. Glen, J.W. (editor), 1982. Hydrological aspects of alpine and high-mountain areas. IAHS Publication 138, International Association of Hydrological Sciences, Washington, D.C. 350 pp. Grey, B.J. and O.K. MacKay, 1979. Aufeis (overflow ice) in rivers. pp. 134-165 In: Canadian Hydrology Symposium 79--Cold Climate Hydrology. National Research Council of Canada, Ottawa. Hare, F .K., 1971. Snow-cover problems near the Arctic treeline of North America. Report Kevo Subarctic Research Station 6: 31-40. Hartman, C.W. and P.R. Johnson, 1978. Environmental atlas of Alaska. Institute of Water Resources/Engineering Experiment Station, Universizy of Alaska, Fairbanks, Alaska. 95 pp. Haugen, R.K., C.W. Slaughter, K.E. Howe, and S.L. Dingman, 1982. Hydrology and climatology of the Caribou-Poker Creeks Research Watershed, Alaska. CRREL Report 82-26, U.S. Army Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire. 34 pp. Kane, D.L., S.R. Bredthauer, and J. Stein, 1981. Subarctic snowmelt runoff generation. pp. 591-601 In: The Northern Communit¥: A Search Faria Qual it¥ Environment. Proceedings of the Specialty Conference. American Society of Civil Engineers. Kane, D.L. and R.F. Carlson, 1977. Analysis of stream aufei s growth and climatic conditions. pp. 656-670 In: Proceedings, 3d National Hydrotecnnical Conference, Canadian Society for Civil Engineering. Quebec, Canada. Kane, D.L., R.F. Carlson, and C.E. Bowers, 1973. Groundwater, pore pressures adjacent to subarctic streams. pp. 453-458 In: Proceedings, North American Contribution, 2nd International Conference on Permafrost. National Acadany of Sciences, Washington, D.C. Kane, D.L., J.N. Luthin, and G.S. Taylor, 1978. Heat and mass transfer in cold regions soils. IWR-65, Institute of Water Resources, Universi zy of Alaska, Fairbanks, Alaska. Kane, D.L. and C.W. Slaughter, 1972. Seasonal regime and hydrological significarx:e of stream icings in central Alaska. pp. 528-540 In: Proceedings, The Role of Snow and Ice-rn Hydrology. Banff, Alberta, Canada. UNESCO/WMO/IASH. 108 Kane, D.L. and J. Stein, 1983. Water movement into seasonally frozen soils. Water Resources Research 19: 1547-1557. Kane, D.L. and J. Stein, 1984. Plot measurements of sn~Nmelt runoff for varying soil conditions. Geophysica 20{2): 123-135. L i stan, G. , 1986. Study of the seasonal snow cover in the Upper Kuparuk drainage on Alaska•s Arctic Slope. M.S. Thesis, Universit¥ of Alaska, Fairbanks. McKay, G.A. and B. Findlay, 1971. Variation of snow resources with topography and vegetation in Canada. Proceedings, 39th Annual Western Snow Confererx:e: 17-26. Meiman, J.R. {editor), 1969. Proceedings of the Workshop on Snow and Ice Hydrology. U.S./IHD Snow and Ice Work Group. Fort Collins, Colorado. 142 pp. t~ichel, B., 1971. Winter regime of rivers and lakes. Monograph 111-Bla, U.S. Army Cold Regions Research and Engineering Laboratory Hanover, New Hampshire. 130 pp. National Acadany of Sciences, 1974. Advanced concepts and techniques in the study of snow and ice resources an i nterdi sci pl inary symposi urn. Nati anal Academy of Sciences, Washington, D. C. 789 pp. Osterkamp, T.E., 1975. Tanana River ice cover. pp. 201-209 In: Proceedings International SympoSTum on Ice Problems, IAHR, Hanover, New Hampshire. Osterkamp, T.E., 1978. Frazil ice formation: a review. J. Hydraulics Division, American Society of Civil Engineers 104 {HY-9): 1239-1255. Pruitt, W.O., 1970. Some ecological aspects of snow. pp. 83-89 In: Ecology of the subarctic regions: PrOceedings of the Helsinki Symposium. UNESCO. Santeford, H. S., 1979. Snow-soil interactions in interior Alaska. pp. 311-318 In: Colbeck, S.C., and M. Ray {editors}. Proceedings, r~odeling of snow cover runoff. U.S. Anny Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire. Seifert, R.D., 1981. A solar design manual for Alaska. University of Alaska, Institute of Water Resources, Fairbanks, Alaska. 163 W· Slaughter, C.W. and D.L. Kane, 1979. Hydro 1 ogi c ro 1 e of shall ow organics soils in cold climates. pp. 380-389 In: Canadian Hydrology Symposium 79--Cola Climate Hydrology. National Research Council of Canada, Ottawa. )laughter, C.W., 1982. Occurrenceand recurrence of aufeis in an upland taiga catchment. pp. 182-188 In: Proceedings, 4th Canadian Permafrost-conference. The Roger J. E. Brown ~1emori al Vo 1 ume. National Research Council of Canada, Ottawa. Slaughter, C.W., 1981. Snowmelt runoff estimation with time lapse photography. Pacific Northwest Section, American Geophysical Union. Central Washington Universi zy, Ell ens burg, Washington. Sloan, C. E., C. Zenone, and L.R. Mayo, 1975. Icings along the trans-Alaska pipe 1 i ne route. Open-Fi 1 e Report 7 5-87, U.S. Geological Survey, Anchorage, Alaska. 39 pp. fabler, R.D., 1975. Estimating the transport and evaporation of blowing snow. pp. 85-104 In: Snow Management on Great Plains Symposi urn Proceedings. Publication 73, Great Plains Agricultural Council, Bismark, North Dakota. rrabant, D., C.S. Benson, and G •. Weller, 1970. Physical-thermal processes in the seasonal snow cover of northern Alaska. pp. 328-332 In: Proceedings of the Twentieth Alaska Science Conference, College, Alaska. Alaska Division, American Association for the Advancement of Science. frabant, D.C. and c.s. Benson, 1972. Field experiments on the development of depth hoar. pp. 309-322 In: Doe, B.R., and O.K. Smith (editorST. Studies of 109 Mineralogy and Precambrian Geology. GSA Memoir 135, Geological Society of America, Boulder, Colorado. U.S. Army Corps of Engineers. 1956. Snow Hydrology. Summary report of the snow investigations. North Pacific Division, U.S. Army Corps of Engineers. Portland, Oregon. 437 p. van Everdingen, R.O., 1982. Management of groundwater discharge for the solution of icing problems in the Yukon. pp. 212-226 In: Proceedings, 4th Canadian Permafrost Conference. The Roger J.E. Brown Memorial Volume. National Research Council of Canada, Ottawa. Viereck, L.A., 1973. Wildfire in the taiga of Alaska. Quaternary Research 3(3): 465-495. Wahrhaftig, C., 1965. Physiographic divisions of Alaska. Professional Paper 482, U.S. Geological Survey. Wendler, G., 1981. Solar radiation data for Fairbanks. Geophysical Institute Report, Solar Energy ~teorological Research and Training Site, University of Alaska. Fairbanks, Alaska. lU NOI.LGH'I'IOG V.LVG GNV NOI.LV.LN3W!lli.LSNI JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 WATER REDISTRIBUTION IN PARTIALLY FROZEN SOIL BY THERMAL NEUTRON RADIOGRAPHY Michael A. Clark 1, Dr Roger J Kettle 1, Giles D'Souza2· ABSTRACT: Thermal neutron radiography has been utilised as a technique for determining relative water movements in partially frozen soil. The method is non-destructive and determinations of local water content can be made at any time throughout the test. The experimental test rig is designed so that the soil matrix can be uniformly irradiated by a thermal neutron beam, to obtain radiographs, which upon interpretation yield information on relative water I ice contents. Instrumentation is incorporated in the soil matrix in the form of psychrometer I thermocouples to record water potential, electrical resistance probes to enable ice and water to be differentiated on the radiographs and thermocouples to record the temperature gradient. Interpretation of the radiographs is accomodated using image analysis equipment capable of distinguishing between 256 shades of grey. Image enhancing techniques are then employed to develop false colour images which show the flow of water and development of ice lenses in the soil matrix. Water content determinations can then be made and plotted against potential measurements for each radiograph. From these graphs, using relevant theory, pore water distributions can be obtained and combined with water content data to give hydraulic conductivities. (KEY TERMS: thermal neutron radiog!aphy; image analysis; hydraulic conductivities; partially frozen soil.) Hydraulic conductivity is a fundamental parameter of a soil. In an unfrozen soil it can be determined using _a permeameter where water enters and leaves the specimen via inlet and outlet reservoirs. However if the soil is subject to freezing conditions the reservoirs have to be maintained at sub-zero temperatures which results in the reservoir water freezing. One way of preventing this is to add lactose to tne water, this however limits the temperature range which may be examined and there is a danger of lactose molecules moving into the soil by molecular diffusion or by transportation by the flowing water.(Williams and Burt 1974). In recent years techniques which dispense with the involvement of a permeameter have been developed. These have often incorporated the use of various forms of radiation to monitor the moisture flux within the soils. X-rays have been used to depict lead shot, imbedded in a soil subject to partial freezing, to give information on water redistribution during freezing (Yoneyama et al, 1983 ) and by using dual gamma-rays, water and density are monitored simultaneously and used to determine water flow in a freezing soil ( Fukuda et al, 1980 ). The techniques reported here have been introduced in earlier papers ( Clark & Kettle, 1985a, 1985b ). Neutron radiography is used to depict, on photographic plates, water I ice in a soil subject to partial freezing. Neutrons are absorbed by atoms according to their atomic weight. Hydrogen, the main constituent of water, has the lowest atomic weight of the elements and is the greatest absorber of neutrons. The cell and matrix are therefore made up of elements with low coefficients of neutron absorption so that the degree of exposure of the photographic plate is directly related to the water I ice content of the matrix. Using the images and data, from instrumentation incorporated in the soil, information on water movements and hydraulic potentials can be obtained and developed to give hydraulic conductivities. 1: Department of Civil Engineering, Aston University, Aston Triangle, Birmingham, England. 2: Remote Sensing Unit, Bristol University, Bristol, England. 113 EXPERIMENTAL SETUP The Snowcal (powdered chalk) matrix is housed in a rig made entirely of PTFE sheet and aluminium. Snowcal PTFE and aluminium have low neutron absorpti~n coefficients and are hence virtually 'transparent' to the thermal neutron beam. The matrix is compacted, at approximately 1 0% water content, into six rectangular cross:section PTFE rings within a rectangular ~ross-sect1on ce~l. The cell made of PTFE for insulation and encased m alumini~m for rigidity, forms the central portion of the experimental set up with similar lower and upper cells completing the main apparatus. The lower cell contains a water bath and water temperature regulation unit and the upper cell con~ains a freezing head. Once the three cells are f1tted together the apparatus is free fr~m leaks and hence all water entering the central cell v1a the water bath ( fed by a Mariette vessel ) remains in the matrix as water or ice. End temperatures during the test are lowered to +4°C and -1°C by circulating ethylene glycol I water solution through the water temperature regulation unit and freezing head using cryostats. The central and upper cells are modified so that psychrometer I thermocouple probes, used to measure water potential, fit into the PTFE rings and protrude into the matrix. Any heave that occurs during the test is allowed to develop by the free vertical movement of the rings. The three cells comprising the main apparatus are illustrated in Figure 1. The apparatus is located so that the flat sided central cell is irradiated by the neutron beam. To detect neutrons which pass through the cell and matrix an x-ray film plate and gadolinium intensifying screen are used, this is housed in a Light Tight Container (LTC) positioned behind the central cell. A reference cell with three compartments containing snowcal at zero water content, 1 00% water content and the initial water content of the matrix is also in the line of the beam and appears on all the photographic plates. Attached to the reference cell is a watch on which the hands and hours are marked with small lengths of plastic tube, the plastic tubing has a high hydrogen content and hence shows up clearly on the radiographs, thereby providing a permanent record of the time of exposure . By comparing the grey shades of the matrix with the grey shades of the reference cell direct determination of water contents can be made. Incorporating the reference cell in this way elliminates any errors in developing or exposure time as each radiograph has its own built in reference standards. As water and ice are depicted identically on the radiographs, it is necessary to distinguish between these two phases by determining the relative electrical resistance of the soil at various locations. 114 0 0 11 v ~ '--2. 0 0 Upper Cell 0 .... ~ .... i 0 ... ~ ~ II ~7 0 • I- 0 p f- ~ Q F reezing Head p ~ TFE Sleeve t$ H ,\ fl PTFE Rings With Holes For Probes lA r/QJ. HI ~ r;:;::v ll II. .f 1-.I .II I I'~ 0 0 Central Cell ~ 0 0 ~ 0 Q v 1.~~:~::~~:::;~:~~:::~~~:~::~~::~;:~=~::::.:::~:;~~::;~;~ . .-.:;·~.:;~ ~ b',d J 1-l .. - 0 0 17 tl tl c Lower Cell 0 0 0 0 ~ ..._. ~ Figure 1 (Scale 1: 2) Porous Brass Cover Water Bath The neutron beam has a large neutron intensity and requires a high degree of shielding. This is achieved using concrete blocks arranged to form a cavity for the aooaratus. The apparatus is surrounded by a series of copper pipes through which ethylene glycol/water mixture, at reduced temperature, is circulated to reduce the air temperature. Access to the apparatus for positioning and removing the LTC is made possible by a narrow slot cut in the shielding. All connections to the instrumentation for the apparatus pass through a channel, formed in the shielding and re-filled with blocks of lead and paraffin wax. The exposure time for each radiograph is accurately maintained using a paraffin wax beam stop to seal off the beam at the end of the exposure period. The radiographs obtained show the formation of the frozen fringe and as they are one to one images of the cell any heave associated with the partial freezing of the matrix can be measured directly from them. INSTRUMENTATION The water supply to the matrix is from a marriote vessel. Within the vessel a polystyrene float, connected to a linear motion potentiometer , rests on ~he water. In this way any water uptake by the matrix IS represented by a movement of the float which is registered by the potentiometer and recorded on a chart recorder. Wescor PST 55-30 psychrometers and a Wescor PR55 psychrometric microvoltmeter are used to measure water potential in the soil matrix.The psychrometers incorporate a small chromel constantan thermocouple concentrically mounted at the centre of a hollow cylindrical stainless steel bulb. This design is chosen as it minimises the effects of temperature gradients which adversely affect the sucti~n readings. The single junction thermocouple perm1ts measurment of these thermal gradients to enable correction of the suction readings.A calibration model for this type of thermocouple psychrometer has been developed by Brown and Bartos (1982) and has been used in this test. Water potentials are measured with the psychrometers using the peltier effect. Water from the matrix is conducted through the bulb to the inner suface where it evaporates until the humidity approaches 100%. An electric current is passed through the thermocouple for fifteen seconds cooling the sensing junction slightly below ambient temperature so that water from the atmosphere surrounding the couple condenses on the thermocouple. 115 After cooling the condensed water evaporates which again cools the junction. This time however the cooling is a function of the rate of evaporation which is a function of the vapour pressure of the atmosphere which in turn is a funtion of the water potential of the matrix. Copper-constantan thermocouples are located alongside the psychrometers to record the matrix temperature throughout the test. The electrical resistance of the soil between the psychrometer probes is determined throughout the test to identify ice in the matrix. This is achieved through connections to the stainless steel shields of the psychrometers. The shields are insulated from the rest of the probe by a plastic casing and from each other by the PTFE rings which house them. The work has been carried out at the Universities of Manchester and Liverpool Research Reactor using a thermal neutron beam from a horizontal access hole, which gives a maximum neutron flux of 7.5 x 107 n cm-2 s-1 at the radiographic position. An exposure time of 18 seconds is used for each radiograph at a reactor power of 1 OOkw. Loading the Matrix With the three cells of the main apparatus in position and all instrumentation in place the matrix is loaded, via the upper cell, into the six 12.5 mm dep PTFE rings of the central cell., The matrix is compacted in ten equal layers. Each layer contains a specified quantity of matrix (dependent on the water content) and is compacted to a thickness of 7.5 mm. In this way the compaction is standardised and no compaction plane is incident with a ring intersection. Test Procedure With the apparatus positioned in the cavity formed in the concrete shielding, and all connections made to instrumentation and cooling equipment, cooled ethylene glycol/water solution is circulated through the copper tubing and left on overnight to lower and stabilise the matrix temperature. An initial radiograph is then taken before freezing is initiated. Freezing is achieve by circulating cooled~thylene/glycol solution through the freezing head. Subsequent radiographs are obtained at approximately half hour intervals throughout the test. The same process is used to obtain each radiograph. With the LTC positioned behind the cell the beam stop is withdrawn, allowing the thermal neutron beam to fall on the cell, and replaced after the exposure time of eighteen seconds has elapsed. When the LTC has radioactively 'cooled down' it is removed and the photographic plate developed. Radiographic Materials FILM 'KODAK' INDUSTREX CX, Medium Speed, fine-grain, high -contrast, direct exposure film. I:BIEI..CflER 'KODAK' LX 24 X-RAY DEVELOPER. STOP BATH -'KODAK' LX INDICATOR STOP BATH RXER 'KODAK' FX-40 X-RAY LIQUID FIXER HARDENER -'KODAK' HX-40 X-RAY LIQUID Ht\RI:E'.ffi RADIOGRAPH ENI-W'JCEMENT AND ANALYSIS Contact prints are made of the radiographs (negatives} and 35 mm slides (transparencies} made of these prints. The transparencies are used in a scanning microdensitrometer.The size of the soil matrix on the original negative is 72.5 mm x 84 mm and on the 35 mm negatives this is reduced to 15 mm x 17.5 mm. The 35 mm negatives are scanned with a spot size of 100 J.Lm (0.1 mm} and hence the soil matrix area is covered by 150 x 175 pixels. Thus each pixel represents an area of 0.5 mm x 0.5 mm on the original negative. The negatives are scanned at 3-D optical density, levels of 0 -255 are assigned to the degree of brightness . The scanner sets white on the negative to zero and black to 255. Therefore the reference cell of 1 00% water content has higher digital count values than those of the 0% water content reference cell. The scanned data are displayed on an 12s Image Analyser. To account for any differences in exposure, printing and scanning, the frames are individually normalised using the 0% and 1 00% water reference cells. For any one frame, the digital counts less than those within the 0% water reference cell are set to zero (black} and those equal to or greater than those within the 100% water reference cell are set to 255 (white} to produce an enhanced black and white image. Any values within these extremes are then linearly stretched so that the new digital count becomes:- 116 Xn = ((X-Xo} I ( X 1 OO -Xo }} x 255 where: = New normalised digital count = Old digital count = Digital counts of 1 00% water content reference cell x0 = Digital count of 0% water content reference cell The soil matrix areas with water contents of 0% · 1 00% are thus enhanced to give a range of 256 grey levels. In order that the water content variations can be more easily seen the frames are density sliced in colour to give false colour images. Different colours are assigned to steps of 25 digital counts, ranging from red for levels of 0 -25 inclusive and blue for levels of 225-255 inclusive. RESULTS The preliminary results are in the form of three radiographs and their enhancements taken at various representative times throughout the freezing test. The corresponding data are given in the following table. TIME 10.24 Psychrometer Water Potential Position (Bar) 6.1 5.6 4.4 4.0 4.9 0.0 TIME 11.17 Psychrometer Water Potential Position (Bar) 12.7 5.6 3.0 4.8 5.1 0.0 TIME 14.02 Psychrometer Water Potential Position (Bar) 15.3 7.8 5.0 4.8 5.2 0.0 EJecRes Approx Water Content (Kll) (%) 28 35 26 23 22 25 24 25 35 26 100 Elec. Res Approx Water Content (Kll) (%) 89 31 26 27 25 20 26 35 26 28 100 Elec. Res. Approx Water Content (Kll) (%) 1100 480 25 29 25 95(1CE) 20 12 16 18 100 TIME 10.24 Radiocrraph taken at tirre 10.24 Black and Nhi te enhancerrent of radiooranh 117 False colour enhanc:EITEnt o-r: radiooranh FRAME 5 TIM~ 11.17 1 2 3 G Radiocra~h taken at time 11.17 Black and \•Jhi te enhancerrent of radionranh 118 False colour enhancer:Ent of radioorar:h VRAME 9 TIME 4 01 Fadioaranh taken at tirre 14.01 Black and Nhite enhancerrent o-F radiocrranh 119 False colour enhanceJTent o-F radiOC"Tanh From the three sets of preliminary results -taken at representive times throughout the test, there is evident correlation between the measured parameters. The complete set of results are at present being collated to give temperature, potential and water content profiles for the matrix throughout the test. The enhanced false colour pictures are in need of further corrections and normalisation before accurate water contents can be detemined. Once this has been achieved a false colour video loop of the experiment will be constructed to show the development of ice lenses in the matrix and will hopefully be shown at the conference. The experimental set up is in need of further development as the temperature gradient achieved in this experiment is unsatisfactory. In a future experiment the apparatus will be housed in an insulated wooden box fitted into the concrete shielding cavity, in order to further reduce the air temperature around the test rig and so facilitate a suitable temperature gradient in the soil matrix. The sides of the box are plywood sections containing paraffin wax for insullation. At the front of the box a PTFE window provides access for the neutron beam and at the side there is a curved opening just large enough to allow the LTC to pass into the box. Access to the box for positioning and removal of the LTC is made possible by a narrow slot cut in the shielding. As the LTC is moved through the shielding and into the box it disturbs rubber strips, suspended from the roof of the slot, which drop behind it to form an effective temperature insulated seal. The authors wish to thank the University of Aston for funding this research through a Faculty grant and Dr Bates and the staff of the Universities of Manchester and Liverpool Research Reactor for their continued support. 120 Williams, P.J., and Burt, T.P., 1974. Measurement of Hydraulic Conductivity of Frozen Soils -Canadian Geotechnical Journal, Volume II, pp 647-650. Yoneyama, K., lshizki, T., and Nishio, N., 1983 . Water Redistribution Measurements in Partially Frozen Soil by X Ray Technique. Proceedings of 4th International Conference on Permafrost, Fairbanks, Alaska, 1983, pp Fuhruda, M., Orhon, A., and Luthin, J.N., 1980. Experimental Studies of Coupled Heat and Moisture Transfer in Soils During Freezing. Cold Regions Science and Technology. 3 (1980) pp 223-232. Clark, M.A., Kettle, R.J., 1985a. A Thermal Neutron Radiography for Studying Mass Transfer in Partially Frozen Soil. Proceedings of 4th International Symposium on Ground Freezing, Sapporo, Japan, 1985. Volume II, pp 168 -173. Clark, M.A., and Kettle, R.J., 1985b. Modifications to Equipment and lmprovments in Facilities, used in the Study of Mass Transport in a Partially Frozen Soil by Thermal Neutron Raiography . Proceedings of 2nd National Symposium on Ground Freezing, Nottingham, England, 1985. Brown, R.W., and Bartos, D.L., 1982. A Calibration Model for Screen -Caged Peltier Thermocouple Psychcrometers. U.S Dept Agroculture, Forest Service, Research Paper INT -293 July 1982 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 THE DEVELOPMENT AND USE OF "HOT-WIRE" AND CONDUCTIVITY TYPE ICE MEASUREMENT GAUGES FOR DETERMINATION OF ICE THICKNESS IN ARCTIC RIVERS David A. Sherstone, Terry D. Prowse, Harry Gross1 ~TRACT: Two ice measurement gauges were developed to assist in the study of ice growth and decay on arctic rivers. the gauges permitted sucessive measurement of ice thickness without the need to drill through the ice cover after initial install- ation. The first type of gauge is a "hot-wire" design, modified from a gauge used in southern Canada. It was proven to provide low-cost, accurate readings in cold climates. Between 1982 and 1986 modifications to the initial design and reuse of components reduced per gauge cost to less than $20 per season. Measurement accuracy was approximately 0.5 em. The second gauge, a conductivity probe was designed to speed data collection and permit remote reading of ice thickness. Parallel electrodes are installed vertically through the ice and allowed to freeze into the ice cover. The magnitude of electrical current carried between the electrodes is proportional to the length of electrode in the water while current carried by the ice is negligible. Ice growth is therefore proportional to the reduction in current flow. Four conductivity gauges were test- ed in the 1984-85 winter. Accuracy was comparable to that of the hot-wire gauges except in the re- freezing period. While more expensive than the hot- wire type, the conductivity gauges is cost-effective in remote areas where transportation costs are high. Although approximately equivalent in accuracy both gauges were tested only ender arctic conditions in a calm flow regime, and performance problems related to frazil ice formation and mid-winter melt periods should be considered prior to use in more temperate regions. (KEY TERMS: ice measurement; freshwater ice; arctic rivers.) INTRODUCTION Within the nordic countries, the most common and standardized method of measur- ing the thickness and basic stratigraphy of freshwater ice involves drilling of the ice sheet (Adams and Prowse 1986). Drilling has a number of drawbacks; it is time consuming, unpleasant and risky in extreme cold, and may alter subsequent ice growth. The drilling procedure can be a problem in obtaining measurements, especially at very low temperatures (-25°) where it often involves a hazardous strug- gle with recalcitrant, gas-powered dril]R. Furthermore, drill holes often lead to surface slushing which can create a travel hazard in areas, such as the Mackenzie Delta, where the ice cover is extensively used for road transport. The effects of slushing produced by drilling can also significantly affect the temporal and spatial patterns of ice thickness. Because the underlying water is often under hydrostatic pressure, either from surface snow loads or increas- es in upstream flow, drilling can initiate a slushing event which would not natural- ly occur. Under cold temperatures, the slushed layer of snow refreezes into an upper stratum of white ice thereby augment- ing the total ice thickness. Although Adams and Prowse (1981) have found that artifical slushing of the snowcover does not affect maximum ice thickness on lakes because of a compensation process in the growth rates of black and white ice, it can produce significant aberrations in short-term growth rates and in spatial patterns of ice thickness. The effects of artifical slushing are probably much greater on rivers, because of the greater possibility of frequent and quite large overflows. Hence, measurement procedures which employ drilling preclude obtaining a record of "natural" ice growth on both 1Respectively, D. Sherstone, Inuvik Scientific Resource Centre, P.O. Box 1430 Inuvik, Northwest Territories, Canada, XOE OTO, T. Prowse and H. Gross, National Hydrology Research Institute, Environment Canada, Ottawa, Ontario, Canada, KlA OE7 121 lakes and rivers. To obtain an "undistrubed" sample, standard practice usually involves drill- ing a new hole at a slightly different location each time a measurement is made. However, the thickness of freshwater ice, especially river ice, is characterized by a considerable degree of local variability. In addition to variability introduced by white ice growth at the surface, the bottom topography of river ice can be quite vari- able due to, for example, frazil accumula- tions and melt-induced wave patterns. To minimize the effect of such spatial varia- bility in the record of ice growth, measure- ments should be made at the same point each sampling period. In view of the problems associated with drilling and the need to obtain repetitive samples from a single site, a number of measurement devices have been developed over the years. The approaches and instruments used by various countries are reviewed in Adams and Prowse (1986). This paper reviews the development and test- ing of two of these gauges under arctic conditions on channels of the Mackenzie Delta. One of the primary design criteria for the gauges was that they provide an ice measurement without inducing artifi- cal slushing. They were also to be inex- pensive enough to be expendable. The first is a "hot-wire" resistance gauge similar to that used by Ramseier and Weaver (1975) in southern Canada, but redesigned for use under arctic conditions. The second gauge or "conductivity probe" was designed to speed data collection and offer the possi- bility of remote reading of ice thickness, either through inter-connecting a number of gauges to a central recorder or through a satellite data collection platform. Hot-wire gauges have been used suc- cessfully in the Mackenzie Delta since 1982. The conductivity probe has been tested under temperate conditions on the Ottawa River in 1983-84, under arctic conditions in the Mackenzie Delta in 1984- 85, and is now being evaluted under sub- arctic conditions in the Schefferville region of Quebec. Data from the Mackenzie Delta program are presented in Sherstone and Prowse (1986). 122 HOT-WIRE GAUGES Design and Construction The earlier device by Ramseier and Weaver (1975) was modified to make it less expensive, easier to assemble in the field and more compact (Figure 1). Major components of the probe include: a stand and base plate, electrical and heating (resistance) wires, contacts, and a weight. The wooden base plates are fabricated from common 2.5 X 10 em lumber with each leg approximately 60 em long. A wooden 2 em thick plywood pad reinforces the junction of the support legs and is used to fix a threaded pipe flange to the assembly. The "L"-shaped pipe stand is made of two lengths of 2.5 em mild steel pipe. The vertical section, 90 em long, is joined to a 60cm horizontal pipe by a 90° elbow fitting. The wire is support- ed at the end of the horizontal bar through a slot in the pipe and retained by a threaded cap. /~ INSULATED RETURN WIRE ___,_...-(} 0 WEIGHT Figure 1. Schematic of Hot-Wire Ice Thickness Gauge (After Ramseier and Weaver, 1975) Details of Material and Construction are Provided in the Text. The resistance wire (Chromel A; 18 gauge) is cut to a length slightly great- er than the expected maximum ice thickness plus the height of the stand above the ice surface. A 2 7 5 -300 em wire is used in the Mackenzie Delta. The electrical cir- cuit is created by fixing one end of an insulated wire to the horizontal pipe and joining (crimp-solder) the other end to the resistance wire at the weight assembly. The weight is currently constructed of steel rod, but any heavy material could be used. Initial cost of the gauges is less than $20 CDN. , but many of the parts are retrievable which can further reduce costs. In the Mackenzie Delta, for example, parts recovered near the end of one winter season reduced the unit gauge price by 50% for the wbsequent season. fustallation and Operation Initial installation involves removing snow from an approximate 2 m2 area. A 15 ~hole is drilled through the ice and the resulting slush used to freeze the base plate around the hole, with the end of the stand directly over the hole. The wire and weight assembly are lowered through the hole and the insulated wire is secured to the pipe frame with electrical "tie-wraps". Ideally, snow should not be replaced over the hole ®til the ice surface has refrozen. Other- wise, the mixture of water and snow will refreeze into a hummock of white ice not representative of surrounding ice strat- igraphy and makes subsequent measurements difficult. At installation,notes are made of the ice thickness, both white and black ice, and the amount of wire which can be pulled up before the weight strikes the. base of the ice sheet. Subsequent measurements involve heating the resistance wire by rttaching a battery, lifting the weight to the bottom, recording the amount of extended wire, and then lowering the weight again. Under cold conditions, the battery can be temporarily disconnected between raising and lowering the weight, thereby freezing the wire in a "measurement" position. The amount of ice growth at the base of the ice sheet between sampling dates is equal to the difference in the lengths of wire which ~n be exposed at the surface (Figure 2). The amount of white ice growth at the 123 surface is obtained by measuring the upward growth on the vertical portion of the pipe frame. Gradations on the pipe assist in measurement. 1 ) j A' B' Figure 2. Measurement of Ice after Freeze-back of Installation Hole. Distance, A, and Length of Resistance Wire, B, are known at the time of installation. During use Measurements A', and B', are taken. Total Ice Thickness is thus B-B'; White Ice Thickness can be obtained from A-A'. In southern Canada sufficient power can be supplied by a 12 volt battery to free the wire in a short period of time. However, under arctic conditions with ice thickness often exceeding 1.5 m and tem- peratures below -25°C, a 12 volt system requires 5 to 10 minutes to free the wire. A series electrical connection of two batteries (24 volts) can reduce the waiting period to less than 5 seconds. Gauge Performance In 1982-83 sixteen gauges were install ed in the Mackenzie Delta (Sherstone 1984a). Four were located in a lake near Inuvik, ten in central Delta locations and two in a channel expected to experience thicker than normal ice. All gauges functioned flawlessly throughout the season. In the following three years, twenty gauges were installed on major Delta channels with one gauge failure in each of the first two years (Sherstone 1984b, 1985). In 1983-8 4 a gauge wire could not be melted free for approximately six weeks during the coldest part of the winter . The cause of this partial fail- ure is unknown, but a problem in one of the two crimp-solder junctions is sus- pected. In 1984-85, a gauge failed soon after installation probably due to a broken wire below the ice. To assess the accuracy of the hot- wire gauges, test holes were drilled in close proximity at various times through- out the growth season. Gauge readings were within 3 to 4 em of the values obtained by drilling. In view of the possible errors associated with the spatial variability of ice thickness, the gauge readings are considered satisfactory. CONDUCTIVITY GAUGES Design and Construction The conductivity gauge was develop- ed by the National Hydrology Research Institute, with final engineering and production undertaken by Richard Brancker Research Ltd. The theory of operation is that a pair of parallel electrodes immersed in water will pass an electrical current proportional to the length of the elect- rode in the water . As the water freezes downward, the ice forms an insulating layer thereby reducing the electrical current flow. Measurement of current flow yields the length of the electrodes still exposed to water and by simple subtraction from the known length of the electrodes, the ice thickness can be obtained . Such a simplified approach ignores possible changes in the conductivity of water with temperature and solute load. To compensate for such variations , a second, shorter pair of reference elect- rodes is positioned below the measurement 124 electrodes and below expected freezing levels. The ratio of current flow between the measurement and reference electrode pairs will be the same as the retio of their lengths in unfrozen water. The gauge consists of two basic parts, an ice probe and a portable reader unit (Figures 3a, 3b. and 4). The bottom of the probe is a hollow plastic pipe which supports two (0.05 em) diameter stainless steel tubes (AISI type 304) running down each side of the pipe. Plastic spacers separate the tubes from the pipe. All electrodes are wired to an electrical coupling head contained within a housing on the ice surface. The top of the housi is threaded and capped to protect the wiring from the elements . --.... -~ .. - . . .. ~--.. . Figure 3a. NHRI Conductivity Type .Ice Th i ckness Probe. Note Clear Plastic Collar Base Plate, required because of large diameter hole in Ice cut in the ice . Figure 3b. Electrical Coupling Head. When not in use a Protective Cap covers the Connectors. Polyvinylchloride (PVC) plastic was used for the probe body because it can withstand low temperatures without becom- ing brittle, is easy to machine, inex- ~nsive and readily available. Stainless steel tubing was used for the electrodes because it does not form an insulating oxide surface, as would aluminum, or cor- rode as copper or ferrous electrodes. As long as the water is not completely oxygen free, the stainless steel retains a stable passivated surface. It has suff·icient electrical conductivity for this appli- cation and low thermal conductivity to minimize vertical heat transfer. Hollow tubing, rather than solid rod, contributes to the low thermal conductivity. The reader unit contains electronic circuitry that automatically balances a Wheatstone bridge by means of precision resistors and reed relays. The results are displayed on LED's on the front panel. Tne use of resistors and a bridge reduces variations resulting from temperature 125 extremes. By making the bridge self- balancing, operator time is considerab- ly reduced. Current costs of the probes are approximately $200 CDN and $1,000 CDN for a reader unit. Design simplifica- tion and construction modifications are expected to reduce these costs, especial- ly in the case of the probes. ICE THICKNESS READER UNil EXTERNAL POWER RU-10 ICE THICKNESS em S IGNAL A POWER SIGNAL B ~ B CD OUTPUT <S) ."· ....... ~ .. ~a••••••' Manufactured by RIChard Brancker Research Ltd.Ot t awa .Canae!c. Figure 4. Reader Unit. Box measures 24.0 em X 12.5 em X 9.0 em and weighs approximately 1.7 kg with batteries. The Reader Un it is connected to the Electrical Coupling Head by two 1.4 m Co-axial Cables. Design Measurement Errors There are four design and construction factors which might affect measurement accuracy: a) "staircase" errors, b) probe geometry error, c) water temper- ature gradient error, and d) ionic con- centration gradient error. Space does not permit the derivation of each error but the results of some "worst case" analyses are provided below. a) Staircase error The measurement electrodes are held in place by a series of spacers. Assum- ing perfect contact beween the electrod- es and spacers, there will be no electri- cal flow from tile region of contact blJL the ice thickness will be underestimated by an amount equal to the total thickness of spacers in the unfrozen water. For example, in the case of a 150 em probe with spacer thickness 0.5 em and separa- tion 10 em, an inital measured ice thick- ness of 90 em would have to be reduced by 3 em (six 0.5 em spacers lie within the unfrozen water). In practice, the magnitude of this error will be less because there is unlikely to be perfect contact between the spacers and the elect- rodes, thereby allowing some electrical conduction from this region. b) Probe geometry errors Firstly, it can be assumed that any bending or distortion in the reference electrodes will be small compared to that in the longer measurement electrodes. Noting this assumption, two other geometry related errors are possible: a constant offset error, where the spacing of the measurement electrodes differs from the reference electrodes and, b) a bowing error, where the spacing between the two measurement electrodes varies along the length. For the small displacements expected, any change in the electrical current flow can be taken as inversely proportional to the change in electrode spacing. Thus, any change in electrical current flow from irregular spacing will produce a corresponding change in the measured length of the water column. With the 0.5 em tubing, manufact- uring tolerances for an electrode spacing of 3.8 em is approximately 0.08 em. The expected maximum offset and bowing errors in this case are 2.1% and 1.3% respective- ly. c) Water temperature gradient error The electrical conductivity of water increases with temperature. In a worst case scenario, the probe would encounter 126 a gradient of 4°C (freezing point to the temperature of maximum density for water) over the length of the electrodes. This results in a maximum error of approx- imately 4%. Much smaller temperature gradients and errors are normally to be expected. d) Ionic concentration gradient Changes in ionic concentration along the electrodes will affect measure- ment readings in the same way as a temp- erature gradient. Changes in conductiv- ity resulting from the process of freeze- out are currently being examined at the Schefferville test site. For freshwater, it is expected that the effects of vari- ations in ionic concentrations will not exceed that due to temperature variations (less than 4%). Under most conditions, the combined error from the above four sources is expected to be less than 5%. Installation and Operation Installation of the conductivity probe is decidedly simpler than that for most other types of ice measurement devices. Site preparation involves clearing a small patch of snow and drill- ing a single hole through the ice which is sufficiently large (approximately 5 em) to permit lowering of the probe and electrodes. In practice, a larger diameter hole will probably be drilled because of equipment availability and the need for a larger hole to permit inspec- tion of the ice stratigraphy. Where large diameter holes are drilled, a collar can be placed between the probe and upper housing to support the device while refreezing occurs (see Figure 3a). Notes are made of the initial thick- ness of white and black ice, and of the collar if one is used. Subsequent growth of white ice is noted on the upper housing To obtain a measurement of ice thickness beneath the surface, the housing cap is removed and the reader unit attached to the electrode terminals. A complete reading requires less than one minute. Gauge performance In January 1985, four conductivity gauges with 1.5 m measurement electrodes were installed in the Mackenzie Delta. The gauges were located within 0.3 - 1.0 m of some previously installed hot- wire gauges. Initial reading of both gauge types was made 14 days after install- ation and then at 10 day intervals when weather permitted. Removal of the gauges proved diffi- cult. Even close to break-up, solar heat- ing had not melted the gauges free. One complete gauge was recovered but this required extracting an approximate 90 kg ice block. Unfortunately, the other three gauges had to be severed at the housing-probe connection. Records obtained from adjacent hot-wire and conductivity gauges are shown in Figure 5. In general, the data from the two types of gauges show sim- ilar trends in ice growth and ice thick- ness. The differences between readings can be ascribed to a number of factors related to gauge installation and spatial variations in sub-surface ice topography. However, because there was no way to visually inspect the ice sub- surface, it is only possible to speculate on the relative importance of these factors. Firstly, differences in gauge readings at the start of the study per- iod can be explained by the lengthy time required for refreezing in the holes used for installation of the conductivity probes. The conductivity probe will not provide an accurate reading of ice thick- ness until the entire ice stratum has refrozen. The presence of water within the hole will result in an under-pre- diction of ice thickness, such as that observed for gauge 84-02 during·the re- freezing period. By the time of the second reading at this site, both gauges were producing similar results. If the two types of gauges are installed at the start of the winter season, when the ice sheet is much thinner, the time required for a stabilization of readings will be significantly shortened. Following the initial refreezing period, site 84-11 showed a consistent Mfference between the two gauge read- ings. Although this may have been rel- ated to the electrode circuitry in the 127 conductivity probe, it could also have been due to large local variations in ice thickness. The difference in thickness may be due to natural spatial variations in the sub-surface ice topography or from effects created by ice drilling during gauge installation. 160 150 140 130 i ~ 120 "' "' w 110 z ~ <.> X ... 100 w <.> 90 BO 70 60 -0 z < ~ ' ' ' ' ' ' HOT·WIRE GAUGES CONDUCTIVITY GAUGES --0 0 0 AT~ Figure 5. Graph of Ice Thickness Values obtained from Hot-Wire and Conductivity Gauges located at Gauge Sites 84-02 (East Channel), 84-07 and 84-08 (Oniak Channell and 84-11 (Middle Channel). Winter 1984-85 > < ~ During the spring period, hydrothermal melt at the base of the ice sheet can amplify local variations in ice thickness. Although water temperatures may only be a fraction of a degree above freezing, the river heat flux is often sufficient to produce large ripples on the base of the ice sheet. For example, Marsh and Prowse (1986) observed, on another part of the Mackenzie River system, ice ripples at the base of the ice sheet with amplitudes of up to 6 em and wavelengths of approximately 20 em. Under such conditions, differences of 12 em could be expected between gauges, despite their close proximity. Further- more, these ripples may be a transient feature which could further complicate the gauge comparison. These type of effects may be responsible for the late season fluctuations in the data from site 84-02. One final gauge problem is ap- parent for site 84-07 in Figure 5. By April 07 the readings from the con- ductivity probe stabilized near 155 em. This was simply due to the ice thickness exceeding the length of the measurement electrodes. Although the original objective of this project was to test the per- formance of the gauge under field conditions, the complications created by various hydroclimatic factors indicates that future calibration of the devices should be undertaken in a controlled laboratory environment. The field results, however, suggest that reliable readings can be obtained. Periodic confirmation of the readings by drilling is recommended, at least until variations in the record can be adequately explained. DISCUSSION The hot-wire gauges have been used in the Mackenzie Delta for four winters and can now be considered operational devices. The gauges are inexpensive, easily manufactured, large- ly recoverable and can be quickly read. They have resulted in increased product- ivity and reduced the danger to field personnel from exposure. The present network of 20 gauges, spread along 115 km of ice roads, can be monitored in one day, of which about two hours is spent in data acquistion. Previously, only five to eight holes could be completed by drill- ing in one day over the same route. Further- more, drilling required almost twice as much time spent by field personnel exposed to the elements. The conductivity gauge is still in a design stage but appears to hold consid- erable promise for reducing field time and speeding data collection, especially when integrated with data-loggers or satellite DCP's. It is hoped that design simpli- fication will reduce the cost of the probes. However, even at the current price, the conductivity gauge can be cost-effective for application in remote areas where transportation costs are high. 128 Both gauges appeared to provide reli- able ice thickness measurements. However, these tests were primarily restricted to arctic conditions within the calm flow regime of the Mackenzie Delta. A number of other performance problems may arise under different climatic and flow condi- tions. For example, more turbulent flow on other rivers may lead to the production of active frazil which will adhere to portions of the gauges exposed beneath the main ice sheet. Frazil accumulations could quickly encase the gauge and render it useless. This problem does not develop on lakes or on rivers without open water zones necessary for frazil production. In case of the conductivity probe, turbulent water does, however, eliminate the errors produced by temperature and ionic grad- ients. These are most likely to be a problem in the case of lakes where there is little mixing beneath the ice sheet. Mid-winter melt periods, which are common in more temperate climates, may also limit the usefulness of the conduct- ivity probe. Absorption of solar radia- tion by the exposed gauge housing and heat conduction along the probe may result in the accumulation of water beside the elec- trodes. This water will increase current flow and result in an underprediction of the thickness of the surrounding ice sheet. Subsequent refreezing, however, should make this only a temporary problem. REFERENCES Adams, W.P. and T.D. Prowse, 1981. Evolution and Magnitude of Spatial Patterns in the Winter Cover of a Temperate Lake. Fennia, 33, 117- 128. Adams, W.P. and T.D. Prowse, with M.A. Bilello, E. Eliassen, S. Freysteinsson, 0. Laasanen, T. Pangburn, B. Raab, E. Tesaker and A. Tvede, 1986. Techniques for Measurement of Snow and Ice on Fresh water in Nordic Countries. Proceed- ings of the Sixth Northern Research Basins Symposium, Houghton, Michigan (in press). Marsh, P. and Prowse, T.D., 1986. Water temperature and heat flux to the base of river ice covers. National Hydrology Research Institute report, Environment Canada, Ottawa, 42 p. ~mseier, R.O. and Weaver, R.J., 1975. "Floating Ice Thickness ans Struct- ure Determination -Heated Wire Technique" Technical Bulletin No. 88, Inland Waters Directorate, Environ- ment Canada, Ottawa, 16 p. Sherstone, D.A., 1984 a. "Ice Thickness in the Mackenzie Delta, Winter 1982- 83: Data Obtained From Hot-Wire Gauge Systems", Report 84-1, Inuvik Scientific Resource Centre, Dept. of Indian Affairs and Northern Development, Inuvik, Feb. 1984, 16 p. Sherstone, D.A., 1984 b. "Ice Thickness in the Mackenzie Delta, Winter 1983- 84: Data Obtained From Hot-Wire Gauges", Report 84-4, Inuvik Scientific Resource Centre, Dept. of Indian Affairs and Northern Development, Inuvik, October 1984, 16 p. Sherstone, D. A., 1985. "Ice Thickness in the Mackenzie Delta: Winter 1984-85". Report 85-2, Inuvik Scientific Resource Centre, Indian and Northern Affairs Canada, Inuvik, Oct. 1985, 13 p. 129 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 RECENT DEVELOPMENTS IN HYDROLOGIC INSTRUMENTATION Vito J. Latkovich and James C. Futrell Il ABSTRACT: The programs of the U.S. Geological Survey require instrumentation for collecting and monitoring hydrologic data in cold regions. The availability of space- age materials and implementation of modern electronics and mechanics is making possible the recent developments of hydrologic instrumentation, especially in the area of measur- ing streamflow under ice cover. Material developments include: synthetic-fiber sounding and tag lines; polymer (plastic) sheaves, pulleys, and sampler components; and polymer (plastic) current-meter bucket wheels. Electronic and mechanical developments include: a current-meter digitizer; a fiber-optic closure system for current meters; noncontact water-level sensors; an adaptable hydrologic data acquisition system; a minimum data recorder; an ice rod; an ice foot; a handheld sediment sampler; a lightweight ice auger with improved cutter head and blades; and an ice chisel. (KEY TERMS: hydrologic instrumentation; electronics; water-level sensors; data acquisition system; microprocessor- based instruments.) INTRODUCTION The U.S. Geological Survey is the principal Federal organization responsible for providing water resources information. Through a system of data-collection programs, measurements of surface water, ground water, and water quality are made at more than 15,000 individual remote field locations throughout the nation. A large number of these locations are located in cold regions; necessitating measurements under extremely cold temperatures and even under ice cover. Continual modernization and automation of instrumentation is highly desirable and essential as a means of increasing overall organization effectiveness and efficiency. The purpose of this report is to briefly describe a few developments that are helping to keep the Survey up to date in hydrologic instrumentation. DEVELOPMENTS BASED ON NEW MATERIALS Synthetic-fiber sounding and tag lines are presently being field tested and evaluated. The material used is Kevlar (the use of brand names in this report is for identification purposes only and does not constitute endorsement by the U.S. Geological Survey), an aramid fiber that is extremely strong and lightweight. The sounding line will incorporate a 4-conductor wire internally for electrical signal trans- mission between the sounding sensor and recording/ listening device. The tag lines vary in size and use from a 1116-in.-diameter wading measurement line to 3/16-in.- diarneter boat measurement lines that either float or are lead- weighted depending on the depth of submergence required. The wading line has a 300-Ib breaking strength and a 600-ft length of line can be pulled taut with a 25-lb force. Lennite, an ultra-high molecular weight polymer (plastic), has been tested and evaluated as a replacement for metal pulleys, sheaves, handles, and water-quality sampler com- ponents. This polymer can be machined in the same manner as metals. It is as durable and much less expensive than the metals. Price AA and Pygmy current-meter bucket wheels have been developed using Lexan, another polymer that is very durable and easily molded. It has been observed during cold weather field tests that ice-covered polymer wheels warm up much faster in flowing water than the existing metal ones. The polymer wheels have solid cups (buckets) that have been laboratory-tested and shown to be almost nonresponsive to the vertical component of velocity. Also, ice buildup does not occur on the solid cups as it always does in the metal open-cups. Polymer wheels are extremely cost effective; an AA polymer wheel may cost about $5 per copy versus $125 for the metal version and a Pygmy polymer wheel may cost about $3 per copy versus $84 for the metal version. 1Supervisory Hydrologist and Engineering Technician, Hydrologic Instrumentation Facility, U.S. Geological Survey. WRD, NSTL, MS 39529. 131 ELECTRONIC AND MECHANICAL DEVELOPMENT A current-meter digitizer (CMD) was developed to automate streamflow velocity measurements . .The CMD is a microprocessor-controlled unit that automatically records the current-meter revolutions, keeps track of time, and con- verts the results into real-time velocities. Meter-rating equa- tions are preprogrammed into the CMD's memory for instant recall, dependent on the mode of recording (automatic or manual) and the type of meter being used. In the read-velocity mode, the hydrographer can rapidly assess variations in velocity by observing if the computed velocity numbers are changing or steady. The CMD straps to the hydrographer's waist or arm and may be operated while wearing mittens in temperatures down to -40° C. The fiber-optic closure system, a near frictionless tech- nique for monitoring current-meter bucket wheel revolu- tions, functions by rotating two U-shaped, fiber-optic bundles in conjunction with the movement of the bucket wheel. An infrared light-emitting diode and a photo tran- sistor are used to sense light changes as the rotating fiber- optic bundles pass and block light transmission. Four pulses are generated for each bucket wheel rotation. An elec- tronically clean, square wave is generated. The spacing between the rotating bundles and infrared sender/receiver is not critical. The upper housing of the system resembles an inverted cup that is sealed with 0-rings to retard the entry of water and silt. Traditionally, the U.S. Geological Survey has collected water-level (stage) data using floats, pressure devices, surface followers, and other devices. All of these sensors/devices require a physical connection from the stream to the data- recording system/shelter, which is not always practical due to restrictions on locating gaging stations and sensors. The 1980 eruption of Mt. St. Helens and the ensuing prob- lems of obtaining water levels in the surrounding streams due to the very unstable channels, prompted the Survey to explore the feasibility of using noncontact water-level sensors; i.e., no physical "connection" between the sensor/ recorder and the stream. To date three different noncontact technologies have been researched and developed: ultrasonic, laser, and radio fre- quency. Ultrasonic sensors are presently employed as early- warning flood and mudslide detectors as well as backup units for primary sensors. They do not operate within the Survey's required accuracy limits (±0.01 ft) and cannot be used as primary water-level sensors. The units are interfaced to digital water-level recorders and/or data-collection plat- forms for near real-time data transmission. A laser sensor prototype was developed, but rejected as a primary sensor because it also could not meet the Survey's required accuracy limits. A radio-frequency sensor prototype has been developed and is awaiting field test and evaluation. This unit 132 holds the most promise of sensing water levels to an accuracy of ±0.01 to ±0.02 ft through a range of 5 to 50ft, and it does not require compensation for temperature changes. Although the cost per unit may be high (up to $9000 for the radio-frequency version), the long-term benefits should be worth the development effort and investment. The Survey's current hydrologic field data-acquisition instrumentation lacks the capability, flexibility, and ability to utilize, and thereby interface with, state-of-the-art elec- tronic sensors, and be software controlled. Present instru- ments are not adaptable to new operations and updated components. They are difficult to integrate into new system configurations and will not operate throughout the required ranges of environmental conditions. Reliability varies from instrument to instrument and site to site. The Adaptable Hydrologic ·Data Acquisition System (AHDAS) will be the Survey's solution to replacing the existing and obsolete punched-paper tape analog to digital recorder (ADR). AHDAS will be a microprocessor-controlled recorder/ retriever system utilizing nonvolatile, solid-state data memory and intelligent control features. Memory capacity can be adjusted to fit specific measurement and parameter requirements. A Field Component (recorder) and Portable Field Interrogator (retriever) are the system's basic com- ponents. The system will interface to numerous sensors and be capable of multiple transmission modes. It will have a satellite communication interface capability, thereby requiring only a transmitter for operation. This development is a major effort by the Survey to take advantage of state- of-the-art electronics to modernize their data-collection instrumentation. The Survey expressed a need for an inexpensive ($400 per unit) ground-water-level recording system to monitor levels in small-diameter (2-in.) observation wells. The system should be easily installed, operated, and maintained for unattended periods up to 1 year. A minimum data recorder (MDR) is being developed that must meet the following per- formance specifications: (1) fit inside a 2-in. borehole or casing; (2) battery powered; (3) record the daily average water level and maximum and minimum (time-tagged) levels for the period of record; (4) operate unattended for at least a 12-month period; (5) employ one of a family of pressure transducers to cover a total range of water levels from 0 to 120 ft; (6) retrieve recorded data using an off-the-shelf portable computer; and (7) monitor water levels to an accuracy of ±0.1 ft. The MDR will be a microprocessor- controlled recorder utilizing solid-state data memory. Future ground-water software development plans include an aquifer test program. Other potential applications for the MDR are recording flood hydrographs for small drainage-area discharge determination and short-term, reconnaissance-type streamflow appraisals. The ice rod under development has several unique features. It is made from 7075-T6 aluminum, has a 7/8-in. outside diameter, is hollow, and each 4-foot section locks together with two screws that back out to fasten into mating holes in the second rod. This fastening method yields a smooth rod with no projections. The mating holes are drilled through the rod to allow ice that has formed during a velocity measurement to be pushed out of the opposing holes when the rod is disassembled. The total assembled length of the rod is 18.25 ft. Internal to the rod are mating, miniature, underwater connectors that carry the electrical signal from the meter to the CMD. The anodized rod is black with white markings and numbers, making visual observations very effective and efficient. An aluminum tail fin attaches to the rod immediately above the ice foot. This fin fits through a 6-in. ice hole when attached to the rod and can be used to check misalignment of the ice meter in flowing water when flow velocities beneath ice cover exceed approximately 1.0 ft 3/sec. A stainless steel pointer-protector attaches to the top of the rod to assist in alignment of the meter into the flow streamline and prevents the entire assembly from falling through the ice hole due to accidental slippage. This device is keyed and may only be attached to the rod in one position that lines up with the buckets of the current meter as attached to the ice foot. A 20-ft cord attaches to the top of the rod with an underwater connector and is held by a restraining clamp on the pointer- protector. This cord resembles a large telephone handset cord and is rated to -40° C. The other end of the cord attaches to the CMD with a military-type threaded connector. The ice foot was developed that attaches to the lower part of the ice rod to accept a current meter and to measure stream depth and ice thickness. An ice thickness gage is part of the ice foot and a circular plate mounted on this gage protects the cups of the ice meter when raised or lowered through slush ice. The zero of the rod is located at the hole used to mount the ice meter in the ice foot and aligns with the centerline of the ice-meter cups. The base of the ice foot is 0.25 ft below rod zero and the top of the ice thickness gage is 0.25 ft above rod zero. Spacing of the ice thickness gage and the base of the ice foot from the buckets of the ice meter were determined by laboratory experiments. Movable plates were mounted above and below the meter while rating in a towing tank to determine the distance of noninterference of the plates with the performance of the meter. Six plate distances were tested from 0.100 to 0.183 ft. When the plates were spaced more than 0.150 ft from the centerline of the ice meter cups, no significant change in meter efficiency (meter revolutions divided by distance traveled) was measured. A spacing of 0.250 ft for the base and the ice thickness gage was selected for the convenience of the field person and is greater than the plate simulation testing for these items that might cause interference with performance of the ice meter. A lightweight, handheld sediment sampler has been developed to operate through a 6-in.-diameter hole in the 133 ice. The sampler, a modification of the US DH-75 system, incorporates a tilting mechanism to allow insertion through the hole. The main section of the sampler is a modified ice rod; therefore, extensions can readily be added to accommodate streams of different depths. Once the desired length of rod is in place, the handle to the pivot strap is pushed down and the sampler tilts to a vertical position. While in this position, the sampler will pass through a 5-in. augered hole. However, a 6-in. hole is more desirable for sampling ease. Once the sampler body is through the ice, the handle is released and the spring mechanism pulls the sampler to a horizontal position for sampling. The sampler is removed after tilting it back to the vertical position. Field tests, including one on the Peace River in Alberta, Canada, have demonstrated that the tilt sampler functions at low temperatures and is a timesaving device. A commercially available, lightweight, easily trans- portable Aqua-Bug ice auger has been tested and evaluated. The auger can be carried to remote sites where access is limited to backpacking, snowmobile, or aircraft and is used where ice thickness does not exceed 2.5 to 3.0 ft. The weight of the auger is approximately 30 lbs. It starts easily and runs smoothly at temperatures down to -40° C. Holes of 6 and 8 in. in diameter, 3.0-ft deep, have been augered with the unit, and it did not show any sign of "killing out" or work- ing beyond its capacity/capability. Cutter heads and teeth were modified for field servicing and portability, as well as increasing cutting efficiency and effectiveness. Cutting rates varied from 0.3 to 0.9 in./ sec depending on which prototype cutting head was used. The original auger had a gear reduc- tion ratio of 35:1; since then the manufacturer has made several design changes including a new gear reduction ratio of 41:1. Also, the name has changed from Aqua-Bug to Tanaka. As a result of the redesign, four new Tanaka augers were purchased and are undergoing field test and evalua- tion. Results of the tests will be available in July 1986. The extension flights, cutter heads, and blades used for the Tanaka can also be used on the heavier Stihl Model 4309 auger which is the mainstay system in the United States and Canada. The Stihl auger is for heavy duty ice work (thicknesses greater than 3 ft). A prototype ice chisel has been developed that has a hickory handle and is lightweight in comparison to existing steel-handle chisels. This chisel has a hardened steel blade that does not dull rapidly in silt-laden ice. The blade is made from type A-2 tool steel. The cutting edge is hardened to Rockwell C54 to C58, the approximate hardness of com- mercial cold chisel. After hardening, the cutting edge is surface-ground with water cooling to avoid altering the temper. This edge is ground at a 17 .5-degree angle and honed to razor sharpness. The blade may be dressed with a file if severe nicks are not present. The weight of this chisel is concentrated at the blade for easier cutting. The hickory shaft is a replacement shovel handle that is commercially available with the D-handle installed. It is 1.5 in. in diameter and is 6 ft long, not including the blade. The fasteners used to hold the shaft to the blade and the D-handle do not protrude and thus cannot snag the operator's mittens. The hickory shaft makes the chisel lighter and does not transmit severe cold to the hands like the steel handle of the existing chisel. For this reason the prototype chisel is easier to operate for extended periods in severely cold temperatures. The D-handle is useful in thick ice and helps to prevent the chisel from slipping through mitten- covered hands through the ice hole. The basic design of this chisel is approximately 20 years old. It was originally conceived by Monty Alford. Water Survey of Canada, for use in the Yukon Territories and has 134 been perfected over many years of actual field use. The chisel was tested at Fairbanks, Alaska, by a Survey hydrographer with 25 years of ice-chiseling experience. He preferred the prototype chisel to the one he designed and has been using for many years. SUMMARY Space-age materials and proven state-of-the-art electronics (especially microprocessor technology) are making possible the many recent developments of hydrologic instrumenta· tion, especially in the area of measuring stream-flow under ice cover. The Survey is taking advantage of those develop· ments, while providing field personnel the proper methods to operate and maintain the new instrumentation. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 PROBLEMS ENCOUNTERED AND METHODS USED IN THE U.S. GEOLOGICAL SURVEY FOR THE COLLECTION OF STREAMFLOW DATA UNDER ICE COVER Ernest D. Cobb and Bruce Parks ABSTRACT: The U.S. Geological Survey is the principal Federal agency responsible for the collection and dissemination of streamflow data in the United States. An important part of this effort in cold climates is the collection of data under winter conditions when ice covers streams and rivers. Standard cup-type as well as vane-type current meters are used with various types of suspension equipment to make discharge measurements under ice. Safety hazards and a variety of ice con- ditions add to the difficulty in making winter streamflow measurements. Stilling wells and intake systems or manometers connected to gas-purge systems are com- monly used to obtain winter stage records. Water in stilling wells tends to freeze and various methods are used to prevent this from happening. Current activities related to providing better tools and in documenting acceptable methods are pre- sented. (KEY TERMS: discharge measurement; ice cover; streamflow; anchor ice; ice break- up; frazil ice.) INTRODUCTION The Water Resources Division of the u.s. Geological Survey is the principal Federal agency responsible for the col- lection and dissemination of streamflow data. In cold climates, an important part of this effort is the collection of streamflow data under winter conditions when ice covers streams and rivers. These data are used for water-supply studies, for waste-allocation studies, and for a number of other purposes. It is, there- fore, important that the best efforts pos- sible, within funding and manpower con- straints, be used in the collection and analysis of winter streamflow data. This paper discusses some of the prob- lems in obtaining winter streamflow data and presents methods used in the collec- tion and analysis of these data. Equip- ment developments and acceptable methods of documentation also are discussed. The wide range of elevations and lati- tudes within the United States results in a large variety of winter conditions. Ice thicknesses of 2 to 3 meters are common in northern Alaska, while thicknesses of only a few centimeters may be found in streams in the more southern latitudes. Some streams are affected by ice for only a few days while those on the Arctic coast of Alaska are frozen from mid- September to early June. Because of this diversity in winter ice conditions, there are corresponding variations in problems in obtaining winter streamflow data in the United States. Many of these prob- lems are discussed in the following sec- tions. DISCHARGE MEASUREMENTS Sites for the measurement of discharge under ice cover usually are selected Respectively, Office of Surface Water, 415 National Center, u.s. Geological Survey, Water Resources Division, Reston, Virginia 22092; and Office of Water Data Coordination, 417 National Center, U.s. Geological Survey, Water Resources Division, Reston, Virginia 22092. 135 during periods of open water. These sites are selected on the basis of channel and flow conditions; if the sites have been used in previous years, experience gained from ice conditions would also assist in site selection. It is usually difficult to determine horizontal angle correc- tions, so a site is selected where hori- zontal angles are minimized. Prior to the discharge the exact location of the channel must be determined. measurement, flow in the In small streams, an entire cross section may be cleared of ice and open-channel tech- niques used to make the discharge meas- urement. In larger channels, or if the ice is quite thick, holes are drilled or chiseled across the section to determine the location and characteristics of the flow. If the flow characteristics are not suitable for a measurement at the site, a different site is chosen and the process is repeated. After the site has been selected, 20 or more holes are drilled or chiseled across the section, prior to the start of the discharge measurement. Additional holes may be made during the measurement to better define the channel or flow characteristics. Most discharge measure- ments are made using a wading rod or an ice rod to suspend the current meter and to obtain depths in the section. The wading rod is used in shallow streams and rests on the bottom of the channel. The meter is raised and lowered on the rod. If an ice rod is used, the current meter is fixed at the bottom of the rod and the rod is raised and lowered from the ice surface and held at the proper location in the vertical. Cable-suspended meas- uring assemblies are used for larger streams. Some use is made of handlines for the suspension of the measuring as- sembly. Data are obtained for the vertical distance from the top of the water in the hole to the bed of the channel, and for the distance from the water surface to the bottom of the ice or slush. The effective depth of flow is the difference in the two readings. The bottom of the ice or slush is determined by raising the current meter in the flow until the meter quits turning or by the use of an "L" shaped stick or other object that can be lowered through the hole in the 136 ice and raised until resistance is felt or sensed from the object coming into contact with the ice or slush. Velocity measurements are made using a variety of current meters. Most current meters used by the Geological Survey are of the vertical-axis type. Electromag- netic velocity meters have been used a few times but have not proven to be reli- able under some conditions (Wellen and Kane, 1983). Most measurements made un- der ice use a standard AA current meter, a pygmy current meter, or an ice meter. The AA and the pygmy meters are both cup-type meters while the ice meter is a vane-type ~eter. The ice meter is used primarily with rod suspension. With rod suspension, velocities are obtained at 0.2 and 0.8 of the effective depths when depths are equal to or great- er than 0. 76 meters and at the 0.6 depth when depths are less than 0.76 meters. No coefficients are applied to the mean of the observed 0.2 and 0.8 depth velocities to obtain a mean velocity in the verti- cal. When the 0.6 depth method is used under ice cover, a coefficient of 0.92 is applied to the observed velocity to obtain the mean velocity in the vertical. The depth at which the 0.2 and 0.8 depth method is used with cable suspension is dependent on the distance above the bot- tom of the weight at which the current meter is set. Figure 1 shows a typical vertical-velocity curve under ice cover. There are many problems associated with discharge measurements under ice cover that are not encountered at other times of the year. Difficulties with access to the station site may be one of the first problems encountered. Win- ter road conditions can be quite hazar~ ous: roads can be icy or deep snow or drifts can cover the road. Helicopters or airplanes may be needed to get to so~ stations. Air travel has added hazards in the winter with high winds or poor visibility. When temperatures are below -40 °C, equipment problems increase and efficiency decreases. Under these con- ditions, waiting for another day when weather conditions are better for working may be the best alternative. The field person is given the discretion to make decisions on what is safe or not safe to do relative to winter field work. "")( '\ ror-----r-----,_-----+------r-----x ~~ \ ~t ~ ... ~0 \ .,_, ffig~r-----r-----,_-----+------r-----,_x ~~ I ~~ )( ~... I ... ~ ~~~r-----r-----,_-----+------r---------~ .... l ~~ /1 ~~ /)( 5! ~r-----r-----,_-----+----x~--~.x---+----~ 100 ~o----no~.2----~o.74~~~~ot.6~--~o~.a----~ •. o~--~ •. 2 VELOCITY, IN FEET PER SECOND Figure 1. Typical Vertical-Velocity Curve Under Ice Cover (Rantz and others, vol. 1, 1982). Although discharge measurements may be desirable during ice formation or break- up, the safety of the field personnel is the overriding concern during these po- tentially dangerous periods. During for- mation, the ice cover may not be complete or it may be too thin or weak to support field personnel and their equipment. If the stream is too deep or swift to wade, it is often better not to attempt to make a measurement. Discharge measure- ments can be difficult or impossible to make at times during ice breakup. Often, large pieces of ice will be carried by the current and can make measurement con- ditions dangerous to the field person or hazardous to equipment. As a result, it is sometimes impractical to make a dis- charge measurement during parts of the ice breakup. Time-saving measures, such as half counts and reducing the number of verticals, help reduce the time the current meter is in the water and exposed to potential damage from floating ice. Alternative methods of flow measurement, such as the timing of floating ice, and returning at a later time to measure the cross-sectional area and to determine the relation of surface velocities to the mean velocity, are considered. Floating slush or frazil ice will col- lect in the cups of the AA and pygmy 137 current meters. This is usually not a problem with the vane-type ice meter. Floating slush ice is more of a nuisance than it is a hazard but is important be- cause when it collects in cups of current meters, it causes the meter to indicate an erroneous velocity. Slush ice can accumulate, sometimes to great depths, under a solid ice cover. When this occurs, a pole may be needed to move the slush aside so that the current-meter assembly can be lowered into the flow. Slush ice can enter and be trapped in current-meter cups as the meter is lowered through the slush. If it is realized that the slush is in the cups, it can sometimes be easily dis- lodged. If it is not detected, the meas- urement of velocity in that section prob- ably will be in error. Aufeis occurs when the stream freezes down to the streambed at a downstream point and water in the stream is backed up and flows over the top of the ice and freezes adding a new layer of ice on top. The process can be repeated many times until the ice can become several meters thick. When this condition exists, it may be very difficult to find the flowing water, especially if the flow is small. A great deal of chiseling or drilling in the ice may be required to find the flow. Flow can sometimes occur between la- yers of ice. This creates a very diffi- cult situation in which to make a dis- charge measurement, and the field person usually tries to find another location for the measurement. If a suitable site cannot be found, the field person may not be able to make a measurement. Sometimes, the flow between each layer can be meas- ured and added to the flow measured in the other layers to obtain the total dis- charge. The accuracy of such measure- ments is usually poor. Care must be used in finding the edges of flow under ice. If augers are used, the cutting blade can be easily damaged if it cuts into bed material near the edges of flow. Therefore, chisels often are used to cut through the ice near the banks. Sand often is imbedded in river ice. This sand will dull the cutting teeth on an auger. Therefore, the teeth must be readily sharpened by filing or be readily replaced with a new set of sharp-cutting teeth. When holes are drilled or chiseled in the ice, the water under the ice may flow up through the hole onto the surface of the ice. When this occurs, a variety of alternatives are considered. Usually, the flow will quit after a few minutes and the measurement can be made. Under some conditions, a snow dam can be formed around the hole keeping the water from flowing out of the hole. Sometimes, holes can be drilled about a stream width downstream that allow the pressure under the ice to be relieved. The system is allowed to stabilize before the dis- charge measurement is made. When pressurized flow is found, there may be a modification in the shape of the vertical-velocity curve until the pres- sure is relieved. Making measurements at places where the water flows upward out of holes in the ice, during extreme cold, is very difficult because the water quickly turns to slush and freezes on the equipment and the field person. When the current meter is moved from one hole to the next, there is a tendency for ice clinging to the meter to freeze, thus immobilizing the meter. The tend- ency for this to happen is reduced if the meter is quickly moved from one hole to the other. If ice has formed on the meter, the meter can be lowered into the water in the next hole and the ice will usually thaw after a while. Sleds on which a compartment is mounted in which the meter may be placed and that is heated using a propane heater may be used to keep the water on the meter from freezing while the meter is being moved between sections. On large rivers when the ice is thick and the flow is swift and deep there are, at present, no practical ways of obtaining good discharge measurements. Some of these streams are 18 meters deep and have velocities of 3 meters per second. Wet-line corrections may be needed but cannot be readily determined, making depths difficult to accurately measure. The meter assembly may also move about excessively in the flow. In high velocities and extreme depths, the current-meter assembly tends to be forced downstream by the current, away from the vertical position. The observed 138 depth is then incorrect and a wetline correction based on the observed angle of the suspension line relative to the vertical is used to adjust the depth readings and to adjust the meter position in the section. In a situation where thick ice exists, the angle of the sus- pension line from the vertical usually cannot be determined. The observable part of the line is stopped by the ice walls of the hole from taking the position it would assume if the meter assembly were suspended in open water. The cost of obtaining discharge meas- urements in the winter can be consider- ably more than at other times. If haz- ardous conditions are anticipated or if large streams are to be measured, two or more field people are required to make a measurement that may normally be made by one person. It often takes more time to get to a gaging site in the winter than at other times and special transporta- tion, such as aircraft or snowmobiles, may be needed. When extremely low tem- peratures are experienced, the field persons are less efficient than in milder weather. More time is needed to make a discharge measurement under ice cover be- cause snow must be cleared, holes must be drilled in the ice, and additional infor- mation must be collected. Not only are winter discharge measurements more cost- ly, but they also have greater uncertain- ty than those made under open-water con- ditions. STAGE DATA Stream stage data may be of limited value for streams that are affected by ice for extended periods of time. The data are of value during periods when anchor ice is forming and releasing and are also of value in determining when ice cover is forming and when it is breaking up. Most offices of the Geological Survey, except in parts of Alaska, at- tempt to collect stage data throughout the year. Stage records are obtained · throughout the winter, in most areas, because thaws occur that result in open- water periods. Ice will also sometimes bridge over the water surface as the flow recession continues through the winter season, providing open-water flow condi- tions. Both of these conditions may pro- vide periods when a stage-discharge rela- tionship is stable and defined. Data obtained during these periods can be extremely valuable in helping to analyze the data when ice cover exists. Stage data usually are collected using stilling wells and intakes or manometers and gas-purge systems. Stilling wells often are set in the streambank with in- takes from the well to the stream. Some- times the well may be attached to the downstream side of a bridge or other structure and holes placed in the well to allow the water level in the stream to be reflected in the well. Where mano- meters are used, the gage shelter may be placed in a safe location some distance back from the stream. Freezing of the intakes or of the water in the well is a major problem in collecting stage data under severe cold conditions. Attempts are made to locate the intakes low in the water column to prevent freezing. At times it is not possible to place the intakes in a posi- tion where they will not freeze. Heat tapes are sometimes wrapped around the intakes. If line current is not avail- able, a generator can be connected to the heat tape to thaw the intakes if they have become frozen. Often the heat tape will become damaged after a period of time and will have to be replaced. The replacement of the heat tape can be a significant effort. In some places, the heat tape is used to thaw the intakes primarily so they will be open when the spring ice breakup occurs. Oil cylinders were used inside still- ing wells in the United States for many years to maintain a liquid column inside the well from which to obtain a stage record. This is no longer recommended because of the potential for spilling the oil and contaminating the stream. A variety of other means are used to try to keep the water in the stilling well from freezing. Some use is made of styrofoam pellets poured into the well. This creates an insulating barrier be- tween the colder air and the water. This is a messy and sometimes troublesome so- lution to the problem. Nitrogen gas is sometimes bubbled into the well to pre- vent ice formation. Where possible, an 139 insulated subfloor often is installed in the well below the frostline. This can be a very effective means of keeping the water in the well from freezing. Propane heaters are used in some wells. Where electricity is available, heat lamps or electric floating-tank heaters are some- times used. In areas where the stream has a steep gradient and where the stream water tem- perature is above freezing, such as down- stream from a dam or a spring, stream water is brought into the stilling well through a pipe with its intake at an upstream point. This water is allowed to circulate through the well and is dis- charged through the stilling-well in- takes. The pipe from the upstream should be smaller than the stilling-well intake so that the level of the water in the well can remain at the same level as the level in the stream at the gage. The warmer circulating water helps keep the water in the well from freezing. This method is best suited to gages having marginal well-freezing problems. Stations using manometers and gas- purge systems sometimes have problems from ice forming over the orifice, es- pecially if the flow of gas is cut off for a period of time. A more common problem occurs when moisture collects in a low spot of the gas-feed line and freezes, but this can be avoided by elim- inating low points in the gas line. Recorders Recorders will often fail for a vari- ety of reasons in severe cold weather. Clocks may quit because the lubricant gets stiff or battery efficiency is re- duced because of the low temperatures. Various parts in the recording system may not be capable of operating in ex- treme cold. Some wires or other parts may become brittle and break during ex- treme low temperatures causing failure. Recording systems operating in extreme conditions should have component parts that are specified and designed to func- tion in those conditions. Many pieces of equipment currently available are not designed for the extreme conditions that are often experienced in the field. Gage houses or recorder components often are insulated. This will not only protect equipment, at times, from the most severe temperatures, but it also will reduce the shock due to large and fairly quick temperature changes. In addition, where electricity is available, heat lamps may be used in the shelter to keep temperatures from dropping too low. Otherwise, propane heaters may be used. Unless the propane heater contains fea- tures to stop the flow of gas if the flame goes out, the gage house can fill with gas and be a safety hazard to the next field person visiting the gage. Stream Controls When the control freezes, the stage record may lose much of its value because the stage-discharge relationship is no longer valid. In some of the more moder- ate winter areas, attempts are, there- fore, made to keep the controlling sec- tion of the stream free from ice to main- tain the stage-discharge rating through- out the winter. This is generally only practical on small streams. One approach to this is to create turbulence in the control section by installing a weir. Ice may form upstream and constrict the approach section, creating a shift to the rating. A shelter is sometimes constructed over the control section of the stream. The lower part of this structure must be made of material that will move with the current if the stream rises while the shelter is in place. The shelter itself should be allowed to float off to one side of the control and be retained on the other side. This prevents loss of the shelter should a major rise of the stream take place while the shelter is still over the control. The shelter can be used to provide a small amount of in- sulation from the colder air outside of the shelter. This, together with the small amount of energy released as the water goes over or through the control section, may prevent the control from freezing. A propane heater also can be installed under the shelter to provide added heat. When the control is a weir, ice will sometimes freeze over the weir and will freeze to a depth on the upstream side of the weir that is lower than the weir crest. When this occurs, siphoning of 140 water out of the pool upstream from the weir may take place until the upstream pool is lowered sufficiently to break the suction. The pool will then fill up and the cycle may repeat itself many times. This will produce a series of stage drops and rises on the stage record. The pattern is easily recognizable but the discharge during this time must be determined using other methods. INTERPRETING STAGE AND FLOW RECORDS The accuracy of flow records during winter months will generally be less than those coll~cted during open-water peri- ads. The accuracy will depend on the type of ice, the areal extent of the ice on the river, the duration of the ice cover, the number of discharge measure- ments obtained during periods of ice affect, the available information from surrounding streams and weather stations, and on a variety of other information that may be available to the analyst. Usually, the records at stations least affected by ice are analyzed first. Those that are most affected, or most diffi- cult, are analyzed last. This allows area flow trends to be better understood prior to analyzing the more difficult records. Probably the most difficult periods of record to interpret are during ice-cover formation and during the spring breakup of ice. There is considerable variation in backwater during these periods. Large volumes of water may be discharged during the spring breakup in a short time. These added uncertainties during ice formation and breakup may significantly affect the error in the annual flow determinations. The stage record collected during periods of anchor ice development and re- lease is generally clearly recognizable. As anchor ice forms on the streambed during the night, it may produce an in- crease in water stage. In the morning when the streambed is warmed by shortwave radiation from the sun, the anchor ice will release and allow the water level to drop to its previous stage before the anchor ice formed. A line drawn between ice-free periods on the stage record, under these conditions, will allow a good estimate of the flow to be made (Rantz and others, 1982, pp. 361-362). The rise due to anchor ice formation is slow com- pared to the fall when the ice is re- leased. The flow records during these periods can be computed from the stage record with little loss of accuracy. The most common method of computing discharge during periods of ice is the hydrographic and climatic comparison method. This involves the use of dis- charge measurements, comparisons with appropriate open-water records at other stations and climatic records, observer notes, power generation data, and other information that might help the analyst estimate flow. If no stage record exists and where there are no nearby records with which to make hydrographic comparisons, an av- erage recession curve may be used, so long as the fall recession continues through the winter without interruption. All discharge measurements and· flow re- ~rds for the period of record are plot- ted on a single hydrograph, and a curve is drawn that most approximately repre- sents the suite of curves. The shape of this curve is then used to estimate a winter's record by drawing a line of the same shape through any measurements made during the period in question. CURRENT ACTIVITIES For the past several years, represent- atives from the Water Resources Division of the Geological Survey have been meet- ing with representatives from the Water Survey of Canada of the Water Resources Branch and with others to compare equip- ment and methodology and to look for better ways of collecting streamflow data under ice cover. Equipment from one agen- cy has been, and is being, t~sted and tried by the other agency and vice versa. As a result, some equipment has been modified to improve data collection. Another joint effort between the Geological Survey and the Water Survey of Canada is the preparation of a manual for the measurement of flow under ice cover. This manual will come out in the United States as a chapter in the "Tech- niques of Water-Resources Investigations" series. 141 There have been a number of recent de- velopments in the area of equipment that have implications in the collection of flow data under winter conditions. Fol- lowing is a brief discussion of a few of these developments in the United States. The Geological Survey has been devel- oping solid-cup bucket wheels for current meters. The solid cups will not allow slush ice to become trapped in the cups as it presently does in the hollow cups. It is expected that in the near future, the solid cups will be commonly used, es- pecially for measurements under ice cover, in the United States. Optic heads for counting current-meter revolutions have been developed. This current-meter option will allow recently developed digital counters to count and record meter revolutions, time, and velo- city. Due to reduced drag resulting from use of the optic heads, the meters have a lower stall velocity, allowing lower stream velocities to be measured. Fiberglass wading rods have been de- veloped. These are lighter, easier to handle in cold weather, and have built-in electrical circuitry for the counting of current-meter revolutions. Ice rods for suspending a current meter through a hole in the ice have been redesigned to im- prove the readability of the distance markings on the rod and to improve the means of adding and disassembling sec- tions to the rod. Weight-suspension systems are being standardized. Considerable testing has been done on various systems in towing tanks to determine good suspension sy- stems for use through ice. The Water Survey of Canada has developed a "tear drop" suspension assembly for use in small diameter holes in ice. This assem- bly is being examined by the Geological Survey for adoption in the United States. Various power heads for ice augers, auger assemblies, and cutting bits are being tested. These tests are evaluating the weight of the units, reliability, ability to cut through ice, ease of assembly and disassembly, and of main- tenance of effective cutting teeth. Ice chisels, for use in the United States, have been redesigned based on work done in Canada. Wooden handles re- place the steel handles previously used. The chisel balance has been improved by placing the center of gravity nearer the lower end of the chisel. The angle of the cutting edge has been modified so that it cuts into the ice better while still retaining the strength of the blade. Acoustic velocity meters are being tested in ice-covered rivers. This meter depends on the reception of an acoustic signal that has been transmitted across a stream at an angle, other than 90°, to the flow direction. The time of travel of the signal across the stream in the upstream and downstream directions are determined and the stream velocity com- puted from the time differential. This metering system should be of value for a number of stream-gaging sites. The use of transducers is being inves- tigated for the purpose of replacing the manometer with a transducer in connection with a gage-purge system. The transducer will be placed in the gage shelter and will measure the pressure transmitted by the gas-purge system. The use of trans- ducers will allow for smaller gage shel- ters to be used and should make insulating or heating the shelter or equipment easier. The temperature of the trans- ducer may have to be measured and recorded to make corrections to the transducer readings. Dilution measurements of streamflow under ice have been successfully made on a few streams. Details of the method are described in the report by Kilpatrick and Cobb (1984). More testing is needed to determine the conditions where this method can be effectively used. CONCLUSIONS There are numerous problems with col- lecting streamflow records in the winter months in many parts of the United States. Nevertheless, these streamflow records are needed and every effort possible, within funding and manpower constraints, must be exerted to collect them. The collection of accurate winter records is highly dependent on the commitment of the people who collect them. Work is progressing on providing better equipment for the field person and on expanding documentation of accepted procedures for the collection of winter flow records. 142 REFERENCES Kilpatrick, Frederick A. and Ernest D. Cobb, 1984. Measurement of Discharge Using Tracers. u.s. Geological Survey Open-File Report 84-136. 73 pp. Rantz, S. E. and others, 1982. Measure- ment and computation of streamflow: Volume 1. Measurement of stage and discharge. u.s. Geological Survey Water-Supply Paper 2175. 284 PP• Rantz, S. E. and others, 1982. Measure- ment and computation of streamflow: VoluJV,e 2. Computation of discharge. U.S. Geological Survey Water-Supply Paper 2175. pp. 285-631. Wellen, Paula M. and Douglas L. Kane, 1983. A comparison of velocity meas- urements between cup-type and electro- magnetic current meters. Report pre- sented to the Annual Meeting of the Alaska Section, American Water Resour- ces Association, Fairbanks, Alaska, November 10-11, 1983. 12 pp. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 SIMPLIFIED METHOD OF MEASURING STREAM SLOPE Jacqueline D. LaPerriere and Donald c. Martin 1 ABSTRACT: For use in measuring stream slopes in remote areas, surveyor's instruments such as a level or telescopic alidade are relatively cumbersome and delicate; furthermore, standard surveying methods require that an unobstructed line of sight be found or cleared. We describe a simple "carpenter's water level" that saves considerable effort. In comparisons with a surveyor's level to measure slopes over 10 m, measurements with our instrument were within 2\\'i of those made with a surveyor's level for slopes of 2.46\\'i and within 0.5\\'i for slopes of 5.31\\'i; and, within 2\\'i for slopes of 4.24\\'i over 20 m. {KEY TERMS: stream slope; field instrument.) INTRODUCTION To test a hypothesis that stream gradient is associated with the concentration of drifting benthic macroinvertebrates (David Zimmer, 1977, personal communication), one of us {J.O.L.) needed to measure stream slope {LaPerriere, 1983). During the first field season, stream reaches were specifically chosen because they had a clear line of. sight that would enable surveying of the slope. In the following field season, sites were chosen for other characteristics, and an alternative method of measuring stream slope was sought. Rough surveying with a clinometer yielded unsatisfactory results. The use of a "carpenter's water level" (here called a slope-meter) suggested by William w. Mendenhall (Professor of Civil Engineering, University of Alaska, 1979, personal communication), was then constructed, and used in the field. We describe the construction and use of this field instrument, and compare its accuracy with that of standard methods of measuring slope with a level and stadia rod. METHODS AND MATERIALS An ordinary level and stadia rod were used in tests of the accuracy of the field instrument. The slope-meter is made inexpensively from a length of transparent plastic tubing, two stoppers that fit the inner diameter of the tubing, two metersticks, some fiber-strengthened adhesive tape, and a length of unstretchable cord (polypropylene 1-cm in diameter is suitable). The cord is taped to the base of the two metersticks so that they are set a pre- determined distance apart when the cord is pulled taut. When this distance does not exceed 10 m, the cord and tubing do not readily become entangled with each other, and the tube can be easily inspected visually for bubbles that might cause errors. The transparent tube (2 m longer than the pre-determined distance) is then placed parallel to the out stretched cord and taped to the top and base of each meterstick so that graduations on the metersticks can be seen through the tubing. Water is then added to the tube until it rises from 1/3 to 1/2 of the height of the metersticks when they are held side-by-side. (Water may have to be added or removed to facilitate readings). The stoppers are used 1Respectively, Assistant Leader and Research Assistant, Alaska Cooperative Fishery Research ~it, 138 Arctic Health Research Building, University of Alaska, Fairbanks, Alaska 99775- 0110. The Unit is jointly supported by the University of Alaska, Alaska Department of Fish and Game, and the u. S. Fish and Wildlife Service. 143 to hold the water between measurements. In use, each of two persons places their meterstick at the wetted edge of the stream, laying the cord along the edge of the stream between them. Each person then reads the height of the water level on the meterstick and the difference between the two readings is the rise. Dividing the rise by the set distance between the two metersticks gives the slope. If slope needs to be surveyed for a longer specific distance, multiple readings are made in sequence. The downstream operator walks to the position of the upstream operator, who then walks upstream the set distance and lays the line on the wetted edge; the reading is then taken again. This procedure is repeated until the total distance has been covered. The design of a more sophisticated carpenter's water level is presented by Walkotten and Bryant (1980). They also provide photos and sketches of their instrument in use. RESULTS We checked measurements made with this field instrument against those made with a surveyor's level and stadia rod. In measurements made by both methods in triplicate at three locations along halls with built-in wheelchair ramps in the Fine Arts Building of the University of Alaska, Fairbanks, (Table 1) the slopes determined were nearly identical (within 2% at slopes 144 of 2.46% and within 0.5% at slopes of 5.3U over a 10 m distance; and, within 2% at slopes of 4.24% over 20 m). DISCUSSION Slopes measured with the "water level" were not sign! ficantly different from those measured with a surveyor's level and stadia rod, even when two successive measurements (extending 20 m) were made at one location. We have used this instrument successfully in the field, where it can be laid out along a stream through riparian vegetation that would block a line of sight measurement with surveying instruments. The instrument is light and durable for use in remote field locations common in Alaska. REFERENCES LaPerriere, J. D. 1983. Alkalinity, Discharge, Average Velocity, and Invertebrate Drift Concentration in Subarctic Alaskan Streams. Journal of Freshwater Ecology 2:141-151. Walkot ten, W. J. and M. D. Bryant 1980. An Instrument to Measure Stream Channel Gradient and Profiles. u. S. Forest Service. Paci fie Northwest Forest and Range Experiment Station. Research Note. 5 pp. TABLE 1. Measurements of Slope at Three Locations by Two Methods METHOD Level and Rod Car2enter's Water Level location Rise (ft. ) Distance (ft. ) Slope (") Rise (em) Distance (m) Slope (") 1 0.79 32.8 2.41 24.5 10.0 2.45 0.79 32.8 2.41 24.5 10.0 2.45 0.79 32.8 2.41 24.7 10.0 2.47 2 1. 74 33.0 5.27 52.8 10.0 5.28 1. 75 33.0 5.30 53.1 10.0 5.31 1. 74 33.0 5.27 53.4 10.0 5.34 3 2.78 66.5 4.18 84.7 20.0 4.24 2.78 66.5 4.18 84.6 20.0 4.23 85.0 20.0 4.25 145 WATER QUALITY 147 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 WATER QUALITY-DISCHARGE RELATIONSHIPS IN THE YUKON RIVER BASIN, CANADA Paul H. Whitfield and W. G. Whitley ABSTRACT: Weekly and biweekly water quality samples were collected at ten sites throughout the Canadian portion of the Yukon River basin during 1982-1983. The relationships between the water quality variables and discharge are examined. Most of the relationships between the variables and discharge are either opositive or negative and exhibit hysteresis, a few variables appear to be independent of discharge. A system of classification of the relationships, with potential causal mechanisms, is proposed. KEY TERMS: Hysteresis, water quality-discharge relationships.) INTRODUCTION Water quality-discharge relationships are characteristic of the processes which result in the movement of materials into rivers and streams. Whitfield and Schreier (1981) found that such relationships formed open loops which persist from year to year and become progressively open with downstream distance. These relationships are described as exhibiting hysteresis, the lagging or leading (in time) of a physical effect in relation to its cause. This paper describes these processes and relationships throughout the C,anadian portion of the Yukon River basin. These relationships can lead to an understanding of how water quality is determined (eg. Davis and Zobrist, 1978; Hall, 1971; Johnson, 1979; Rieger and Olive, 1985; Toler, 1965; Whitfield and Schreier, 1981). Such an understanding is pertinent in the north where a general lack of data requires extrapolation to unsampled or poorly sampled basins. Understanding the dynamic processes which give rise to these relationships requires a knowledge of the entire seasonal nature of these systems. Water quality data were gathered at ten sites at one week intervals (7 sites) and two week intervals (3 sites) during the Yukon River Basin Study. These data describe a basin largely unaffected by human activities. They were gathered to provide information about the seasonal and spatial variation in water quality, and are described by Jack et al. (1983). This paper proposes a classification system to describe the relationship between water quality variables and discharge. These relationships exhibit varying degrees of hysteresis and dependence on discharge. The degree of hysteresis is approximated by measurement of the lag 1 autocorrelation coefficient The clas~ification system was developed from a sorting and grouping exercise. Hard to classify cases were also identified. A mechanism for classifying new cases on a statistical basis is demonstrated. Possible processes which result in water quality variables behaving in the described manner are proposed. STUDY AREA The Yukon River basin is the fifth largest river basin in North America, draining northwestern Canada into the Bering Sea through central Alaska. At the end of the Canadian portion of the basin the drainage area is 294,000 km 2 (113,500 sq miles) with an average annual flow of 2340 m 3 /sec (82,660 cfs) at Eagle, Alaska (U.S.G.S., 1984). Discharge data were obtained from published records (U.S.G.S., 1984; Water Survey of Canada, 1983; 1984). The drainage basin consists of two distinct portions, the western portion drained by rivers dominated by the glaciers of the St Elias Mountains (Wrangells) and the Coast Mountains; the remainder of the basin drains the Yukon Plateau and the Mackenzie mountains. The physiography t:>f the basin 1Respectively, Inland Waters Directorate, Environment Canada, 502-1001 West Pender Street, Vancouver, B.C., V6E 2M9; and Northern Affairs Program, Indian and Northern Affairs Canada, 200 Range Road, Whitehorse, Yukon Territory, Y1A 3Vl. 149 is described in detail by Bostock (1948). The water quality da:.a described here were collected between August. 1982 and September,l983. The procedures used to collect and analyse the samples are described by Jack et al. (1983) and are typical of those used by most water quality agencies. Whitfield (1983) evaluated a number of water quality sampling sites in the lower portion of the Yukon River and found that sampling at a one week interval provided the most representative data. The ten stations which had sampling frequencies of weekly or every two weeks were selected from the 21 sites sampled by Jack et al. (1983) and are shown in Figure 1. Samples were collected once every two weeks at three sites (Nordenskiold, Takhini at Highway #2, and Yukon River below Whitehorse) and once each week at the other seven sampling sites. ·, ~ Yukon Temtory \ 0 I \"koo "''"' ot Eoote,Aioska L Northwest ( Terntones Alaska I I I I 0 \ ; " jl "-------, ,----.'-PellyRiverat Pelly Crossmg __ /_____ \ ,~ -~---,"' I ) ."' ' I , /' ~~ ~ //\ Yukon Rrver at CarmacKs\ Ueily River ot Faro /-~ " '-. ~..._ -~---~ ~ j"J: Nordenskrold Rrver at Ccrmocks ~\""" :::-, ~ r ',, /~ \, ~ Tokhinl Rrver ot Hlghwayll2_. )' y k R b 1 Wh t h '------/"" --.-\ u on rver eow r e ors~ ' ~ '' ~ ~ _/ ~' ~on ~et\at o"tlet of Matsh Lake '~~ ~ ' -.... --"'";. ;".,.,;'ci1t{j)~ - -~'-- Pacific Ocean Figure 1. Study area showing-sampling locations. RESULTS AND DISCUSSION Jack et al. (1983) collected data for some 37 water quality variables. We examined these records and eliminated those variables which had little or no variation, those variables where a substantial portion was either missing or less than detection limit, and those variables which provided redundant information. We then had 26 variables to evaluate for the ten 150 rivers. The variables that were examined were total alkalinity, total hardness, calcium (dissolved), sodium (dissolved), potassium (dissolved), chloride (dissolved), fluoride (dissolved), pH, specific conductivity, turbidity, colour, silicate (reactive), sulphate (dissolved), nitrate+ nitrite, non-filterable residues, extractable arsenic and selenium, total dissolved nitrogen, total organic carbon, total inorganic carbon, total phosphorus, and total copper. lead, iron, manganese and zinc. Two additional data sets were subsequently eliminated for one river because of missing data. We then prepared log-log plots of each water quality variable against discharge, although for pH we used linear-log plots. The 258 plots were roughly grouped on the basis of visual similarities. This generally resulted in similar stations and variables being grouped together. The grouping was complicated by a lack of agreement within variables across all stations. It became evident that the ten stations were best considered as two main groups; the first consisted of six "river" sites, while the second consisted of three "lake-fed" sites. One site, the Yukon River at Carmacks, was often difficult to classify into either of these groups. We believe that this site is substantially affected by Teslin and Labarge Lakes which are upstream of this site, and reflects a transition between the "lake-jed" tyPe sites and the "river" type sites. Variables from the "lake-fed" sites appeared to be independent of discharge and were either random or dependent on other processes. This view was subsequently modified when it was realized that what had appeared to be independence of discharge was simply low slope relationships, when compared to the "river" systems. Thus one of the important differences between the II river" sites and the "lake-jed" sites is that the "river" sites show more concentration variation in relation to discharge. . This difference is due to the buffering and trapping processes in lakes. The relationships between water-quality variables and discharge are much more evident for the six 11 river" sites. Here we found that every case could be placed into one of four groups: (1) having a positive relationship with discharge, and exhibiting clockwise hysteresis, (2) having a negative relationship to discharge, and exhibiting counterclockwise hysteresis, (3) having no relationship to discharge, but exhibiting episodic or sequential behavior. (ie. some order exists), or ( 4) appearing to be entirely random. We observed that the degree to which hysteresis was evident increased with distance downstream or with the magnitude of the discharge, confirming the result of Whitfield and Schreier (1981). Yang et al. (1983) observed a reversal of the direction of the hysteresis for sediment-discharge relationships in the Yangtze River. The lagging of sediment was attributed to the timing of flooding and '.raining of rice fields. Rieger and Olive (1985) :xamined sediment-discharge relationships during torrn events in five small watersheds. They found everal types of relationships; sediment led discharge n 41% of the cases, lagged in 13%, was in phase in %, and 41% of the cases had no identifiable pattern. Their results show multiple types of relationships Nithin and between watersheds and storms. Rieger md Olive (1985) suggest that these complex ·esponses are the result of several sediment transport nechanisms. Our results show indications of some of hese complex responses. Since the basins that we :onsidered are relatively large these complex :esponses at the storm event level are generally nasked. The grouping of the variables suggests that ;ommon mechanisms cause the observed behavior. The variables that exhibit the positive relationship with discharge, and the clockwise hysteresis are: lon-filterable residues, turbidity, colour, total iron, :opper, zinc, manganese, and lead, extractable arsenic, and total phosphorus. These variables are sediment related and suggest a relationship to erosional processes. The postulated mechanism can be described as follows: freshet snowmelt and surface runoff carry surface material over seasonally frozen soils into the streams, while during post-freshet the thawed soils allow precipitation to replenish groundwater thus contributing less sediment to the streams. Figure 2 is a log-log plot of total iron concentration against discharge. The broad open loop shown there is characteristic of this group of variables. Although t.~e Yukon River at Carmacks was often difficult to classify in parallel to other stations, Figure 2 shows that this was not always the case. Figure 2 is typical of the hysteresis reported by Whitfield and Schreier (1981). The variables that exhibit a negative relationship with discharge and a counterclockwise hysteresis are: total alkalinity, calcium, total hardness, sodium, specific conductivity, silicate, sulphate and total inorganic carbon. These variables are major ions. generally conservative, and have a relationship to groundwater (Whiffield and McNaughton, 1986). The postulated mechanism which results in the observed hysteresis is described as follows: snowmelt and surface runoff dilute groundwater during freshets; during post-freshet conditions concentrations increase with the increasing dominance of groundwater. These relationships are more evident in the Yukon Basin than Whitfield and Schreier (1981) observed in the Fraser River Basin of British Columbia because of the increased dominance of groundwater and the dominance of snowmelt as the source of water in this generally arid area. Figure 3 is a log-log plot of sulphate concentration against discharge for the Pelly River at Pelly Crossing. This figure is the classic case of the negative relationship of concentration with discharge and the counterclockwise 151 Yuvon River at Carmacvs. Y.T. 10,----------------r-----------, ' ~ • • • • • ~ ~ 0> i ' ' E ~~· . ~ .~ . ': c tj1 + ·!,z· l; i "'' 0 I* + .. ' ;§ t·~ ~ 0.1 • . + I .if-4- $ 0.01 ~----------------------------1 100 1000 Discharge in m' /second 10000 Figure 2. Clockwise hysteresis with a pos111ve relationship to discharge. Points are joined in date order. [In Figures 2-9 the symbols indicate; c -March-April, • -May-June, o -July-August, • -September-October, ~ -November-February.] ~ 0> E .~ ~ 0 ..c Q_ :; (/) Pelly River at Pelly Crossing, Y.T. 100 ~-----------------------------, •• Ill • .. ---.: • 10+--------------------------~ 10 100 1000 Discharge in m' /second 10000 Figure 3. Counterclockwise hysteresis with a negative relationship to discharge. Points are joined in date order. Symbols are as given in Figure 2. hysteresis. We observed that the "openness" of these loops increased with downstream distance, and/or increased magnitude of discharge. The variables that exhibit no relationship to discharge, but show episodic or another form of ordered behavior are: chloride, fluoride, potassium, and total organic carbon. As a group, these variables are an odd assemblage. It is unlikely that the behavior of all four of these variables would be due to the same mechanism. Chloride patterns are typified as shown in Figure 4. These patterns are generally open loops with an apparent independence of discharge. Similar patterns were observed for potassium, suggesting a similar mechanism is active. Fluoride patterns were also highly characteristic, an example of which is given in Figure 5. The pattern shown in Figure 5 is typical of all fluoride series, each plot showed a crossover point in October-November forming a figure 8. Total organic carbon exhibited a pattern which tended to be bimodal in ·a fashion somewhat similar to fluoride. The first three variables are postulated to be controlled by a process which consists of constant input concentrations from atlhospheric inputs which are subsequently modified by contact with soils. Thus the highest concentrations occur at the beginning of the spring thaw, and the lowest values at the time of freeze-up. · The variables that appeared to be independent of discharge and generally behaved in a Pelly River at Pelly Crossing, Y.T. 10~---------------, 0.1+--......................... ---....... .....----~ 10 100 • 1000 Discharge in m /second 10000 Figure 4. A typical pattern for chloride and potassium with discharge. Points are joined in date order. Symbols are as given in Figure 2. 152 "-. 0> E c ·;; 0.1 -o ·;;: 0 :::s G: Pelly River at Faro, Y.T. o.o1+-_....,...., ........ .....,.--.--....... .....---.......... ........t 10 100 • 1000 10000 Discharge in m /second Figure 5. A typical pattern for fluoride with discharge. Points are joined in date order. Symbols are as given in Figure 2. random manner are: total dissolved nitrogen, nitrate + nitrite, extractable selenium, and pH. Of these we observed two distinct patterns. The first, shown by pH and selenium was an apparent lack of variation, Figure 6 is typical of the results we obtained. The two nitrogen variables behave in very similar manners, partially due to their interrelated nature. Typical plots of both nitrate + nitrite and total dissolved nitrogen showed an independence of discharge but a great deal of variation with a definite episodic pattern. Figure 7 shows total dissolved nitrogen concentrations in the Yukon River at Eagle, Alaska. The high variation in concentration as well as several periods of episodic behavior are evident To illustrate the differences that exist between "river" and ... lake-fed" systems we have included a plot of sulphate against discharge (Figure 8) which can be compared with Figure 3. Even though the range of sulphate concentrations is one to ten in Figure 8 as opposed to 10 to 100 in Figure 3, there is only the slightest hint of a decrease in sulphate concentration with increasing discharge. This property was common to all of the "lake-fed" systems, and for this reason we have considered them separately. With the exception of these differences in the slope of the relationship with discharge, variables such as sulphate behaved in generally the same manner as in "river" systems. One of the common characteristics of the "lake-fed" sites is the behavior of the variables: colour, turbidity, phosphorus and iron. These variables were observed to exhibit a pattern which we considered to be of a non-random nature. These variables have a large peak in spring (March-April) and sometimes a secondary peak in the late summer. The coincidence of these peaks with peaks in activity of biological processes in the lakes suggests that these variations are due to primary production within the lakes. Figure 9 is a plot of turbidity at the outlet of Marsh Lake which shows the peak in the early portion of the year and a secondary peak occurring at peak discharge in the late summer. This plot is typical of these four variables in the three "lake-fed" sites. Although the classification scheme proposed has clear distinctions between the various groups, some cases remain difficult to place in single groups. These cases generally have a more closed than open structure and often contain crossovers and secondary loops. While this was particularly true for many variables for the Yukon River at Carmacks, several other cases also existed. Whitfield and Schreier (1981) designated many of the cases they considered as mixed (clockwise and counterclockwise hysteresis). The data which they considered consisted of many fewer observations (8) than we had available to us (26 -55). We believe that although cases do exist which contain secondary loops, each and every case contains a dominant structure which allows it to be placed into one of the groups described. Klondike River at Dawson, Y.T. 14-r----------------, !' 'E :J 12 10 '0 8 ._ 0 '0 c 0 iii .s :i a. o+---~~---~~----~~ 1 m , mo Discharge in m /second moo Figure 6. A typical pattern for pH and selenium with discharge. Points are joined in date order. Symbols are as given in Figure 2. 153 .;;:::_ m E s c ., m _g Yukon River at Eagle, Alaska 10-r-----------------, z 1 "0 ., > 0 "' "' 0 :E ,2 1000 ' Discharge in m /second 10000 Figure 7. A typical pattern for nitrogen variables with discharge. Points are joined in date order. Symbols are as given in Figure 2. Yukon River at outlet of Marsh Lake 10'-r---------------~ 1+----------------~ 100 Discharge in m' /second moo Figure 8. The relationship between sulphate and discharge at a "lake-fed" site. Points are joined in date order. Symbols are as given in Figure 2. We were concerned that since the classification of individual cases required substantial judgement that a general application was not immediately practical. We then considered the basis for our proposed scheme on a more mechanistic basis. The first criterion in the classification is the behavior in relation to discharge and we had found three possibilities: (1) increasing concentration with increasing discharge, (2) decreasing concentration with increasing discharge, and (3) concentration independent of discharge. This property can be measured or estimated in a number of ways: simple correlation was thought to be the simplest, limiting the range to -1 to 1 and avoiding the problems of scale inherent in statistics such as slopes etc. The second classification criterion which was used has been alluded to as an ordering of sorts or a lack of randomness. This property is common in most sampled time series, such as those being described here. Whitfield (1984) examined the time series properties of these same series of data and found many of them to be highly autocorrelated. In its simplest form, autocorrelation at lag 1 is the correlation of each point in a series with the one immediately following, it also varies from -1 to 1 and also is dimensionless. These two classification criteria, or statistics, should be sufficient to place the classification of cases on a more rigorous basis. We then estimated both the correlation of each variable with discharge, and the autocorrelation of each ::i ,_: -; .£: Yukon River at outlet of Marsh Lake 10-r-----------------., 0.1+--------------.......... 100 Discharge in m 3 /second 1000 Figure 9. The relati~nship between turbidity and discharge at a "lake-fed" site. Points are joined in date order. Symbols are as given in Figure 2. 154 c ., 0.75 ~ ., 0 u c 0.50 .Q a ~ "-0 0.25 u E ::J <( (J) 0 0 _j -0.25 -1 River Sites D •• r: ,..,. .. '1[c iE ••• rn v §ri : • • l: LJ<>OOc-:-0 fE • 00 I • ill ...... 0 0 ~I fE • • CJ 0 • ., • ~~0 ill i 0 0g 1 .-.-: <> •• o• c I oo • • ----·--r • • • -0.75 -0.50 -0.25 0 0.25 0.50 0. 75 Correlation Coefficient r • Positive ~ Negative J Cl and K · • pH and Se EE F and TOC • Nitrogens Figure 10. Relationship between correlation with discharge and autocorrelation within the grouped variables at "river" sites c ., 0. 75 ~ ., 0 u c 0.50 .Q a ..., "-"-0 0.25 u 0 "':i <( (J) 0 _j Lake-fed Sites fi I :::: D • .lEE.' ~ oil'~ r·· Ill [J ill •• 0 <> ill oo• ' . OJ~ = • r::::J • ~ • .• • : a;:----· t • -• -1 -0.75 -0.50 -0.25 0 0.25 0.50 0.75 Correlation Coefficient r a Positive c_-=: Negative 0 Cl and K <· pH and Se ffi F and TOC e Nitrogens Figure 11. Relationship between correlation with discharge and _autocorrelation within the grouped variables at "lake-fed" sites variable. We then prepared two plots, one for the "river" sites (Figure 10), and one for the "lake-fed" sites (Figure 11). Variables were grouped according to the classification system described previously and common symbols used for each group. These figures both show that the proposed classification scheme fits fairly well with a description that would use the statistics describing the correlation to discharge, and the amount of autocorrelation within each series. It is not, however a perfect fit There are several instances where a case would be classified differently, if only the statistical properties were used. These cases were also difficult to classify on a visual basis, and may be due to anomalous data. There remain a number of limitations. concerns, and questions that need further consideration. One area that we are particularly concerned about is the lack of data from rivers that are directly glacially fed. Data from that type of system is needed to fully understand the range of variation present in the Yukon Basin. Some superimposed loops occur at high discharge rates. that might be attributable to the late summer glacial outflow. There are limitations to the classification scheme using either the visual or the statistical approach. One of these is a problem of validation, ie. how can we confirm that a specific case is correctly classified? In the process of the original classification exercise we examined a number of cases which were odd. Classification of these is often not possible without an in-depth knowledge of the variable being considered. For example, during a portion of the year the data were at or near the detection limit and the loops therefore were flattened at one end, or the slope of the relationship to discharge appeared to change. Superimposed loops, related to storms and/or delayed discharges from higher elevations also confuse the relationships. Similarly, Rieger and Olive (1985) found numerous variations on hysteresis of sediment and discharge in storm runoff. Perhaps of most concern is that the present classification system does not provide a feeling for the concentration disparity between rising and falling limbs of the annual hydrograph. In some cases the differences observed were of only minor importance, whereas in some, such as those shown in Figures 2 and 3. the differences are a factor of 2 to 10. The use of rigorous statistical techniques, such as time series methods will alleviate most of the inherent estimation problems (eg. Rieger and Olive. 1985). This type of analysis has a number of important applications. First, it aids understanding of the seasonal variation of water quality variables with discharge. Variables can be grouped with others that behave similarly and distinguished from groups with different behavior. It identifies variables which are poorly understood such as fluoride, chloride, and other variables that appear to be independent of discharge. Secondly, with an understanding of the underlying processes and variation with the size of the stream. tentative extrapolation can be made to other areas and other variables. This is of particular value in northern Canada where data and understanding are limited. Thirdly, sampling can be directed towards variables and seasons in a knowledgeable way. In northern systems sampling is often most intensive in the spring and summer whereas many variables reach limiting values in the fall and winter. 155 CONCLUSIONS The classification system described here is consistent with mechanisms postulated for each grouping, and these mechanisms are relevant processes in determining water quality in this subarctic region. Visual and statistical analysis both resulted in the same groupings. Variables associated with erosion demonstrate a positive relationship to discharge and a strong leading hysteresis. The dominance of groundwater in controlling the concentrations of many variables is reflected in broad open loops with negative slopes. Some variables were observed to be independent of discharge and some of these also demonstrated hysteresis. This taxonomy is consistent with that proposed by Whitfield and Schreier (1981), and the causal mechanisms are believed to be similar. The classification system proposed may allow the characterization of other areas with more limited sampling. UTERA TURE CITED Bostock, H.S. 1948. Physiography of the Canadian Cordillera, with special reference to the area north of the fifty-fifth parallel. Geological Survey of Canada Memoir 247. Davis, J.S., and J. Zobrist 1978. The interrelationships among chemical parameters in rivers -analysing the effects of natural and anthropogenic sources. Progress in Water Technology. 10:65-78. Hall, F.R. 1971. Dissolved solids-discharge relationships. 2. Applications to field data. Water Resources Research 7:591-601. Jack, M., B.E. Bums, and T.R. Osler. 1983. Water Quality -Yukon River Basin. Yukon River Basin Study Water Quality Work Group Report No. 1. Indian and Northern Affairs Canada. Johnson, A. 1979. Estimating solute transport itt streams from grab samples. Water Resources Research 15: 1224-1228. Rieger. W.A., and L.J. Olive. 1985. Sediment responses during storm events in small forested watersheds. In: Proceedings of the Statistical Analysis of Water Quality Monitoring Data. October 1985. Toler, LG. 1965. Relation between chemical quality and water discharge in Spring Creek, Southwest Georgia. U.S. Geological Survey Professional Paper 525-C:C209-C213. U.S. Geological Survey. 1984. Water Resources Data for Alaska. Water Year 1983. U.S. Geological Survey Water-Data Report AK -83-1. Water Survey of Canada. 1983. Surface Water Data, Yukon and Northwest Territories, 1982. Environment Canada. Water Survey of Canada. 1984. Surface Water Data, Yukon and Northwest Territories, 1983. Environment Canada. Whitfield, P.H.. 1983. Evaluation of water quality sampling locations on the Yukon River. Water Resources Bulletin 19: 115-121. Whitfield, P.H.. 1984. Optimization of water quality monitoring the Yukon River Basin. Yukon River Basin Study Water Quality Work Group Repon No. 2. Inland Waters Directorate, Environment Canada, Pacific and Yukon Region. Whitfield, P.H. and H. Schreier, 1981. Hysteresis in relationships between discharge and water chemistry in the Fraser River basin, British Columbia. Limnology and Oceanography 26:1179-1182. Whitfield, P.H., and B. McNaughton. 1986. Depression of dissolved oxygen under ice in two Yukon Rivers. Water Resources Research in press. Yang, Z.-S., J.D. Milliman, and M.G. Fitzgerald. 1983. Transfer of water and sediment from the Yangtze River to the East China Sea, June 1980. Canadian Journal of Fisheries and Aquatic Sciences 40(Supplement 1):72-82. 156 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 THE ROLE OF SNOWCOVER ON DIURNAL NITRATE CONCENTRATION PATTERNS IN STREAMFLOW FROM A FORESTED WATERSHED IN THE SIERRA NEVADA, NEVADA, USA Jonathan J. Rhodes, C. M. Skau, and D. L. GreenZee 1 ABSTRACT: In-basin hydrology was moni- tored intensively concurrent with chemi- cal sampling for N0 3 -N concentrations in snow, vadose water, groundwater, overland flow and streamwater in a small headwater stream in the Sierra Nevada. Denitrifi- cation rates from a meadow within the watershed were measured in the field. Mean N0 3 -N concentrations were: 0.037 mg/1 for snow; 0.001 mg/1 for ground- water; and 0.097 mg/1 for overland flow. N0 3-N concentrations in the stream remained <O .001 mg/1 except during the generation of hydrograph peaks by snow- melt. When the watershed was snow- covered, peak nitrate concentrations in the stream coincided with discharge ~xima. As snowcover was melted off the lower watershed, nitrograph maxima became lagged behind peak discharge, a common pattern during snowmelt in the Sierra Nevada. The lagged pattern results from the operation of biological N0 1 -N removal mechanisms which are light-and/or temperature-sensitive. The removal mechanisms are denitrification, uptake by plants and in-stream uptake by periphy- ton; all have diurnal patterns of uptake after snowcover removal due to diurnal heating and irradiance. The· removal mechanisms are dormant or steady beneath the insulating snowcover which excludes light, hence lagging does not occur during complete snowcover. By con- trolling both runoff mechanisms and light and heat delivery to the watershed, snowcover strongly affects concentration patterns of N03-N during snowmelt. (KEY TERMS: snowmelt chemistry; nitrate transport; aquatic biology; nutrient cycling; snowmelt runoff; small streams.) INTRODUCTION Runoff mechanisms generating stream- flow from snowmelt are known to vary seasonally and areally in forested water- sheds. The dominance of the particular mechanism in generating snowmelt runoff is dependent on antecedent hydrologic conditions, basin topography and the intensity and duration of snowmelt (Dunne and Black, 1971; Price and Hendrie, 1983). The mechanisms, such as sub- surface and overland flow, involve different paths and widely divergent travel times prior to entering streamflow (Dunne, 1978). Water pathway and resi- dence time are the strongest controls on water chemistry on both the geochemical and biological levels. Therefore, it is logical that water from different runoff regimes would have different chemical signatures. This is especially true when dealing with a non-conservative constitu- ent, such as N0 3 -N, which is strongly affected by biologic reactions. Stream chemistry should be a reflection of the type of mechanisms contributing to streamflow generation at a point in time and space. Snowcover greatly influences both runoff mechanisms and biology, hence it exerts a strong control on stream chemistry. 1Respectively, Graduate Research Assistant, College of Forest Resources, University of Washington, Seattle, WA 98195; Professor of Watershed/Forestry, Department of Range, Wildlife and Forestry, University of Nevada Reno, Reno, Nevada 89512; and Hydrologist, Tahoe Regional Planning Agency, P. 0. Box 1038, Zephyr Cove, NV 98448. 157 In 1982, a field study was initiated to investigate the hydrologic processes controlling nitrate export in streamflow from a high-elevation, forested watershed in the central Sierra Nevada, Nevada, USA. Nearby Lake Tahoe is currently undergoing cultural eutrophication due to accelerated nitrogen loading (Goldman, 1985). The ultimate objective of the research was to provide a scientific foundation for watershed management policies in the Lake Tahoe Basin. The more immediate goal of the study was to characterize the nitrate chemistry of precipitation, vadose water, groundwater, overland flow on hillslopes, and stream- water both temporally and areally. Concurrent with water quality sampling, intensive hydrologic monitoring was performed in order to determine how in-basin hydrologic processes affected the nitrate load in streamflow. The initial phase of the research identified denitrification within the subsurface system as a possible important component of N0 3 -N removal. Subsequently, summer and autumn denitrification rates in a wet meadow were measured. Results of the study help to elucidate the effect of snowcover and mechanisms of snowmelt runoff generation on chemical constitu- ents strongly affected by biology and flowpath. DESCRIPTION OF THE STUDY SITE The experimental watershed is located in the Toiyabe National Forest on the east side of the Sierra Nevada, in Washoe County, Nevada, USA, and is approximately 3 km from Lake Tahoe (Figure 1). Elevations in the basin range from 2012 m at the stream outlet to approximately 2500 m at the upper divide. North-aspect slopes dominate the east- west trending watershed. Slope inclina- tion varies from 20 to 50 percent. A first-order stream drains the 0.80 km 2 watershed. The stream is fed by perennial springs and numerous ephemeral springs which flow during snowmelt and after heavy rain events. The flowing segments of stream channel expand and contract rapidly in response to snowmelt and drought. Snowmelt runoff is the dominant generator of streamflow. Peak 158 streamflow ranged from 3 1/s in late summer to 30 1/s during spring snowmelt. Mean annual precipitation in the area is estimated to be 76 em with approximately 85 percent falling as snow. The remainder of the precipitation usually falls in autumn rain events; summer storms are infrequent. Snowcover typically blankets the area from October to May. 0 Location of 111---+-t-Lake Tahoe and vicinity. 10 km CARSON CITY ~ STUDY AREA Figure 1. Location of the study area. Soils are comprised of coarse, decomposing granodiorite. Infiltration rates and saturated hydraulic conductiv- ity values in the soils are high (Rhodes, 1985). As in most forested, mountainous watersheds, overland flow rarely occurs except when the soil profile is complete- ly saturated. Soil depths vary from 4 m on ridge axes to 0.5 in the channels. Vegetation is a mosaic of riparian zones, wet and dry meadows, and old growth conifers. The watershed outlet is dominated by a wet meadow flanking the riparian zone. The riparian zones contain alder (Alnus tenuifolia), willows (Salix spp.), horsetail (Equisetum arvense), and ferns (Athyrium felin- femina). The upper slopes support virgin stands of ponderosa pine (Pinus ponder- ~), Jeffrey pine (Pinus jeffreyii), and red and white fir (Abies magnifica and concolor). Alder is known to be a nitrogen fixer. INSTRUMENTATION AND METHODS Hydrology Precipitation quantities were measured with a recording precipitation gage located in the lower wet meadow and augmented by a snow board network. Streamflow discharge at the watershed outlet was measured with a continuous stage recorder coupled with an 0. 5 m H-flume. Groundwater levels were moni- tored via piezometers constructed of 5 em ID PVC pipe screened at the terminus. Forty-six piezometers were located throughout the watershed with terminus depths ranging from 0. 3 to 4 m. Depth to groundwater in the piezometers was measured in a non-recording manner; ~asurement frequency ranged from every 4 hours diurnally to monthly dependent on instrument location and season. Water Chemistry Soil water samples were collected by means of suction cells attached to porous ceramic cups. Details of sampler con- struction, preparation and installation can be found in Linden (1977). Forty samplers were used; they were located in 37 sites at depths ranging from 0. 3 m to 3.3 m. The samplers were located in tandem with the piezometers. Measurement of groundwater levels during sampling allowed differentiation of samples into the vadose or groundwater regimes. The site was visited approximately twice weekly during the winter, and four times weekly during snowmelt. Samples of the stream chemistry at the outlet were taken during the visits and augmented by diurnal sampling during the melt events. Samples of precipitation were taken from snowboards within 24 hours of storm cessation. All soil water samplers were sampled monthly except during December, 1982, through February, 1983, due to instrument malfunction caused by the heavy snowpack. Prior to sampling, the soil water samplers were purged and a 159 vacuum was applied. Samples were taken from the soil water samplers within 48 hours of the application of a vacuum. Additional surface water samples were taken from the snowpack base, ephemeral channel segments, overland flow on saturated areas, and springs during the spring snowmelt. All water chemistry data presented is from sampling done October, 1982, through June, 1983. This period corresponds to the period of snowcover in the watershed during the year. All samples were collected in polyethylene bottles which had been rinsed with diluted HC1 and three sample aliquots prior to sample collection. Samples were immediately taken to the EPA-certified Water Lab of the Desert Research Institute for analysis. Samples were analyzed by ion chromatography. The method has been shown to give excellent agreement with liquid chemistry methods (Fishman and Pyen, 1979) and is accurate to + 0.001 mg/1 with a lower detection limit of 0.001 mg/1 N0 3 -N. Duplicate sample results were always _! 0.001 mg/1 N0 3 -N. Denitrification Denitrification was measured direct- ly in-situ in the lower wet meadow using acetylene blockage techniques (Ryden, et al., 1978; Ryden, et al., 1979a; Ryden, et al., 1979b) using a soil chamber system constructed as recommended by Denmead (1979). Six pairs of sampling sites were arranged in a circle approxi- mately 6 m in diameter and 8 m from the stream. Each sampling pair was sampled simultaneously during the sampling runs. Additionally, the following were also measured at the sampling area during each sampling run: 1) soil temperature at the 10, 25 and SO em depth with thermistor/ thermometers; 2) depth to groundwater using piezometers; and 3) soil moisture content at 15 em intervals to 2 m using a neutron soil moisture gage and aluminum access tubes. Displacement of N2o gas from the samples and analysis with gas chromatography was done based on the procedures outli~ed by Ryden, et al. (1978). Measurements were taken during the summer of 1983 and summer through autumn of 1984. RESULTS Hydrology Precipitation amount over the watershed area was 103.7 em of water equivalent. The snowpack was the great- est in over 100 years of record in the region for depth, water equivalent and persistence. Some snowcover remained in the watershed through June. Groundwater was the dominant genera- tor of streamflow both in time and space (Rhodes, et al., 1984). There was no indication of interflow; the dominant pathway of precipitation was through the vadose zone into groundwater and then to the stream (Rhodes, 1985). Soil tempera- tures at the 10 em depth ranged from 2 to 3.5°C from December 1982 through January, 1983. Groundwater levels and soil moisture remained high through mid- winter; they were maintained by snowmelt at the pack base generated by heat flow from the soil (Rhodes, 1985). During spring snowmelt, piezometer data indi- cated areas of profile saturation ex- panded and overland flow was observed. Overland flow contributions to streamflow increased during spring snowmelt and were greatest during daily peak runoff (Rhodes, 1985) • Water Chemistry Precipitation had a weighted mean concentration of 0.037 mg/1 NO -N. Groundwater had consistently low Nd 1 -N concentrations. Of 158 groundwater samples, 112 were below detection limits of 0.001 mg/1. Mean concentration in the groundwater was 0.001 mg/1 NO -N. The N0 3-N concentrations in the va~ose zone were highly variable temporally and areally, ranging from < 0.001 mg/1 to 1.06 mg/1 N0 3-N in 45 samples. Samples with high concentrations in the vadose zone were always restricted to areas with bare soils near the watershed boundaries (Rhodes, 1985). While alders are notori- ous nitrogen fixers, soil water samples from alder thickets did not have high N0 3-N concentrations. Eight samples of 160 soil water were taken from the alder thickets in April and May, 1983. Only two samples had concentrations above detection limits, they were 0.001 and 0.004 mg/1 N0 3 -N. Four samples of meltwater taken from the base of the snowpack gave a mean concentration of 0.093 N0 3-N, which is a reasonable enrichment given the meltwater fractionation of contaminants in the snowpack (Johannessen and Henriksen, 1978). Three grab samples of overland flow on saturated areas were taken in May 1983.. These overland flow samples had a mean concentration of 0.097 mg/1 N0 3-N. Sixteen samples from flowing stream segni.ents in the upper watershed yielded a mean concentration of 0.011 mg/1 N0 3 -N with a range of 0.002 to 0.026 mg/1 N0 3 -N. Concentrations in the stream at the outlet remained low except during snowmelt and rain events. Of 99 samples, 33 had concentrations > 0.001 N0 3-N mg/1. Peak concentration in the stream at the outlet was 0. 007 mg/ 1 during the peak snowmelt period on May 18, 1983. The nitrate budget for the period of investigation indicates uptake and/or removal of N0 3-N is highly efficient. The NO -N load to the watershed in precipiBation was 450 g/ha from Sept. 1982 to June 1983. During this period, 48 g/ha was exported from the watershed in streamflow (Rhodes, 1985). Actual retention or removal of N0 3-N within the system may actually be nigher because inputs of N0 3 -N via dryfall and plant fixation were not quantified. Also, inputs of other nitrogen forms were not quantified and these forms, such as N~y' NH 4-N, can be converted to N0 3-N nitrification. Hydrograph and Nitrograph Relationships From October, 1982, through March, 1983, streamflow generation was dominated by groundwater. Streamflow hydro graphs exhibited no peaks and N0 3-N concentra- tions were always < 0.001 mg/1. A rain-on-snow event penetrated the snowpack on March 11, 1983. The event raised initially high groundwater levels and generated a streamflow peak subse- quent to the development of profile saturation (Figures 2a and b). The stream was sampled diurnally during this event. The No 3 -N concentration in the stream peaked coincident with discharge and dropped rapidly afterwards (Figure 2a). 004 40 0 MARCH 1963 '"j 0 c Cal "' 003 30 l N03-N -g a -u c I "< ~ ·e ' ' It "' ~ II E w It 'I .<:: ; 002 ~ 20 It ' E " « It ~ 0 II :c z u II z (/) I I 2 Q i5 I I f! 001 I 10 :I I a: I u w a:: a_ 00 0 3 a::-wE (b) f---~('50 0<:{ zu:: :::>0:: o=> o::C!l <Do oz f--<:{ _J :r: f--:;:: D._Q wo:: au._ 111 12 13 14 15 16 17 18 20 TIME (days) Figure 2. (a) Storm hyetograph, stream- flow hydrograph and streamwater nitro- graph, (b) Piezometric data from inten- sively monitored site for March, 1983 rain on snow event • A pattern of elevated N0 3 -N concen- trations coincident with peak daily snowmelt discharge emerged with the onset of snowmelt. When snowmelt generated daily discharge peaks, N0 3 -N concentra- tions in the stream rose above the detection limits; during periods of groundwater-generated flow, there were no hydrograph peaks and N0 3 -N concentrations remained below detection limits (Figure 3). This pattern of N0 3-N concentration and stream discharge continued through May, 1983. With ongoing snowmelt, the extent of saturated area in the watershed expanded, concomitantly increasing both runoff and NO -N concentrations at the stream outlet ~y increasing the overland flow component of runoff. However, peak No 3-N concentrations became lagged about 10 hours behind discharge peaks once snowcover was removed from the stream channel and the lower wet meadow in mid-May (Figure 4) • Nitrograph and 161 hydrograph minima remained roughly coincident (Figure 4). The lagged nitrograph pattern is pervasive during peak snowmelt in streams in the Sierra Nevada (Coats, et al., 1976; Leonard, et al., 1979). .010 50 APRIL 1963 .008 'U 40 5l ' ... ... ' t5 30 .006 ~ a:: z ~ ' :r: (.) .004 ~"' U'J 20 0 ' 'N03 -N II '' 10 /I II .002 /lt 1 • • 0o 2 4 6 8 10 12 14 16 18 20 22 24 26 Figure 3. streamwater 50 'U 40 :!! TIME (days) Streamflow hydrograph and nitrograph, April, 1983. MAY 17-19, 1963 .008 ' ... 30 .006- ~ w <.!) a:: "' E ~ :r: .004 z 20 I "' ~ 0 0 z 10 .002 ~-L~~-L~~~~,8~~~-L~~~~7r,f o,7 TIME (days) Figure 4. Streamflow streamwater nitrograph sampling, May 17-19, snowmelt. Denitrification hydrograph and for diurnal 1983, at peak During the 1984 sampling season, 109 samples of N2o gas were taken from the meadow. Measured denitrification rates ranged from 0.14 to 2.94 g N0 3 -N/ha/hr with a mean of 1.26 g N0 3 -N/ha/hr (Green- lee, 1985). The rates had a strong diurnal variation (Greenlee, 1985). Rates peaked at about 13.30 PST and then dropped (Figure 5). The diurnal fluctua- tion of denitrification rates appears to be associated with soil temperature (Figure 5). Denitrification rates have been found to be highly correlated with soil temperatures by other researchers (Freney, et al., 1978; Ryden, et al., 1978; Ryden, et al., 1979a; Ryden, et al., 1979b; Denmead et al., 1979). Of all the variables measured, soil tempera- ture at the 10 em depth gave the best correlation with denitrification rates. However, the correlation of temperature with denitrification was not particularly strong with r = 0.66 at the 0.01 confi- dence level (Greenlee, 1985). Denitrifi- cation rates had little diurnal variation on days with minimal variation in soil temperature (Greenlee, 1985). -~ 2.0 0 ..c ........ z 1.0 z 0 o~~~~~--~~~~~~~~~ o6:oo 12:oo IS:oo TIME (hours) Figure 5. Denitrification rates (DN) for a 12 hour period on 9-18-84. The denitrification rates are averages for the two simultaneously sampled plots and are plotted in the middle of each sampling period. The temperature at 10 em below the soil surface (T10) is plotted for approximately each 1/2 hour period. DISCUSSION In light of the sampling results of water chemistry, the pattern of N0 3-N concentration with discharge during early spring snowmelt (Figure 3) is as expected. When only groundwater gener- ated the streamflow flow, stream concen- trations were similar to groundwater 162 concentrations. During the generation of hydrograph peaks a greater fraction of overland flow over saturated margins contributed to streamflow. Water reach- ing the stream by overland flow had higher NO -N concentrations because it bypassed e\e soil system, thus avoiding plant uptake and denitrification in the anaerobic subsurface environment. Due to the higher concentrations of N0 3-N in overland flow, in-stream N0 3 -N concentra- tions increase with increasing overland flow contributions and peak coincident with maximum discharge (Figure 3). This also explains why N0 3 -N concentrations increase ~ith the increasing snowmelt runoff; i.e., because saturated area within the watershed expanded. This conclusion is strengthened by the snow- melt and water chemistry data from the water year '83-'84. During that year the shallow snowpack melted and retreated rapidly. Overland flow was not observed, snowmelt peaks were not generated and N0 3-N concentrations in the stream never rose above the detection limits of 0.001 mg/1 N0 3-N (Melgin, 1985). The lagging of nitrograph peaks after the discharge peaks in mid-May (Figure 4) is initially perplexing. Groundwater and vadose water during this period had concentrations less than the peak value in the stream; they could not have accounted for the lagged peak. The lagging only occurred after the lower wet meadow and stream channel were snow-free, allowing light penetration and diurnal warming. The lagged nitrograph pattern is hypothesized to result from the operation of N0 3 -N removal during the day by biologic processes which are light- and temperature-sensitive and consequent- ly have a diurnal pattern. Three bio- logical processes are believed responsi- ble for the removal of nitrate from water in the watershed: anaerobic denitrifica- tion in wet meadows; uptake by vascular macrophytes; and uptake by in-stream periphyton. Denitrification is temperature- sensitive. While denitrification rates were not measured during the winter, the consistently low groundwater N0 3-N concentrations indicate denitrificat1on was active. Results of the denitrifica- tion rate measurements also indicate rates would have been steady because soil temperature varied little beneath the insulating snowcover (Rhodes, 1985). Upon partial snowcover removal, the upper 10 em of soil in exposed areas heated from 1 °C to rc from May 15 through 19, 1983 (Rhodes, 1985). Given the tempera- ture and denitrification rate results, it is likely the nitrification rates in- creased dramatically due to the tempera- ture increase during this period and that the rates had a diurnal fluctuation. The daily maximum denitrification rates in exposed areas probably occurred near solar noon as is the case with the rates measured during summer and autumn, 1984. Uptake by terrestrial vascular macrophytes largely coincides with evapotranspiration rates, hence it is both light-and temperature-sensitive. Peak rates coincide with solar noon. In the Sierra Nevada, plants typically exert their maximum seasonal uptake after removal of snowcover. Hence, plant uptake was probably exerting a maximum effect during period when the stream nitrographs began to exhibit lagging. N0 3 -N uptake by periphyton has been shown to be proportional to solar irradi- ance (Triska, et al., 1983) and exhibits a periodicity of activity similar to that of phytoplankton in len tic water (Toetz, 1976). Most past research into nitrate uptake by periphyton and diel patterns of N0 3-N concentrations has been conducted during summer baseflow conditions (Manny and Wetzel, 1973; McColl, 1974; Duff, et al. 1983; Triska, et al., 1983; Sebetich, et al., 1984). However, it should be possible to extrapolate the research results on periphyton N0 3 -N uptake to the study site because many periphyton species are well adapted to cold environ- ments (Hynes, 1970). Research has shown N0 3-N uptake by periphyton is at a ~ximum near solar noon and rapidly drops to a minimum with the onset of darkness (Triska, et al., 1983). Active periphyton communities have been shown to exert a powerful control on nitrate transport in streams (McColl, 1974; Duff, et a., 1983; Triska, et a., 1983; Sebetich, et al., 1984). Streams affect- ed by periphyton have NO -N concen- trations with a distinct dtel pattern; the maximum N0 3 -N concentration occurs near midnight and the minimum during the day (Triska, et al., 1983; Sebetich, et 163 al., 1984). N0 3 -N uptake appears to be greatest at low concentrations when the algae is nitrogen "starved" (McColl, 1974; Triska, et al., 1983; Sebetich, et al., 1984). Actively growing algal mats have higher uptake rates than mature mats (Duff, et al., 1983; Triska, et al. , 1983) • The algae have been shown to be opportunistic, rapidly increasing N0 3 -N upt~ke with increasing N~3 -N supply (Tr1ska, et al., 1983; Sebet1ch, et al., 1984). Periphyton communities in the stream were observed by early May, appearing within a few days of snowcover removal from the stream channel. Sluggish, diffuse streamflow and open canopy conditions in the riparian zone prior to leaf emergence provided an ideal location for periphyton colonization. Given the above review of past lite,rature, it is likely the algal mats exerted a pro- nounced effect on the nitrograph in mid-May. The communities were actively growing, shading was minimal, and the N0 3 -N/P ratio in the stream was always less than 1 indicating intense nitrogen starvation. In-stream uptake is likely. Chemistry-based hydrograph separation techniques (Pinder and Jones, 1968) and nitrate concentration data for overland flow and groundwater indicate the peak discharges would have been 99% ground- water generated. The estimate of the fraction of streamflow generated by groundwater is unrealistically high in light of the extent of saturated area and overland flow observed at the time (Rhodes, 1985). The hydrograph separa- tion technique assumes an absence of uptake mechanisms. The higher NO -N concentrations in the upper waters~ed also suggest NO -N removal along the stream length. ~owever, in the absence of accurate discharge measurements in these reaches, this evidence for N0 3 -N removal is inconclusive. All three biological removal mechan- isms would have a diurnal peak near solar noon, which corresponds to the time of peak discharge and the anticipated nitrograph peak if N0 3 -N removal were not occurring. The nitrograph without the operation of N0 3 -N removal is hypothe- sized to appear as in Figure 6. Thus, the timing the the observed peak appears to be due to diurnal mechanisms superimposed graph. NO -N removal on 3 the nitro- c-~ Actual N03-N observed 50 ........,. Hypothehcal concentrations w1th removal /uptake mechanisms dorrnant -Streamflow discharge .008 40 u 5l ' -< 30 z w (!) a:: .004 c5' <t I u (f) Ci 20 .002 10 oiL7 .!.-JL....!..~-'--L-L.__J_..L-J.........l.--,1'::-8 ....i.-J.._l.--'---'--J.__J~-'---'--'-;-;'.19 o TIME (days) Figure 6. Stream hydrograph, actual streamwater nitrograph and hypothetical nitrograph without uptake/removal mechan- isms. Nitrogen undergoes a wide variety of biological transformations. In order to conclusively use N0 3-N as flow path indicator, all forms of nitrogen in the nitrogen pool must be monitored. Plant fixation of N2 and subsequent nitrifica- tion can contribute N0 3-N to streamflow. However, our soil water data do not indicate such subsurface sources. All forms of nitrogen were not monitored. However, our combined hydrologic, chemi- stry and denitrification rate data strongly suggest that the lagging of peak NO -N concentrations in streams behind daily discharge peaks is due to diurnally ;arying N0 3-N removal and uptake mechan- 1sms. The nitrate removal mechanisms would have been nearly dormant under the snowpack. Periphyton were probably absent due to the exclusion of sunlight by the snowcover. Certainly diurnal fluctuations would have been essentially non-existent. Thus, an explanation is provided for the absence of lagging of peak nitrate concentrations earlier in the season when snowcover was complete. CONCLUSIONS Runoff mechanisms obviously exert a strong control on stream chemistry, especially in the case of reactive z 164 constituents. Such control is evident in seasonal and diurnal patterns of stream flow and its NO -N concentrations. Beneath a snowpack ~iological mediation of water chemistry is minimized by cooling the system and excluding light. Diel biological activity is steadied by the snowcover shielding the sun and insulating the soil system. Under snowcover, diurnal patterns of nitrate concentration are mainly a function of the type of mechanism generating runoff. However, with partial removal of snow- cover within a watershed, light-and temperature-sensitive biological pro- cesses can alter the concentration patterns in streamflow generated solely by runoff mechanisms. Thus, on both the hydrologic and biological levels, the presence of snowcover exerts a pronounced influence on N0 3-N concentrations which shift seasonally. Other chemical con- stituents which undergo biological reactions may be affected in a similar manner. In most areas, No 3-N concentrations in streamflow increase with increasing discharge. In mountainous areas, peak annual discharge occurs during spring snowmelt. Streams in the Sierra Nevada export the major part of the annual N0 3-N load during spring snowmelt (Leonard, et al., 1979; Rhodes, 1985). Denitrifica- tion and N0 3-N uptake by periphyton can greatly affect N0_1 -N concentrations in streamflow and tfieir distribution in time. Presently, data on these processes during complete snowcover and spring snowmelt is lacking. Considering the importance of the snowmelt period in annual hydrologic and chemical budgets of mountainous areas, research needs to be conducted to investigate the role of these N0 3-N removal mechanisms in shaping snowmelt runoff chemistry. ACKNOWLEDGMENTS This study was supported in part by a grant from the Office of Water Research and Technology, U.S. Department of Technology, and in part by a grant from the Mcintyre-Stennis program of the U.S. Department of Agriculture. LITERATURE CITED Coats, R.N., R. L. Leonard, and C. R. Goldman, 1976. Nitrogen Uptake and Release in a Forested Watershed, Lake Tahoe Basin, California. Ecology 51:995-1004. Denmead, O.T. 1979. Chamber Systems for Measuring Nitrous Oxide Emission from Soils in the Field. Soil Science of America Journal 43:89-95. Denmead , 0 • T. , J • Simpson. 1979. Oxide Emission Soil Science 43:726-728. R. Freney, and J. R. Studies of Nitrous from a Grass Sward. of America Journal Dunne, T. and R. D. Black. 1971. Runoff Processes During Snowmelt. Water Re- sources Research 7:1160-1172. Dunne, T. 1978. Field Studies of Hill- slope Flow Processes. In: Hills lope Hydrology, M. J. Kirkby (Editor). John Wiley & Sons, New York, New York, pp. 227-293. Duff, J. H., K. C. Stanley, F. J. Triska, and J. A. Avanzino. 1983. The Use of Photosynthesis-Respiration Chambers to Measure Nitrogen Flux in Epilithic Algal Communities. Verh. Int. Ver. Limnol. 22:1436-1443. 'ishman, M. J. and G. Pyen. 1979. Determination of Selected Anions in Water by Ion Chromatography. In: Proceedings of the 21st Rocky Mountain Conference on Analytical Chemistry, July 30, 1979, Denver, Colorado. Freney, J. R. , 0. T. Denmead, and J. R. Simpson. 1978. Soil as a Source or Sink for Atmospheric Nitrous Oxide. Nature 273:530-532. Goldman, C. R. 1985. Lake Tahoe: In- creasing Fertility and Decreasing Transparency During a Quarter-Center of take Basin Development. Eos 66:114. Greenlee, D. L. 1985. Denitrification Rates of a Mountain Meadow Near Lake Tahoe. M.S. Thesis, University of Nevada Reno, Reno, Nevada. 165 Hynes, H. B. N., Running Waters. Press, Toronto. 1970. The Ecology of University of Toronto Johannessen, J. and A. Henriksen. 1978. Chemistry of Snow Meltwater: Changes in Concentration During Melting. Water Resources Research 14:615-619. Leonard, R. L., L. A. Kaplan, J. F. Edler, R. N. Coats and C. R. Goldman. 1979. Nutrient Transport in Surface Runoff from a Subalpine Watershed, Lake Tahoe Basin, California. Ecolo- gical Monographs 49:281-310. Linden, D. R. 1977. Design, Installation and Use of Porous Ceramic Samplers for Monitoring Soil Water Quality. U.S.D.A. Technical Bulletin 562 10 pp. Manny, B. A. and R. G. Wetzel. 1973. Diurnal Changes in Dissolved Organic and Inorganic Carbon and Nitrogen in a Headwater Stream. Freshwater Biology 3:31-43. McColl, R. H. A. 1974. Self-Purification of Small Freshwater Streams: Phos- phate, Nitrate and Ammonia Removal. New Zealand Journal of Marine and Freshwater Research 8:375-388. Melgin, W. 1985. The Influence of Hillslope Hydrology on Nitrate Trans- port in a Forested Watershed Near Lake Tahoe. M.S. Thesis, University of Nevada Reno, Reno, Nevada. Pinder, G. F. and J. F. Jones. 1968. Determination of the Ground-water Component of Peak Discharge from the Chemistry of Total Runoff. Water Resources Research 5:438-445. Price, A. G. and L. K. Hendrie. 1983. Water Motion in a Deciduous Forest During Snowmelt. Journal of Hydro- logy 64:339-356. Rhodes, J. J., C. M. Skau, and w. Melgin. 1984. Nitrate-Nitrogen Flux in a Forested Watershed--Lake Tahoe, USA. In: Recent Investigations in the Zone of Aeration. Proceedings of the International Symposium, Munich, West Germany, October 1984, P. Udluft, B. Merkel, and K. H. Prosl (Editors), pp. 671-680. Rhodes, J. J. 1985. A Reconnaissance of Hydrologic Nitrate Transport in an Undisturbed Watershed Near Lake Tahoe. M.S. Thesis, University of Nevada Reno, Reno, Nevada. Ryden, J. C., L. J. Lund, and D. D. Focht. 1978. Direct In-field Measure- ment of Nitrous Oxide Flux from Soils. Soil Science of America Journal 42:731-737. Ryden, J. C., L. J. Lund, and D. D. Focht. 1979a. Direct Measurement of Denitrification Loss from Soils: I. Laboratory Evaluation of Acetylene Inhibition of Nitrous Oxide Reduction. Soil Science of America Journal 43: 104-110. Ryden, J. C., L. J. Lund, J. Letey, and D. D. Focht. 1979b. Direct Measure- ment of Denitrification Loss from Soils: II. Development and Application of Field Methods. Soil Science of America Journal 43:110-118 Sebetich, M. J., V. C. Kennedy, S.M. Zand, R. J. Avanzino, and G. W. Zellweger. 1984. Dynamics of Added Nitrate and Phosphate Compared in a Northern California Woodland Stream. Water Resources Bulletin 20:93-101. Toetz, D. W. 1976. Diel Periodicity in Uptake of Nitrate and Nitrite by Reservoir Phytoplankton. Hydrobiolo- gia 49:49-52. Triska, F. J., V. C. Kennedy, R. J. Avanzino, and B. N. Reilly. 1983. Effect of Simulated Canopy Cover on Regulatjon of Nitrate Uptake and Primary Production by Natural Peri- phyton Assemblages. In: Dynamics of Lotic Ecosystems, T. D. Fontaine and S. M. Bartell (Editors). Ann Arbor Science, Ann Arbor, Michigan, pp. 129-159. 166 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 RESERVOIR WATER QUALITY SIMULATION IN COLD REGIONS C.Y. Wei and P.F. Hamblinl ABSTRACT: The requirements for environ- mental impact analysis associated with the license application of the proposed ~usitna Hydroelectric Project have moti- 'ated a program of data collection, development and testing of a reservoir simulation model in two Alaskan basins. The model developed for simulating the temperature and suspended sediment dis- tributions and ice and snow cover in lakes and reservoirs is based on the dynamic reservoir simulation model called DYRESM. The model was first applied to a known water body, Eklutna Lake, which had a number of properties in common with the proposed reservoirs. Comparison of simu- lations with field observations are pre- sented. Excellent results were obtained and the general applicability of the model to the proposed reservoirs in a subarctic region is demonstrated. The model was then applied to the proposed Watana and Devil Canyon reservoirs. The influence of water quality requirements for downstream temperatures and suspended sediments on the reservoir behavior are discussed. Finally, the model's capability to simu- late operation of the proposed J?ltllti-level intake structures under different meteor- ological and project operating conditions is demonstrated. The effect of the oper- ating conditions on two water quality parameters, temperature and suspended sediment concentration, provides informa-' tion required to evaluate the influence of the project operations on fishery resources. The predicted ice-cover thick- ness served as input to terrestrial wild- life management planning. (KEY TERMS: Eklu t na Lake; glacially fed lake; hydroelectric project; multi-level intake; reservoir simulation, stratifica- tion; subarctic Susitna basin; suspended sediment concentration; temperature and ice.) INTRODUCTION The Susitna Hydroelectric Project has been proposed by the Alaska Power Author:... ity in order to provide a dependable and economical source of energy for the Alaska Railbelt area. The project would consist of two dams, a 197 m (646 ft) high con- crete arch dam at the Devil Canyon site and a 270 m (885 ft) high earth and rock- fill dam at the Watana site located on the Susitna River about midway between Anchor- age and Fairbanks (see Figure 1) the prin- cipal load centers of the Railbelt region. It would be built in three stages and would ultimately provide 6,900 Gwh/year of energy with a firm capacity of 1,620 MW. The project and the basin are described in more detail by Gemperline (1986). !Respectively, Senior Hydraulic Engineer,Harza Engr. Co., 150 S. Wacker Dr., Chicago, IL 60606; and Research Scientist, National Water Research Institute, Canada Center for In- land Waters, 867 Lakeshore Road, Burlington, Ontario L7R 4A6. 167 j I 0 40 MILES SCALE.~ 0 80 KILOMETERS Figure 1. Location of the Proposed Susitna Hydroelectric Project and Eklutna Lake. The potential effects of the Susitna Hydroelectric Project on the fishery resources of the Susi tna River are an important concern in the development of the project. The potential effects in- clude possible changes in both the quan- tity and quality of water in the Susitna River with respect to the fish and other wildlife habitats. The potential changes in river flows, water temperature, sus- pended sediment concentration, and ice conditions must be evaluated carefully in an environmental impact sutdy associated with the license application of the project. Such concerns and needs have motivated a program to collect data at Eklutna Lake and in the Susitna Basin, to develop and test the DYRE~~ reservoir model and to apply the model to predict the temperature and suspended sediment regimes in the proposed Watana and Devil Canyon reservoirs. The DYRESM reservoir temperature simulation model has been extended and enhanced by adding capabili- ties to simulate suspended sediment con- 168 centrations in the reservoir, the winter ice-cover formation, and the operation of a multi-level intake for applications to the Susitna Project. Testing of the model with the Eklutna Lake data showed good agreement between predicted and observed outflow temperatures and winter ice-cover thicknesses. Following successful tests of the DYRE~ model, the model was applied to predict the temperature regimes of the proposed Watana and Devil Canyon reservoirs. The selective withdrawal capability of the proposed multi-level intake structures was simulated for different project operating conditions. To assist in the determina- tion of the outflow turbidity the extended DYRESM model was applied to simulate the suspended sediment concentrations in the proposed reservoirs and the Eklutna Lake data collection program was expanded to provide sediment data for testing the model. Satisfactory results were obtained on the suspended sediment concentrations of the outflow. Further refinements of the anlaysis are also needed to improve the simulation of supended sediment regime of the lake. THE DYNA}liC RESERVOIR SIMULATION MODEL The dynamic reservoir water quality simulation model, DYRESH, was originally developed by Imberger and Patterson (1981) and has been modified and enhanced by Harza-Ebasco Susi tna Joint Venture to include simulations of multi-level intake operations, frazi 1 ice inflow, and sus- pended sediment concentration. A snow. and ice-cover algorithm as developed by Patterson and Hamblin was also incor- porated in the DYRE~~ model. The basic features of the model are summarized as follows. In the formulation of the modelling strategy of the model, the principal phys- ical processes responsible for the mixing of heat and other water quality components are parameterized. This approach is in contrast to other simulation models which are largely empirically based. While the modelling philosophy employed in DYRESM requires a reasonable understanding of the key processes controlling water quality, so that they may be parameterized correct- ly, this process related approach to modelling has the advantage that the resulting model may require less cali bra- tion and is more generally applicable than the empirically based methods. A second major consideration in model formulation has been to keep the computational over- head as low as possible in order to keep the running costs of the simulation of a number of variables over time periods of .· up to three years within reason. Thus, · the basic time step of the model is one day although sub-daily time steps as short as one-quarter hour are allowed when required for model stability and accuracy. The influence of ice and snow on the heat transfer across the water surface of a reservoir is taken into account by cal- culating the area of snow and ice cover and their thicknesses as a function of time. The effect of snow and ice cover on the heat transfer is to reduce the amount of short wave radiation reaching the upper layers of the reservoir through the re- flective and absorptive properties of ice and snow and to reduce the cooling of the reservoir surf ace that would otherwise occur by providing a covering layer of reduced thermal conductivity. The reflec- tive properties of the snow cover on short wave radiation are varied in time accord- ing to well established empirical rela- tions involving snow age and snow tempera- ture. A novel feature of the ice cover simulation model is the turbulent trans- port of heat from the water to the ice due to the under-ice flow field generated by i~lows and outflows (Patterson and Hamblin, 1986). Additionally, since snow thickness is li)lli ted by the bearing capac- ity of the ice cover, snowfall beyond this limit is added to the water surface. The input of a volume of frazil ice is provid- ed as a percentage of the daily inflow. ·The mass of ice contained in the inflow is computed and added to the existing ice thickness in both the cases of full or partial ice. In the latter case the frazil ice input is added to the ice cover until full ice cover is reached. The horizontally averaged profiles of suspended sediment are changed in the model by three processes, namely by ver- tical mixing, by inflows and outflows and 169 by settling. The supended seiment concen- trations are solved based on a sediment transport equation and the water density calculations include the contributions from the suspended sediment as well as temperature. Density inversions resulting from suspended sediments are checked and mixed if required to establish a stable density distribution. Settling velocities are input appropriate to the size range of the inflowing sediments. The sediments are deposited on the bot tom as they reach the bottom layer of the reservoir. The effective difusivities of the sediment fractions are ass~ed to be identical with that of temperture. APPLICATION TO EKLUTNA LAKE To facilitate the testing of the DYRESM model on a existing glacially fed lake, Eklutna Lake was selected for an intensive field study (R&M Consultants, 198S(a)& (b)). Eklutna Lake, shown in Figure 2 is located approximately 48 kilometers (30 miles) northeast of Anchorage and 160 kilometers ( 100 miles) south of the project site. The purpose of testing the DYRESM model was to demonstrate the appli- cability of the model to the proposed reservoirs. Both the proposed reservoirs and Eklutna Lake are located in the south- central region of Alaska. Eklutna Lake is also operated for hydroelectric production and has a similar average residence time of 1.65 years to that of the Watana reser- voir (1.77 years). The testing of the DYRESH model was performed in two phases. In the initial phase, the basic DYRESM model for tempera- ture and ice simulation was tested. In the second phase, the suspended sediment option was added to the model and the testing of the model was conducted in conjunction with the expanded sediment sampling and turbidity data collection program. A meteorological and limnologi cal data collection program was initiated in June 1982 which was designed specifically for model verification purposes. A major component of the field study was a weather station established at the south-east end of the lake. Besides the pertinent mete- ' ' LEGEND • Observation Station A Weather Station 9 Stream Gauge 0 1 Km ~ SCALE: 0 1 Mile I Figure 2. Location of the Observation Stations, Eklutna Lake orological data such as solar radiation, air temperature, relative humidity, and wind speeds etc, these data also include daily inflow and outflow discharges and their temperatures, and suspended sediment concentrations as well as approximately monthly to semi -monthly surveys of lake temperature and turbidity profiles. In addition, occasional ice thicknesses were measured. E w r-- 1;! J 1982 Testing of the DYRESM Model: Temperature and Ice The DYRESM model was applied to simulate the average temperature distribution in F.klutna Lake for the period starting June 1, 1982 and ending May 30, 1983. ~ analysis of the initial results (Harza- Ebasco Susitna Joint Venture, 1984) led to several improvements of the model. These improvements included application of the Anderson incor.rl ng long-wave radiation equation instead of the Swinbank formula during the sub-freezing conditions (Tennessee Valley Authority, 1972) and incorporation of intake design and wind effects ort temporal thermocline displace- ment in estimation of outflow tempera- tures. The simulated and observed outflow tem- peratures are shown in Figure 3. In gen- eral, the differences between the pre- dicted and observed winter outflow tem- peratures were within 0.5°C (1 °F). How- ever, stnnmer deviations of up to about +2 .0°C (3 .5°F) also occurred, especially during and after high wind periods. The surf ace wind shear effects and the inter- nal wave motions near the intake struc- ture are extremely difficult to model with a one-dimensional approach and three- dimensional modeling is not considered practical. The results also show an excellent cor- respondence between measured ice thickness and predicted ice thickness, being within the standard error of the observatiom except in March. The accuracy of the March prediction is difficult to evaluate., The only ice measurement available was made at Station 13 which is located near' the north end of the lake (Figure 2). The M 1983 Figure 3. Observed and Simulated Eklutna Lake Ice Thicknesses and Outflow Temperatures. 170 relatively thick ice measured was probably due to local accumulation of rafted ice caused by persistent downlake winds. In addition to the outflow temperatures, the observed and simulated time history of t~perature versus depth are shown in Figures 4(a)&(b). The general character- istics of the observed temperature regime are simulated reasonably well by the DYRESM model. The results of the study demonstrate the applicability of the DYRESH model to simulate the hydrothermal behavior of a reservoir in the specific region of the Susitna Project. The study has also demonstrated the need for accurate cli- matic data to apply the model properly. The accuracy and reliability of field measurement instruments and data collec- tion procedures must be maintained for the results to be useful. Testing of the Extended DYRESM Model: Suspended Sediments The extended DYRESM model was tested using the Eklutna Lake data to determine 70 ~ 60 l I 0 \ ~ 50 \ 0 \ m 40 7 6 ~·c 4 'V \ ' w I / > / 0 30 \ I m ~ 6 \ 3.~ ( I \ " 20 \ \ I I I 0 j. .. ,. I iii 10 \ I I R&M 1985(a) \ I I 0 M J J A s 0 N D J F 1982 1983 (a) OBSERVED (Station 9) 70 ~ 60 l 0 ~ 50 0 m 40 w > ' g 30 \ ( \ ;: 20 4 0 \ iii 10 I M J J A S 0 N D J F 1982 1983 (b) SIMULATED Figure 4. Observed (a) and Simulated (b) Isotherm Distribu- tion in Eklutna Lake. 171 its ability to simulate suspended sediment concentration of the outflows from the Susitna project. The hydrological and meteorological data collection program was continued (R&M Consultants, 1985(a)&(b), and Coffin and Ashton, 1986) with special emphasis on suspended sediment sampling for the period from May to November 1984. The total incoming suspended sediments were measured twice weekly. An empirical relation was established using these observed semi -weekly inflow suspended sediments and the corresponding dis- charges. This relationship was applied to estimate the daily total inflow suspended sediment concentration. These suspended sediment concentrations ranged from 0.15 to 570 mg/1 in the inflow streams, and from 0.50 to 36 mg/1 in the outflow. Peak values in the inflow occurred in late July or early August, and in the outflow in late July to mid-August. The winter in- flow and outflow suspended sediment con- centrations were on the order of 0.1 mg/1. The total suspended sediment influent from both Glacier Fork and East Fork streams (Figure 2) were first divided into three groups representing three different particle size ranges. The particle size ranges selected were 0-3 microns, 3-10 microns and greater than 10 microns. Initial test runs indicated that particles greater than 10 microns would settle rapidly to the bottom of the lake and have little effect on the average suspended sediment concentration profiles. The greater than 10 micron sediments were therefore ignored in the study. The total incoming suspended sediments of each part- icle size range were then estimated based on the weighted particle size distribu- tions determined from the samples taken from East Fork and Glacier Fork streams. These samples were obtained in three field trips made on July 21, August 28, and October 23, 1984. The daily particle size distributions were interpolated from these three basic distributions. To apply the extended DYRESM model, it was necessary to specify an initial vertical distribution of suspended sediment, the particle set- tling velocity, and the average density of the particle. Based on Stoke's law, a settling velocity of 1.5 x 10-6 m/s was used for the 0-3 micron sediments and 4.0 x 10-5 m/s for the 3-10 micron sediments. A particle specific gravity of 2.60 was used in the study while the measured specific gravity varied from 2.50 to 3.00. The DYRESM simulations were made for Q-3 micron sediments and 3-10 micron sediments separately. The resulting outflow sus- pended sediments of these two separate analyses were then combined to yield the total outflow suspended sediment concen- trations as shown in Figure 5. The pre- dicted outflow suspended sediment concen- trations were in reasonable agreement with data obtained from the powerhouse tailrace especially considering the coarse resolu- tions of the inflow concentrations. On two occasions, the field data show tempo- rary increases in suspended sediment con- centrations that were not predicted by the DYRESM model. Prior to these events, there were relatively heavy rains. A small stream which flows into the lake near the intake may have carried signi- ficant amount of sediments and caused the suspended sediment concentration to in- crease locally and temporarily at the intake area. The suspended sediment influents were, later in a test study, neglected and the simulation was repeated. The resulting temperature distributions showed the in- fluence of the suspended sediments to be insignificant. Hence, the suspended sedi- ment inputs to a reservoir may be ignored in most cases in this region when only the temperature analysis is needed. APPLICATION TO SUSITNA RESERVOIRS Simulation of Temperature and Ice Following the successful testing of the DYRESM model with the Eklutna Lake data, the model was applied to determine the temperature regimes in the proposed \Vatana and Devil Canyon reservoirs. Fifteen years of hydrologic and meteo- rological data have been assembled and analyzed. The data collected at the Hatana and Devil Canyon weather stations since 1982 (R&M Consultants, 1985(c)) were incorporated. In this paper, the hydro- thermal and ice regimes of both reservoirs analyzed for the Case E-VI flow require- ments are discussed. The Case E-VI flow requirements are considered as an optimal operating condition with respect to the 172 ' ~ 150 :i 0 0 ci ~ 100 c w c z w ... ~ 50 rJ) 0 N ¢> Outflow (Observed) --Outflow (Simula1ed) -----Inflow (Observed) 0 J F M 1983 ~-/~-,t A M ~ '" I' I o '' '' I o ''' ~ I I •" •" ,., ~ n: -··' 11111 It I ~ '•' '•' '•' '•' :·: i ~ ',, '' ,, ' I~ : ... ;u ' I •' ,, •' ,, '' ': I : : ' I : \ r. \r-· \:: ,. ::: ~ I: I I,' ,, ' ~· \ 0 ' Figure 5. Observed and Simulated Outflow Suspended Sedi· ml!nt Concentrations from Eklutna Lake. energy and instream The 1981-1982 inflow flow requirements., and treteorological conditions, which represent average to wet' year conditions were used in the analysis., The energy demands considered included energy demands predicted for all three• stages of project operation. Each study, case was simulated for a period of two years starting in November. The completed l~atana reservoir would have a maximum depth of about 220 m (720 ft) with a total volume of about 1.16 x• 1ol 0 m3 (9 .4 x 106 acre-ft) The Watana dam would be built to an intermediate height of 214 m (702 ft) above foundation in. Stage I (scheduled for operation from 1999 to 2004) and raised to the final height in' Stage III (scheduled for operation from 2012 onward). The Watana reservoir would have a maximum depth of about 165 m (540 ft) and total volume of 5. 24 x 109 ml (4 .25 x 106 acre-ft) in the first and second stages. The Devil Canyon reservoir would be completed in Stage II (scheduled for operation from 2005 to 2011) and the maximum depth would be about the same as that of Stage I Watana reservoir. How- ever, the total volume of the Devil Canyon reservoir would be about one-quarter of the Watana reservoir in Stage I or II and about one-tenth of the completed Watana reservoir. The Watana control structure which would be located on the right abutment of the dam at about 46 m ( 150 ft) below the nor- mal maximum reservoir level would include a multi-level powerhouse intake and a single-level outlet works (cone-valves) intake near the bottom of the structure. The two-level powerhouse intake of the Devil Canyon reservoir would also be located near the normal maximum reservoir level, however, the outlet works intakes would be located at the center of the dam, about 168 m (550 ft) below the normal maximum reservoir level. The outlet works are provided to release flow, (a) to meet envi rornnental minimum flow requirements when powerhouse flows are not adequate, and, (b) to release flood flows when the reservoir( s) are full and the powerhouse does not have the capability to pass the inflow. The outlet works are normally operated in July and August. Two intake operation policies were considered, name- ly, inflow temperature matching and warm- est possible release. A reservoir opera- tion study (Harza-Ebasco Susitna Joint Venture, 1985 (a)) was carried out to determine the inflow, outflow, and reser- voir level of each reservoir for various project stage, energy demand, and the d~nstream flow requirements. The temperature and ice simulations thus performed using the DYRESM model indicate that, regardless of the project operation schemes and project status (stages), both reservoirs would develop stratification in the stnnmer months of June, July, August, and September. Overturns would occur in spring and fall followed by the formation of ice-cover in winter. At Watana, the general aspects of the t~perature behavior are similar to Eklutna Lake. The ice-cover would form in November and a total meltout would occur inMay. A maximum ice thickness of 1.0 to 1.5 meters can be expected in March (Figures 6 & 8). The multi-level intake structures pro- posed for the Watana reservoir for dif- ferent project stages provide the project capability to release water selectively from various levels of the "stratified water body in the reservoir. The llatana s~mer release temperatures can be approx- imately controlled to satisfy a predeter- mined objective. In the summer, the river inflows are more responsive to variations in the meteorological conditions than the reservoir due to the shallowness of the river. The river inflow warms up in the early summer and cools down in the late s~ner more rapidly than does the reser- voir. Hence, the Watana discharge water would be colder in the early summer and 173 ~] ~'] 1 ul 2 > 3 ~ 4 -II Ill 11.11 I II -~· ••• I -~-...... (Top Level) I -5 • Ill ·-• -(Outlet Works) WATANA -I (Outlet Works) 15r-~---r--,---r--.---.--.---.--..--r--.---. ~ w a: 10 :I 1-< a: w 5 a. :::E w 1- ', ' ', --WATANA ----DEVIL CANYON --------Inflow ' ~~~=~~--~~--~ OM J J AS 0 1981 J F M A 1982 Figure 6. Simulated Susitna Project Intake Operations, Ice Thicknesses and Outflow Themperatures- lnflow Temperature Matching (Stage II). 1 ...I 2 w 3 ~ 4 ...I 5 w ~ < 1- ~ So> u.i 15 ~ 10 1-< a: w 5 a. :::E w 1- 0 ---• M J (Top Level) l--End of Simulation (Outlet Works) WATANA (Outlet Works) I--End of Simulation --WATANA ----DEVIL CANYON --------Inflow J A SO NO J FMA 1981 1982 Figure 7. Simulated Susitna Project Intake Operations and Outflow Temperatures-Warmest Possible Release (Stage II). warmer in the early fall than the natural river conditions. However, in most of the summer months (June, July and August) the Watana discharge temperatures can be regu- lated to approximate inflow temperatures through operation of the multi-level intake except in times when large releases are made through the outlet works (Figures ~'] fl -0- ~ 1~~~~~~~~~ .W~-~~~~~~~~~(T~o~p~L~e~v~el~)~ W 2 •• II i::j 3 -•• •11 I ~ 4. -1 I •• -• w ~ 10 1-< a: ~ 5 ::E w 1- -(Outlet Works) WATANA --WATANA ----DEVIL CANYON --------Inflow Figure 8. Simulated Susitna Project Intake Operations, Ice Thicknesses and Outflow Temperatures- Inflow Temperature Matching (Stage Ill). 6 & 8). In the winter, inflow tempera- tures -would be near 0°C (32°F) and the reservoir temperatures in the reversed stratification zone would range from near 0°C (32°F) at the contact surface with the ice-cover to approximately 4°C (39°F) at the top of the hypolimnion. Therefore, the Watana discharge temperatures would be slightly warmer during the winter than natural river conditions. As a ·result, the discharge temperatures would range from approximately 5 to l2°C (41 to 54 °F) in the summer and approximately 0.5 to 3°C (33 to 3rF) in the winter depending on the project and meteorological conditions, and energy demand level. When the Devil Canyon reservoir is com- pleted, it will receive inflows from the Watana reservoir and the tributaries down- stream of the Watana dam. Hence, the Devil Canyon main inflow would be cooler in early summer and warmer in early fall than the natural project inflows. In addition, relatively large summer dis- charges releasing water through deep low- level outlet works and the smaller size of the reservoir relative to Watana would reduce the summer average residence time and result in a larger variation in the 174 temperature of the hypolimnion than in Wa tana. The hydrothermal regime of the Devi 1 Canyon reservoir especially in the hypolimnion would therefore, be more sen- sitive to the operation of the outlet works than Watana. The temperature in the hypolimnion would vary from 4 to l0°C (39 to 50°F) in the summer. With relatively shallow warmer surface layer (epilimnion), the operation of the Devil Canyon intake would be less effi- cient in terms of selective withdrawal than Watana. Comparison of the Watana and Devil Canyon outflow tanperatures are shown in Figures 6-8. In Figure 6, out- flow temperatures predicted for Stage II conditions are shown. The operation of the powerhouse intake structures was simu- lated to match the release temperature with the natural inflow temperature. The summer outlet works operations and the relatively thin epilimnion reduced the effectiveness of the selective withdrawal using the nulti-level intakes in both reservoirs. Similar effects were noted when releasing warmest possible water from the Devil Canyon reservoir in Stage II as demonstrated in Figure 7. In mid-June of 1981, the De vi 1 Canyon intake releases were changed from the top-level ports to the bottom level ports due to decreased reservoir level and caused a 4°C (7 °F) reduction in outflow temperature while the Watana release temperature rose steadily. Thus, the Devil Canyon releases can be up to 5°C (9°F) colder than the Watana releases in June and July. The effective· ness of the intakes is improved when the project is fully developed, the reservoir levels are more stable and the operations of the outlet works are less frequent in Stage III as shown in Figure 8. Effect of Intake Operation on \-linter Release Temperture The ice-cover formation in a reservoir is strongly dependent upon the meteorolog- ical conditions prior to the surface freeze-up. At the fall overturn, the reservoir destratifies and becomes iso- thermal with a relatively uniform vertical temperature distribution. Mixing and further cooling would continue in a cold climate until the surface of the reservoir freezes. The presence of ice-cover pre- vents further mixing and hence conserves the heat remaining in the reservoir. By changing the operating policy of the multi-level intake, the amount of heat stored in the reservoir prior to the freeze-up can be altered, to some extent, and hence the timing of the reservoir surface freeze-up and the subsequent win- ter release temperature can also be modi- fied. As an example, by releasing warmest possible water from the Watana reservoir in the summer less heat would be preserved in the water body in the period prior to freeze-up and the surface water tempera- ture would be reduced to the freezing point sooner. Hence a freeze-up of the surface water my be induced, in some cases, about two weeks earlier than in the case of operating the intake by matching the inflow temperatures. With ice-cover formed two weeks sooner, more heat would be preserved for the remaining winter. Therefore, an increase of the outflow temperature of up to about 1 °C (2 °F) may be obtained. Simulation of the Suspended Sediment Concentration In addition to the outflow temperature of a reservoir, the turbidity level of the reservoir releases may also have an impact on downstream fisheries. Because of the 'nature of inflows to the gracially fed Eklutna Lake, the turbidity is mainly , derived from the suspension of the glacial flour. To provide basic information for further assessment of the turbidity effects, the extended DYRESM model was applied to predict the suspended sediment concentrations of the Watana and Devil Canyon reservoir au tf lows. Based on the Eklutna Lake tailrace sediment ·data, the suspended sediments of the Wa tana reser- voir outflow are expected to be comprised primarily of particles of size less than 3-4 microns. Larger size particles would generally settle out rapidly to the bottom without significantly affecting the aver- age concentration levels in the reservoir and outflows. In the study, only sedi- ments of up to 10 microns were analyzed and the solution procedure applied is 175 identical to that used in the Eklutna Lake study. The total sediment influent to the Watana reservoir was estimated fran the USGS data at Gold Creek gaging station based on the drainage areas. The average particle size distribution curve of the river suspended sediments available at the project area from the Cantwell station was used to determine the suspended sediment influent of each sediment group from the total suspended sediment influent. Fif- teen percent of the total suspended sedi- ment influent was assigned to 0-3 micron sediments and 12 percent to the 3-10 micron sediments (Harza-Ebasco Susitna Joint Venture, 1985(b)). The 1970, 1981, and 1982 flow conditions which represent the low sediment influent, high sediment influent, and average sediment influent years, respectively, were considered. The operation of the multi-level intakes was simulated to withdraw the near surface water since it allCMs for withdrawal of water with the lowest level of suspended sediment concentration. Figure 9 shCMs the predicted project outflow suspended sediment concentration from the Devil Canyon reservoir for the Stage III condi- tion. These analyses indicate that the suspended sediment concentration level of the summer release flows from the project would be decreased from the pre-project condition of about 60 to 3000 mg/1 to about 50 to 200 mg/1. In the winter, the suspended sediment concentration level would be increased from a range of 1 to 80 mg/1 to a range of 10 to 100 mg/1. :::: 0 E. 400 z 0 0 300 ci w (J) 0 200 w 0 z ~ 100 --Outflow ------Inflow (J) ~---- ::J (J) 0 N D 1981 J A S 0 Figure 9. Simulated Susitna Project Outflow Suspended Sediment Concentrations (Stage Ill). CONCLUSIONS A dynamic reservoir simulation model suitable for water quality simulation in environmentally sensitive northern regions has been briefly outlined and applied to two Alaskan Basins. Comparisons of model simulations with field observations in Eklutna Lake are considered to be suffi- ciently satisfactory to permit utilization of the model for the design and environ- mental assessment of the proposed Susitna Hydroelectric Project in Alaska. A number of novel aspects of the model- ing of the behavior of northern lakes and reservoirs have been found in the study; namely, the insensitivity of the lake-wide heat budget to river-borne frazil ice, the possibility of controlling the date of freeze-up in reservoirs by altering the fall heat storage, the need for improved estimation procedures for incoming long- wave radiation, the requirement for the accurate specification of heat transfer between water and ice and the requirement for highly resolved measurements of sus- pended sediment loading in time. The model has been applied to yield quantitative estimates of the thermal behavior of the proposed Watana and Devil Canyon reservoirs. The reservoir strati- fication and release temperatures simu- lated by the model have demonstrated the feasibility of utilizing the proposed multi-level intakes to control the down- stream temperatures in consideration of the environmental needs most of the time. The outflow suspended sediment concentra- tions and ice-cover thickness have also been simulated with reasonable degrees of confidence to assist in the planning of the fishery and wildlife resources manage- ment. ACKNOWLEDGEMENTS The supports provided by the Alaska Power Authority, Harza-Ebasco Susitna Joint Venture, R&M Consultants, and the Manager of the Hydrologic and Hydraulics Studies of the Joint Venture, E.J. Gemperline in carrying out this study are deeply appreciated. The following engi- neers participated in the study at various stages: T.H. Hsu, D.L. Muirhead, 176 J.S. Kuo, M.F. Rogers, and J.H. Lin, Mr. W. Dyok also contributed in the 1m· tial stage of the study. Their contribu- tions are acknowledged. REFERENCES Coffin, J.H., and U.S. Ashton, 1986, Sediment Budget of a Glacier Lake, Eklutna Lake, Alaska. In: Proceedings of the Cold Regions Hydrology Sympo- siun, American Water Resources Associa· tion. Gemperline, E.J., 1986. Hydrology and Hydraulic Studies for the Licensing of the Susitna Hydroelectric Project. In: Proceedings of the Cold Region Hydro- logy Symposium, American Water Re- sources Association. Harza-Ebasco 1985(a). Sus i tna Case E-VI Joint Venture, Alternative Fl01o1 Regime. Susitna Hydroelectric Project. Prepared for Alaska Power Authority. Harza-Ebasco Susi tna Joint Venture, 1985(b). Effects of the Proposed Proj- ect on Suspended Sediment Concentra- tion. Susitna Hydroelectric Project. Prepared for Alaska Power Authority. Harza-Ebasco Susitna Joint Venture, 1984, Eklutna Lake Temperature and Ice Study- with six months simulation for Watana Reservoir. Susitna Hydroelectric Pro- ject. Prepared for Alaska Power Au- thority. Imberger J., and J.C. Patterson, 1981. "A DynaMic Reservoir Simulation Hodel - DYRESM:S," Transport Hodels for Inland and Coastal Waters, Chapter 9, Academic Press. Patterson, J .C., and P.F. Hamblin, 1986 Thermal Simulation of a Lake with Win- ter Ice Cover. To be published. R&M Consultants, Inc., 1985(a). Glacial Lake Physical Limnology Studies: Eklutna Lake, Alaska. Volunes 1 & 2. Susitna Hydroelectric Project. Pre- pared under contract to Harza-Ebasco Susitna Joint Venture for Alaska Power Authority. R&M Consultants, Inc., 1985(b). Pro- cessed Climatic Data, October 1983 - December 1984, Eklutna Lake Station. Volume 7. Susitna Hydroelectric Proj- ect. Prepared under contract to Harza-Ebasco Susitna Joint Venture for Alaska Power Authority. R&~1 Consultants, Inc., 1985(c). Pro- cessed Climatic Data, October 1983 - December 1984, Watana Station and Devil Canyon Station. Volumes 4 & 5. Susitna ~droelectric Project. Prepared under contract to Harza-Ebasco Susitna Joint Venture for Alaska Power Authority. Tennessee Valley Authority, 1972. Heat and Mass Transfer between a Water Surface and the Atmosphere, TVA Report No. 0-6803, Tennessee Valley Authority. 177 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 TROPHIC LEVEL RESPONSES TO GLACIAL MELTWATER INTRUSION IN ALASKAN LAKES J. P. Koenings, R. D. Burkett, Gary B. Kyle, Jim A. Edmundson, and John M. Edmundson 1 ~STRACT: Many large lake and riverine systems in Alaska are influenced by melt- ~ter intrusion from glaciers and/or snow ~cks. Occluded within the ice or snow are sediments ranging in size from cobbles ~colloids. Our interest centers on the colloidal sized particles since they are in large part responsible for the turbid ~ture of lakes. We found that turbidity, caused by inorganic particles, initiates significant impacts on aquatic production beginning with effects on temperature regimes, 1 i ght profi 1 es, and nutrient inputs. Specifically, glacial lakes with ~rbidity levels above 5 NTU are charac- terized by seasonally eleva ted phosphorus levels, but lowered areal primary produc- tion, reduced zooplankter densities, a restricted diversity within the macro- zooplankton community, and 1 owe red fish yields. In addition, because glacial lakes ~hibit characteristics unique from clear water systems many of the cause/ effect models derived for such lakes e.g. phos- phorus input versus ch 1 orophyll ~ response, ~st be corrected for light restriction and for forms of non-biologically available phosphorus prior to application on turbid systems. Finally, as both reduced zoo- plankton production and fish yield are tied to increased turbidity, a compressed ~photic zone needs consideration as a negative modifier of areal fish production estimates from glacial lakes. (~Y TERMS: glacial silt, euphotic zone, turbidity, smolt yield, sockeye salmon.) INTRODUCTION Many of the freshwater lakes of the ~cific Rim support populations of anadromous salmonids. In particular, the young or juveniles of the sockeye salmon (Oncorhynchus nerka) spend one to three years of in-lake rearing before emigrating as smolts to the ocean. Lakes classified as sockeye nursery areas have widely varying capacities to rear juveniles to smolts. The ability to understand and then predict, through empirical models, the juvenile rearing or carrying capacity of lake systems is important for either stock management or enhancement purposes. A successful theory predicting poten- tial fish yield must contain terms descri- bing the effects of physical (e.g. solar radiation, temperature) and chemical (e.g. nitrogen and phosphorus), and of other groups of aquatic organisms in the water body (Mathisen 1972, Rigler 1982). Much effort has recently been spent on predic- tions of aquatic primary production from nutrient levels (Vollenweider 1976, Smith 1979), and then relating primary produc- tion to fish yield (Nelson 1958, McConnell et al. 1977, Hecky et al. 1981, Jones and Hoyer 1982, and Koenings and Burkett 1986). As turbidity imparted by inorganic parti- cles has been described as a modifier of primary production (Goldman 1960, Oglesby 1977, Edmundson and Koenings l985a, Grobbelaar 1985, and Koenings and Burkett 1986), we began to question the effect of increased turbidity on other levels of aquatic production in cold regions (Lloyd et al. 1986); and on the potential alter- ation of trophic linkages leading to fish yield in lakes turbid with silt. Glacial lakes, located throughout Alaska, are a sub-set of oligotrophic sockeye nursery lakes that are distin- guishable by characteristics derived from the large scale intrusion of glacial lAlaska Department of Fish and Game, FRED Division, P. 0. Box 3-2000, Juneau, Alaska. 179 meltwater. These systems are turbid (>5 NTU) for much of the year by virtue of colloidal sized inorganic particles which remain suspended in the water column. The effect of such particles on the autoctho- nous production of natural lakes within cold regions is largely unknown, and is described here through comparisons with the well described limnological charac- teristics (Hutchinson 1957, Wetzel 1975, Likens 1985) and sockeye rearing capaci- ties (Foerster 1968) of clear water systems. Herein, we develop cause/effect relationships at each trophic level (1°, 2°, and 3°) that act to separate turbid, cold-water lakes from geographically adjacent clear and organically stained systems. Such trophic level assessments are necessary because of the ever increas- ing pressure on the salmonid resources of Pacific Rim nations, and the concomitant need for alternative management approaches (Koenings and Burkett 1986). METHODS AND MATERIALS Limnological data were collected and analyzed by the State of Alaska, Department of Fish and Game, Limnology Program. The field and laboratory techniques are descri- bed in detail in the Fish and Game Limnology Manual (Koenings et al. 1986). In general, the sampling protocol was conducted at three week intervals during the ice-off to ice-on period, and at least once during mid-winter. Physical para- meters included measurements of lake tem- peratures, light penetration, turbidity, and Secchi disk transparencies. Lake temperatures and dissolved oxygen levels were recorded at 1 m intervals to the lake bottom or to a depth of 50 m using a YSI model 51 temperature/dissolved oxygen analyzer. Both incident and reflected solar radiation (footcandles) were re- corded at 0.5 m intervals from the sur- face to a depth equivalent to 1% of the sub-surface reading using a Protomatic submersible photometer. The algal light compensation point (depth) was defined as the depth at which 1% of the sub- surface active light [photosynthetically available radiation (400-700 nm)] penetrates (Schindler 1971). 180 Nutrient Chemistry Water quality samples were collect~ from the 1 m depth and mid-hypolimnetic zone (minimum of two sampling stations) fi the analysis of algal nutrients (phospho- rus, nitrogen, silicon, and carbon) and other water quality parameters (e.g., pH, conductivity, alkalinity, and turbidity). In general, reactive phosphorus (P) analyses were determined using the molybdenum-blue method as modified by Eisenreich et al. (1975). Total-P analys1:' utilized the same procedure following ac1a·l persulfate digestion. The various com- ponents used to characterize tota 1-P i.e., particulate, colloidal, and dissolved were1 analyzed after Koenings et al. (1986). Turbidity measurements were determined on samples collected from the 1 m strata using a model DRT-100 nephelometric turbidimeter. Primary Production Autochthonous primary production (algal standing crop) was estimated by chlorophyll a (chl a) analysis after the fluorometric-procedure of Strickland and Parsons (1972). We used the low strength acid addition recommended by Reimann (1978) to estimate phaeophytin. Water samp 1 es ( 1-2 l i ters) were f i 1 tered throug~ 4.25 em GF/F filters to which 1-2 ml of a saturated MgC03 solution were added just prior to the completion of filtration. The filters were then stored frozen in separate plexi-slides for later analysis. Primary productivity (rate of carbon uptake) within the euphotic zone was determined through carbon-14 isotope experiments using three incubation depths i.e., surface (l m), mid-euphotic, and t~, light compensation point. We utilized t~ light-dark bottle technique after Saunden' et al. (1962) using 2-4 stations per la~ with 6-8 sampling dates over the May- October period. Replicate clear and lightproof bottles containing 100 ml of lake water were labeled with 5.2 micro- curies of NaHcl4o3 and incubated in situ 4-6 hours. Samples were fixed with Lugol 's acetate solution and filtered through a 2.5 em GF/F filter. Filters were stored frozen in 20 ml polyethelene vials. Prior to analysis, filters were thawed, acidified, and then dispersed inb1 a toluene based scintillation cocktail. ~gal carbon incorporation was quantified using a Packard model 3255 scintillation spectrometer. Volumetric uptake rates (mg Cfm3fday) were determined using the cumulative percent time productivity curve after Vollenweider (1965). Mean daily rates were determined from integrated areal productivity versus time plots. Secondary Production Zooplankton were enumerated from replicate 50 m or bottom to surface vertical tows using either a 0.2 m or 0.5m diameter, 153 J.l mesh conical zoo- ~ankton net. The net was pulled at a constant 1 m/s, and \'lashed well before r~oving and preserving the organisms in 10% neutralized formal in (Haney and ~11 1973). At least two stations per lake were sampled 6-8 times during the open water period. Identification within the genus Daphnia followed Brooks ( 1957); of the genus Bosmina after Pennak (1978); and of the copepods after Wi 1 son ( 1959), Yeatman (1959), and Harding and Smith (1974). Identification and enumeration consisted of counting triplicate 1 ml subsampl es taken with a Hensen-Stempel pipet in a 1 ml Sedgewick-Rafter cell which was subdivided into transects. Finally, descriptions of laboratory experiments used in monitoring Daphnia reproductive success and survivorship under varying turbidity regimes are detailed in Edmundson and Koenings {1985b). Tertiary Production The Tustumena Lake sockeye fry stocking, smolt production, and fry dis- tributional information were taken from re~rts by Flagg (1985), and Thomas et al. (1984). For other lakes, sockeye smolt ~~uction information was obtained from ~chnical data reports available through the Alaska Department of Fish and Game. The rearing sockeye fry distributional data were obtained through the appl ica- tion of hydroacoustic assessment tech- niques (Kyle 1985). In short, this technique consists of recording fish signals on tape, in digital format, from a series of cross-lake transects using a scientific echosounder. The number of &~es represent individual fish which 181 can be converted to fish densities for each transect at various depth intervals. Finally, lake temperature profiles were measured to relate fry rearing temperatures to fry rearing depths. RESULTS AND DISCUSSION Primary Production Primary production has been correlated with solar radiation (Brylinski and Mann 1973), temperature (Grobbelaar 1985), and nutrients (Schindler 1978) with Smith (1979) stressing the specific role of phosphorus in limiting primary producti- vity. In addition, phosphorus-chlorophyll (P-C) response models have in recent years provided lake investigators a useful tool in evaluating primary production responses to changes in nutrient loading (Sakamoto 1966, Dillon and Rigler 1974, Vollenweider 1976, and Prepas and Trew 1983). In general, lakes in the same geographic area with similar morphometric features and nutrient levels should have similar capa- cities for autochthonous primary produc- tion. Since factors limiting primary productivity could largely determine total system production (Hecky 1984) including the tertiary trophic level (Nelson 1958, McConnell et al. 1977), we were interested in the cause(s) for the decreased chloro- phyll ~ (chl ~) levels in glacial lakes compared to the clear water lakes. Thus, our investigations centered on processes by which glacial silt particles affect temperature regimes, light profiles, and the phosphorus cycle. First, the most noticeable result of glacier meltwater intrusion is an increase in turbidity. Recent studies describing the nature of turbidity derived from glacier melt (Edmundson and Koenings 1985a) stress the strong correlations between turbidity and both particle size and con- centration. In addition, we found that turbidity levels of from 5-60 NTU were largely due to particles (silt) within the 20-40 micron size class. Furthermore, electron micrographs of these sized glacial particles revealed irregular shapes with smooth planar surfaces. Particles with this structure are characterized by large surface area to volume ratios; and would be less likely to settle out, thereby acting to enhance turbidities. Large concentrations of these planar- structured particles reflect or backscatter a larger proportion of incident solar radiation compared to non-turbid lakes (Figure l ). We found the ratio of inci- dent to reflected light in two clear water systems to equal 30:1 and 17:1 (Hidden and Leisure Lakes respectively) compared to a ratio of 3:1 in the glacial lakes. Moreover, in Packers Lake (organically stained) this light ratio equalled 60:1. Thus, our data suggest that glacial lakes reflect or backscatter significantly more light than occurs in both types of non- turbid lakes. The effect of light reflec- tion or backscatter may also have impli- cations regarding temperature regimes. Hecky (1984) found that sediment induced turbidity, as a result of impoundment activities in Southern Indian Lake, not only decreased light levels, but also lowered seasonal mean lake temperatures l-2°C. As penetrating solar radiation can directly warm lower layers to a depth of 10m, the action of silt particles to reduce the depth of light penetration reduces deep heating (Ragotzkie 1978). Therefore, evidence suggests that lake temperatures (seasonal) can be both lower in glacial lakes and higher in stained lakes compared to clear water lakes solely as a result of increased/decreased light backscatter within surficial strata. Since photosynthesis is determined by a combination of temperature, light, and nutrients, the effect of turbidity induced backscatter in lowering glacial lake temperatures may in itself result in decreased primary productivity. Second, a strong correlation was found between turbidity (NTU) and depth of light penetration expressed as euphotic zone depth (EZD) (Edmundson and Koenings l985a, Lloyd et al. 1986). That is, turbidity levels were inversely correlated to EZD (Log EZD = l .23 -0.66 Log NTU: r2 = 0.94) and more importantly, even low turbidities (5-10 NTU) severely restrict light penetration (EZD). Edmundson and Koenings (1985a) established that turbidity levels >5-10 NTU and an EZD <6-4 m could be used to distinguish clear from glacial lakes. In turn, the adverse effect of turbidity restricted light regimes on areal aquatic productivity through various carbon-14 182 algal transfer and silt addition experi- ments has been shown by Goldman (1960, 1961), Tizler et al. (1976), and more recently by Edmundson and Koeni ngs ( l985a In essence, areal carbon uptake rates within oligotrophic lakes decreased line- arly with increasing turbidities i.e., reduced light levels (Koenings and Burket! 1986). Moreover, we found gross differ- ences between clear and glacial lakes wh~ comparing euphotic volume (EV) as a pro- portion of the total lake volume among various lake types e.g., clear, organical· ly stained, and glacial (Table 1). For clear lakes (n=ll) the EV ranged from 41· 9m~ of the tota 1 with a mean va 1 ue of 64%, whereas ~n the organically stained lakes (n=5) the EV ranged from 8-32% and averaged 21%. However, in the glacially influenced lakes (n=lO) the EV ranged from a low of <1%, in a heavily glacial system (45 NTU), to 23% in semi-glacial lakes (5-10 NTU) yet averaged only 6.7%. Thus, the EV as a percentage of the total lake volume is significantly less in glacial lakes compared to non-turbid lakes. This severely restricts the volu~ of water capable of authocthonous produc- tion. -(/) (I) =o c: CIS 0 -0 0 --0 0 0 ,.... X 1--:c (!:) :J 1--z w 3 2 1 x Packers o Leisure +Hidden • Kenai CLEAR • Crescent . • Tustumena 0 00 0 0 • • Ptarmigan. .. .. .. GLACIAL c 0~------~----~-------+------~ () 0 z 150 300 450 600 REFLECTED LIGHT (footcandles) Figure 1. The relationship bet~&een incident solar radiation (400-700 nm) and reflected radiation (backscatter) from two clear water lakes (Hidden and Leisure), an organically stained syst~ (Packers), and four glacially influenced lakes. Table 1. A comparison of total surface area, water residence time, light compensation depth, and euphotic volume (EV) (expressed as a percent of total lake volume) between clear, organically stained, and glacially influenced Alaskan lakes. Water Light Volume Lake type Surface Area residence compensation Eu~hotic volume Lake (m3 x 10+6) (water clarity) (m2 x 10+6) (acres) time (yr) level (m) (m3 x 10+6) (% of total) Tustumena 37,000 Glacial 295.0 55,597 17.2 1.1 325.0 <1 Ski 1ak Glacial 99.0 24,463 1.5 148.5 <l(esl.) Kenai Glacial 56.0 13,837 4.0 224.0 <1 (est.) Tons ina 726 Glacial 13.7 3,394 2.0 1.3 18.0 3 Tazlina 10,584 Glacial 155.9 38,528 3.8 1.6 249.4 Trail Lakes 124 Glacial 8.1 1,754 0.2 0.5 3.6 K1utina 3,020 Glacial 67.1 16,587 3.1 3.8 255.0 8 ------------------------------------------------------------------------------------------------------------------ Eklutna 512 Semi-Glacial 14.0 3,458 1.8 3.0 42.0 8 Ptarmigan 125 Semi-Glacial 3.0 750 1.1 6.0 15.6 17 Crescent 389 Semi-Glacial 16.2 4,002 0.3 5.5 89.1 23 -------------------------------------------------------------------------------------------------------------- Bakewell 67 Organic Stain 2.8 692 0.4 4.5 12.6 19 Hugh Smith 198* Organic Stain 3.2 800 1.1 5.0 16.0 8 Packers 26 Organic Stain 2.1 519 3.0 4.0 8.4 32 McDonald 197 Organic Stain 4.2 1,035 0.7 7.5 31.5 16 Falls 30 Organic Stain 1.0 254 0.5 9.5 9.0 29 ------------------------------------------------------------------------------------------------------------------ Bear 19 Clear 1.8 Hidden 138 Clear 6.8 Upper Russian 122 Clear 4.6 Karluk 1,920 Clear 39.0 Tokun 38 Clear 1.8 Eshamy 122 Clear 3.6 Leisure 23 Clear 1.1 Larson 29 Clear 1.8 Sea Lion Cove <1 Clear o. 1 Nunavauguluk 4,489 Clear 79.0 *Volume of the mixolimnion Third, we were interested in the ef- fects of glacial silt particles on nutrient (e.g., phosphorus) cycles. Our fraction- ation studies have determined that glacial ~rticles are composed of significant amounts of inorganic particulate phosphorus (IPP) or rock phosphorus (Edmundson and Koenings 1985a). An example is the effect of glacial silt on the phosphorus cycle in Crescent Lake, a semi-glacial system which has a seasonally defined on/off glacier- melt cycle (Figure 2). Total phosphorus 445 0.8 9.5 17.1 90 1,680 11.2 15.0 102.0 74 1,137 1.1 13.0 51.0 42 9,637 6.0 20.0 780.0 41 448 1.0 16.0 25.0 65 890 2.7 20.0 72.0 59 259 0.3 18.0 19.8 86 445 1.9 9.5 17.1 59 19 0.3 6.9 0.3 80 19,513 6.0 25.0 1 ,975 44 183 (TP) levels increased from 4.5 ~g/1 in the spring (April-June) at a time of minimum glacier melt to 8.5 ~g/l during the summer (July-September) at maximum glacier melt, and then decreased to 4.5 ~g/l in winter (Figure 2a). In turn, total par- ticulate phosphorus (TPP) which is com- prised of inorganic (IPP) and organic (OPP) particulate fractions (Koenings et al. 1985) followed a similar trend i.e., increasing from 2. 0 ~g/ 1 to 7. 5 ~g/ 1 in mid-summer and decreasing to 1.5 ~g/1 A 10 9 .A .. ..... 8 ..J ' 7 C) :I ..... 6 en 5 ::) 0: 4 0 :c 3 a. en 2 0 :c 1 a. 0 A M J J A S 0 N TIME (months) 8 -Q. Q. .... .. 0 ?P.· ..... en ::) a: 0 :c a. (/) 0 ::r: a. D J F 100 90 80 70 60 50 40 30 20 10 0 A inorganic (I PP) : · : : : : • • : : : ~ : ~ organic : : : : .· : : : . : : • • . '··· ··:::::::(OPP) :::·:: . . . . . . . . . . . . . . . . . ................ . . .. . . . . . . . . . . . . . . . . . . . . . . . . MJ JASON OJ F' TIME (months) Figure 2. Seasonal phosphorus cycles in Crescent Lake (semi-glacial): (A) total phosphorus (TP) levels compared to total particulate phosphorus (TPP) and organic particulate phosphorus (OPP); and (B) seasonal fluctuations in inorganic and organic' particulate phosphorus expressed as a percent of TPP. in winter. However, seasonal fluctuations of OPP levels were minimal as concentra- tions remained between l-2 ~g/1. That is, while OPP comprised 80% of TPP during the spring and winter, IPP comprised 80% at the height of meltwater intrusion. Thus, the increase in TPP, and indeed TP, was due to increased IPP resulting from glacial melt during the summer (Figure 2b). Because of glacier melt, a sig- nificant amount of TP in the lake during the peak-growing season is comprised of non-biological IPP i.e., rock phosphorus. Finally, phosphorus-chlorophyll a (P-C) regression analysis of glacial lakes initially showed little correlation be- tween chl a levels and phosphorus concen- trations. -However, Edmundson and Koenings (l985a) observed that several semi-glacial lakes overlapped into the regression used to derive the clear-water lake P-C model. When these systems were excluded, a significant P-C relationship was derived for glacial lakes which showed the same basic trend established for clear water Alaskan lakes except for a signifi- cantly lower chl ~response at equivalent 184 phosphorus levels (Figure 3a). To explair' lower chl ~levels, at particular P con-, centrations, in glacial lakes we felt corrections for IPP and light restrictio~ were appropriate. Using IPP corrected total phosphorus levels, the semi-glacial lakes fell entirely within the clear lake, P-C response model, and justifiably, were excluded from the glacial P-C model.' vJhile a consideration of IPP resulted in, a significant reduction between observed and predicted chl a levels in glacial lakes (Figure 3b),-a substantial differ-, ence remained in lakes >5-10 NTU which could not be explained by IPP alone. Following the principle of Verduin et al., (1978) which considered the ratio of EZD to mean lake depth in P-C relationships, ' Edmundson and Koenings (l985a) found that, this light correction factor further re- duced the amount of chl a expressed per ' unit phosphorus concentration. That is, corrections for both IPP and restricted ' light combine to reduce 92% of the over- all differences between observed chl a, and chl ~predicted using clear water- P-C models. A 1.o -.J .5 ..... / / Ill :I ;, 0.0 / / // // /,/' -----::/: ...---.----:::,.// .c u C) -.5 0 .J //glacial -1.0 1,-----.-_.:./::..__~/-.------,r-----.-1 0.0 .5 1.0 1.6 -.J ...... 6. Ill LOG TP ( u g I L ) 2.0 :I 111.0.0 :c z / 7 / clear / / .--- // ----u C) -6 0 . .J /~...---_.., ~/ --///glacial -1.0-4r------~-~---,.----r----.-l 0.0 .5 1.0 1.5 2.0 LOG TP ( u g I L ) Figure 3. Phosphorus-ch 1 orophyll ~ (P-C) response models, with 95% confidence limits, derived for clear water (-) and glacial lakes (---); (A) excluding the semi-glacial lakes, and (B) exclu- ding the semi-glacial lakes and after correcting for IPP. Secondary Production A comparison of seasonal (May-October) ~an macro-zooplankter densities from ~veral oligotrophic lakes shows a d&rease in euphotic zone depth and mean seasona 1 macro-zooplankton dens i"ty with increased glacial influence (Figure 4). The clear water lakes (Karluk-Bear Lake) have an average EZD of 16 meters, and an average macrozoopl ankton density of 475,000 organismsjm2. The organically stained lakes, along with those with some glacial influence (Ptarmigan-Kenai Lakes), have an average EZD of 5 meters and average macro-zooplankton density of 205,000 organi smsjm2 over the same time period. Finally, Tustumena and Upper 185 Trail Lakes, which are heavily influenced by glacial runoff, have an average EZD of 0.5 meters and average macro-zooplankton density of 23,500 organisms/m2. While light restrictions from glacial meltwater intrusion are associated with reduced macro-zooplankton densities, it is also evident that glacial silt affects macro-zooplankton diversity as well (Table 2). The macro-zooplankter species compo- sition is conspicuous in the total absence of the non-discriminate filter feeding cladoceran species, specifically Bosmina and Daphnia, from the macro-zooplankton community in glacial lakes (6-45 NTU) while they are well represented in non- turbid (0-5 NTU) lakes. In contrast, the selective feeding species, represented by the copepods, are well represented in both lake types. Because both the turbid and non-turbid lakes possess similar oligo- trophic lake features, i.e., low primary productivity, it appears that glacial meltwater produces additional conditions unfavorable for the survival of cladoceran species thereby enhancing the success of the copepods, particularly Diaptomus and Cyclops. Laboratory tests designed to mimic the effect of in-lake glacial silt intru- sion on Daphnia survival and reproduction, indicate that Daphnia are able to thrive under conditions of high turbidity (Figure 5). In these tests, Daphnia were exposed to varying levels of turbidity ranging from 0 to 60 NTU by the addition of increasing concentrations of glacial silt. The surprising result was that Daphnia survival and reproduction over time was most successful under the highest turbidity level of 60 NTU. This was attributed to silt particles providing sites for bacterial growth, thus, pro- viding an unlimited food source which ultimately enhanced Daphnia survival and reproduction (Edmundson and Koenings l985b). These results indicate that silt by itself is not detrimental to Daphnia survival and reproduction when food levels are high. That Daphnia cannot survive and therefore are not found in turbid Alaska lakes (Edmundson and Koenings l985b) whereas they thrive under turbid conditions in the laboratory, appears to be contra- dictory. However, the contradiction 0 0 0 .. "' E .. a. "' ;! :: clear water lakes " c: c: £ ~ "' c: ~ "' ~ g_ ~ "' 1 0 Jj §~ .... I c: .. S!> f ~ "' .. "' Q ()J organically stained or glacially Influenced c "' " : " ~.~ ... " "' " c: 0 ::>/; .. " :Z:CI) a. "' I heavy glacial Influence .. c: " E " .,_ ;;; "·- " <>:! ... ::ll- E :1: ... Q. w Q w z 0 N 20 (J 16 i= 12 ~ 8 g; 4 w Figure 4. Comparison of seasonal mean macrozooplankton densities among clear lakes (mean EZD =16m), semi-glacial and organically stained lakes (mean EZD = 5 m), and lakes with heavy glacial influence (mean EZD = 0.5 m). disappears when we consider the presence of glacial silt together with low level food conditions that exist in oligotrophic Alaskan lakes. Arruda et al. (1983) has shown that cladocerans survive well in turbid, southwestern United States reservoirs, where, under warm water con- ditions bacteria are able to colonize silt particles, thus providing an addi- tional food source for foraging Daphnia. In colder Alaska lakes; however, condi- tions are not conducive to this type of association (Stockner 1983), thus cladocerans gain no benefit from ingested glacial silt particles. Moreover, studies conducted by Patalis and Salki (1984), before and after impoundment of Southern Indian Lake, a cold water, north- ern latitude lake, have shown a link between decreased water temperature and decreased macro-zooplankton density. In addition, McCabe and O'Brien (1983) have suggested that the ingestion of silt par- ticles interferes with Daphnia survival and reproduction. Indeed, Edmundson and Koenings (1985b) have shown from both in situ and laboratory tests, where chl a levels were manipulated as well as 186 turbidity levels, that Daphnia survival and reproduction fared best under conditions~ high chl a levels and low silt concentra- tions. Those exposed to 1 ow chl a 1 evels and high silt concentrations suffered in· creased mortality rates and repressed reproduction severe enough to result in extinction. 70 Adult Daphnia SurviYorshlp 0 NTU o--o 30NTU ~ 60 60 NTU A---i:l "' 0: w 50 m •o ::;: :> z ..1 .. 30 ... 0 ... 20 New Production (Including eggs, embryos and newly hatched young) 0 NTU --~-.. 30 NTU ____ _. 80 NTU A---• r ,, I I ' -' ' I ' ' ' '' ~ ~~--c.---Ail=....:: ,', ...L 10 I I ' I I T ' t '• ,, I- T ' • 'I ,.._ 0 24 48 72 88 120 144 168 192 2115 240 TIME (hours) Figure 5. Laboratory experiments at 13 ( sho~Jing adult Daphnia survivorship and reproduction when exposed to varying turbidities. Values are means of thr~1 replicates with the latter three dat~ rep res en ted by means ±. 1 S. E. Investigations by Allen (1976) and Richman and Dodson (1983) have shown th~ under conditions of low food abundances copepod species have a competitive adva~ tage over cl adocerans. Cl adocerans, whic' are non-discriminate filter feeders, mus filter a large amount of water in order t obtain a small amount of food, wasting energy. In contrast, copepods, with a selective feeding strategy, inspect and ingest a lower number of individual par~~ cles per ~nit_t~me, expen~ing ~e~s energy~ The non-d1scr1m1nate and 1neff1c1ent · foraging strategy of cladocerans is ade- quate for survival and reproduction in Alaskan oligotrophic, clear water lakes (Table 2). However, high levels of sus- pended silt relative to food particles, · results in the filtering and ingestion oft Table 2. Comparison of mean seasonal zooplankton con~unity composition between glacial (>6 NTU) and non-turbid (~5 NTU) lakes. Lakes exemplify an extensive data set [glacial (n = 18), non-turbid (n = 78)] that includes systems located throughout Alaska. Relative densities are: absent (-}, ~33% (+), ~34% ~66% (++), and ~67% (+++). Glacial Lakes (6-45 NTU} Taxa/Lake Tustumena Kenai Crescent Cl adocera: Bosmina sp. D:1phnia sp. Holopedium gibberum AZona sp. Polyphemus pediauZus Copepoda: Cyclops sp. + ++ + Diaptomus s p. + + Epischura s p. Rotifera: K8llicottia Zongispina + + + Aoplanahna sp. +++ K8rateUa sp. ConochiZoides s p. glacial silt particles of the size range of algal material normally consumed by cladoceran species. Under this regime, cladocerans are unable to obtain, from the total amount of material ingested, the required amount of energy necessary for long term survival. Finally, we also suggest that the ~rthenogenic reproductive strategy of cladocerans is adversely influenced by glacial meltwater intrusion. The rapid reproduction of many broods of young is severely disrupted by the ingestion of silt particles which reduces the ability of cladocerans to meet the energy demands for this reproductive strategy. In contrast, the selective filter feeding and s 1 ow ~ced sexual reproductive strategy of the wpepod species, requires less energy. This is especially germane under condi- tions of low food availability and high concentrations of suspended silt. Thus, it is the presence of suspended glacial silt particles in combination with low algal densities that results in the ~imination of the non-discriminate filter feeding cladocerans from the macro-zooplankton community. Grant ++ + + 187 Non-Turbid Lakes (0-5 NTU) Ptarmigan Hidden Packers Lei sure Karluk Badger + + +++ + + + + + + + + + + ++ + + + +++ +++ ++ + ++ + + + + + + ++ + + Tertiary Production Density-dependent relationships between forage density, food type, and fish abun- dance infer differences in juvenile growth potential in lakes having different capa- cities for food production. In general, we would expect that sockeye salmon nursery lakes having higher capacities for forage (zooplankton) production would show higher sockeye growth rates (at comparable fry densities) than would lakes with lower capacities for zooplankton production. In addition to a high overall productive capacity for zooplankters, sockeye rearing lakes need to produce preferred zooplankters (i.e., cladocerans) for the most efficient energy transfer. While we acknowledge that density-dependent factors such as zooplankton abundance and type do effect sockeye production, recent com- parisons between glacial and clear lakes reveal that density-independent factors may be as, or even more, important in limiting fish production. A compilation of sockeye smolt sizes from clear, organically-stained, and glacial lakes, indicates that age 1 and age 2 smolts produced from glacial lakes are consistently small (Table 3). In fact, the age 1 and age 2 smolts produced from clear or stained lakes were 16-104% and 22-135% greater in length respectively, than from glacial lakes. Considering the consistent production of near threshold- size smolts (Koenings and Burkett 1986) from differentially sized glacial lakes located in divergent geographic regions in Alaska, we believe that cooler juvenile rearing temperatures characteristic of glacial lakes, as well as limited forage, influences the size of smolts. Table 3. Representative mean length and weights of sockeye smolts from clear, organically stained, and glacial lakes located throughout Alaska. Values were obtained over the entire outmigration period. Lake Clear Hidden Larson Leisure Tokun Eshamy Karluk Frazer Chil kat McDonald Stained Geographic Region Cook Inlet Cook Inlet Cook Inlet p .w.s. P.W.S. Kodiak Kodiak Southeast Southeast Packers Cook Inlet Hugh Smith Southeast Glacial Tustumena Crescent Kenai Tazlina Klutina Tons ina Chil kat Cook Inlet Cook Inlet Cook Inlet P.W.S. P.H.S. P.W.S. Southeast Age 1 Length Weight (11111) (g) 143 86 80 72 76 101 76 100 70 97 69 70 68 62 72 64 64 65 27.3 5.1 4.0 2.5 3.4 10.7 3.1 2.6 8.5 2.9 2.8 2.8 2.1 P.W.S.-Prince William Sound Age 2 Length Weight (nm) (g) 200 123 97 101 113 103 110 79 132 78 85 76 72 73 72 64 70 83.9 16.5 9.0 7.8 14.1 7.9 4.5 21.2 4.4 5.0 3.8 3.1 To illustrate the effect of a density- independent factor (rearing temperature) on sockeye production, we focused on smolts produced from Tustumena Lake (glacial); and fish growth relative to rearing temperatures in Hugh Smith Lake (stained). Rearing distribution studies were under- taken in Hugh Smith Lake because the sockeye smolts displayed only small dif- ferences in size or age structure in response to increased forage production resulting from nutrient enrichment. Oiel patterns of vertical fry migration, ob- tained by hydroacoustic surveys, showed 188 that the majority of fry remained in cool rearing temperatures of 7°C or less (Pelb 1985). Furthermore, using the gro~Jth mode of Iwama and Tautz (1981) to predict age 1 smolt size based on lake rearing tempera- atures, the length of growing season, and the assumption of food satiated growth; Peltz (1985) found that over four years U1 predicted weight of age 1 smolts deviated from the observed by only 1-11%. These results indicate the potential for fry rearing temperatures to either control or at least strongly influence the growth of juvenile sockeye. In glacially turbid Tustumena Lake, the 7.8 million sockeye fry initially stocked in 1979, along with the recruits from the 1978 escapement of 110,000 adulb produced an estimated 400,000 + age 1 smolts in 1980. However, with the ever increasing fry rearing pressure on the zooplankton forage base; the smolt age compositions, and in particular, the smolt sizes have remained essentially unchanged. For example, from the production of only 2.3 million smolts in 1981 to 16.7 million in 1985, the age composition of age 1 smolts decreased by only 8%, while the siz1 of age 1 smolts was identical, and the siz1 of age 2 smolts decreased by about 5% (Figure 6). Considering that the initial 1981 smolts were near threshold-size, a density-dependent response to a seven-fola increase in the total number of smolts conceivably would have caused a shift in age structure to predominantly age 2 smolts. The lack of a density-dependent response suggests that a density-independa factor of cool rearing temperatures has a major influence on juveni 1 e sockeye growtn to smolt in this glacial lake. These detailed seasonal and diel hydroacousti c surveys suggested that 1 imit! on juvenile fish growth rates within a season caul d be set by accumulated temper- ature units. Thus, both the rate of juvenile growth and the total growth within an annual period can be heavily influenced by in-lake rearing temperatures and the length of the open water period respectively. Considering the potential impact of an abiotic density-independent factor on the age and size of smolts (the end product of freshwater rearing); we hydroacoustically surveyed several sockeye~ nursery lakes for juvenile rearing distri-l butions relative to temperature (Figure 7). Smolt Age composition Smolt SIU ('K,) (mm) (g) Age 1 82 70 2.8 Age 2 18 88 5.1 Age 1 80 69 2.9 19 Age 2 20 82 4.8 Age 1 80 70 2.9 Age 2 20 83 5.1 Age 1 80 73 3.3 Age 2 20 85 5.2 Age 1 76 70 2.6 Age 2 24 84 4.8 2 4 6 8 10 12 14 16 18 SMOL T PRODUCTION (•10 6 ) Figure 6. Summary of annual sockeye smolt production from Tustumena Lake (1981-1985) showing the constancy in smolt size and age composition despite significant increases in smolt numbers. Such unresponsiveness in smolt patterning (Koenings and Burkett 1986) to density increases is characteristic of a density-independent controlling factor( s). Using the 7°C isopleth as a guide, lakes were ranked by actua 1 fry rearing temper- atures during July and/or August. \~e found that low rearing temperatures for juveniles were not confined solely to glacial systems, but that low rearing ~mperatures were consistent only within the glacial lakes surveyed. Despite the possibility that over a longer period of time, variations could occur which would wrrespondingly affect the growth rate, other investigators have found similar temperatures affecting sockeye growth (Clarke 1967, Brett et al. 1969). Al- though Brett et al. (1969) and Shelbourn et al. (1973) demonstrated that the opti- mum temperature for growth shifts to a lower temperature with a decrease forage base (which is applicable to glacial lakes), the findings of Biette and Green (1980) and Clarke (1978) showed at a reduced to moderate forage base, sockeye ~owth was enhanced under cyclic rearing ~mperatures. This is because to a large 189 extent energy expenditure is governed by activity (e.g. feeding) and temperature. Since metabolic rates are closely corre- lated to feeding rates in fish (Paloheimo and Dickie 1966) it is reasonable to assume that temperature will act as a major independent variable. In addition to temperature being a factor in sockeye juvenile growth, it appears that juvenile feeding behavior is influenced by reduced light conditions in glacial lakes. Normally, in clear lakes sockeye fry are distributed near- surface at night while being dispersed through deeper strata during the day. This pattern of feeding behavior has been hypothesized to be a result of predator avoidance and/or the requirement of light for sockeye fry to locate zooplankton prey (Narver 1970, Eggers 1978). In glacially-turbid Tustumena Lake, rearing juveniles were distributed near surface during the day while being dispersed through deeper strata at night (Thomas et al. 1984). In addition, the surface affinity ~f the fry was much greater in the fall than during the summer. Thus, the reduced hours of sunlight in the fall combined with the input of glacial silt over the summer, forced a greater percentage of foraging fry nearer the surface during the day. Although fish visibility is reduced by glacial silt, the presence of rearing fry nearer the surface during the day rather than at night allows for increased risk of capture by predators. Hence, in glacial lakes with substantial predator populations, the fry are at a great survival disadvantage having to balance the need to feed on sparse zoo- plankters with predator avoidance. Finally, as discussed earlier, the intrusion of suspended glacial silt parti- cles lowers areal primary productivity; a direct result of reduced light penetra- tion. In turn, Koenings and Burkett (1986) showed that mean annual areal rates of photosynthesis for several oligotrophic lakes were highly correlated to euphotic zone depth (EZD). By combining lake surface area (m2) with EZD (m) an index of potential system production, euphotic volume, was calculated. It was natural to assume that a comparison of fish production for different types of lakes (i.e., clear, organically-stained, and glacial) should be compared on the basis of euphotic volume rather than on an areal basis, to incorpor- ate the affects of light (Figure 8). Smolt production based on lake surface area indi- cates a significant difference between lake types; however, the same smolt pro- duction based on euphotic volume shows that smolt production was more consistent across lake type. This emphasizes the concept that when sockeye lakes are defined as rearing area limited, they fall within the same general range of production when euphotic volumes are considered. CONCLUSIONS The immediate result of glacier melt- water intrusion is an increase in turbidity, a function of varying particle concentra- tions as well as size distributions. Suspended glacial silt has been found to dominate the phosphorus cycle, and to determine light regimes in glacially turbid (~5 NTU) lakes. Specifically, our 190 DAY REARING TEMPERATURES (RANGES FOR HIGHEST ONE-THIRD FISH DENSITY BY DEPTH INTERVAL) tS t6 .... () 14 ~ w t2 a: ~ tO < a: UJ I • ~ 6 .ti+------'1'-1:-em-pe-ra-:--lu-re -:-:-tlm-::11-:-lno-t-' -.,-0--1 W 0 I 1- NIGHT REARING TEMPERATURES (RANGES FOR HIGHEST ONE-THIRD FISH DENSITY BY DEPTH INTERVAL) IS t6 § 14 w t2 a: ~ tO 1-< a: ~ 6 f ::::; w 1- 2 I 9 0 I I t temperature llmltlng I ~ I Figure 7. Diurnal in-lake juvenile rear- lng temperatures showing the greater potential for cooler temperatures to effect sockeye gro\\lth rates in glacial lakes compared to clear or stained systems. fractionation studies have shown that rock phosphorus contained within the particulate inorganic fraction dominates the phosphorus cycle in glacial lakes; a~ the light compensation depth was found to be inversely related to turbidity levels thereby acting to define the extent of the euphotic zone. As a result, the euphotic volume, as a percent of the total lake volume, is significantly less in glacial systems compared to the non-turbid syste~ we examined. As a consequence, algal cell can spend considerable time in an aphotic zone that extends well beyond the critica\ mixing depth used to estimate the point · where such mixing becomes a major primary production regulator in turbid waters (Grobbelaar 1985). Moreover, as light is attenuated rapidly in turbid waters, only' a small amount of the energy flux is 500 400 ;; e N n=4 ~ 300 r ' • " " 0 l 20()-• . 0 n.4 n=5 I n:5 n=2 I 0 40 30 20 .. ~ > w ' ., !:i 0 :I .. .. 0 a: w .. :I ::> 100 n: 2 10 z I z < w :I Figure 8. Sockeye smol t production from turbid, stained, and clear water lakes based on surface area (km2) and euphotic volume (EV) showing the reduced variation in smolt numbers between different lake types after correcting for the depth of light penetration. available for photosynthesis (Grobbelaar 1985). The high proportion of reflected light acts to compress the euphotic zone, ~t because of the abiotic nature of the ~rticles involved a shallow euphotic zone does not indicate increased volu- metric production. This is unlike clear water systems where increased turbidity is caused by dense concentrations of algal cells; and lost euphotic depth is compensated for by increased volumetric production (Koenings and Burkett 1986). The effects of glacial silt were evident both on primary productivity (carbon uptake rates) and on primary pro- duction (chlorophyll a). That is, volume- tric measurements of carbon uptake de- creased with an increase in glacier melt influence. Moreover, when productivity ~tes were integrated over the euphotic zone, areal productivity was considerably reduced especially at the height of glacial meltwater influence. ~Je also observed that summer period chlorophyll ! levels were significantly less than those predicted using phosphorus- chlorophyll response (P-C) models derived ~om clear water lakes. The inconsistency ~tween P-C models was found to be par- tially explained by the dominance of inor- ganic particulate phosphorus in the total 191 phosphorus cycle, and by a consideration of the ratio of euphotic volume:total volume. Thus, the magnitude of glacier meltwater intrusion results in decreased autochtho- nous production, and the increased oligotrophy of glacial lakes. Lowered primary production also results in reduced densities of herbivorous macro-zooplankters which forage on the algal community. We have also found that filter feeding cladocerans e.g. Bosmina and Daphnia are uniquely absent from the zooplankton community of glacial lakes. While primary production and summer temper- atures in glacial lakes are low, both are not beyond the lower limit we have observed for geographically similar clear water systems that contain robust populations of cladocerans. In addition, we have found cladoceran species to be absent from the zooplankton community of fishless but glacially turbid lakes. The overlapping size ranges of algal material and glacial silt allow ingestion of the glacial parti- cles by non-discriminating filter feeders. Such an inefficient foraging strategy, especially when particle concentrations are high and algal numbers low, results in the eventual elimination of Bosmina and Daphnia Thus, the macro-zooplankton community of glacial lakes consists entirely of the selective herbivore Diaptomus and the raptorial feeding Cyclops. The adverse consequences of lowered autochthonous primary productivity is also felt at the tertiary level. A consequence of the restricted species composition withi the sparse zooplankton community of glacial lakes, is a lowered rearing potential for sockeye salmon fry relative to clear water lakes. Goodlad et al. (1974), o•Brien (1979), Vinyard (1982), Koenings, and Burkett (1986) have found that plankti- vorous fish, given a choice, prefer cladocerans as prey species over copepod species. Since salmon fry, rearing in glacial lakes, must expend more time and energy capturing evasive copepod species; the areal rearing potential of these lakes is limited compared to clear water lakes. Thus, the restricted light regime of glacial lakes becomes an important factor in correcting areal rates of fish yield. LITERATURE CITED Allan, J. D., 1976. Life History Patterns in Zooplankton. Amer. Natur. 110:165-180. Arruda, J. A., R. G., Margolf, and R. T. Faulk, 1983. The Role of Suspended Sediments in the Nutrition of Zoo- plankton in Turbid Reservoirs. Ecology. 64:1225-1235. Biette, R. M. and G. H. Green, 1980. Growth of Underyearling Sockeye Salmon, Oncorhynchus nerka, Under Constant and Cyclic Temperatures in Relation to Live Zooplankton Ration Size. J. Fish. Res. Bd. Canada 37:203-210. Brett, J. R., J. E. Shelbourn, and C. T. Shoop, 1969. Growth Rate and Body Composition of Fingerling Sockeye Salmon, Oncorhynchus nerka, In Rela- tion to Temperature and Ration Size. J. Fish. Res. Bd. Canada 26:2363-2394. Brooks, J. L., 1957. The Systematics of North America Daphnia. Mem. Conn. Acad. Arts. Sci. 13:1-180. Brylinski, M. and K. H. Mann, 1973. An Analysis of Factors Governing Produc- tivity in Lakes and Reservoirs. Limnol. Oceanogr. 18:1-14. Clarke, W. C., 1978. Growth of Underyear- ling Sockeye Salmon, Oncorhynchus nerka, On Diel Temperature Cycles. Fish. Mar. Serv. Tech. Rep. 780: 19 p. Dillon, P. J. and F. H. Rigler, 1974. The Phosphorus-Chlorophyll ~Relationship in Lakes. Limnol. Oceanogr. 19:767-773. Edmondson, W. T. and G. G. Winberg (eds), 1971. A Manual for the Assessment of Secondary Productivity in Fresh Waters. International Biological Program Handbook. 12:358 p. Edmundson, J. A. and J. P. Koenings, 1985a. The Effects of Glacial Silt on Primary Production, Through Altered Light Regimes and Phosphorus Levels, in Alaska Lakes. p. 3-19. In: L. P. Dwight [chairman] Proceedings, Resolving Alaska's Water Resources Conflicts. Report IWR-108. University of Alaska-Fairbanks, U.S.A. Edmundson, J. M. and J. P. Koenings, 1985b. The Influences of Suspended Glacial Particles on the Macro-Zooplankton Community Structure Within Glacial Lakes. p. 21-35. ~: L. P. Dwight [chairman] Proceedings, Resolving Alaska's Water 192 Resources Conflicts. Report IWR-108. University of Alaska-Fairbanks, U.S.A. Eggers, D. M., 1978. Limnetic Feeding Behavior of Juvenile Sockeye Salmon in Lake Washington and Predator Response Limnol. Oceanogr. 23:1114-1125. Eisenreich, S. J., R. T. Bannerman, and D. E. Armstrong, 1975. A Simplified Phosphorus Analysis Technique. Environ Letters 9:43-53. Flagg, L. B., 1985. Sockeye Salmon Smolt Studies, Kasilof River, 1985. Alaska Department Fish and Game, FRED Tech. Data Report (In Press). Foerster, R. E., 1968. The Sockeye Salmo~ Oncorhynchus nerka. Fish Res. Board of Canada.Bull. 162:422 p. Goldman, C. R., 1960. Primary Producti- vity and Limiting Factors in Three Lakes of the Alaskan Peninsula. Ecol. Monogr. 30:207-230. Goldman, C. R., 1961. Primary Productiv- vity and Limiting Factors in Brooks Lake, Alaska. Verh. Int. Ver. Limnol. 14: 120-124. Goodlad, J. C., T. W. Gjernes, and E. L. Brannon. 1974. Factors Affecting Sockeye Salmon (Oncorhynchus nerka) Growth in Four Lakes of the Fraser River System. J. Fish. Res. Board of Can. 31:871-892. Grobbelaar, J. A., 1985. Phytoplankton Productivity in Turbid Waters. J. Plankton Res. 7:653-663. Haney, J. F. and D. J. Hall, 1973. Sugar Coated Daphnia: A Preservation Tech- nique for Cladocera. Limnol. Oceanogr. 18:331-333. Harding, J. P. and W. A. Smith, 1974. A Key to the British Freshwater Cyclopoid and Calanoid Copepods. Sci. Publ. Freshwater Biol. Assoc. 18:1-54. Hecky, R. E., E. J. Fee, H. J. Kling, and J. W. M. Rudd, 1981. Relationship Between Primary Production and Fish Production in Lake Tanganyika. Trans. Amer. Fish. Soc. 110:336-345. Hecky, R. E., 1984. Thermal and Optical Characteristics of Southern Indian Lake Before, During, and After Im- poundment and Churchill River Diver- sion. Can. J. Fish. Aquat. Sci. 41 :579-590. Hutchinson, G. E., 1957. A Treatise on Limnology. I. Geography, Physics, and Chemistry. John Wiley and Sons, Inc., New York. 1015 p. Iwama, G. K. and A. F. Tautz, 1981. A Simple Growth Model for Salmonids in Hatcheries. Can. J. Fish. Aquat. Sci. 38:649-656. Jones, J. R. and M. V. Hoyer, 1982. Sport Fish Harvest Predicted by Summer Chlorophyll a Concentration in Midwestern Lakes and Reservoirs. Trans. Amer. Fish. Soc. 111:176-179. ~enings, J. P. and R. D. Burkett, 1986. The Production Patterns of Sockeye Salmon, Oncorhynchus nerka, Smolts Relative to Temperature Regimes, Euphotic Volume, Fry Density, Forage Base Within Alaska Lakes. Proceed- ings, 11 Sockeye 85 11 , International Sockeye Salmon Symposium, Nanaimo, British Columbia, Canada. November 18-22, 1985. (In Press.) ~enings, J. P., R. D. Burkett, G. B. Kyle, 1985. Limnology and Fisheries Evidence for Rearing Limitation of Sockeye Production in Crescent Lake, Southcentral Alaska, (1979-1982). Alaska Department of Fish and Game, Juneau, AK. FRED Technical Report No. 57. 113 p. ~enings, J. P., J. A. Edmundson, J. M. Edmundson, and G. B. Kyle, 1986. Limnology Field and Laboratory Manual: Methods for Assessing Aquatic Produc- tion. Alaska Department of Fish and Game, Juneau, AK. FRED Technical Report No. (In Press) . ~le, G. B., 1985. Abundance and Distri- bution of Juvenile Sockeye Salmon in Five Lakes Throughout Southcentral Alaska During 1982 and 1983 with Comments on the Use of Hydroacoustic Survey Information. Alaska Department Fish and Game, Juneau, AK FRED Technical Report No. (In Press). Lloyd, D. S., J. P. Koenings, and J. D. LaPerriere, 1986. Effects of Turbidity in Freshwaters of Alaska. North American J. of Fish. Man. (In Press). Likens, G. E. [ed.], 1985. An Ecosystem Approach to Aquatic Ecology. Springer- Verlag, New York, N.Y. 516 p. ~thisen, 0. A., 1972. Biogenic Enrich- ment of Sockeye Salmon Lakes and Stock Productivity, Verh. Internat. Verein. Limnol. 18:1089-1095. ~Cabe, G. D. and J. W. O'Brien, 1983. The Effects of Suspended Silt on Feeding and Reproduction of Daphnia pulex. American Midland Naturalist 110:324-337. 193 McConnell, W. J., S. Lewis, and J. E. Olson, 1977. Gross Photosynthesis as an Estimator of Potential Fish Production. Trans. Amer. Fish. Soc. 106:417-423. Narver, D. W, 1970. Diel Vertical Move- ments and Feeding of Underyearling Sockeye Salmon and the Limnetic Zoo- plankton in Babine Lake, British Columbia. J. Fish. Res. Board Can. 27:281-316. Nelson, P. R., 1958. Relationship Between Rate of Photosynthesis and Growth of Juvenile Red Salmon. Science. 128:205-206. Oglesby, R. T., 1977. Phytoplankton Summer Standing Crop and Annual Pro- ductivity as Functions of Phosphorus Loading and Various Physical Factors. J. Fish. Res. Board Can. 34:2255-2270. Paloheimo, J. E. and L. M. Dickie, 1966. Food and Growth of Fishes. II. Effects of Food Temperature on the Relation Between Metabolism and Body Weight. J. Fish. Res. Board Can. 32:869-908. Patalis, K. and A. Salki, 1984. Effects of Impoundment and Diversion on the Crustacean Plankton of Southern Indian Lake. Can. J. Fish. Aquat. Sci. 41: 613-637. Peltz, L., 1985. Evidence for Temperature Limitation of Sockeye Salmon, Oncorhynchus nerka, Growth in Hugh Smith Lake, Southeastern Alaska. Alaska Department of Fish and Game. Unpub- lished Report. 23 p. Pennak, R. W., 1978. Fresh-Water Inverte- brates of the United States, 2nd Edition. John Wiley and Sons, New York 803 p. Prepas, E. E. and D. 0. Trew, 1983. Evaluation of the Phosphorus-Chlorophyll Relationship for Lakes Off the Precambrian Shield in Western Canada. Can. J. Fish. Aquat. Sci. 40:27-35. Ragotzkie, R. A., 1978. Heat Budgets of Lakes, p. 1-18. In A. Lerman (ed.) Lakes, Chemistry,IGeology, Physics. Springer-Verlag, New York, N.Y. Reimann, B., 1978. Carotenoid Interference in the Spectrophotometric Determination of Chlorophyll Degradation Products From Natural Populations of Phytoplankton. Limnol. Oceanogr. 23:1059-1066. Richman, S. and S. Dodson, 1983. The Effect of Food Quality on Feeding and Respiration by Daphnia and Diaptomus. Limnol. Oceanogr. 28:948-956. Rigler, F. H., 1982. The relation Between Fisheries Management and Limnology. Trans. Amer. Fish. Soc. 111:121-131. Sakamoto, M., 1966. Primary Production by Phytoplankton Community in Some Japanese Lakes and its Dependence on Lake Depth. Arch. Hydrobiol. 62:1-28. Saunders, G. W., F. B. Trama, and R. W. Bachmann, 1962. Evaluation of a Modified 14C Technique for Shipboard Estimation of Photosynthesis in Large Lakes. Univ. Michigan, Great Lakes Res. Div., Spec. Rep. No.8. 61 p. Schindler, D. W., 1971. Light, tempera- ture, and Oxygen Regimes of Selected Lakes in the Experimental Lakes Area, Northwestern Ontario. J. Fish. Res. Bd. Canada 28:157-169. Schindler, D. W., 1978. Factors Regula- ting Phytoplankton Production and Standing Crop in the Worlds Freshwaters. Limnol. Oceanogr. 23:478-486. Shelbourn, J. E., J. R. Brett, and S. Shirahata, 1973. Effect of Temperature and Feeding Regime on the Specific Growth Rate of Sockeye Salmon Fry, Oncorhynchus nerka, \-lith a Consider- ation of Size Effect. J. Fish. Res. Board Can. 30:1191-1194. Smith, V. H. 1979. Nutrient Dependence of Primary Productivity in Lakes. Limnol. Oceanogr. 24:1051-1066. Stockner, J. G., 1983. Personal Communi- cation. Department of Fisheries and Oceans, Fisheries Research Branch, West Vancouver Laboratory, 4160 Marine Drive, West Vancouver, British Columbia V7N 1N6. Strickland, J. D. H. and T. R. Parsons, 1972. A Practical Handbook of Seawater Analysis, 2nd Ed. Fish. Res. Board of Canada Bull. 167:310 p. Thomas, G. L., R. E. Thorne, J. McClain, and D. Marino, 1984. Hydroacoustic Measurements of the Density and Distri- bution of Juvnile Sockeye in Tustumena Lake, Alaska During 1984. University of Washington. Fish Res. Inst. Contract No. 14-16-0007-83-5271. Tizler, M. M., C. R. Goldman, and R. C. Richards, 1976. Influence of Sediment Inflow on Phytoplankton Primary Pro- ductivity in Lake Tahoe (California- Nevada). Int. Revue Ges. Hydrobiol. 61:169-181. 194 Verduin, J., L. R. Williams, V. W. Lambou, and J. D. Bliss, 1978. A Simple Equation Relating Total Phosphorus to Chlorophyll Concentration in Lakes. Verh. Internat. Verein. Limnol. 20:352. Vinyard, G. L., 1982. Feeding Success of Hatchery-Reared Kokanee Salmon When Presented With Zooplankton Prey. Prog. Fish Culturist. 44:37-39. Vollenweider, R. A., 1965. Calculation Models of Photosynthesis-Depth Curves and Some Implications Regarding Day Rate Estimates in Primary Production Measurements. Mem. !st. !tal. Idrobiol. Suppl : 18:425-457. Vollenweider, R. A., 1976. Advances in Defining Critical Loading Levels for Phosphorus in Lake Eutrophication. Mem. !st. !tal. Idrobiol. 33:53-83. Wetzel, R. G., 1975. Limnology. W. B. Saunders Co., Philadelphia. 743 p. Wilson, M. S., 1959. Calanoida. p. 738-794. In: W. T. Edmondson [ed.], Fresh-Water-Biology, 2nd ed. John Wiley and Sons, New York, N.Y. Yeatman, H. C., 1959, Cyclopoida. p. 795-815. In: W. T. Edmondson [ed.], Fresh-Water-Biology, 2nd ed. John Wiley and Sons, New York, N.Y. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 DEEP-LYING CHLOROPHYLL MAXIMA AT BIG LAKE: IMPLICATIONS FOR TROPHIC STATE CLASSIFICATION OF ALASKAN LAKES Paul F. Woods* ABSTRACT: Concerns over potential eutrophication of Big Lake have arisen because the lake's shoreline is extensively developed with residences, most of which have individual, on-site septic systems. Such concerns led to a study of the lake's physical, chemical, and biological characteristics during 1983- 84. The range in concentration from 418 chlorophyll a samples was 0.05 to 46.5 micrograms per liter. The east limnological station had significantly higher chlorophyll a concentrations in both years. The highest concentrations during each year at each station occurred in deep-lying chlorophyll maxima within the hypolimnion. Time-depth plots of chlorophyll a were used to illustrate the influence of sampling design on the calculation of the lake's mean chlorophyll a concentration. Three sampling designs that ranged from reconnaissance-level to in vivo fluorometric profiling yield mean chlorophyll a concentrations ranging from 1.16 to 4.96 micrograms per liter. The largest mean concentrations were calculated from samples collected throughout the euphotic zone; the other two sampling designs were restricted to the epilimnion and hence failed to sample the deep-lying chlorophyll maxima. (KEY TERMS: chlorophyll; trophic level; eutrophi- cation.) INTRODUCTION The proximity of Big Lake to the rapidly develop- ing communities of Anchorage, Wasilla, Eagle River, and Palmer has encouraged residential development and intensive, year-round recreational use of this 1,213-hm 2 lake in the Matanuska-Susitna Borough of southcentral Alaska (Figure 1). Maintenance of the lake's water quality is essential because Big Lake receives about 16,000 person-days per year of fishing effort for salmonids such as rainbow trout (Salmo gairdneri) and Dolly Varden char (Salvelinus malma); in addition, the annual escapement of sockeye salmon (Oncorhynchus nerka) from Big Lake has reached nearly 200,000 fish in recent years (Engel, L.J., Alaska Department of Fish and Game, oral commun., 1985 ). The rapid residential development of the lake's 27 -km shoreline and its numerous islands has created concern over the potential for nutrient enrichment, or eutrophication, of Big Lake. Most of the approximately 1,000 permanent lake residents dispose of their wastewater via individual, on-site septic systems. Such concerns prompted the U.S. Geological Survey and the Alaska Department of Natural Resources, Division of Geological and Geo- physical Surveys to conduct a cooperative study of Big Lake during 1983-84 with the following objec- tives: (1) to determine the lake's trophic state; (2) to interpret spatial and temporal trends in physical, chemical, and biological variables; and (3) to investi- gate the relation between phytoplankton primary production and selected limnological variables. An extensive data base of chlorophyll a concentrations was assembled during 1983-84 because this water- quality variable is germane to each of the three study objectives. This paper describes trends in depth and time for chlorophyll a concentrations at the east and west limnological stations of Big Lake during 1983-84. These data are then used to show the influence of sampling design on the trophic state classification of Big Lake. The 232-km2 drainage basin of Big Lake is char- acterized by low-relief terrain with elevations ranging from approximately 40 to 168 m. Glacial deposits of sand and gravel underlie much of the area; muskegs are common in low-lying areas. Mean annual pre- cipitation near Big Lake is 395 mm with July as the wettest month. July is also normally the warmest month (mean temperature = 14.4°C) and January is the coldest (mean temperature = -11.4° C). Ice as thick as 1 m covers the lake from late October through late May. Daylength varies from 5.5 hr on the winter solstice to 19.5 hr on the summer solstice. The east basin of Big Lake has a maximum depth of 15 m and contains the lake's major inlet, Meadow Creek, and the lake outlet into Fish Creek (Figure 2). The lake's deepest point, 27 m, is in the west basin. The 1,213-hm2 lake contains 112 hm 3 and thus has a mean depth of 9.2 m. *Hydrologist U.S. Geological Survey, 1209 Orca Street, Anchorage, Alaska 99501. 195 0 I 0 5 I I 10 10 15 20 MILES I I I I I 20 30 KILOMETERS Figure 1. Location of Big Lake within the Matanuska-Susitna Borough of Southcentral Alaska. METHODS Sampling for chlorophyll a was done from January 1983 through December 1984 at the two primary limnological stations (Figure 2), one each in the lake's east and west basins. Samples were collected biweekly during the open-water season and approxi- mately monthly when ice covered the lake. During 1983, numerous discrete-depth samples were obtained with an opaque Van Dorn sampler. In 1984, in vivo fluorometry was used to profile the water column for a qualitative description of chlorophyll a. Discrete-depth samples were then obtained at representative depths. The samples were analyzed fluorometrically for chlorophyll a, with correction for pheophytin, per methods in Wetzel and Likens (1979). Numerous other variables were also sampled during this study; these data are available in publications of the U.S. Geological Survey (1984, 1985). RESULTS AND DISCUSSION Chlorophyll a concentrations, in JlgL·1 (micro- grams per liter), at the east limnological station 196 ranged from 0.05 to 46.5 (mean = 3.9, n = 101) in 1983 and from 0.05 to 9.2 (mean = 2.6, n = 102) in 1984 (Figure 3). At the west limnological station, chlorophyll a concentrations ranged from 0.07 to 8.0 (mean = 1.8, n = 108) in 1983, and from 0.14 to 6.4 (mean = 1.7, n = 107) in 1984 (Figure 3). Para- metric statistical tests indicated that chlorophyll a concentrations at the east limnological station were significantly higher than those at the west limno- logical station for both years. Chlorophyll a con- centrations at the east limnological station alone did not differ significantly from 1983 to 1984; the same result was obtained for the west limnological station. The lowest concentrations of chlorophyll a were measured between October and March, under ice cover. In mid-March, chlorophyll a concentrations began to increase in the upper 5 m of the lake even though an ice cover nearly 1 m thick still covered the lake. Chlorophyll a concentrations within the upper 5 m of Big Lake did not exceed 3.5 and 5.5 J.lgL·1 , respectively, at the west and east limnological stations throughout the study. The annual maximum chloro- phyll a concentration was usually attained by the end of June, within a month of the melting of the lake's ice cover. During late July to early September 1983, however, the east limnological station experienced a second pulse of chlorophyll a that reached a peak concentration of 46.5 .J.lgL-1 at a depth of 9 m. The yearly maximum concentration of chloro- phyll a at each limnological station occurred near the lower depth limit of the euphotic zone, defined here as the depth at which photosynthetically active radi- ation is 1 percent of that incident upon the lake's surface. Big Lake's deep-lying chlorophyll maxima were situated within the hypolimnion, in that the depth of the euphotic zone normally exceeded the depth of the epilimnion during late May through mid- September. Deep-lying chlorophyll maxima have been reported to occur under similar conditions in many lakes (Fee, 1976; Priscu and Goldman, 1983; Pick et al. 1984). These chlorophyll maxima have frequently been composed of flagellated, colonial algae of the subphylum Chrysophyceae (Fee, 1976; Pick et al. 1984). The 1983-84limnological study at Big Lake included sampling for phytoplankton species composition and biomass at selected depths. The phytoplankton samples that were taken at the 10-m depth during the occurrence of the deep-lying chlorophyll maxima were dominated numerically by unidentified microalgae. The largest contribution to algal biomass was from the Chrysophyceae which were commonly represented by genera such as Chromulina, Chrysochromulina, Dinobryon, Kephyrion, and Mallomonas, all of which are flagellated organisms. The occurrence of large concentrations of chloro-, phyll within a hypolimnion, that also contains part of the euphotic zone, may reveal the presence of photosynthetic bacteria; however, anoxic conditions also must exist because such bacteria are obligate anaerobes ( Rheinheimer, 197 4 ). The deep-lying EXPLANATION • Primary limnological station \1 1 Secondary limnological station and number 'oJ(),. '¢' Solar radiation monitoring station Bathymetric contour interval 3 meters 61°32'30" ~~~.5~~~0-----------=========~2KILOMETERS Y:iio. -==--==..,..o============:::::i1 Ml LE Figure 2. Bathymetric map of Big Lake and location of the limnological stations. chlorophyll maxima in Big Lake developed in well- oxygenated water; thus photosynthetic bacteria did not account for the large concentration of chloro- phyll a. The sampling design for chlorophyll a at Big Lake was only one of a myriad of possible designs which might have been employed. The three project objec- tives required a large amount of chlorophyll a data in order to adequately investigate Big Lake's physical, chemical, and biological limnology. If the project objective had been a reconnaissance level survey, then perhaps only single, near-surface samples would have been obtained at selected dates during the ice-free seasons. If the project had been designed along the lines of a Phase 1 Diagnostic-Feasibility Study of the U.S. Environmental Protection Agency's Clean Lakes Program, then chlorophyll a samples would have been obtained at several depths within the epilimnion on a biweekly basis from May through August (U.S. Environmental Protection Agency, 1980). The sampling design actually employed at Big Lake was based in part on the results of reconnaissance sam- 197 pling conducted by the author at Big Lake in 1982 that revealed metalimnetic dissolved-oxygen maxima. Additionally, the author has sampled other south- central Alaskan lakes and has found evidence of vertical stratification of chlorophyll a concentrations. The distribution of chlorophyll a in Big Lake (Figure 3) can be used to illustrate the influence of sampling design on the mean concentration of chloro- phyll a obtained under three different sampling designs. The first sampling design (I) is that actually employed at Big Lake during 1983-84. The data for the other two sampling designs are subsets of data taken from the first sampling design. The second sampling design (II) employs three samples taken biweekly from the epilimnion. Single near-surface samples on a biweekly basis constitute the third sampling design (III). The period analyzed for the three sampling designs was restricted to late May through September. A statistical comparison of results obtained under the three sampling designs is shown in Table 1 for the east and west limnological stations. The mean concentrations of chlorophyll a .... = = ~ 5 w ..... w :::; :!': :r.· t;: w 0 1 J East Station West Station M A M A 1983 1984 Figure 3. Time-depth plot of chlorophyll a concentrations, in micrograms per liter, during 1983-84 at the east and west limnological stations. D (/) a: w ..... w :::; z :r." t;: w 0 (/) a: 10 ~ w :::; :!': :r.· t;: w 15 ° obtained with the second and third sampling designs are different for each station and year; however, the 95-percent confidence limits overlap and thus the mean concentrations are not significantly different. The mean concentrations of chlorophyll a obtained with the first sampling design are significantly larger than those obtained under the other two sampling designs except in two instances. There is a slight overlap in the 95-percent confidence limits for sampling designs I and II at the east limnological station during 1984. A similar overlap exists between sampling designs I and III at the west limnological station during 1983. A wide disparity also exists for the ranges of chlorophyll a concentrations obtained with the three sampling designs. The significantly larger mean values and the wide ranges in chlorophyll a concentrations obtained under the first sampling design are largely attributable to the deep-lying chlorophyll maxima in the hypolimnion that were not located by the other two sampling designs. The mean concentration of chlorophyll a in a lake during the summer has often been used to categorize the lake's trophic state as either oligotrophic, meso- trophic, or eutrophic. An international workshop on the control of eutrophication recently concluded that chlorophyll a is a widely cited and accepted indi- cator of lake trophic state (Rast, 1981). Numerous trophic state categorizations based on chlorophyll a have been proposed; one of the more recent is that of Hern et al. (1981). These authors classify a lake as oligotrophic if its mean chlorophyll a concentration within the euphotic zone is less than 2.3 pgL-1 ; it is eutrophic if chlorophyll a exceeds 6.4 pgL-1 ; and the lake is mesotrophic if it contains between 2.3 and 6.4 pgL·1 . Application of this trophic state categoriza- tion to the chlorophyll a data in Table 1 yielded the following results. Sampling designs II and III cate- gorized both limnological stations as oligotrophic during 1983-84 because the mean concentrations of chlorophyll a did not exceed 1.93 ).lgL·1 . Sampling design I also yielded an oligotrophic ranking for the west limnological station in both years; however, the east limnological station was mesotrophic in both years. The differences in the results obtained by the three sampling designs have important implications for limnological studies in Alaska because deep-lying chlorophyll maxima do not appear to be unusual features. Woods ( 1985) reported the occurrence of metalimnetic dissolved-oxygen maxima, which may indicate deep-lying chlorophyll maxima, in six of nine small lakes in southcentral Alaska that he studied in 1982-83. Deep-lying chlorophyll maxima were verified by discrete-depth sampling in four of the nine lakes. A 1985 study of the primary produc- tivity of Wasilla Lake in Wasilla, Alaska found a deep- lying chlorophyll maximum at the lower depth limit of the euphotic zone (Vaught, K.D., University of Alaska, written commun., 1985). The Alaska Depart- ment of Fish and Game has also discovered deep-lying chlorophyll maxima in southcentral Alaskan lakes (Koenings, J.P., Alaska Department of Fish and 199 Game, oral commun., 1985). Given the frequent incidence of deep-lying chlorophyll maxima, the trophic state of many Alaskan lakes may be under- estimated if chlorophyll a sampling is restricted to the epilimnion or near surface region such as might occur in a reconnaissance-level study of numerous lakes. Table 1. Comparison of chlorophyll a concentrations obtained by three different sampling designs during late May through September at Big Lake Chlorophyll a concentrations, in micrograms per liter, Sampling design at indicated limnological station and year and statistics East 1983 West 1984 DESIGN I 1 Mean 4.96 1.89 2.81 1.92 95% C.I. 3.13-6.79 1.56-2.22 2.38. 3.24 1.69. 2.15 Range 0.32-46.5 0.24-8.05 0.34-9.17 0.53. 6.41 n 71 78 71 76 DESIGN 112 Mean 1.92 1.26 1.93 1.41 95% C.l. 1.37. 2.47 1.10-1.42 1.45-2.41 1.20 -1.62 Range 0.93-8.52 0.33-2.04 0.79-7.17 0.53. 2.88 n 30 30 27 27 DESIGN III a Mean 1.48 1.20 1.39 1.16 95% C.I. 1.07-1.89 0.84-1.56 1.10. 1.68 0.90-1.42 Range 0.93-2.67 0.33-2.04 0.79 -1.83 0.53-1.49 n 10 10 9 9 1 Numerous samples (5 to 10) from euphotic zone taken biweekly. Profiles of water temperature, photosynthetically active radiation, and in vivo fluorescence used to select depths. 2 Three samples from epilimnion (top, mid, bottom depths of epilimnion) taken biweekly. 8 Single, near surface sample taken biweekly. REFERENCES CITED Fee, E.J. 1976. The Vertical and Seasonal Distribu- tion of Chlorophyll in Lakes of the Experimental Lakes Area, Northwestern Ontario: Implications for Primary Production Estimates. Limnology and Oceanography 21(6): 767-783. Hem, S.C., V.W. Lambou, L.R. Williams, and W.D. Taylor. 1981. Modifications of Models Predict- ing Trophic State of Lakes: Adjustment of Models to Account for the Biological Manifesta- tions of Nutrients. EPA-600/3-81-001. Pick, F.R., C. Nalewajko, and D.R.S. Lean. 1984. The Origin of a Metalimnetic Chrysophyte Peak. Limnology and Oceanography 29(1): 125 -134. Priscu, J.C. and C.R. Goldman. 1983. Seasonal Dynamics of the Deep-Chlorophyll Maximum in Castle Lake, California. Canadian Journal of Fisheries and Aquatic Sciences 40(2): 208 -214. Rast, W. (ed.). 1981. International Workshop on the Control of Eutrophication. International Institute for Applied Systems Analysis, Laxen- burg, Austria. October, 1981. Rheinheimer, G. 1974. Aquatic Microbiology. J. Wiley and Sons, New York. U.S. Environmental Protection Agency. 1980. Clean Lakes Program Guidance Manual. U.S. Environ- mental Protection Agency, EPA-440/5-81-003. U.S. Geological Survey. 1984. Water Resources Data, Alaska Water Year 1983. U.S. Geological Survey Water-Data Report AK-83-1. U.S. Geological Survey. 1985. Water Resources Data, Alaska, Water Year 1984. U.S. Geological Survey Water-Data Report AK-84-1. 200 Wetzel, R.G. and G.E. Likens. 1979. Limnological Analyses W.B. Saunders Co., Philadelphia. Woods, P.F. 1985. Limnology of Nine Small Lakes, Matanuska-Susitna Borough, Alaska, and the Survival and Growth Rates of Rainbow Trout. U.S. Geological Survey Water-Resources Investi- gations Report 85-4292. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 FACTORS INFLUENCING THE QUALITY OF SNOW PRECIPITATION AND SNOW TBROUGHFALL AT A SIERRA NEVADA SITE Sheri Woo and Neil Berg 1 AlmRACf: Regression analysis showed precipitation type (rain, snow, mixed rain and snow, hail/graupel) and latitudinal stom path to be iuportant factors influencing both precipitati~ conductivity and nitrate solute concentration (r = 0.88 and 0.~. respectively) at a subalpine site near Lake Tahoe, California. Weaker relationships existed between both precipitation type and !Ycipitation pH and sulfate solute concentration (r = 0.50 and 0.32, respectively). StOJlll duration, interstOJlll period, and precipitation amant and intensity were relatively 'IJiliDportant explanatory variables. Vegetative influence on snow throoghfall under forest canopy was minimal, coopared to unaffected precipitation. Solute concentrations of H, ~ 4 , Cl, and Ca in snow drip fran no mjor conifers, lodgepole pine and red fir, were generally greater than precipitation in an open site. Contrary to studies in rain-daninated environments, snow drip cm.dstries between fir and pine were statistically different only for Ca. (KEY 'IIBfS: ata>spheric deposition; lodgepole pine; throughfall chemistrr, red fir.) Aquatic ecosystem responses to acidic deposition have been documented in Scandinavia, Canada, and the northeastem United States. Water quality degradation has resulted in fish kills 8Dd the inhibitioo of aDphibian reproductioo in snomelt pools (Leivestad 8Dd Muniz, 1976; Pough, 1976). Terrestrial ecosystem responses have also been documented; declines in forest growth 8Dd increases in tree mrtality have been associated with atDDspheric deposition (Linzoo 8Dd Stelly, 1985). Processes regulating ata>spheric chemical loading differ in teaperate versus frigid climates. Theoretical and eupirical research shows evidence of an acidic ••pulse" or ''spike" during snowmelt (Colbeck, 1981; Johannessen 8Dd Henriksen, 1978). Because concentratioo of solutes during snow met811X)rphism is likely, research m the effects of " acid rain" camwt be directly extrapolated to effects of acidic snow. The extent of acidic deposition in upland areas in California is largely UllkDown but could be critical. Currently, in the Sierra Nevada of California, precipitation chemistry is roatinely mnitored at seven sites by no governmental agencies, the National Ata>spheric Deposition Program 8Dd the California Air Resoorces Board. However, both agencies mnitor no sites that lie in proximity, so m a large scale, only five locations represent f montainous region of approximately 486,600 kni • For cooparison, this area is greater than the states of New llaDpshire and VelliiJD.t camined. Although the anthors realize that other mnitoring sites my exist, known sites have not been located to detemine chemical spatial variation. Although the data base is inadequate to allow a cooprehensive analysis of wetfall chemistry by precipitation type, the available data suggest that the pH of rainfall is lower than that of snowfall (Table 1). All Dllllitoring stations in the Sierra Nevada saDple precipitatim in forest clearings or "open sites." Standardization of saDpling is an advantage inherent in Dllllitoring nenorks. However, if detemination of chemical loading into soil and water is a research or management objective, then precipitation saDpling in forested areas is necessary. Abou.t 451fo of the snow zone on the westem slope of the Sierra Nevada is forested. Therefore, at ground level, little mre than half of the area is currently Dllllitored for acidic depositim. Without accounting for forest influences, chemical loading estimates, used to predict water quality effects, are inaccurate. ~spectively, Hydrologic Teclmician 8Dd Supervisory Hydrologist, Pacific Sollthwest Forest and Range El:periment Station, Forest Service, U.S. Department of Agriculture, P.O. Box 245, Berkeley, California 94701. 201 TAaE 1. Suamary of Acidic Precipitation Research in California Location .Mean !if Coostituents Saaple type Reference Berkeley 4.66 004 m3 m4 Ca Rain McColl19fO Hopland 5.10 II II II Tahoe City 5.17 II II II Davis 5.20 II II II Menlo Park 5.20 004 m3 Ca Rain Kennedy et al. 1979 Los Angeles 4.49 004 m3 m4 C'A Rain &brgan and Liljestrand 19f0 Riverside 4.97 II II II Pasadena 4.41 II II II Sierra Nevada (51 sites) 5.8 004 Ca Snow . Feth et al. 1964 Sierra Nevada (26 sites) 5.6 004 N>3 Ca Snow Brown and Sk:au 1975 Los Angeles 3.6 N>3 004 Cl m4 Na Mist ~r et al. 1983 II 3.3 and K Ca )fg Pb Fe Ni Fog " Warner Springs 3.74 m3-N m4-N Ca )fg Rain mlis et a1 1983 Central Sierra Nevada Cl P N>3 004 Rime Nachlinger 1984 Want Valley 5.1 Soow? Leonudeta1.1981 Eastern Sierra Nevada 3.7-4.9 m4 Na Ca )fg K N>3 Rain Melack et a1 1982 II 5.7 and 004 Cl Snow II Central Sierra Nevada 5.3 N>3 004 Cl A1k Na Snow Berg and Woo 1985 II 4.9 and Ca Mg K Rain II San Nicolas Island s.o N>3 004 Rain Calif. Air Resources ~ 4.6 II II Board 1984 Pasadena II II II II m Monte II II II " Mt. Wilson 4.8 II II II a pH values are presented as ~ appear in the references. Salle are averaged pH, and sane are converted to H as J.leq/L, averaged, then reconverted back to pH. Studies of precipitatim chemistry in forests are DOIDeroos, but mst research relates to mtrient cycling problems, not acidic depositim. Hydrogen ion is an lDli.Dportant mtrient, and mst early studies did not Dlllitor pH. Hence, few studies have investigated the additim of hydrogen ion to snow surfaces UDder a forest canopy. Though teminology in the literature varies, the authors prefer that of Zinke (1967) and Kittredge (1953). "Precipitation" is wetfall saupled in a forest clearing. "Forest tbrooghfall" is that precipitation s&Dpled UDder the forest canopy during a st01m event. In a snowfall event, "drip11 refers to intercepted snow precipitation that falls or drips from vegetatim after the st01m event. 202 Most througbfall studies doctmmt precipitatim as rain. A few studies differentiate snow fran rain, but their collection _,thods are lDlClear (Feller; 1977, Verry and TiDmlns, 1977; Jones, 1984) (Table 2). No studies involving the chemistry of snow precipitatim, tbrooghfall, and drip are known for the 110011tainous areas of California. This paper examines the chemistry of two C<JilKXUmts of the hydrologic cyt:le: snow precipitatim or snow tbrooghfall and drip. Further, explanations for observed chemical differences are sought. Specific objectives of this study are • To detemine which, if parameters are related chemistry. Parameters any, _,teorologic to precipitatim considered are TAIIB 2. S1mmary of 'Ihrooghfall pH Studies by Forest Type. Forest type ~ip. = Other constituents Reference Alder Conifer Mixed N>3 lfi4 org.N II II Tarrant 1968 II II b Sugar DBPle 6.1 6.1 6.1 4.06 6.0 6.1 6.1 4.85 Ca .Mg K Na N>3 lfi4 org .N P04 004 Cl Eaton et al. 1973 Yellow birch 4.06 II II Beech 4.06 II II Eucalyptus 5.0 Red oak.c 5.6 Live oak. 5.6 5.30 5.22 5.4 4.45 4.10 3.95 5.7 4.5 5.28 4.4 K Na lfi4 Ca .Mg Cu Fe Mn :zn N>3 Il2.ro4 C1 IID3 004 Total reducing sugars. Total N. polyphenols ~11 and lllsh 1978 Malcolm et al. 1968 II Longleaf pine 5.6 II II Beech 5.2 Na K Ca .Mg Mn Cl 004 P04 Total N Nihlgard1970 II II Spruce d 5.2 Douglas-fir 5.20 Mixed westem hem-4.5 lock. redcedar (snow) N>3 Kjeldahl-N Na K Mg Ca Alk K Na Mg Ca Fe .&h lfi4 Cl Il2.ro4 N>3 004 IID3 SiOl Sollins 19~ Feller 1977 : Precipitation is rain except for final entry. Converted to pH lDlits fran uw/L. c Picked leaves 11ere swirled in distilled water for 5 min. to sinul.ate canopy drip. d Not standanl procedure. Converted to pH 1D1i ts fran J.18/L. precipitation amount. intensity. density. and type; stonn duration and type; and length of interstonn period. To examine the vegetative influence on precipitatioo chemistry by c<Jit)Uing chemical differences between precipitation and forest throughfall. precipitation and. drip under red fir. precipitation and drip UDder lodgepole pine. and red fir and lodgepole pine drip. The study was perfomed at the liD\ Forest Service's Central Sierra Snow Laboratory (CSSL). elevation 2103 m. fran 18DIJ&1"Y 1 to August 31. 1984. CSSL is located 2S km northwest of Lake Tahoe m the 11est side of the Sierra crest. Precipitation events are strongly influenced by maritime Eteorologic cooditions. Vegetation at this location consists of red fir (Abies magnifica) and lodgepole pine (Pinus contorta). with manzanita and ceanotlms understory. Trees range up to 38 m in 203 height and are approximately 100 years old. Soils are slightly acidic. naturally well-drained coarse loams having moderately high subsoil pemeability and very low substratum pemeability. Available moisture holding capacity ranges fran 3 to 7 em. )brltoring was conducted at two sites. a 0.3 ha clearing immediately west of CSSL and a forest site located abou.t 2SO m north of CSSL. Doring each stonn event. precipitation and throughfall samples were collected fran the open and forest sites. After each event. drip samples were collected UDder red fir and lodgepole pines. Multiple samples were collected througboot each stonn. and subsamples (or "replicates 11 ) were also taken. When precipitation occurred as snow. precipitation and throughfall samples 11ere taken fran the snowpack surface at each 10 to 12 em snow acClJIIIllation interval. 'Ihe samples 11ere packed into linear polyethylene (Im) bottles. capped. and later melted at roan teq,erature. For snow events depositing less than 10 em depth and for precipitation types other than snow. samples 11ere collected fran plastic lined collector boxes. not the snow surface. Sample collectors 11ere not exposed to dryfall except as it may have occurred corumrrent with wetfall or drip. After a snow stOJ.'IIl, drip was collected using similarly lined plastic boxes. Two boxes were used for each tree species, and boxes were placed under the tree driplines, 1 to 1.5 m away fran the trunks. Sauples continued to be collected lDltil snow was no looger visible in the trees. Tree selection depended on age and proximity to trees of other species. Chemical contributions fran old, very mature trees are often greater than fran younger trees (Mecklenburg and Thkey, 1964). Therefore, boxes were placed in small stands of pure fir or pine having trees of medilDD height and age. Precipitation, throughfall, and drip samples were analyzed for pll, cooductivity, and alkalinity at CSSL within 24 hours after collection. Acidity was measured with a Fisher Analog pH meter equipped with a coomination flowing junction reference electrode. (Trade names and cc:mnercial products are mentioned solely for infomation. No eodorsement by the U.S. Department of Agriculture is iDplied.) Conductivity was measured with a Weather-Measure Wheatstone bridge cooductivity meter. Alkalinity was determined by color titration. In addition, precipitation, throughfall, and drip samples were analyzed for major cations and anions (Ca, Mg, Na, K, m3 , oo4 , and Cl). Samples were stored frozen in rm bottles and cation solute concentrations were later determined by atomic absorption spectrophotauetry. Nitrate and sulfate solute concentrations were analyzed by mlecular absorption spectroscopy. Chloride solute concentrations were determined by color titration. Precipitation 810001lt, duration, and time fran previoos stOJ.'IIl were read fran the recording charts of a weighing bucket precipitation gauge. Precipitation intensity was calculated fran the same charts, using the time span of each sample as the base interval. Snow density was measured by two snow boards. Precipitation type was visually observed and recorded. Sto.on types, classification of stOl'IDS based on latitudinal source area and path, were assigned by observing the track of low pressure areas from satellite pbotos (MOnteverdi, 1976). Snowmelt volume was calculated fran stream discharge records for a tributary of Onion Creek, 7 Ian fran CSSL. Snowmelt timing was determined through analysis of runoff into snownelt lysimeters located in the open site at CSSL (Kattelmann, 1984). Variance in Chemistry Data Low solute concentrations of all monitored constituents led to uncertainty about the quality of 204 the chemistry data set. Analytical accuracy was determined by cmparing the SlDD of cations with the S1DD of anions, i.e. , by an ionic balance (Figure 1). A regressi~ using all 100 observations iDplies fair accuracy (r = 0.86), but the points are not evenly distributed throughout the range of the ionic sums. When a separate regression is run on those points with a CBfion sum of less than 200 Jleq/L (93 observations), r = 0.63. ~ .......... o- Q) 700 600 500 ..3--400 ::::;: :;:) (/) z 300 Q z <( 200 .·. . 100 ': ... . .:'\,' ·;i: .. . IONIC BALANCE . . (N = 100) ·tt,= ... • : 0~~~----r----.----.----.----.---~ 0 100 200 300 400 500 600 700 CATION SUM (~eq/Q Figure 1. Ionic balanc~ the S1DD of anions versus ~ S1DD of cations. r for all points is 0.86; r for points less than 200 Jleq/L is 0.63. High accuracy is iDplied by a straight line of slope = 1 and intercept = 0. To check analytical precision, 66 subsamples, or replicates, were analyzed. Unfortunately, standard deviations varied with the magnitude and range of each element (Table 3). For example, calcimt in drip samples exceeded calcilDD in precipitation (mean drip calcium = 49 Jleq/L; mean precipitation calcium = 6 J.leq/L). The standard deviation for drip calcium was correspondingly lower than for precipitation (g) drip calcilDD = 6 J.leq/L; ID precipitation calcilDD = 9 Jleq/L). In general, higher analytical 1Dlcertainty exists in precipitation than drip data. This is not surprising, if one considers that precipitation values were very close to analytical detection limits. Chemical data in low concentrations require careful analysis, given the lack of precision and accuracy inherent in the methods of current technology. Fxtreme care llllSt be taken in sampling and handling to prevent contamination. Also, the inherent variation in the data, which increases with decreasing concentration, sanetimes requires sophisticated statistical analyses. One should recognize that in IIBI1Y water chemistry data sets, uncertainty is high when concentrations are very low. The variability found within this data set is taken into accOWJ.t in all further analyses. TAIIE 3, Sabsllq)le Analytical Variation. by Saq;Jle Type, Samle tvoe If Alkb Coale M> 3 S)4 Cl Ca Mg Na 1: Open precipitation v;a 4.6 22.0 4.5 0.7 4.1 13.1 11.0 9.8 10.7 2.5 0.5 3.0 1.3 0.3 6.3 15.3 9.4 24.7 11.8 2.1 Til 21 16 21 18 13 16 15 14 7 16 Forest throngbfall !lean 4.6 15.6 5.4 0.8 7.8 19.1 13.3 22.4 13.1 3.5 S) 0,5 2.1 1.5 0.5 6.5 10.7 12.1 28.9 5.7 1.9 IF 15 12 14 12 9 11 11 10 6 10 Drip fran fir llean 20.0 1.0 36.0 o.o 66.0 43.3 49.4 37.4 41.1 23.8 S) 2.6 1.9 4.7 0.1 4.8 5.9 6.3 17.6 48.8 8.2 IF 8 5 8 5 5 5 4 4 4 4 Drip fran pine Mean 12.5 1.8 22.1 0.9 25.5 28.0 25.3 12.9 31.6 10.1 S) 1.6 2.0 0.1 2.4 9.4 2.8 3.0 10.0 3.8 IF 9 3 10 6 6 6 6 6 6 6 : Units are 11eq/L llDless otherwise noted. Alkalinity in 11eq/L as bicarlxmate. ~Conductivity in J.!S/cm at 25°C. Standard deviation. 0 Degrees of freedan. Meteorological Parameters Related to Precipitation Chemistry Multiple regression was used to evaluate the · relative iq>ortance of varioo.s meterological parmreters governing precipitation chemistry. Given the large standard deviations in chemical analyses, · regressions were run using only pH. conductivity, ID3, and ~ 4 as the dependent variables. COOductivity and pH are broad, general indicators of water quality. m3 and ~4 are suspected primary precursors to acidic deposition. Of the reterological parameters, precipitation intensity, duration. IIIOOWlt, density and type were chosen because they may characterize a stonn' s ability to scavenge pollutants. Stom type, a variable 205 classifying stoms by latitudinal path, was chosen because source area detennines the potential IIIOOWlt of pollutants entrained. Stom type represents source area and t~rature, and chemical differences are expected when these parameters vary. The time between storms, the interstonn period, is a measure of attoospheric loading before rai.nont or washout processes occurred. Lastly, a randan IIIJIOOer variable was included to evaluate the statistical relevance of each variable. Of the eight independent variables, all coobinations of five variables were run against each dependent variable (pH, conductivity, N> , and ~4 >. Two hundred eighteen coobinatimJ were possible, and subsets of 11 best fits" were selected. The frequency of occurrence of each independent variable in each regression was noted, and illportance evaluated. A variable occurring less frequently than the randan :number variable was considered 11 unillportant" (Table 4). TAIIE 4. Frequency of Meteorological Variables Related to Precipitation Clemistry. Meteorological Deoen.,t variable variables pH Coal. M>3 S)4 Intensity 8 16 9 s Aaomt 14 16 16 s Duration 11 16 9 s Jnterstom period 10 16 16 s Snow density 8 10 9 s Precipitation type 23 42 27 16 Stom type b 18 16 9 s Randall Jllllli)er 7 16 9 s 2 0.45-<>.50 o.83-o.ss 0.75-().fk> r 0.27-il.32 SE 0.17 3.09 0.81 6.28 c 25 42 16 n 27 : Cooductivity in J.!S/cm at 25°C. c Randan Jllllli)er inclllded in the independent variable set. t«miler of values in the "best fit" subset. r 2 values were high for conductivity and N>~. but low for pH and ~ 4 • For pH, only SOlO or the variability could be accOWJ.ted for by the regression. Twenty-five cad>inations of the ~~ variables were within 0.05 of the highest r (0.50). Precipitation type and stom type were the two variables IDOst used, and therefore evaluated fS "iq>ortant." Conductivity had the highest r value (0.88); the only inportant variable was precipitatim type. m3 regressions were similar to the conductivity results, with precipitation type the mly iq>ortant factor. <Dl.y 32Jt of the ~ 4 variation could be accounted for by the regression. Precipitation type and stonn type were again the variables occurring JIX)St frequently. Precipitation type and. to a lesser extent. stom. type were i.Dp>rtant variables related to precipitation chemistry. Regression analysis serves to identi~ relationships between variables. it cannot ensure cause and effect between variables. However. it is not unreasonable to speculate that different precipitation types have different scavenging capabilities. and that precipitatioo type thereby influences precipitation chemistry. In this data set. four types of precipitation were classified: snow. rain. mixed rain and snow. and hail/graupel. Upoo inspectioo. we observe that rain and hail/graupel are associated with lower pH and higher conductivity. m3. and 004. Precipitatioo type may. alternatively. be the product of atm>spheric conditions that affect precipitation chemistry. Atm>spheric teaperature and humidity are factors governing precipitation type. perhaps one of these factors directly governs precipitation chemistry. That stom type is a parameter of secoodary i.Dp>rtance supports this interpretation. Low latitude stol:IDS are generally wanner. with low pressure paths over Hawaii and soothem Califomia. High latitude stOl:lllS generally ccme fran the Gulf of Alaska. and are usually quite cold. Mid-latitude stOl:lllS. as the tem iq)lies. are in the middle range of stom. paths and teaperatures. Vegetative Influences oo Precipitation Chemistry Field o.enatu.s. Precipitation. throogbfall. and drip were collected for 11 stol:IDS. A significant delay occurred (scmetimes 2 days) between the collection of open precipitation and forest throughfall during a stom.. and fir and pine drip collected after a stom.. The separation of forest tbroughfall and drip was justified; snow generally stayed oo vegetation ooce in cootact with it. Therefore. snow falling to the forest floor during an event had little or no contact with vegetation. This generalizatioo is true because snow in this area is often heavy and wet. wind rarely was strong enough to blow snow off vegetatioo. Drip sauples contained water in both solid and liquid fom.. Differences in drip phase were apparent between tree species. Fir needle and branch configuratioo held Dl>re snow longer relative to pine. Fir drip was often liquid in fom. while pine drip tended to be in snow "clUDps." Often. fir drip was yellow. seeming to correspond to the 8111)0Jlt of liquid in the drip sauple. Drip which fell as liquid was yellower. Yellow drip was never observed tmder pines. Four ]zypotheses were fOl:llled oo the basis of the reviewed literature. The statistical significance of chemical differences between precipitation and 206 forest throogbfall. precipitatioo and red fir drip, precipitatioo and lodgepole pine drip. and fir and pine drip were examined by paired t-testing. Jl«> chemical trends within stol:IDS were observed. so coostituent coocentrations within stOl:lllS were averaged to obtain a representative chemical signature for each stom.. By averaging values for each stol:lll. data files of 10 coostituents for 4 sauple types for 11 stOl:lllS were created. Jlllsalts &::a. ~is n.ti.Dc. Results fran paired t-tests for the four ]zypotheses are SlJIIIIIIU'ized (Table 5) • Each colODD corresponds to one of the tested Jzypotheses: note that one-and two-sided t-tests were euployed. Delta refers to the mean differences calculated fran n tlllllber of stol:IDS. Altboogh n is always small ( < 9 in all . cases). there were fran 2 to 7 sauples averaged for each stom.. An arbitrary significance level was chosen ( ot = 0.05). but the reader is encouraged to assign "significance" based oo the mean differences and standanl errors. and oo management objectives. The uncertainty in scme of the analytical methods casts doubt oo the precisioo and accuracy of mean differences and significance levels. In the discussioo of the four ]zypotheses that follows. the cooclusions are teapered by the magnitude of analytical variation. TAIU!: S. Mean Differences in Cbemistry between SDow Precipitatim, Forest 'Diroughfall, and Drip frtm Rad fir and Lodgepole pine. a Precip-t~ Precip-dripfir c Precip-drip pine d Dripfi..r;line e Element delta n delta n delta n delta n H -il.18 6 -12.20 8. -9.05 8. 3.16 8 A1k. -1.39 6• 10.27 6. 8.41 s. -{). 79 s Cood. -{).83 6 -19.SO 8• -1S.16 8• 4.31 8 tm -{).18 s 0.3S 6 -{).66 6 -1.01 6 !1)4 -4.10 6 -23.70 6. -7.63 6. 16.00 6 a. -11.40 6 .Jll. 70 6. -8.60 6 13.10 6. CA -7.20 6 ~.10 s. -7.30 6 16.20 s. )G ~.40 6 -17.10 s -1.45 6 14.20 s NA -1.22 4 -23.20 s -10.SO 6 13.lll s IC -1.30 6 -16.10 s. -4.40 6. 10.lll s a Delta = aritbaetic differeuce, n = lllllliler of stOI:IIIS, • = significance at ""'= 0.05, unita: allalinity in 11eq/L as IID3, cooductivity in !18/ao, all others in 11eq/L. b lb: DO chemical differeuce between open precipitation and forest throughfall. Hl: difference exista. c lb: DO chemical differeuce between open precipitation and drip frcm red fir. Hl: precipitatioo alkalinity > red fir drip alkalinity. Hl: precipitation H, cooductivity, tm, !1)4, Cl, Ca, Mg, Na, IC < those of red fir drip. d lb: DO chemical differeuce between open precipitation and drip frcm lodgepole pine. Hl: precipitatioo alkalinity > lodgepole pine alkalinity. Hl: precipitation H. cooductivity, tm, !1)4, Cl, ea. Mg. Na, IC < those of pine drip. e lb: DO chemical differeuce between drip frcm fir and pine. Hl: fir drip allalinity < pine drip alkalinity. Hl: fir drip H, cooductivity, tm, !1)4, Cl, Ca, Mg, Na, IC } those of pine drip. Plecipitatiaa. ..t Fonst '.lllroaaJafaU. At o~-= 0.05, oo statistical differences existed between precipitation fran the 11 open'' site and forest throaghfall for any constituents except alkalinity (Table 5) • Standard deviations for precipitation mi throoghfall were small for H, alkalinity, and cooductivity with respect to their mean values, bot were large for m3 , 00 4 , Cl, Ca, Mg, Na, and t (Table 3). lberefore, we cannot definitively cooclude that no differences exist for the anions mi mjor cations. However, all mean differences are fairly small (Table 5, col'UIIIl 1), and in very few cases wonl.d a difference of the mgnitude cited have any practical meaning (note that units for cations and anions are IJeq/L) • We cooclude that no practical difference exists between precipitation : and forest drip H, COIIductivity, and alkalinity, and that indications point to no difference in N> , 9>4, Cl, Ca, Mg, Na, and K as well. 3 A difference between open precipitation and torest throoghfall was not anticipated, because forest throoghfall had little or no contact with vegetation. 'Ibis result agrees with that of Janes (1984). hecipitatiaa. ..t Jlrl4 Fir Drip. Statistically significant differences between precipitation and drip fran red fir are iq>lied for H, alkalinity, conductivity, 00 4, Cl, Ca, and K (Table 5) • Analytical prec1sion, indexed by standard deviations, for fir drip was good for H, cooductivity, 00 , Cl, and Ca (Table 3). Precipitation st~ deviations are SDBll only for H, alkalinity, and conductivity. Also taking into account the mgnitude of the mean differences, we ~lude that practical and statistical differences exist between precipitation and fir drip for H, alkalinity, and conductivity, and that H and conductivity are greater in fir drip than in precipitation. Alkalinity is less in fir drip than in precipitation. Sulfate, Cl, Ca, and K differences were statistically significant, bot IIIICertainty in measurement of these constituents for precipitation makes a definitive. cooclusian inpossible. Poor precision for Mg and Na prevented any definite conclusions. The results of these tests agree with those fOUDd in the literature. Rain throoghfall and drip mer conifer species have been found to be mre acidic than the anbient precipitation (Nihlgard, 1970; Malcolm and McCracken, 1968) • 'Ibis is probably due to leaching and/or washing of organic acids fran conifer vegetation surfaces. Washing and leaching WOD.l.d also cause an increase in drip cooductivity. Differences in 004 between precipitation and fir drip are not surprising because 004 is a mjor mtrient and fairly DDbile. 207 Pncipitatiaa. ..t~Adppole PiM Drip. ~sults fran chemical ca1p1risan between precipitatioo and drip fran lodgepole pine are similar to results fran precipitation and fir drip (Table 5, col'UIIIl 3). Balancing the iDportance and mgnitudes of analytical standard deviations, mean differences, and significance levels, we definitely cooclude that H and COIIductivity are greater in pine drip than in precipitation. Alkalinity is less in pine drip than in precipitation. Statistical difference is iq>lied for 004 and K, bot mean differences are not large, and stindard deviations for these constituents in precipitation are also large. No conclusions are drawn for the other constituents. Discussion of expected results for the above hypothesis parallels the discussion of results for chemical differences in precipitation and fir drip. Briefly, pine drip was likely to be DDre acidic than precipitation due to leaching and 11ashing processes. Fir aa4 PiM Drip. Few statistically significant differences were implied between red fir and lodgepole pine drip chemistry (Table 5), bot mean differences for the mjor anions and cations ranged fran 10 to 16 !Jeq/L, except for m3 • Except for Na and Cl, analytical precisian was gOod for fir and pine drip. A practical and statistical difference between drip fran fir and pine is indicated for Ca. No difference is indicated for H, alkalinity, and COIIductivity. No conclusions are atteupted for the other constituents. Field observations led to expectations of practical differences between fir and pine drip. Snow stayed loager an the fir canopy than oo the pine canopy. And as intercepted snow became drip, pine drip had a greater fraction of solid ice than fir drip. Because liquid drip canes into contact with wgetation DDre thoroughly than ice, fir drip s8Dples were expected to have lower pH, and higher conductivity. Although trends were toward lowered pH and elevated COIIductivity in fir drip, the mgnitudes of the analytical standard deviations my have masked the differences. The literature review also led to expectations that throoghfall and drip chemistry wonl.d be species dependent. However, in DDSt studies cited, events of rain and throughfall, not snow and drip, were observed (Parker, 1983). Rain throughfall canes into DDre intimte contact with trees than snow drip, perhaps a difference exists between fir and pine throughfall, bot only a small difference exists between fir and pine drip. Fran this study, the following conclusions are drawn: • Precipitation type (rain, snow, hail/graupel) is related to pr~ipitation conductivi~ and nitrate concentration (r = 0.88 and O.a>. respectively). A weaker relationship exists between prec!Pitation type and PI and sulfate concentration (r = O.SO and 0.32. respectively). Storm type is a related variable of secondary i.Dp.n:tance. • Variabili~ in chemical analysis often DBSks trends or relationships in water sauples with very low concentrations. This variability DDSt be taken into account during stibseqnent statistical analysis. • NO practical difference exists between precipitation fran an open site and forest throughfall for caoductivi~. alkalini~. and H. For the major cations and anions, no difference is inplied, but analytical error is high, and a conclusive statement is not possible. • Acidity, coodnctivi~. and SO concentrations are greater in fir drip than fn precipitation. Alkalini~ is lower in fir drip than in precipitation. Statistical analyses and the magnitude of mean differences inply that Cl. Ca. and K are also greater in fir drip than in precipitation. • Acidity and conductivi~ are greater in pine drip than in precipitation, and alkalini~ is less in pine drip than in precipitation. The SO difference was statistically significant, but Je mean difference was small. Potassilllll difference was also statistically significant, but analytical variation for K in precipitation was high. • NO statistical or practically significant differences were found between fir and pine drip for H, coodnctivi~. and alkalini~. A difference in Ca is indicated, but nothing can be said for the other constituents. The results of this study suggest further investigation of precipitation chemistry by precipitation type. Vegetative influences on precipitation chemistcy may also be uxxlified by precipitation type. If estimation of chemical loading into surface water and soil is an objective, then the influence of different vegetative species on throughfall and drip chemistry should be studied. This research was supported in part by the Division of Atm:>spheric Resources Research, Jllreau of Reclamation, U.S. Department of Interior. 208 Without the attention to detail and careful S811l>le handling procedures followed by R. Osterlmber, T. Mihevc, and R. Kattelmann, this study would not have been coopleted. We are also thankful for the field assistance provided by D. Azurm and M. Pack. Berg, N.H •• and S. Woo. 1985. Acidic Deposition and Snowpack Chemistcy at a Sierra Nevada Site. In: Proc. West. Snow Conf.. Boulder. Colorado, Colorado State University, Ft. Ooll~, Colorado. pp. 76-frl. Brewer, R.L •• R.J. Gordon, L.S. Shepard, and E.C. Ellis, 1983. Otemistry of Mist and Fog fran the Los Angeles Urban Area. Atmospheric Fn.viron. 17(11):2267-2270. Brown, J.C •• and C.M. Skau, 1975. Chemical Calpositioo of Snow in the East-central Siem Nevada. Cooperative Report Series Publication NO • .AG-1, Center for Water Resources Research, Desert Research Institute, Reno, Nevada. 13 p. California Air Resources Board, 1984. Second Ammal Report to the Governor and the Legislature oo the Air Resources Board's Acid Depositioo Research and Monitoring Program. State of California, Air Resources Board, Sacramento, California. 85 p. plus appendices. Colbeck, S.C.. 1981.. A Sinulation of the Enrichment of Atm:>spheric Pollutants in Snow Cover Runoff. Water Resources Research 17(5) :1383-1388. Eaton, J.S., G.E. Likens, and F.H. Bonnann, 1973. Throughfall and Stemflow Chemistcy in a Northern Hardwood Forest. Journal of Ecology 61:495-508. Ellis. B.A •• J .R. Verfaillie, and J. Kunmerow, 1983. Nutrient Gain fran Wet and Dey Atm:>spheric Deposition and Rainfall Acidi~ in Southern California <llaparral. Oecologia 60:118-121. Feller, M.C.. 1977. Nutrient Movement through Western Hemlock-Western Redcedar Ecosystems in Southwestern British OollliiDia. Ecology 58:1859-1866. Feth, J.H •• S.M. Rogers. and C.E. Roberson, 1964. Chemical Calposition of Snow in the Northern Sierra Nevada and Other Areas. USGS Water Supply Paper 1535-J. 39 p. Hansen, E.A.. and A.R. Harris. 1975. Validi ~ of Soil-\fater Sauples Collected with Porous Ceramic Cups. In: Proc. Soil Science Soc. Azrer. 39:528-536. Johannessen, M •• and A. Henriksen, 1978. Otemistry of Snow Meltwater: Changes in Concentration during Melting. Water Resources Research 14(4) :615-619. Jones, H.G •• 1984. The Influence of Boreal Forest Cover on the Chemical Calposition of Snowcover. In: Proc. East. Snow Conf •• Washington, D.C. Eastern Snow Conference, pp. 126-138. Kattelmann, R.C., 1984. Snowmelt Lysimeters: Design and Use. In: Proc. West. Snow Conf., Son Valley, Idaho. Colorado State lhiversity, Ft. Collins, Colorado. pp. 68-79. Kennedy, V.C., G.W. Zellweger, and R.J. Avanzino, 1979. Variation of Rain Chemistry During Storms at Two Sites in Northern California. Water Resources Research 15(3):687-702. Kittredge, J., 1953. Influences of Forests on Snow in the Ponderosa-Sugar Pine-Fir Zone of the Central Sierra Nevada. Hilgardia 22(1) :1-99. I.eivstad, H., and I.P. Muniz, 1976. Fish Kill at Low pH in a Norwegian River. Nature 259:391-392. leonard, R.L., C.R. Goldman, and G.E. Likens, 1981. ~ Measurements of the PI and Oulmistry of Precipitation at Davis and Lake Tahoe, California. Water, Air, and Soil Pollution 15:153-167. Linzcm, S.N., and J.M. Skelly, 1985. Effects of Gaseous Pollutants on Forests in Eastern North America. In: Abstracts, Muskoka Conference '85, International Syqx>siun on Acidic Precipitation, Musk.oka, Ontario. pp. 277-278. Malcolm, R.L., and R.J. McCracken, 1968. Canopy Drip: A Source of Mobile Soil Organic Matter for Mobilization of Iron and Aluninum. In: Proc. Soil Science Soc • .Arrer. 32:834-838. McColl, J.G., 1980. A Survey of Acid Precipitation in Northern California. Final Report for Contract A7-14~30. California Air Resources Board. Mecklenburg, R.A., and H.B. Tak.ey, 1964. Influence of Foliar Leaching on Root Uptake and Translocation of Calcium:-45 to the Stems and Foliage of Phaseolus Vulgaris. Plant Physiology 39:533-536. llelack, J .M., J .L. Stoddard, and D.R. Dawson, 1982. Acid Precipitation and Buffer Capacity of Lakes in the Sierra Nevada, California. In: Proc. Amer. Water Res. Assoc. Syllp. Hydrallet., Bethesda, Maryland. pp. 465-472. llorgan, J.J., and H.M. Liljestrand, 1980. Measurement and Interpretation of Acid Rainfall in the Los Angeles Basin. W.M.· Keck Lab., Division of Engineering and Applied Science, California Institute of Teclmology, Pasadena. Rept. No • .AC-2-80, February, 1980. 54 p. lkm.teverdi, J.P. , 1976. The Single Air Mass Disturbance and Precipitation Characteristics at San Francisco. Monthly Weather Review 104: 128~1296. Nachlinger, J., 1984. Rime AcCUllll.ation and Chemical Composition in a Subalpine Forest. lfupublished report, lhiversity of Nevada, Reno. 18 p. Nihlgard, B., 1970. Precipitation, its Chemical Ccmposition and Effects on Soil Water in a Beech and Spruce Fresh in Sooth Sweden. Oih>s 209 21:208-217. Parker, G.G., 1983. 'lbroughfall and Stemflow in the Forest Nutrient Cycle. In: Advances in Ecological Research, Volune 13. A. MacFadyen and E.D. Ford (eds.), pp. 58-120. Pough, F.H., 1976. Acid Precipitation and Ed>ryonic MOrtality of spotted Salamanders, AmbY5toma maculatun. Science 221(4235) :68-70. Sollins, P., C.C. Grier, F.M. McCorison, K. Cranack, Jr., and R. Fogel, 1980. The Internal Element Cycles of an Old-Growth Douglas Fir Ecosystem in Western Oregon. Ecological Monographs 50(3) :261-285. Tarrant, R.F., K.C. Lee, and W.B. Bollen, 1968. Nutrient Cycling by 'lbroughfall and Stemflow Precipitation in Three Coastal Oregon Forest 'I)pes. t9lt\ Forest Service Res. Pap. M-54. Portland, Oregon. 7 p. Verry, F.S., and D.R. Tilmms, 1977. Precipitation Nutrients in the Open and Under Two Forests in Mlnnesota. Canadian Journal of Forestry Research 7:112-119. Zinke, P.J., 1967. Forest Interception Studies in the lhited States. In: Forest Hydrology, Proceedings of a National Science Foundation Advance Science Seminar, University Park, Pennsylvania, pp. 137-161. POSTER SESSION 211 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 PRIMARY PRODUCTION, CHLOROPHYLL, AND NUTRIENTS IN HORSESHOE LAKE, POINT MACKENZIE, ALASKA Paul F. Woods and Timothy G. Rowe* ABSTRACT: A limnological study of Horseshoe Lake was started in 1985 to determine the relation between nutrients and primary production in the 64.8 square-hectometer lake prior to and after any influx of dairy wastes produced by the nearby Point MacKenzie Agricultural Project. This paper docu- ments results from data collected from June through September 1985 at Horseshoe Lake's east arm, the part of the lake nearest the agricultural project. Con- centrations of total phosphorus, dissolved inorganic nitrogen, and chlorophyll a were characteristic of oligotrophic lakes as was integral primary produc- tion, which ranged from 44.2 to 104.5 milligrams of carbon per square meter per day. The ratio of dis- solved inorganic nitrogen and dissolved ortho- phosphorus indicated that phytoplankton growth was limited by nitrogen, not phosphorus. Concen- trations of dissolved chloride, instead of nitrate are used as the primary tracer of dairy wastes because the nitrogen budget of Horseshoe Lake may be aug- mented by nitrogen fixed by the dense stands of alder trees (Alnus spp.) .that occupy the lake's 7.5 kilometer shoreline. (KEY TERMS: eutrophication; ground-water pollu- tion; primary productivity; chlorophyll.) INTRODUCTION The Point MacKenzie area within the Matanuska- Susitna Borough of southcentral Alaska (Figure 1) is characterized by undulating, low-relief terrain con- taining numerous muskegs, ponds, and lakes; most of the area is underlain by thick glacial deposits. Several large dairy farms have recently begun operation at Point MacKenzie as part of the 5,666 hm2 Point MacKenzie Agricultural Project. The dairy wastes generated by the farms are placed in excavated manure lagoons where the potential exists for these wastes to leach into the shallow ground-water system and eventually to reach nearby ponds and lakes. Dairy wastes characteristically are rich in nitrogen and phosphorus, two nutrients that are closely associated with eutrophication, or nutrient enrich- ment, of lakes. Horseshoe Lake is immediately adjacent to the west boundary of the Point MacKenzie Agricultural Project and the lake's northeast shoreline is about 1,500 m southwest of a dairy farm's manure lagoon. The first dairy cows arrived at this farm in the autumn of 1984; prior to construction of the farm the area adjacent to Horseshoe Lake was essentially undeveloped. Water-quality data collected at Horse- shoe Lake during 1981-82 indicated that the western arm of the lake had low concentrations of dissolved oxygen in its hypolimnion (Glass, 1983). These results suggest that the lake may have a limited capacity to assimilate additional nutrients because the lake's oxygen deficit will be intensified by significant increases in biological productivity. The U.S. Geological Survey and Division of Geo- logical and Geophysical Surveys of the Alaska Depart- ment of Natural Resources recognized the potential for nutrient enrichment of Horseshoe Lake by an influx of manure lagoon leachate and thus began a cooperative limnological study of Horseshoe Lake in October 1984. A major goal of the project is to assess limnological conditions in the lake prior to the anticipated arrival of manure lagoon leachate. If leachate is detected the project will then study the response of the lake's biological productivity to the increased loading of nutrients. The purpose of this paper is to document the results of the first year of limnological sampling in Horseshoe Lake's east arm, the part of the lake nearest the dairy farm manure lagoon (Figure 2). Horseshoe Lake is U-shaped and has a surface area of 65.2 hm2 , a volume of 1.9 hm3, and a mean depth of 2.9 m. The west and east arms of the lake are, respectively, 7.5 and 5.5 m deep and are connected by a shallow channel. The lake has no defined inlet streams; its single outlet stream drains to the Little Susitna River. Dams have been constructed by beavers at the lake's outlet and between the east arm and an adjacent pond. The Alaska Department of Fish and Game reports that the lake is accessible to anadromous salmonids, particularly silver salmon (Oncorhynchus kisutch ). The climate of Point MacKenzie can be character- ized by climatological data gathered at Anchorage, *Hydrologists, U.S. Geological Survey, 1209 Orca St., Anchorage, Alaska, 99501 213 which is 20 km southeast of Horseshoe Lake (Table 1). Mean annual precipitation is 366 mm, and September is the wettest month. The mean annual air temperature is 1.8°C with July having the highest mean monthly air temperature of 14.5"C. Horseshoe Lake is ice covered generally from mid-October to mid-May. 0 I 0 Boundary of Pt. MacKenzie Agriculture Project 0ost Lake 0 Lake 2 3 KILOMETERS i' 'i I 1 2 3 MILES COOK INLET Pt. MacKenzie Anchorage Figure 1. Location of the Point MacKenzie Agricultural Project and Horseshoe Lake, Matanuska·Susitna Borough of Southcentral Alaska. METHODS A limnological sampling station was located over the deepest part of the lake's east arm (Figure 2). Sampling was conducted six times between early June and mid-September 1985 on a tri-weekly 214 schedule. The 5.5 m deep water column was pro- filed at 0.5 m intervals for water temperature, dissolved-oxygen concentration, and photosyn- thetically active radiation (PAR). The PAR data were measured using a spherical quantum sensor and were used to determine the extinction coefficient and the depth of the euphotic zone, defined here as the ' depth at which in situ PAR is 1 percent of the PAR incident upon the lake's surface. Water column trans- parency was measured with a 20 em Secchi disc. Water samples were collected from 1 m beneath the surface and 0.5 m above the lake bottom. Methods described in Koenings et al. (1985) were used to analyze the water samples for the following constituents: total phosphorus, dissolved ortho· phosphorus, dissolved ammonia, dissolved nitrite plus nitrate, and total ammonia plus organic nitrogen. Dissolved chloride was determined for the 1-m depth samples per· methods in Skougstad et al. (1979). Chlorophyll a samples were collected at depths of 1, 2, 3, and 4 m and were analyzed fluorometrically, with correction for pheophytin, per Wetzel and Likens (1979). Phytoplankton species composition was determined per Greeson (1979) from samples collected at depths of 1, 2, 3, and 4 m. Primary productivity of phytoplankton was measured in situ using the carbon-14 light and dark bottle methodology described in Koenings et a/. (1985). Incubations occurred between 1000 and 1600 hours and were 4 to 5 hours in duration. Incubation depths were 1, 2, 3, and 4 m. Following incubation the 125 mL samples were filtered onto glass-fiber filters and then assayed using a liquid scintillation spectrophotometer. Concentrations of inorganic carbon were analyzed on an infrared carbon analyzer. The amount of. PAR incident at the lake surface was recorded for the incubation period as well as the entire 24 hours of the sampling date. The hourly rates of primary productivity were expanded to daily rates based on the ratio of incubation PAR to daily PAR. Additional limnological data have been collected as part of this study at other stations on or near Horseshoe Lake (Figure 2). These additional data are available within the U.S. Geological Survey's WATSTORE water-quality data computer system. RESULTS AND DISCUSSION The ice cover on Horseshoe Lake was completely melted by May 25, 1985. By the June 4 sampling trip the temperatures in the water column of the east arm sampling station ranged from 10.3 to 13.5"C. The maximum surface and near-bottom temperatures of 19.6 and 17.5° C, respectively were measured on July 19. During the September 18 sampling trip the • east arm was isothermal at 10.3° C. Ice re-formed on · the lake in mid-October. The six temperature pro- files showed that the east arm did not stratify · thermally from June through September 1985. Glass · (1983) also found that Horseshoe Lake's east arm did 150°09'30" EXPLANATION A East arm sampling station 6 Other sampling stations ~ Beaver dam • Spring sampling station 0 150°08'30" 300 400 500 METERS 500 1000 1500FEET Depth contours 1 meter Figure 2. Bathymetric map of Horseshoe Lake and location of the east arm sampling station. not stratify from June through September 1981. This lack of thermal stratification is attributable to the shallowness of the east arm and to Horseshoe Lake's exposure to strong winds, which are character- istic of the area. Water column transparency ranged from 4.2 to 5.2 m. The euphotic zone extended to the lake bottom on all six sampling trips. The penetration of PAR to the lake bottom, even on cloudy days, was a consequence of the low values of extinction coef- ficients which ranged from 0.33 to 0.45 m·1 • 215 Dissolved-oxygen concentrations ranged from 5.4 to 14.6 mgL-1 (Figure 3). Concentrations less than 9 mgL-1 occurred during August in water deeper than 4 m. Percentage saturation of dissolved oxygen was as high as 141 percent and as low as 58 percent (Figure 4). Concentrations of nitrogen and phosphorus and nitrogen to phosphorus ratios for depths of 1 and 5 m are shown in Table 2. There were only minor differences between nutrient concentrations at depths of 1 and 5 m on any given sampling date. Table 1. Air temperature and precipitation characteristics during June through September 1985 near Horseshoe Lake Month June July August September Monthly precipitation 1 (millimeter) Long-term mean :1 I 27.4 50.0 53.6 62.2 1985 25.6 25.2 89.9 80.5 Monthly mean air temperature 1 (oC) Long-term mean 2 I 1985 12.4 1L1 14.5 14.7 13.4 12.9 9.0 8.7 1 Data recorded at Anchorage WSCMO AP (U.S. Department of Commerce, issued annually) 2 Period of record is 65 years. Total phosphorus concentrations ranged from 6.4 to 14.0 JlgL-1 (mean = 9.1) whereas dissolved ortho- phosphorus concentrations ranged from 1.4 to 4.1 JlgL-1 (mean= 2.4). The range and mean of dissolved nitrite plus nitrate concentrations were, respectively, 0.7 to 2.8 and 1.6 JlgL-1 . Dissolved ammonia concen- trations ranged from 1.0 to 18.8 mgL-1 (mean= 5.6). Nitrogen to phosphorus ratios ranged from 1.2 to 6.5 (mean = 2.7) with the lowest values computed for August 28, the date with the lowest combined concentrations of dissolved nitrite plus nitrate and dissolved ammonia. When nitrogen to phosphorus ratios are less than 5, nitrogen limitation of phyto- plankton photosynthesis and growth is indicated (Rast and Lee, 1978). Concentrations of dissolved chloride at the 1 m depth averaged 3.2 mgL-1 within a range of 2. 7 to 3.6 mgL-1. The distribution of chlorophyll a in the east arm is shown in Figure 5; concentrations ranged from 0.9 to 4.1 }lgL-1 with a mean of 2.7 JlgL-1 . The highest concentrations on a given sampling date occurred in the lowermost depths. The mean con- centration of chlorophyll a for each sampling date was inversely correlated (r = -0.91, P less than 0.05) with Secchi disc transparency. The taxomomic composition and percentage occurrence of phytoplankton are listed in Tables 3 and 4, respectively. Based on number of cells per liter, Chrysophyta was the dominant phylum on all sampling trips. The Bacillariophyceae, or diatoms, were most abundant on the first five sampling trips and they were mainly represented by the pennate diatom Nitzchia. Dinobryon, of the Chrysophyceae, dominated the phytoplankton on September 18. Within the Chlorophyta, Ankistrodesmus commonly was the most abundant whereas Anabaena tended to dominate the Cyanophyta. The mean number of cells per liter for the water column ranged from 17 8,875 on July 19 to 262,825 on September 18. The number of cells per liter at each of the four depths was uncorrelated with chlorophyll a concentrations. Integral primary production of phytoplankton in the 5.5 m deep euphotic zone ranged from 44.2 to 104.5 mg C m·2 d-1 (Figure 6) and was uncorrelated with daily PAR, which ranged from 19.6 to 42.7 Em-2 . The lack of correlation was likely due to 216 photoinhibition of primary production in the near- surface samples on sunny days. Primary production on a given day was highest at depths of 3 or 4 m on the four sampling dates that were sunny or partly sunny. The remaining sampling date, August 7, was cloudy with rain; its highest rate of primary produc- tion was at 1 m. Primary production. in mg C m·3 d-1 , ranged from 28.8 at 1 m on August 7 to 1.5 at 2 m on July 19. Hourly specific primary production (primary production normalized for chlorophyll a concentration) at the depth of maximal primary production varied from 0.33 to 0.64 mg C (mg chlorophyll a)·1. Several of the limnological variables measured at Horseshoe Lake have been extensively used to classify the trophic state of lakes. Based on criteria listed by Taylor et al. (1980) and Wetzel (1975) the east arm of Horseshoe Lake is oligotrophic, or nutrient-poor (Table 5). The range in daily integral primary pro- duction falls within ranges cited for other oligo- trophic lakes by LeCren and Lowe-Connell (1980) and Wetzel (1975). The occurrence of nitrogen limitation and the oligotrophy of Horseshoe Lake's east arm support the concerns expressed over the possibility of eutrophica· tion via leachate from the dairy farm manure lagoons. The principal constituents emanating from animal feedlot wastes are the nutrients nitrate and phos- phorus plus chloride and occasionally, heavy metals; however, because of its mobility, nitrate is the only nutrient that enters the ground-water system in sub- stantial quantities (Miller, 1980) Chloride is a con- servative constituent that is also highly mobile in ground water and thus serves as a reliable tracer of contamination (Canter and Knox 1985). Any sig- nificant increases in the dissolved chloride concentra· tion in the east arm would constitute evidence that manure lagoon leachate had entered Horseshoe Lake. Similar increases in nitrate concentrations in the lake would not necessarily provide reliable evidence of contamination because nitrate is a non-conservative constituent in that it is an important nutrient for aquatic plant growth. In addition, nitrate may be added to Horseshoe Lake from the numerous alder trees (Alnus spp.) which inhabit much of the lake's 7.5 km shoreline. Goldman (1961) reported that the Table 2. Nutrient concentrations and nitrogen to phosphorus ratios at depths of 1 and 5 meters in the east arm of Horseshoe Lake during June through September 1985 Depth Nutrient concentrations (micrograms per liter) Nitrogen to Date (meters) Dissolved Dissolved· phosphorus Total phosphorus orthophosphorus nitrite + nitrate Dissolved ammonia ratio 1 June 5 1 7.4 2.4 2.3 4.6 2.9 5 10.1 4.1 2.3 8.3 2.6 June 26 1 10.8 3.2 1.2 18.8 6.5 5 14.0 2.9 2.3 15.4 6.1 July 19 1 7.5 2.2 0.7 3.0 1.7 5 11.1 1.6 0.7 4.3 3.1 August 7 1 9.0 1.4 0.7 2.2 2.1 5 10.4 2.3 0.7 3.3 1.7 August 28 1 7.0 2.0 1.3 1.0 1.2 5 8.6 2.3 2.3 1.0 1.4 September 18 1 6.9 3.0 2.8 4.3 2.4 5 6.4 2.0 1.3 1.2 1.2 1 (Dissolved ammonia and dissolved nitrite plus nitrate): (Dissolved orthophosphorus) Table 3. Taxonomic composition of phytoplankton1 in the east arm of Horseshoe Lake during June through September 1985 Phylum Chlorophyta Phylum Chrysophyta Ankistrodesmus Sub-phylum Chrysophyceae Crucigenia Dinobryon Starastrum Diceras Scenedesmus Sub-phylum Cosmarium Bacillariophyceae Franceia Coscinodiscus other Chlorophyta Nitzschia Phylum Cyanophyta Anabaena Chroococcus Gloeocapsa Oscillatoria other Cyanophyta Phylum Pyrrhophyta Glenodinium Ceratium other Pyrrhophyta Cymbella Tabellaria Pinnularia Amphipleura Gomphonema Synedra other Bacillariophyceae 1 Taxonomic classification per Prescott (1970) orr---.---..--.-.----.--,----.--.---------, SEPTEMBER Figure 3. Time-depth plot of dissolved-oxygen concentration (mgi;1 ) in the east arm of Horseshoe Lake during 1985. 217 0 100 (/) CI: w 1-2 w ::!: ~ ::r:· 3 1-c. 100 w Cl 4 5 JUNE JULY AUGUST SEPTEMBER Figure 4. Time·depth plot of percentage saturation of dissolved oxygen in the east arm of Horseshoe Lake during 1985. or--------,~--------.----------..--.-----, JUNE JULY AUGUST SEPTEMBER Figure 5. Time-depth plot of chlorophyll a concentration (pgi;1) in the east arm of Horseshoe Lake during 1985. Date June 5 June 26 July 19 August 7 Table 4. Percent occurrence of phytoplankton, based on cells per liter, in the east arm of Horseshoe Lake during June through September 1985 Mean number Mean percent occurrence by taxonomic design 1 of cells Chrysophyta per liter Chlorophyta Cyanophyta Pyrrhophyta Chrysophyceae I Bacillariophyceae 223,850 21.9 0 0.5 18.1 59.5 216,150 13.9 8.8 0.9 1.8 74.6 178,875 37.9 14.9 1.1 3.0 43.1 256,625 28.0 28.1 0.6 3.4 39.9 August 28 257,550 36.7 11.8 0.3 12.7 38.5 September 18 262,825 10.4 1.7 0 68.3 19.6 1 Based on four samples from depth of 1, 2, 3, and 4 m. 0 2 3 4 5 Cl) 0 a: w 1-w 2 ~ z 3 ::r:' 4 1-a.. 5 w 0 0 1 2 3 4 5 PRIMARY PRODUCTION, IN MILLIGRAMS OF CARBON PER CUBIC METER PER DAY 0 5 JUNE 4 DAILY PAR= 42.7 E m·2 2 - 3 +-Pz = 0.39@ PAR= 80 1-1E m·2 s·1 4 5 20 25 30 0 JUNE 26 DAILY PAR = 33.0 E m·2 2 -Pz = 0.37@ PAR= 74JJE m·2 s·1 3 pp = 57.8 4 pp = 91.7 5 20 25 30 JULY 19 DAILY PAR = 30.4 E m·2 Pz = 0.64@ PAR = 244 JJ E m·2 s·1 EXPLANATION o = Primary production, in milligrams of carbon per cubic meter per day PP = Integral primary production, in milligrams of carbon per square meter per day 10 15 Pz = Specific primary production, in milligrams of carbon per milligram of chlorophyll a per hour 20 25 30 AUGUST 7 ~ DAILYPAR=19.6Em·2 Pz = 0.33@ PAR= 40JJE m·2 s·1 20 25 30 AUGUST 28 DAILY PAR= 26.4 E m·2 s= 0.59@ PAR= 160JJE m·2 s·1 PAR =Photosynthetically active radiation, in either Einsteins per square meter per day or microEinsteins per square meter per second -= Lake bottom Figure 6. Depth distribution of daily primary production in the east arm of Horseshoe Lake during 1985. 218 Table 5. Trophic state designation for east arm of Horseshoe Lake based on selected variables [source: Taylor et al (1980), Wetzel (1975) Inorganic nitrogen Chlorophyll a Secchi disc Integral primary Total phosphorus production concentration concentration concentration transparency (milligram carbon Trophic state (microgram per liter) (microgram per liter) (microgram per liter) (meter) per square meter Oligotrophic .c:10 .C:200 Mesotrophic 10-20 200-650 Eutrophic >20 500-1,500 Mean value in east basin of Horseshoe Lake, 9.1 7.2 June through September 1985 primary production of Castle Lake, California was enhanced by nitrogen leached from leaf litter and soils associated with alder trees. Burns and Hardy (1975) reported that alders are capable of fixing nitrogen at annual rates between 56 and 156 kg ha·1 . At Horseshoe Lake the leaching of nitrogen from leaf litter and the shoreline area may be enhanced because beaver activity has raised the lake's water level enough to inundate lowlying shoreline areas. Water-quality data have recently been collected from a spring adjacent to the pond at the northeast shoreline of Horseshoe Lake (Figure 2), however, it is not known if this spring delivers ground water that may eventually be contaminated by the manure lagoons. The spring's mean nitrate concentration is 73.2 pgL-1 and its mean dissolved chloride concentra- tion is 1,400 }lgL-1 (Table 6). A recent report by Madison and Brunett (1985) serves to put a perspec- tive on the mean nitrate concentration of this spring. They statistically analyzed the nitrate concentrations from 87, 000 wells in the United States and assumed for classification purposes that concentrations less than 200 pgL-1 represent natural background values whereas concentrations larger than 3,100 }lgL-1 may indicate contamination by human activities. In Alaska only 5.2 percent of the 1,205 wells analyzed had nitrate concentrations exceeding 3,100 pgL-1 ; 60.9 percent had nitrate concentrations less than 200 per day) "'7 >3.7 50-300 7-12 2.0-3.7 250-1,000 >12 .::2.0 :>-1,000 2.7 4.7 66.4 pgL-1. Thus, the spring's nitrate concentrations can be assumed to represent uncontaminated ground water. Water-quality sampling will continue at the spring as well as at the limnological stations to develop an understanding of the spatial and temporal variability of Horseshoe Lake's water-quality characteristics prior to the anticipated arrival of manure lagoon leachate. There are also plans to install a series of wells between the manure lagoon and Horseshoe Lake to allow water-quality sampling of the shallow ground-water system. REFERENCES CITED Burns, R.C. and R.W.F. Hardy. 1975. Nitrogen Fixation in Bacteria and Higher Plants. Springer-Verlag, New York. Canter, L.W. and R.C. Knox, 1985. Septic Tank System Effects on Ground Water Quality. Lewis Publishers. Chelsea, Michigan. Glass, R.L. 1983. Hydrologic Data for Point Mac- Kenzie Area, Southcentral Alaska, October 1983. U.S. Geological Survey Open-File Report 83-142 update. Table 6. Concentrations of dissolved nutrients and chloride in a spring adjacent to Horseshoe Lake during 1985 Concentration of dissolved constituent (microgram per liter) Sample Date Nitrite plus Nitrate Ammonia Orthophosphorus Chloride August 7 62.0 <:1 2.7 1,300 August 28 66.8 <:1 3.9 1,400 September 18 77.0 <:1 5.4 1,500 October 7 86.9 ..:1 6.7 1,400 Mean 73.2 4.7 1,400 219 Goldman, C.R. 1961. The Contribution of Alder Trees (Alnus tenuifolia) to the Primary Pro- duction of Castle Lake, California. Ecology 42:282-288. Greeson, P.E., (ed.). 1979. A Supplement To ·· Methods for Collection and Analysis of Aquatic Biological and Microbiological Samples (U.S. Geological Survey Techniques of Water- Resources Investigations, Book 5, Chapter A4). U.S. Geological Survey Open-File Report 79- 1279. Koenings, J.P., J.A. Emundson, J.M. Edmundson, and G.B. Kyle. 1985. Limnological Methods for Assessing Aquatic Production. Alaska Depart- ment of Fish and Game. Division of Fisheries Rehabilitation, Enhancement and Develop- ment. Technical Report. LeCren, E.D. and R.H. Lowe-McConnell. 1980. The Functioning of Freshwater Ecosystems. Cam- bridge University Press, Cambridge. Madison, R.J. and J.O. Brunett. 1985. Overview on the Occurrence of Nitrate in Ground Water of the United States in U.S. Geological Survey, National Water Summary 1984. U.S. Geo- logical Survey Water-Supply Paper 2275, pp. 93-105. Miller, D.W. (ed). 1980. Waste Disposal Effects on Ground Water. Premier Press, Berkeley, California. 220 Prescott, G.W. 1970. How to Know the Freshwater Algae. Wm. C. Brown Co. Publishers, Dubuque, Iowa. Rast, Walter and G.F. Lee. 1978. Summary Analysis of the North American (U.S. Portion) OECD Eutrophication Project: Nutrient Loading. Lake Response Relationships and Trophic State Indices. U.S. Environmental Protec· tion Agency, EPA-600/3-78-008. Skougstad, M.W., M.J. Fishman, L.C. Friedman, D.E. Erdmann, and S.S. Duncan. 1979. Methods for Determination of Inorganic Substances in Water and Fluvial Sediments. U.S. Geological Survey Techniques of Water Resources Investigations. Book 5, Chapter Al. Taylor, W.D., V.W. Lambou, L.R. Williams, and S.C. Hern. 1980. Trophic State of Lakes and Reservoirs. U.S. Environmental Protection Agency. Environmental Monitoring and Support Laboratory. Technical Report E-80·3. U.S. Department of Commerce. Issued annually. Climatological Data, Alaska. Wetzel, R.G. 1975. Limnology. W.B. Saunders Co., Philadelphia. Wetzel, R.G. and G.E. Likens. 1979. Limnological Analyses. W.B. Saunders Co., Philadelphia. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 WATER QUALITY OF ABANDONED MINE RUNOFF: A CASE STUDY OF ALASKAN SITES David B. Pott, Robert E. Lindsay, and Nicholas Pansicl ABSTRACT: The purpose of this study was to evaluate surface water quality and soil chemistry of two abandoned mined areas in interior Alaska and to use these data to help develop reclamation alterna- tives for the sites. Soil and water chemistry data from the sites indicated that no major pollutants were threatening nearby water courses. Leaching of miner- al salts is occurring at both sites, with Na being the predominant cation found. Spoil materials at both sites are oxidized and neutral to slightly acid. Emphasis in the lower-48 states is upon rapid establishment of vegetation on disturbed lands with barren spoils viewed as reclamation problems. In colder climates advancement of vegetation cover onto barren areas often requires years even with the most optimum soil condi- tions. The reclamation alternatives selected for the two sites discussed in this study emphasized removal of safety hazards rather than extensive regrading and topsoiling which would have destroyed existing vegetation and disturbed soil profiles which had been developed in the years si nee mining. (KEY TERMS: mining; land reclamation; runoff; water quality; sodium; Alaska.) INTRODUCTION Surface mining of coal and hard miner- als disturbs large quantities of earth and the pedogenic processes that have been occurring there for millenia. Mining exposes parent and secondary minerals and greatly accelerates their weathering pro- cesses. Runoff from surface ml.nes is a recognized source of pollution for streams and rivers. Water quality impacts from ml.ning are associated with soil erosion and oxida- tion of pyritic materials, which causes production of acids and leaching of salts and toxic metals (Anderson and Hawkes, 1985; Down and Stocks; 1977; Hadley and Snow, 197 4; Letterman and Mitsch, 1978; Nordstrom, et al., 1979). The types and severities of water quality impacts vary with mine geochemistry and hydrology, receiving water quality and quantity, and mine reclamation measures. There are num- erous reports of mine runoff water quality in temperate and subtropical regions rut, we have found a paucity of infomation on water quality impacts from ml.nes in cold regions. We recently were given an opportunity to perform an environmental assessment and engineering design for reclamation of two abandoned surface coal mines in the interior region of Alaska. This study was funded through the Alaska Abandoned Mined lRespectively, Pott and Pansic, Harza Engineering Company, 150 s. Wacker Drive, ~icago, Illinois 60606; and Lindsay, Harza Engineering Company, 900 W. 5th Avenue, #700, Anchorage, Alaska 99501. 221 Lands Program. This paper presents a case study of two of these sites. The investigation program is described, water quality and soil chemistry data are presented, and site-specific problems are identified. Quantitative analyses of the sites' soils and runoff were made to characterize the severity of the problem at each site. We assessed the feasibility of reclamation of the mines. The nature of the sites, given their soil and hydrologic condi- tions, determine to a large extent the potential to revegetate and reclaim the sites to "acceptable" conditions. I 50 100 KILOMETERS Figure 1. Map Showing Location of Study Areas. 222 STUDY AREAS The mine sites studied are known as Diamond Mine and Dunkle Mine. Their locations are shown in Figure 1. Diamond Mine comprises three surface coal mine areas; the east, central, and west pits (Figure 2). Dunkle Mine f / ( I p t! \ / p II ' /J I cf.f:l~ POND ,,--....... }: : II } \,~ _ _,) ~ { ~/EAST PIT CENTER PIT : <~-' WEST PIT SCALE 0 1000 FEET L....._j SCALE 0 400 METERS Q WATER SAMPLES / / .. · SURFACE DRAINAGE (Eventually to Dry Creek) Figure 2. Diamond Mine Map Showing Water Sampling Locations and Surface Drainages. Soil Samples Were Taken at Each Pit. consists of both a surface mine and under- ground workings. Dunkle is located within the eastern boundary of Denali National Park and Preserve (Figure 3). ;/'···\ I \ I /1 \ l // c"""" u"J \ /~~--,,,, / I .··-----....._ --... y"-Bog Drainage \ 0 ~--. I ~ ... ~ --flY ' r-c y ~~-k -~:£~:,I OOGGY l $/ \ ·'7---~ ~-· \ AREA l ~ \\ "'-... / \ I 0 · ·.Groundwater Seep--Q \ / v~y· __\'\ ~ l' ,./ _/ MINED AREA \ ·. ~ :~~~~H I , ___ _ .0 \\ .PILE 1 / 0 WATER SAMPLES f}. SOIL SAMPLES \ I \. ___ / SCALE 0 100 200 FEET I I I I I SCALE 0 25 50 METERS Figure 3. Dunkle Mine Map Showing Soil and Water Sampling Locations and Surface Drainages. INVESTIGATION PROGRAM The investigation program for the study included the following major· tasks: o Comprehensive literature review; o Field survey and sample col- lection; o Water and soil chemical analy- ses; o Environmental assessment; and o Reclamation evaluation. Published literature was reviewed to provide background for the study effort. Few materials, however, pertained speci- fically to Alaska or cold regions applica- tions. The field survey team included a reclamation specialist, an environmental chemist, a landscape architect, and a geo- technical engineer. 223 METHODS Soil Surface soils were grab sampled and composites were made, each representing barren and vegetated areas in each pit or spoil pile of the mines. In some cases, we took samples of subsurface (0.75 m) materials. Soils were analyzed for the following: 0 Organic matter 0 Cation exchange 0 Phosphorus (P) capacity 0 Potassium (K) 0 Sulfur (S) 0 Magnesium (~ig) 0 Manganese (Mn) 0 Calcium (Ca) 0 Iron (Fe) 0 Sodium (Na) 0 Aluminum (Al) 0 Hydrogen (H) 0 Zinc (Zn) The soils were dried at room tempera- ture prior to being passed through a 2 mm sieve. Na, K, Mg, and Ca were extracted using 1 N ammonium acetate, buffered to pH 7. Zn, Mn, and Fe were extracted using 0.1 N HC1 while exchangeable A1 was extracted with 1 N KC1. The soil extracts were anal- yzed for metals by flame atomic absorption spectrophotometry, according to Black (1965). Organic matter, cation exchange capacity (CEC), total S, total P and solu- ble salts ~vere also measured by methods detailed in Black (1965). A 1:1 water to soil mixture was made for pH measurement with a glass membrane electrode and pH meter. Water Water samples for metal analysis were collected in acid-washed polyethylene bot- tles, and acidified with 2 ml concentrated HN0 3 per liter. These samples were filter- ed through 0.45 micron pore filters and analysed via inductively coupled plasma emission spectrophotometry. The metals we attempted to quantify included sodium, potassium, calcium, magnesium, aluminum, iron, manganese, and strontium. The technique also allowed quantification of the nonmetals phosphorus and silicon. Dissolved oxygen, temperature, and conductivity were measured using field meters (Yellow Springs Instrument Co., Yellow Springs, Ohio). A glass-membrane pH potentiometer was also used in the field (Fisher Scientific Co., Pittsburgh, PA). Total acidity was titrated to the phe- nolpthalein end point (Hach Chemical Co., Loveland, CO). RESULTS AND DISCUSSION Complete results of the field samp- ling and data collection are detailed in the project reports (Harza Engineering Co., 1984a and 1984b). Tables 1 through 4 present some data relevant to this dis- cussion. The water chemistry data of the two mines show various levels of impact from the mining and disturbance of the soils. 224 Toxic Hazard Evaluation The streams draining the Diamond Mine are moderately hard and neutral, not unlike area streams draining non-mined lands. Of the three samples taken, none exceeded Alaska's primary maximum contaminant con- centration standards for metals and in only one case was a secondary maximum contamin- ant concentration standard exceeded. The secondary standard of 0.05 mg/1 Mn was found to be exceeded by a factor of 11.6 in the stream draining between the east and center pits at Diamond Mine (0.58 mg/1 Mn). This level of Mn is not toxic to most aqua- tic life (USEPA, 1976). At Dunkle Hine, we found there to be no violations of primary contaminant standards. There were, however, five in- stances in which the secondary standards were exceeded: three violations of the Fe standard and two violations of the Mn stan- dard. Both of these metals are notoriously associated with mine drainage. The concen- trations we found were not hazardous. Coal Creek below the mined area had 0.37 mg/1 Fe, slightly above the 0.3 mg/1 standard. The other four violations were not likely related to the mining activities. They came from two samples: the bog water above the mine and groundwater seep at the high- wall. These two samples both contained Fe and Mn in excess of the secondary stand- ard. Leaching of Soluble Salts It is apparent that the spoils at both mined sites are leaching salts into the nearby creeks. Much of the data sug- gests this. Sodium seems to be the main cation being leached at both sites. At Dunkle ~line, notice (in Table 3) the dilu- tion of Na from the first order recipient stream, Coal Creek, to the third order Costello Creek. The high levels of extractable Na in the barren spoil piles are the probable source of this leached Na. Spoils of both mined sites are oxi- dized and neutralized according to the mine classification in USEPA (1973). Soluble TABLE 1. Soil Analyses of Dunkle Mine. ·-·-------------------·------- Cation Organic Soil Exchange Soluble SamEle Location Matter pH Capacit~ Na K Ca _!!g_ Al Fe Mn p s Salts (%) (S.U.) (meg/100g) -(Elemental concentrations are in mg/kg) --(mmhos/cm) Undisturbed vegetated 5.6 4.5 3.6 46 98 140 47 267 113 80 5 24 0.2 area north of mine Main spoil pile 7.5 6.0 16.0 780 139 690 774 7 193 49 69 387 4 .1 barren areas N Main spoil pile 4.9 6.5 6.6 69 126 610 290 7 171 59 74 79 0.5 N vegetated areas U'l Main spoil pile 2.7 7.8 6.5 200 113 540 326 8 182 58 77 83 0.8 subsurface South spoil pile 1.9 7.4 7.3 116 113 550 462 7 178 144 73 300 1.6 barren areas South spoil pile 8.1 5.6 9.7 76 142 760 348 7 162 40 66 47 0.5 vegetated areas Caribou lick 3.6 7.5 5.9 42 80 830 169 7 206 120 60 12 0.3 ----- TABLE 2. Soil Analyses of Diamond Mine ------------------------- Cation Organic Soil Exchange Soluble Sample Location Matter pH CaEacity Na K Ca ~ Al Fe Mn p s Salts {%) (S.U.) (meg/lOOg) ---(Elemental concentrations are in mg/kg) ---(mmhos/cm) West pit barren areas 5.9 5.6 9.5 130 130 550 440 8 157 34 27 169 0.7 West pit vegetated areas 7.9 5.2 7.3 51 98 500 217 21 207 37 8 10 0.2 Center pit barren areas 8.4 5.5 13.5 280 145 680 604 9 222 35 26 196 0.8 N Center pit vegetated 8.6 5.7 9.8 141 166 560 466 8 179 31 8 33 0.4 N ~ areas East pit main gob 9.0 4.7 12.6 42 89 660 302 19 180 49 4 20 0.3 pile barren areas East pit vegetated 8.6 5.0 11.7 40 110 850 260 27 175 72 7 15 0.2 areas East pit main gob 8.5 5.6 10.9 46 98 700 433 20 188 53 4 14 0.3 pile subsurface (.75m) ----------------- TABLE 3. Water Analyses of Dunkle Mine. A dash indicates that the datum was not collected. Total Sample Location Conductivity ~-Acidity (umhos/cm) (S.U.) ·(as mg/1 Caco 3 ) Na K Ca _l:YL_ Al Fe Mn P (Elemental concentrations are~mg/~-- Si Sr ------------------------------------------------------ Costello Ck. upstream of Camp Ck. Camp Ck. up- stream of Coal Ck. Coal Ck. 99 44 110 7.5 7.2 7.4 <17 3.1 (1.0 (17 2.0 (1.0 34 6.5 (1.0 19 7.6 (0.05 0.09 (0.05 0.15 1.6 0.17 8.7 2.3 (0.05 0.16 <0.05 (0.05 2.6 0.07 18 2.4 (0.05 0.37 (0.05 (0.05 2.8 0.23 N N ~ Camp Ck. down-48 6.8 <17 2.2 (1.0 9.0 5.0 (0.05 0.17 (0.05 (0.05 2.6 0.08 stream of Coal Ck. Costello Ck. downstream of Coal Ck. Groundwater seepage into mine Boggy surface water entering mine 97 125 28 7.5 (17 2.8 (1.0 18 6.7 (0.05 0.16 (0.05 (0.05 1.8 0.16 1.4 1 .1 40 8.4 0.42 0.68 0.11 0.19 4.4 0.17 6.0 54 0.68 (1.0 4.9 0.93 0.06 4.8 0.22 (0.05 3.8 (0.05 -------------------------------------------------------- I -MI en! I ~I I ~I 1.1'\ 0 . 0 v 0\ . 00 0 0 1.1'\ 0 0 v 0 1.1'\ 0 . 0 v ...... . 0 ...... v 0 . ...... r-.. ...... v ...... . >CJ 0 . 0 N . 1.1'\ 1.1'\ 0 . 0 v 1.1'\ 0 . 0 v ...... N . 0 1.1'\ 0 . 0 v r-.. ...... 0\ . (") 1.1'\ . ...... r-.. ...... v ..;t-. 1.1'\ (") N >CJ ...... . 0 00 . 0\ 0 . 0 00 1.1'\ . 0 1.1'\ 0 . 0 v 1.1'\ 0 . 0 v ...... 1.1'\ ...... ...... N . >CJ . N ...... (") 228 salts are apparently one limiting factor to complete natural revegetation. Comparison of the soluble salt levels in barren and vegetated spoils confinns this. Although we have found no soil and water chemistry data from past studies, we would expect weathering and pedogenic processes to occur much more slowly in colder regions like these sites than in more temperate areas. The Dunkle Mine hasn't been worked since the early 1950's; the Diamond Mine since 1972. Runoff from neither mine site appears to have yet receded to background salt levels. For comparison, Brown et al. ( 1984) found runoff from revegetated Texas Gulf Coast lignite mined spoils to return to natural levels within 4 to 15 months. One interesting point concerns the chemistry of the runoff entering Dunkle Mine from a bog above the site. This water showed the lowest pH and highest acidity of any other sample at the site. This sample also contained a low level of Ca, a high concentration of iron, and low conductivity. Although we did not measure dissolved organic matter, these parameters and the dark tea-colored nature of the water suggested the pre- sence of substantial quantites of humic substances. Humic substances are known for their acidity and ability to pre- ferentially complex Fe over Ca (Schnitzer and Khan, 1972; Alberts and Giesy, 1983). Reclamation Evaluation Traditional reclamation of aban- doned mined lands has emphasized regrad- ing spoil piles to original contour, covering barren or toxic sites with soil, and establishing vegetation to stabilize and reduce erosion and drainage through toxic materials. How- ever, in many cases, results of such practices may be undesirable. Emphasis in the lower-48 states is upon rapid establishnent of vegetation cover on disturbed lands with barren or slowly revegetating spoils viewed as reclama- tion problems. While this may be true in other parts of the country, the same :riteria do not necessarily apply in cold regions. Soil formation along with vegeta- tion establislnnent and growth are often slow, requiring years to accomplish the same degree of cover which can be obtained in one growing season in warmer climates. Regrading of problem spoils identified in this study may actually create worse problems. Soil profiles which have devel- oped through the vegetational influences of aggregate formation and accumulation of organics would be destroyed, and the resulting compaction of the spoil would inhibit infiltration and percolation of water. Exposure of phytotoxic materials which may be present deep within spoil areas would negate any amelioration which has resulted from years of surf ace weathering and leaching. Soil and water parameters of both the Diamond and funkle mines were examined carefully to aid in developing reclamation alternatives which met a basic goal of abating problem conditions in a cost-effec- tive manner. Recommended reclamation alternatives for both sites were limited to the removal of public safety (unsound structures) and aesthetic problems. In the case of the Diamond Mine, water quality and soil parameters did not indi- cate the need for extensive reclamation. The site had several extensive barren areas and many of them were being reclaimed by invasion of surrounding vegetation. The preferred reclamation alternative involved removal of existing or potential public safety hazards while leaving barren spoil areas unreclaimed. A sub-alternative to the preferred alternative included removal of a highwall which was visible from a major highway. Water quality would be adversely affected by transport of sedi- ments if this work were carried out. The Dunkle Nine also exhibited exten- sive barren spoils which, in most tradi- tional reclamation plans, would be candi- dates for extensive regrading and topsoil- ing. However, soil and water quality sampling of the site did not indicate the need for this type of reclamation. The preferred reclamation alternative included work to remove several collapsing struc- tures considered to be safety problems. 229 CONCLUSION In this study, we examined field con- ditions of two abandoned surface mine sites in interior Alaska. These data were evalu- ated to assist in preparation of reclama- tion plans. at both Spoils Mines were t rali.zed. apparently There was salts, rut no problems at soluble metal found. Dunkle and Diamond oxidized and neu- some leaching of evidence of toxic either site was Natural revegetation has begun at both sites. Regrading and seedings of slopes was not recommended. Our reasons for this included the slow development of vegetative cover in cold regions, the possible exposure of unoxidized or acidic materials, and the likelihood of increased soil erosion at the sites. A number of unsafe collapsing structure existed at both mines; the preferred reclamation alternative involved removal of these public safety hazards. ACKNOWLEDGEMENT The authors gratefully acknowledge the support of our client, the Division of Mining of the Alaska Department of Natural Resources, in funding the recla- mation study and allowing us to publish our findings. The comments of J.T. Passage and two anonymous reviewers were most helpful. LITERATURE CITED Alberts, J.J. and J.P. Giesy, 1983. Conditional Stability Constants of Trace Metals and Naturally Occurr- ing Humic Haterials: Application in Equilibrium Models and Verifica- tion with Field Data. In R.F. Christman and E.T. Gjessing (Editors), Aquatic and Terrestrial Humic Materials. Ann Arbor Science, Ann Arbor, HI. PP• 333-348. Anderson, M.T. and C.L. Hawkes, 1985. Water Chemistry of Northern Great Plains Strip Mine and Livestock Water Impoundments. Water Resources Bulletin 21:499-505. Black, C.A. (Editor), 1965. Methods of Soil Analysis Part 2 -Chemical and Hicrobiological Properties. Arne r. Soc. Agron. Inc., Madison, WI. 1572 PP• Brown, K.W., J.C. Thomas, and L.E. Deuel, Jr., 1984. Chemical Characteristics of Surface Runoff from Soils and Revegetated Lignite Mine Spoils. Journal of Soil and Water Conservation 39 (2):146-149. Down, C.G. and J. Stocks, 1977. Environ- mental Impact of Mining. John Wiley, New York, New York. Hadley, R.F. and D.T. Snow (Editors), 1974. Water Resources Problems Related to Hining. American Water Resources Association, Symposium Proceedings No. 18, Minneapolis, Minnesota, 236 pp. Harza Engineering Company, 1984a. Diamond Mine Engineering Design Alternatives and Environmental .~sessments. Alaska Dept. Natural Resources, Division of Mining, Anchorage, AK. Harza Engineering Company, 1984b. Dunkle Mine Engineering Design Alternatives and Environmental Assessments. Alaska Dept. Natural Resources, Division of Mining, Anchorage, A.K. Letterman, R.D. and W.J. Mitsch, 1978. Impact of Mine Drainage on a Hountain Stream in Pennsylvania. Environmental Pollution 17:53-78. Nordstrom, D.K., E.A. Jenne, and J.W. Ball, 1979. Redox Equilibria of Iron in Acid t1i.ne Waters. In: E.A. Jenne (Editor), Chemical Modeling in Aqueous Systems. Speciation, Solubility, and Kinetics. American Chemical Society, ACS Symposium Series 93, Washington, D. C • pp. 51-79 • Schnitzer, M. and S.V. Khan, 1972. Humic Substances in the Environment. Marcel Dekker, New York, N.Y. 327 pp. U.S. Environmental Protection Agency (USEPA), 1973. Methods for Identify- ing and Evaluating the Nature and 230 Extent of Non-Point Sources of Pollu· tants. USEPA, Washington, D.C. 261 pp. (As referenced in Down and Stocks, 1977). U.S Environmental Protection Agency (US EPA), 1976. Quality Criteria for Water. USEPA, Washington, D.C. 501 PP• JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 THAWING OF GROUND FROST ON A DRAINED AND UNDRAINED BOREAL WETLAND SITE L. E. SWANSON AND R. L. ROTHWEL~ ABSTRACT: A study of ground frost levels in a drained and undrained boreal wetland site near Slave Lake, Alberta showed significant delay of thaw as well as lower temperatures at 40 and 60 em depths in the drained area. The undrained area froze to greater depths than the drained area but thawed more rapidly. In the lower hydraul- ic conductivity portion of the drained area, 16.7 % of the sample points retained ice throughout the summer and were still frozen at the time of ground re-freezing in the fall. Thickness of the frozen layer appeared to be related to hydraulic conductivity as well as drainage. Thawing from underneath the frozen layer accounted for 38 % of total thaw in the undrained area. (KEY TERMS: peat land drainage, ground frost, soil temperatures.) INTRODUCTION Alberta contains extensive areas of forested and nonforested organic soils, the majority of which are located in the northern half of the province. These sites are characterized by .surficial organic deposits 0. 3-4.0 m deep and water table levels within 0.1-.3 m of the ground surface throughout most of the year. The thermal properties of surficial peats in these areas are highly variable, with water content being the principal deter- mining factor. Thermal conductivity is very low in dry peat (0.105 J/m s °C), but is much higher in moist peat (0.432 J/m s °C) and frozen peat (Lee, 1978). Moist conditions in autumn and winter promote heat transfer out of the soil, but drying of surficial peat in summer inhibits heat transfer into soil. This lowers mean annual soil temperature and promotes forma- tion of zones of permafrost (Brown, 1977). Discontinuous permafrost occurs in Alberta north of about 570 latitude (Fish- eries and Environment Canada, 1978), however localized patches of permafrost may be found south of this boundary in areas where surface peats dry in summer (Zoltai, 1971; Brown, 1965). Because of the thermal imbalance caused by the insulating effect of dry peat, practices which cause the extreme drying of surficial peats, such as drainage for agriculture and forestry, could be expected to lower mean annual substrate temperatures, possibly promoting the formation of discontinuous permafrost south of established limits. A study was conducted to evaluate the effect of drainage on freeze/thaw cycles and frost depth on a boreal wetland site. Because of the significant differences in thermal properties between wet and dry peat, it was hypothesized that freeze/thaw cycles should be significantly different between drained and undrained areas, with the thaw in the drained area being delayed due to lower soil temperatures caused by the insulating layer of dry surficial peat. It was also desired to know if frost depths and freeze/thaw cycles were related to the spacing of ditches or hydraulic conductivity within the drained area. The site selected for study was a peatland fen complex located 36 km south- east of the town of Slave Lake, Alberta, 1Respectively, Graduate Student and Professor, Department of Forest Science, Univer- sity of Alberta, Edmonton, Alberta, Canada, T6G 2E2 231 550 10' latitude. This area is character- ized by forest cover of black spruce (Picea mariana), tamarack (Larix laricina) and bog birch (Betula glandulosa). Peat deposits range from 0 to 4 m in depth. In the spring of 1984 a peatland drainage project was implemented in the area by the Alberta Forest Service in which an area approximately 0. 5 km2 was drained. Ditch spacings of 25 m and 40 m were used, based on saturated hydraulic conductivity classes of the peat. METHODS Frost recession was determined as described by FitzGibbon (1981). This method involved augering through the frozen peat in spring to determine total frost thickness, then monitoring the recession of the upper frost boundary throughout the summer by probing. Probing was done with a 5 mm diameter rod which passed easily through the peat but did not penetrate a frozen zone. Sampling began May 9-12,1985, prior to the start of frost recession. Measurements were obtained at 64 sites using an ice auger and probe. Depth to the upper frost boundary was measured by probing on a weekly basis throughout the summer season. Frost recession was judged completed when repeated probings and temperature measurements could not detect the presence of frost. Sampling began May 9-12, 1985. Sample points were located in three distinct areas: the 25 m spacing zone; the 40 m spacing zone; and an adjacent undrained zone. Transects of 20 sample points ( 10 hollows, 10 hummocks) were established in each zone. Hummocks and hollows were 16 to 52 em different in elevation and had unique microclimates and site characteristics. Beginning June 15, 1985 a second transect of 36 sample points (18 hollows, 18 hummocks) was established in each of the areas to determine if site disturbance from foot traffic affected earlier observations. Substrate temperature profiles were measured at four hummocks and four hollows in each of the three areas. At each point, thermocouples were installed 232 at depths of 10, 15, 45, and 60 em below the ground surface. Temperatures were recorded biweekly throughout the summer using a handheld digital multimeter and a small electronic reference junction. During July and August a datalogger was used to monitor diurnal temperature fluctuations. Temperatures were recorded at hourly intervals continuously for approx- imately one week at one hummock in each of the three zones. Precipitation was monitored using a Belfort Universal rain gauge placed at the southern edge of the drained area. Water table levels were obtained from a network of recording water wells maintained by the Department of Geology, University of Alberta. RESULTS Water table height was 20 to 50 em lower in the drained area than the undrain- ed. Average water table levels for the summer were 55 and 20 em below the ground surface for the drained and undrained areas respectively. Water table responses to precipitation in all areas were pro- nounced, with the return of the water table to prestorm levels occurring within 4-5 days. Summer rainfall (May to October) for 1985 was 207 mm, 40% lower than the 63 year average of 350 mm. Probe surveys for total frost thick- ness in the three zones on May 9-10 showed average frost thicknesses in the 40 m spacing zone to be significantly less than in both the 25 m spacing or undrained zones (Scheffe's test, 0.05 level). Mean frost thicknesses were 41.6 em in the undrained, 36.7 em in the 25 m spacing and 22.4 in the 40 m spacing. Associated standard errors were 2.63, 2.68 and 2.63 respectively. Ice structure in the 40 m spacing was often granular, while in the 25 m spacing and the undrained zones layers of solid, concrete ice were evident. Frost recession was most rapid in the undrained area. This was most evident in the hollows (Fig. 1). During the period May 10-June 13, 1985, average depth to the frost table decreased 16.9 em in the undrained area hollows, versus 7. 4 em and 7. 8 em for the 25 and 40 m spacings respectively. The corresponding averages for the hummocks were 17.5 em, 11.8 em, and 9.8 em. Frozen soil was detected more frequently in the 25 m spacing zone throughout the summer (Fig .2) than in the other areas. At the onset of frost formation in early October, 16.7% of the sample points in the 25 m spacing still retained ice. In the undrained area frozen soil was found at 2. 3% of the sample points at this date. Frozen soil was not detected in the 40 m spacing zone after September 10, 1985. Transects which were started in May showed a slightly higher thaw rate than those started in June, suggesting some trampling might have occurred while sampling. Thermocouple readings indicated soil temperatures in the surface 10 em were warmer in the drained area than in the undrained area. Midday peat temperatures at 10 em depth in the drained areas were 0 to 4. 4 °C warmer than in the undrained area for most of the summer (Fig. 3). It ias only in late August that midday temperatures at 10 em were warmer in the mdrained area. Soil temperatures below 10 em were warmer in the undrained area than in the ~rained area throughout most of the summer (Fig. 3). The same trend was observed by Pessi (1958) in a drained peatland in Finland. Maximum tempera- ture differences of 4.9 and 4.3 °C at 40 and 60 em respectively occurred in late July. The maximum diurnal temperature flue tuat ion was greater in the drained areas than in the undrained. Fluctuations of up to 20 °C in a single day were recorded in the drained areas at the 10 em depth (Fig .4). In the undrained area amplitudes of only 4 oc were observed at 10 em. At 25 em depth diurnal fluctuations of 8 °C were recorded in the drained areas versus less than 1 °C in the undrained area. Temperature fluctuations at the 40 and 60 em depths in all areas were small, rarely exceeding 1 oc; day. Frost recession occurred from both above and below in all zones, with thaw from below accounting for as much as 38 % of total thaw in the undrained area. Table 1 shows the maximum depth of frost penetration as determined on May 9-10, as well as the total depth of thaw from above, and the resulting percentages of 233 thaw which must have occurred from above and underneath. The depth of thaw from above is the position of the frozen layer one week prior to total thaw. Percent thaw from above is (thaw from surface/initial depthxlOO), and percent from below is (100- percent thaw from above) and represents the proportion of the total thaw which occurred from beneath the frozen layer. The final frost depth a week prior to complete thaw is indicative of the total depth of thaw from above since, in most cases, the frozen layer was so thin at this point that it could be penetrated with the probe using only light finger pressure. Initial Depth of Percent lower frost thaw from of thaw depth (em) above (em) from: ------------------------------- Mean std. Mean std. above below error error Undr. 56.3 (2.98) 35.0 (2.43) 62% 38% 40 m 36.9 (2.69) 26.8 (2.34) 73% 27% 25 m 52.0 (2.93) 35.9 (2.51) 69% 31% ----------------------------------------- Table 1. Thaw from beneath the frozen layer; maximum depth of frost penetration in winter and total depth of summer thaw. DISCUSSION Frost recession was slower in the drained areas than in the undrained area because the dry surficial peat acted as a barrier to conductive heat transport into the soil. The effectiveness of dry peat as an insulator is well documented (Rigg, 1947; Benninghoff, 1952; Brown, 1963; Mac Far land, 1969; Moore and Bellamy, 1974). Thermal conductivity of peat decreases greatly with reduced water con t en t (Lee , 1 9 7 8 ) • Therefore heat transport into the dry surficial peat would be limited to convective processes (vapor diffusion and rainfall), which can be slow and are extremely difficult to measure (Brown, 1963). Frost recession was most rapid in the undrained area, even though its ice thickness was greater than the drained areas. The faster recession was attri- buted to the higher water content and E u £ a. • 0 UJ !,1 :r ~ Ill .... z 5 Q. UJ _J Q. :::; ~ ... 0 .... z UJ u 0:: UJ Q. -II -10 -12 -14 -16 -18 -20 -22 -24 -26 -28 -30 -32 May 10 May 111 May 24 May 31 June 7 c Undrained + o.-40 o D•-2S Figure 1. Frost recession rates in hollows for the undrained, drained 40 m and drained -25 m spacing areas. 100 June 19 to October 3, 1911S 90 c Undrained 110 + D•aln•d, 40 m 70 o Drained, 2:5 m 60 50 40 30 20 10 0 Figure 2. Frost recession shown as percentage of sample points in each area where ice was detected. Percentage is (number of sample points where ice was detected I total number of sample points) xlOO. u .. • I! "' • 0 234 0 May 24 May 31 June 7 -10 b" / ~~ ... -:-·'" ' '? -<· , -20 . ~ 7 • o· u I • "I a -30 1: :I '1: I• :I il 1: •I I• •I 1: :I • ' ·I 0 -40 1t c .. .a a c Undrained ·I :I -so ·I • :I 0 •I •I ~ -60 2 4 6 II 0 2 4 6 II 0 2 4 6 Subetrate temperature (degreea C) 0 June 27 July 28 Augullt 211 -10 '? -20 ~ • u a -30 '1: ~ . • 0 -40 ] ~ Q. -so ~ -eo /' I :f' D . . .. ,;i' , ' ~r )' ;J .:f' 1: ,. "I ,. :I I .I I "I ·' ; I i e! 0 0~ ;: "I I •I ' •I :'I ;I .I Undrained ·I •I "I ·I "I "I ;, Drained, 40 m :I :I .I ·I ·I •I Drained, 2:5 m :: ·I "I •I ;I :I :I L •I ~ 10 IS ~ s 10 15 0 s 10 15 Subatrote temperature (dec.,reea C) Figure 3. Midday substrate temperaturi profiles for various days throughout the summer, 1985. 22 21 /\ , ... -,, 20 I ' I ' ,, I \ 1i I --.. ' ,_ ' Ill I ' I ' 17 I ' I ' 111 I ' I I ' IS I \ 14 I ' I 13 I ' --I ' 12 I ' 11 ·.' I .. , 10 / ., '· a ' ... , __________ ... , , .. , .. II \ 7 6 s -Undrailed 4 ----Drained-40m 3 2 ·· ··· Drained-25m. 0 4 8 12 Ill 20 24 HOUr' Figure 4. Diurnal temperature variation at 10 em depth in the undrained, drained -40 m and drained -25m spacing areas. resultant higher thermal conductivity of the surface peat in the undrained area. Differences in water content of the surface peats between the drained and undrained areas were amplified by the unusually low rainfall levels observed during the study season. In the drained area the surface peat dried out as a consequence of lower water table levels, and rainfall events were too infrequent and too rapidly drained away to have any significant wetting effect. Higher water table levels in the undrained area were sufficient to maintain a moist layer of surface peat in spite of low rainfall levels. Frost recession in the 40 m spacing uea was more rapid than in the 25 m spacing. This occurred ostensibly because of less ice and lower frozen water contents in the 40 m spacing area. The ice in the 40 m spacing was thinner and appeared to be porus in structure, possibly because of the peat's higher saturated conductivity, which with drainage resulted in a rapid removal of water from the peat substrate. The more extensive ice in the 25 m spacing was in part attributed to it's lower saturated conductivity which would lessen drainage and make more water available for freez- ing. Average saturated hydraulic conduc- tivites for the 40 m and 25 m spacings were respectively .35 and .18 m/day (Toth, 1986) • The presence of remnant frost in the undrained area in early October (at 2.3 percent of the sample points) was attribu- ted to low summer precipitation in 1985. S~mer rainfall for 1985 (May-October) was 40% below the 63 year average for the Slave Lake region. As a result, some surface peat material in the undrain- ~ area dried out, albeit to a much lesser extent than occurred in the drained area. In a wetter year frost recession would probably occur earlier and at a faster rate on all sites because of latent and convective heat transfer (e.g. evaporation and precipitation) into the peat sub- strate. Thawing of ground frost from above and below was evident in all of the study areas. As the study area is located in a fen complex, sub surf ace groundwater flow should be expected. FitzGibbon ( 1981) found 40% of thaw to occur from beneath 235 the frozen layer in a fen in Saskatchewan, whereas in a nearby bog (with no sub- surface flow) no thaw from below was observed. FitzGibbon concluded that thaw from below was produced by heat exchange from groundwater. The lesser amounts of thaw from below in the drained areas compared to undrained in the present study (27 .4 % and 31.0 % versus 37.8 % ; see Table 1) possibly was the result of less heat exchange with ground water in the drained area due to a lower water table. Analysis of temperatures in this study reflected the same trends as observed for frost. Surficial peat temperatures in the undrained area at 10 em were generally cooler than in the drained area. This is likely due to greater evaporation rates near the surface in the undrained area. Also, more heat energy is required to warm the peat in the undrained area due to higher water contents and hence higher specific heat and heat capacity. Tempera- tures and the rate of warming at lower depths ( 25-60 em) were greater in the undrained area because of its higher thermal conductivity. The wide diurnal temperature fluctuations at 10 em depth may be attributed to the reduced thermal capacity of the peat in the drained area. The effects of evaporative cooling appeared to be restricted to the upper 10 em. The undrained area was cooler at the surface due in part to greater evaporation rates, however at greater depths the higher thermal conductivity of the moist, albeit cool, surface peat in the undrained area resulted in higher temperatures at depth and more rapid thaw relative to the drained area. Evaporation played a more important role in drying of the surface peat, which lowered its thermal conductiv- ity and restricted the flow of heat into the subsoil. The results of this study contain implications for peatland management and utilization in Northern Alberta. The timing of frost recession and snowmelt runoff are of hydrological interest. Snow melt and associated runoff occur on average from mid to late April, when most peatlands are still frozen (Figure 5). The impermeable frost layers act as barriers to infiltra- tion of snowmelt and recharge of peat- lands. This explains the rapid snowmelt runoff from peatlands compared to more prolonged runoff from adjacent mineral 30 Dis- charge 20 M3 /S 10 100% 50 % ~~~~--~~~--~~~--~~ 0 % J F M A M J J A S 0 N D Percent of area with ice Figure 5. Seasonal runoff and frost occurrence at Saulteaux study area. soil sites where infiltration is possible. Furthermore, it emphasizes the importance of summer precipitation as a source of recharge for peatlands and maintenance of peatland ecosystems. Drainage of peatlands in Northern Alberta must be approached cautiously. A consequence of drainage will likely be the alteration of summer thermal regimen in the area, with surface layers warmer than the undrained condition but subject to much wider diurnal fluctuation, and temperatures at lower depths depressed below those of the undrained condition. Drainage under conditions of shallow humic peats of low hydraulic conductivity could lower substrate temperatures enough to induce localized permafrost. The area chosen for this study is well south of both the established limits of discon- tinuous permafrost and the 0 °C mean annual air temperature isotherm, yet 16.7 % of the sample points in the 25 m spacing zone retained ice year-round. Permafrost occurring naturally in Northern Alberta is confined mainly to peatlands, speci- fically where drying of surface peat occurs in summer (Brown, 1977). One would sus pee t that peat land drainage, which causes the drying of surface peat, could easily induce permafrost formation in this region. Further study is needed to verify specific effects of drainage on thermal regimen, for example with respect to peat type and local hydrology. The production of localized perma- frost would be detrimental to the objec- tives of drainage for agriculture and forestry. A frozen layer near the surface would inhibit root penetration and reduced 236 temperatures would slow metabolic activi· ty. The significance of alterations to thermal regime caused by drainage needs to be investigated relative to silvicultural and horticultural requirements of species planned for management on peatlands. REFERENCES Benninghoff, w. S., 1952. Interaction of Vegetation and Soil Frost Pheno· mena. Artie 5:34-44. Brown, R. J. E., 1963. Influence of Vegetation on Permafrost. Proc. Int. Con£. Permafrost, Lafayette, Ind. Brown, R' J. E., 1965. Permafrost Investi· gations in Saskatchewan and Manitoba. Nat. Res. Council, Dev. Bldg. Res., Tech Paper 193. Brown, R. J. E., 1977. Muskeg and Perma· frost. In: Muskeg the Northern environment in Canada, N. W. Radfortb and C. 0. Brawner ed. University of Toronto press. pp. 148-163. FitzGibbon, J. E., 1981. Thawing of Seasonally Frozen Terrain in Central Saskatchewan. Can J. Earth Sci. 18:14921496. Fisheries and Environment Canada, 1978. Hydrologic Atlas of Canada. Printing and Publishing, Supply and Services Canada. Lee, R. 1978. Forest Microclimatology. ColTh~bia University Press, New York. MacFarlane, I. C. (ed.), 1969. Muskeg Engineering Handbook. Univ. of Toronto press. Moore, P. D., and D. J. Bellamy, 1974. Peatlands. Springer -Verlag, New York. Pessi, Y., 1958. On the Influence of Bog Draining Upon Thermal Conditions in the Soil and in the Air Near the Ground. Acta Agric. Scand. 8: 359-37 4. Rigg, G. B., 1947. Soil and Air Tempera· tures in a Sphagnum Bog of the Pacific Coast of North America. Am. J. Bot. 34:462-469. Toth, J. 1986. Personal communication. Department of Geology, University of Alberta, Edmonton, Alberta. Zoltai, S. C.,l971. Southern Limit of Permafrost Features in Peat Land- forms, Manitoba and Saskatchewan. The Geological Assoc. of Canada, Special paper no. 9, pp. 305-310. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 PROBABILITY DISTRIBUTIONS OF RAIN ON SEASONALLY FROZEN SOILS John F. Zuzel 1 ABSTRACT: A method to quantify daily rainfall amounts associated with different anteredent oondi- ticm and the relationship:> between daily rainfall, frozen soil , snow on the ground and runoff are P"esented. 'I'm P"ocedure inoorJXrates a daily soil frost simulation model which P"edicts the jl"esence cr atsence of soil frost fran <Bily weather reoords with algorithms for extracting rain and rain-on-snow events in the P"esEnce cr aooence of frozen soil fran these saine reccrds. 'I'm jl"esence cr aooence of soil frost was simulated fa' a 30 yr period using National 'lleatrer Service reoords fa' HO"o, Oregon. Model out- rut was used to characterize the occurrence of soil frost at the test site and as in]llt to the algcritlln f~ classifying daily precipitation events. Probability distril::utions fa' rain and rain on sn:w in the presence or absence of soil frost are presented. Rain-on-smw events were associated with 57% of the yearly peak discharges at a nearby crest gage. Daily precipitation amount and sna.~ depth jointly accounted fa' 72.6% of the variance in rU10ff peaks at the crest gpge site. (KEY TEfM>: soil froot, simulatioo models, freqJency analysis, rain on sn:w, wintEr rU10ff .) JNTRODUCI'ICN Occurrence of frozen soils is an extranely im- portant oonsideration in understanding runoff and soil erosion in mid-latitu:ie climates Characterized 'rJj 1& winter temperatures and shallow transient snowpacks. Rain on srow in the jl"esence 0" aooence of frozen soil cr rain on frozen ground have been factcrs in many significant floods in the Pacific Nattwes t regi oo ( Jctmon and HcArthur , 1973) • Soil frost 0" frozen soil occurs when the sur- face freezes and moisture migrates fran tOO deeper soil layers to the freezing front and then freezes . Infiltration rates in frozen soil are determined partly by the structure of the soil frost and the soil water oontent at the time of freezing (Storey, 1955) • Trimble et al., ( 1958) proJU!ed four tmns to describe soil frost strl.()tl..l'e: granular, honeycomb, stalactite, and ooncrete. Coocrete frost !'as a dense strl.()tl..l'e resulting in a very lew j:a"llleability and is the most common type of soil frost fOllled in bare agrirultl..l'al soils (Storey, 1955; Trimble et al., 1958) • In contrast, the loose JXrOUS strl.()tl..l'e of granula-frost allGJS water to infiltrate readi.J..y. In an infiltration experiment in northeast Oregon, Pikul et al., ( 1985) reJXrted m infiltration in a fall seeded Walla Walla silt loam (Typic Haploxeroll, ooarse-sil ty, mixed, mesic) winter wheat (Triticum aesti vum L.) ploc with 13 em of ooncrete frost jl"esent. In oontrast, an infiltration rate of 3 em per day was observed when the soil was not frozen. In a forest soil, Kane and Stein (1983) reported an infiltratioo rate of atout 17 em per cay in a lew total moistl..l'e content, frozen, Fairbanks silt loam. In p3rts of the Pacific Northwest east of the Cascade Hountains and in p3rts of tOO IntErlllOUltain West, ooncrete froot is characteristiCBliy formed on bare agricultural soils; thus frozen soils play a majcr role in rU10ff P"'duction, soil erosion, and sedimentatioo. The stu:iy reJXrted here was oondl.()ted using his- torical and recent field data near Horo, Oregon (Fig.Jre 1), in the southern JXrtion of the Columbia Plateau (USDA, 1981). 'I'm area is characterized by hunid winters and dry summers with relatively low rainfall intensity, ustally less th3n 4 rrm per tour (HO"nEr et al., 1957) . Average anntal P"eci pi tati on fer Moro is 288 mm of which 70% occurs during the NovembEr tlro~ April period. 'I'm winta--climate is also characterized by a sffillew transiEnt Sl'Vrl JE.Ck subject to several accunulatim and melt cycles each 1Hydrologist, U.S. Dep3rtmmt of Agrirultll'e, Agrirultll'al Research Serv1.ce, Columbia Plateau Conservation Research Cmter, P.O. Pox 370, Pendleton, Oregon, 97801. 237 .. I ·-.\ ,,..---, I ._ ... -,... \ I I I I I STATE OF ', / OREGON : Moro I I I I ' .. ___________ .J Sherman County Figre 1 • Looati en of tre stuiy site in mrthcentral Oregon. winter. Frozen soils are oomrnon and occur nearly every year. The winter rmoff fran steeply-sloping croplands and rangelands in tre Pacific NO"ttwest can result in serious soil ancsien and sedimentatien and can be tre majO" flood rroducing mechanism in sane watersheds. Quantitative ooscripticns of tre rain- fall amount and frequency associated with winter runoff events are lacking and are EBSEntial fO" m<re accurate predictions of runoff, soil erosion and sedimentation. The objectives of the research rep::rted mre wane to cpantify daily rainfall amomts associated with diffanent antecedent conditicns and to investigp.te tre relaticnshi{l3 between daily rain- fall , frozen soil , smw en tre g"OU1d cmd rll10ff. METHOffi Daily weatrer data, inclu:ling maximun and mini- mun air tanperatures, precipitation, snowfall and snow on the ground were obtained fran National Weatrer Sanvire records for l1oro, Oregon for 1948 through 1978. Missing da.ta wane estimated fran tre nearest weatrer station record which oontained the necessary variable. Also, crest gp.ge data fO" 1959 t!TOI.J@1 1978 fO" Gordon Hall ow at DEl'1oss Springs, Oregon, were obtained from the U.S. Geological Survey, Water Resources Divisien reoords. The crest gage reoords contained only tre water year peak dis- charge so lesser peaks that occurred during the period of reoord were rot available. Tre crest gp.ge site was located 2 km mrth of MO"o, Oregon, and has a drainage area of approximately 23 km 2 • Elevatien of the basin varies from 475 to 664 m while the elevation of the !1oro weather station is 560 m. Dtring eacn of tre five winters of ·1980 to 1984, plot runoff, soil erosion, weatrer variables, soil 238 tanperattres, and soil frcst were moni tared in farm fields located near l'1oro, Oregon (Figre 1). Eadl year a diffanmt site was instrumented because the cropping practice in this area is to seed winter wheat every otrer year in a IXU"ticulcr field. In al- ternate years, these fields are not tilled after harvest rut remain oovered with crop residue over winter. In tre fall of tre seeding year, tre field> ere tille:i and planted to winter wheat; the surface residue and plant oover is minimal during these winter seasons. This rrocedure p->oduced soil frost data fran plots with similcr strface conditicns fcr between year mniXU"isons. All sites were within 2 krn of the Mora weatrer station. Ead1 site was instru- ll1El1ted to continoously record air temperature and relative humidity at 1 meter above the surface. Precipitation· and soil temperature at about 5 em belo.v the soil strface were also monitO"ed (Zuzel et al. , 1982) . Daily solcr radiati en was monitored at tre CB.libratien site in 3 of tre 5 test years. Fro3t oopth was meastred weekly or tllO"e often using frost tubes (Ricard et al., 1976). VisU3.l oottrminaticn of soil frcst try hand sampling was also performed as needed to rrovioo a check on frcst tube meastranents. Smw oopth and smw water equivalent were measured weekly using procedures for sampling very sh3ll01 smw (USDA, 1973). This field data set was used to verify a daily soil frost simulation model vklidt predicts tre rresence or absenre of soil frost fran daily weather records. The soil frost model~ ooscribed by eery (1982) and tre req.1ired inputs are maximum and minimun air tanperattre, smw oopth and solcr radiatien. Tre model predicts net daily heat flux across the soil strface. Tre cunulati ve daily heat flux is tren used to Jredict tre .{resence or ab- sence of soil frcst. Negp.tive valLEs of cunulative soil reat indiCBte tre .{resmce of soil frost while positive values indicate an unfrozen soil. Sim solcr radiation is not included in daily weather records, a method was developed to predict daily valLES of solcr radiation fran potential radiation and daily air tanperattre range. Tre lJI{)(i:)l req.1ires an initial valLE of soil !'Eat at tre beginning of each yearly rm and an anpiriCBl, site specific constant vhl en accounts fO" tre effects of oover and other soil physical Jroperties. 'fll) site constant was obtained by maximizing the number of days tre model correctly Jredicted tre rresence cr absmce of soil frcst over a 5-year field CB.litraticn period using the 1980-1984 field data set fO" inpJt and valida.tion. Initial soil heat was estimated using the average soil tanperattre and reat cap3.city of tre top 3J an of soil • Soil frost fO" a 3:)-year period (1948-1978) was tren simulated using National Weatrer Sanvire record> for Moro, Oregon. Daily maximum air tanperattre, minimun air tan~rattre and snow on the ground for November trrou@1 ~1arch, 1948 to 1 W8 were used in trn simulatioo. Tl"E field verified site constant and initial soil mat were rot varied fran year to year. Mcrel ootp.1t was used to characterize and quantify soil frcst at trn MO"o site and as inp.lt to trn algo-- rithn fO" classifying p-'ecipitatioo events. Precipi tatioo events were classified as: 1) rain on smw, soil frozen; 2) rain on snow, soil thawed; )) rain on frozen soil, ro srow; and 4) rain on \J1- froZ£n soil, ro srow. Tl"E algcri thn used to classify tre r:recipi tatioo events is p-'esented in Fig..~re 2. Fig.~re 2. Fl<NTchart of trn algcrithn wed to classify types of p-'ecipitatioo events fran w=atrer reoords and cunulati ve soil haat. The conditions which must oo met fO" a rain-cn-srDW event to mve occurred are tmt p-'ecipitatioo h3.s oc- curred but snowfall has not; the maxim1.111 air temperature was greater than 0°C; and snow was present on the ground trn !l"evious d:l.y. Tl"E condi- ticm v.ffich must oo met fO" a rainfall event to have occurred are that precipitation has occurred tut 239 sl'DNfall h3.s rot, the maximl.lll air temperature was greater than 0°C and snow was not present on trn groll1d trn !l"evious 0" current d:l.y. In both cases, trn cunulati ve soil haat is wed to ootEnlline wrntrer trn soil is frozen or mt. The undetermined event type occurs when precipi tatioo was mt reoorded as sl'DNfall and trn maximun air temperature was less than 0°C (Figure 2). This sittatioo am oo inter- preted as freezing rain or sleet which is a fairly common occurrence in the Pacific NO"Uwest. It is also JXOOible th3.t this event t~ simply reflects an errO" in trn hist<riCBl reoord. Probability distritutions fO" trn maximun anntal series of rain on srow a1d rain on bare ground were fit to trnoretiCBl distritutions fO" both the frozen soil and thawed soil cases. Tl"E events p-'odt.Ced by trn algorithn were correlated with the crest gage record fran GO"cbn Holl<NT at DEMcss Spring3, Oregon. Tl"E d:l.tes of peak discharge listed in crest gage reoords are rot a1 ways e.xo.ct , so ootail ed e:xa:ni nati oo of weatrnr reoords was neoossary to ootEnlline or con- finn tl"e d:l.te of trn ~ak in all cases. RESILTS AND DisaJSSICN For the 1980-1984 calibration period, a site constant of 1 .0 and an initial soil heat of 4190 kJ /m 2 prodl.Ced trn maximun nunber of correct pr8dic- tions of trn !l"esence or absence of soil frost when compared to the field observatiorn. Overall, too model correctly !l"edicted trn !l"esence or absence of soil frost 80 percent of the time for the 5 year CBlitratioo ~iod (Table 1). Model output for tre 30 year weatrer record at MO"o, Oregon suggests th3.t frozen soil occurs every winter at this site. Tl"E soil was frozen fran 6 to 116 days per year with an average of 57 days per year. Freeze-tl'Bw cycles (otrnr th3.n ditrnal cycles) varied fran 1 to 7 cycles ~ year. The distribution of predicted soil frost by months shows that historically, January has the highest irdoonce of soil frcst; tm soil was frozen 67% of trn time. Also, model output indicates that trn soil was frozen ffiO"e th3.n 50% of trn time in both December and FetrU3ry. Between 6 and 116 days ~ year of frozen soil can oo expected at MO"o, Oregon. Tl"E results of the simulations for nunber of frozen soil d:l.ys ~ year were fitted to a rormal distritutioo. (Fig..~re 3). The estimated parameters of the distritutioo are also shJwn in Fig..~re 3. In oroor to test trn goodness of fit of trn ror- mal distribution, a correlation analysis was performed using trn eXDeedance !l"Oba.bilities oori ved TABlE 1. Top:>gra!ilic ctlaracteristics of sites near i1<ro, Oregon used to verify the frost !X'ediction model. Distance Site and Frost fran i1<ro winter Elevation Sloj:e Asj:ect jX'edi cti on1.1 weather nunber (meters) Year (percent) (degree:;) (percent) station (km) 1 561 1W9-1980 14 S401t1 83 0 2 567 1980-1981 11 351W 82 .5 3 539 1981-1982* 24 N65E 91 2 4 539 1982-1983* 24 tt15E 63 2 5 539 1983-1984* 24 N65E 80 2 _ll Percent of d3.ys ootween 1 November and 31 I'1arch in which frost CX!Currmce was oorrectly !X'edicted. * Scme site. from model output and the theoretical exceedance JX'OtabilitiEE. This test yielred a oorrelation ooef- ficiEnt of 0. W6. The chi square test in:iiCE.ted that the hyp:>thesis of a rormal distrirution oould not be rejected at the 5% signifiCE.nce level. An obvious differEnce between the distributions exists in the lower tail because the theoretiCE.l distrirution is mtX:)lll1ded while the model derived d3.ta has a lCNier +2 ', ' Z1 RANDOM NORMAL DEVIATE +1 0 -1 I I I 1201-~' ..... ... Ill • > .... • > Ill - e 80- ...I 0 0 z w N ~ 401- LL. 1- ' >-Y • t.J + 0Z1 '-<./' ',· . ,, ... ' .. Parameters .... 57.33 o• 34.08 I 10 ' ' .. ' '· ' ' . ' ' ' · ... ', ·.~, . ~, ' .. , ., I I I I I I '~ 30 50 70 90 EXCEEDANCE PROBABILITY (Percent) -2 99 Figwe 3. Cunulative freq..tency and fitted theoretiCE.l distrirution of the nunber of froza1 soil d3.ys per year at M<ro, Oregon. 240 bol.ll1d of 6. Fran these analyses we concluded that the normal distribution can, with the indicated parruneters (Figure 3), be used to represent the simulation model distribution for the number of froza1 soil d3.ys par year. Analysis of the Moro weather record using tre alganithm shown in Figure 2 produced the results jX'esmted in Table 2. A rain-on-sn:M event with either froza1 or thawed soils occurred in 28 of the 30 years of record at this location; a jX'otability of occurrEnce in any year of 0. 93. Rain on frozen soil in the absmce of srow oover CX!Curred in 25 of ti--t 30 years (P=0.83). Rain on srow with froza1 soil had an occurrence probability of 0. 73, while rain on smw with thawed soils had a !X'Obability of 0.57 (Table 2). ~ annual SEries of maximun precipitation for each year of CX!Currmce were extracted f<r all event types except the tndetermined. Summary statistics are presEnted in Table 3. Because all of the anmal SEries are rotioeably skewed, the data were fit to the lognormal, gamma, and 3-parameter lognormal distrirutions. ~ 3-parameter lognormal distribu- tion !X'oduced tre best fit in all CE.Ses. Goodness of fit was subjectively judged by comparing observed data parameters with p3rameters obtained using data generation techniques. Goodness of fit was statisti- cally assessed using the chi-square statistic with 10 class intervals defined so that the expected number of observations in each clas."3 interval was the scme. ~ hyp:>tresis that too 3-parameter lognormal dis- tribution describes the c:Bta oould rot oo rejected f<r any of the event types at the 5% significance TABlE 2. Re:>ul ts of c:Bily rainfall event classifiretim fer Mero, Oregon fer too Novanber- March time period, 1948-1978. fllent type Nunber of Events FroZEn soil TtB.wed soil Total Rain 39 829 1088 ( .83) (1.00) (1.00) Rairroo-smw CJ5 38 133 (. 73) ( .57) ( .93) urretEfllli ned 39 NunOOr's in p3rmtooses irx:iimte too p->ot:Bbility of occurrrooo in any year. while the distribution fer rain on trowed soils in too aoorooo of smw am be ooscribed by: Y = (e3.23221 + .30419Z _ 10 •525 ) (4) Equatioos smh as 1 through 4 can be used directly to generate daily precipitation amounts usingMante Carlo simulation. This is done by allowing Z to deviate randomly. The simulated mta mn tren be used as inp.1ts fer jilysiCBlly t:Bsed runoff, erosion and sedimrot ffiOOOls. TOO cre:>t gpge mta analysis for Gordon Hollow provided an estimate of the precipitation amount necessary to {l"odme runoff as well as suggesting the relationships between runoff peak, snow on the ground, and mily rainfall amount. TOO ~ tB.d teen in operation since 1959 and of too 20 yearly peak disd'larges recorood tl'rough 1978, five years tB.d 00 TABlE 3. St.mncry statistics fer 8 rain-m-sn:M runoff events at G<roon Hollav at DEMcss Springs, Oregon. Variable Me:m Precipitation (mm) 14.93 Average Air TE!Ilperattre ( °C) 4.87 Maximun Air TE!Ilperattre ( oc) 8.62 Anteoooont srow on ground (an) 6.87 Peak Disd'large (m 3 /sec) 9.1 level. The the or eti cal and ol:sErved distri b.lti ons ' fer rain on srow with ttB.wed soil, rain on snow with frozen soil , rain on froZEn soil , and rain on ttB.wed soil are presented in Figures 4 through 7. The CI.IDlulati ve frequency distrihltian fer rain on srow with th:l.wed soils is: Y = 0 .57 (e2.14767 + 0.61338Z _ 4.B2S) (1) we Y is maximum precipitation (mm)· and Z is a standard normal ooviate. TOO distrib.ltian fer rain on srnw with froZEn soils is: Y = 0 .73 (e2.89757 + 0.39667Z _ 10 .39 ) (2) and fa> rain on froZEn soil in too absroce of smw: Y = 0 .83 (e2.67804 + 0.513272 _ 6 •964 ) (3) Medicn Std. Dev. Max. Min. r 13.97 8.74 32.00 5.3 0.890 4.50 1.% 8.00 3.0 -{), 318 8.00 2.88 12.00 4.0 -0.147 3.00 7.72 3.4 2.5 0.637 4.3 10.4 27.9 1.2 1.00 241 flav, one peak disd'large occurred in Jme and 14 oc- curred during too November tl'rough March period. T're analysis sh:>wed ttB.t 8 of toose 14 events (57%) were the result of rain on snow while the remaining 6 events wa"e too result of rain with no snow cover. Mer rover, too soil was froZEn during 7 of too 8 rain- m-sn:M events and froZEn during 1 of the raim.'all events. Peak discharges ranged fran 1 • 2 to 27.9 m3 /sec with a me:m of 9.1 m3 /sec, while p->ecipitation ranged from 5 to 32 rrm with a mean of 15.0 mm. Statistics of the rain-on-snow runoff events are shown in Table 3 as well as tre oorrelation ooeffi- cimts fer each variable with peak discharge. A regression analysis using anteoooont srow on too gromd and {l"ecipitation amount as independent variables and peak discharge as too oopenoont vari- able {l"odt.red too re:>ults sh:>wn in Table 4. Average air temperattre a1d maximun air temperattre wEre also irclt.ded to sh:>w the relative importance of these variables in rain-m-sn:M events. It sh::>uld be mted that while both antecedent snow on the ground 45~------------------------------, 35 Figure 4. Cunulati ve frecpency and fitted thaoretiali distribltioos of tre annwl series fer rain on sn:M with thawed soils. 45r-------------------------------~ • 49 ~ 35~ I I"\ I •sa ... \ II! r-I v ' ' ' ' ~ ' 5 ' ., ' ' ... . .... -.... '• .... • . '""' ..... ............. " ................ ..... ~~ . ·~ .. ~ 9~~~~~~~~~~~~~~~r~~~,~~~a··~-~· 9.9 9.2 9.4 9.8 9.8 t .9 EXCEEDANCE PROBABILITY Figure 6. Cunulati ve frecpency and fitted thaoreti ali distribltions of tre annwl series fer rain on frozen soils in tre aoomce of Sn:M oover. 242 45r-------------------------------~ 35 Figure 5. Cunulati ve frecpency and fitted thaoretiml distribltion of tre annwl series fer rain on srow with frozen soils. 45~----------------------------~ 49 • 9~~~~~~~~~~~~~-~·~~ 9.9 9.2 9.4 9.8 9.8 t .e EXCEEDANCE PROBABILITY Figure 7. Cunulative frecpency and fitted thaoretiCBl distribltion of tre annwl series fer rain on thawed soils in tre aoomce of sn:M oover. and rrecipitatien each e~lain a signifimnt jH'rent- age of the observed variation in peak flow, the combination of both does not increase explained variance to a signifimnt degree. This is JTOi:ably OO<Euse tre values of snow on the ground are con- oontrated at 3 to 5 em. If rocre srow d:ita in tre mid and higtler ranges WEre available, srow en the ground would JTOi:ably assune ma'e signifimnoo. TABlE 4. Coefficia1ts of detff'Ulinatien fO" 2 i~pandent variables en p3ak discharge at GO'cbn Hollcw near DEMcss Springs, Oregon. Variable Anteredent smw en gromd (em) Preci pi tati en (rrm) B::>th Average Air Tanperattre ( °C) Maximun Air Tanp3rattre (°C) SlM-11\RY AND CDNU.lBICNS r lj() 79 00 10 2 A daily soil frost simulation model which predicts the presmoo 0" a.tsmoo of soil frost fran tE.ily weatrer records was combined with algorithms for extracting rain and rain-cn-smw events in tre l!'e5Enoe and absmoo of frozen soils fran these same records. Overall tre soil frost JXrtim of tre mOO:::ll rorrectly JTedicted tre jresffioo or absence of soil froot 00 peroont of tre time fa' a 5-year (1 Novanber to 31 March) calitratien JH'iod. Soil frost for a 30-year jH'iod was simulated using National Weather Service historical daily weather records fa' MO'o, Oregon; within 2 km of tre field m.lil:ratien site. Daily maximun air tempera- tll'e, minimun air tanperattre, and smw en tre ground f1r Novanber trrougTJ. t1ard1, 1948 to 1978 WEre wed in tffi simulatioo. Tre field verifioo site ronstant and initial soil reat WEre rot varia:! fran year to year. Hodel output was used to jredict soil frost at tre ~l1ro site and as inp..1t to tre algcrithn f.cr classify- irg daily jrecipitatien events. Rreul ts srowed trat hista'ia3lly, frozen soils occurred every year and the soil was frozen an average of 57 days JH' year. Tre highest incidence of frozen soil occurs during Jmuary wren tre soil was frozen 67 pa"oont of tre time follcwed clcsely by February (53%) and December (51%). Tre nunber of freeze-th3.w cycles otrer tran ditrnal cycles varied fran 1 to 7 per year with an average of 3 per year f1r tl'e simulatien period. Frecpency distributions 243 of tre nunber of cays !H' year with frozen soils WEre fit to a normal distribution. Event analysis jrodLCed 133 rain-cn-sn::M events of which 95 occurred while tre soil was frozen. Rain on snow occurred in 28 of tre 30 years of rerord at this location (P=0.933). The average number of events per year was four. However, tre anomt of ITecipitatien was less tran 2.5 rrm for 50% of these events. Rain-cn-SI'XYvT events WEre associated with 57% of tre yearly p3ak discharges at a near'Q[ crest gage in operation from 1959 through 1978. Daily jrecipitatien anot.rlt and smw depth jointly acromted for 72.6% of the variance in runoff p3aks at this site. Prob3.bility distributic:m fa' rain on smw and rain on frozen ground were fit to t!Tee p3ranetEr log10r'Ulal distributic:m fa' roth tre frozen soil and trawed CBSes. Tre mOO:::ll described in this paper is simple, easy to use and reqJires c:nly tre readily available inp..1t cata of d:iily maximun and minimun air tanpera- ttre, precipitatien, srowfall and smw en tre gromd. It mn be wed in any area wtere a <Xll!plete histori- cal weather record is available to estimate the paranetErs of tre jrob3.bility distributions. Model outp..tt should prove useful in runoff and ercsicn jrediction and in the design of erosion control, water conservation and drainage facilities. Equaticns 1-4 can be wed directly to genErate daily precipitation inputs for physically based rU10ff, ercsicn, and sedima1t jrodLCtion mOO:::lls. Freeze-traw cycle infO'maticn mn be wed in plmning strLCttres 0' roadbeds in areas ~re frcst reaving is a poten- tial jrOblan. REFERENCES Cary, J. W. 1982. Arroult of soil ioe jredicted fran weatrer obsErvatic:m. Agric. Meteorology 2'7: 35- 43. HO'nEr, G. M., W. A. Starr, and J. K. Patterson. 1957. The Pacific Northwest wheat regicn. In Soil, Tre Yearbook Of Agriculture. U.S. Dept. Agr., Washington, D.C., pp. 475-481. Jomson, c. W. and R. P. McArthur. 1973. Winter storm and flood analyses, NO'ti"Mest IntEriO'. In Proc. Hyd. Div. Spec. Ccnf. Hydraulic Engineering and the Enviroment. Am. Soc. Civil Eng., pp. 359-369. Kme, D. L. and J. Stein. 1983. WatEr movemEnt in seasonally frozen soils. Water Resources Res. 19:1547-1557. Pikul, J. L., Jr., J. F. Zuzel, and R. N. Greenwalt. 1985. Water infiltration in frozen soil. In Columbia Basin Agl"icul tural Research, Oregon Research Experiment Statim Special Reprt 738, pp. 6-11. RiCE.rd, J. A., W. Tobiasson, and A. Greutcrex. 1976. ~ field assembled frcst gp.ge. Te:::h. Note. Cold Regions Res. Eng. Lab. CO"p5 of Eng., U.S. Army, Hanover, Nev Hemp., 7 pp. Storey, H. C. 1955. Frozen soil and spring and winter floods. In Water, The Yearbook of Agriculture, U.S. Dept. N9'., Washington, D.C., pp. 179-184. Trimble, G. R., Jr., R. S. Sartz, andR. S. PiErce. 1958. How type of soil frost affects infiltration. J. Soil and Wattr Ccnserv. 13:81- 82. u.s. Dept. of Agrictilttre. 1973. Srow strvey san- pling guide. Agr. Handbook 169. Washington, D.C., pp. 26-27. U.s. Dept. of Agric. 1981. Land Reso~rce Regicns and Major Land Use Resource Areas of the United States. Agr. Handl:xx>k 296, Washington, D.C., pp. 6-7. Zuzel, J. F. , R. R. Allmaras, and R. Greenwalt. 1982. RU1off and soil trooim on froza1 soils in mrtheastern Oregon. J. Soil and Water Cons. 37(6):351-354. 244 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 EVIDENCE OF GROUNDWATER RECHARGE THROUGH FROZEN SOILS AT ANCHORAGE, ALASKA 1 James A. Munter ABSTRACT: Three water-level observation wells located within the Municipality of Anchorage, Alaska have exhibited distinc- tive and anomalous water-level rises during January 1985, in conjunction with mid-winter thaws accompanied by precipita- tion events. One of the wells exhibited similar behavior during January, 1984. The magnitude of the rises is about 20-60 percent of the total annual water level fluctuation of each well. The surficial geologic materials near each well are at least moderately permeable under thawed conditions but are frozen near the land surface during the January events. The data suggest that the absence of available water may be the primary reason for the absence of groundwater recharge in areas ~ere soils are frozen during part of the year, rather than the reduced permeability of frozen soils. (KEY TERMS: recharge; frozen soils; groundwater; Anchorage; Alaska.) INTRODUCTION Groundwater recharge is poorly under- stood in many geological and climatolog- ical environments, yet is of importance for many applications, including water- supply assessment, runoff or flood fore- casting, and waste-disposal evaluation. The rate of groundwater recharge is influenced by seasonally frozen soils in areas where several months of subfreezing temperatures occur. Zenone and Anderson (1978) suggest that frozen soils can act "as a relatively impermeable barrier that temporarily restricts ground-water re- charge". Kane and Stein (1983) demon- strated that the rate of infiltration is based on soil texture and total soil moisture (water and ice) conditions, with high ice-content soils having a relatively low permeability. Because low ice-content or fractured frozen soils may occur in many settings, the process of groundwater recharge through frozen soils is of considerable interest. In this paper, the geologic and climatologic conditions surrounding anomalous mid-winter episodes of rising water levels in three observa- tion wells within the Municipality of Anchorage, Alaska, are presented. Exam- ination of these conditions may prove useful for evaluating groundwater recharge processes in other areas. OBSERVATION WELLS Three water-level observation wells within the Municpality of Anchorage are described in this paper (fig. 1). The Potter Marsh wells were drilled in 1984 as observation wells for the Alaska Division of Geological and Geophysical Survey (DGGS) under a contract with a local water-well driller, and the Ravenwood well was drilled in 1983 as a water-supply exploration hole by a driller contracted by the Anchorage School District. 1Hydrogeologist, Alaska Department of Natural Resources, Division of Geological and Geophysical Surveys, P.O. Box 772116, Eagle River, Alaska 99577. 245 ALASKA (\ Hili /,ah•' 4ffll~~~ 6130 ~ ./6/ ••• • Study area 61 '00 0 I I·, •• I ••• •• I 0 8 16km 1 soo oo· 149"00' Figure 1. Locations of observation wells. Ravenwood Well The Ravenwood well was drilled on a bench-like ground moraine deposit composed principally of diamicton (Schmoll and others, 1980) that is 118 m in elevation above the Eagle River at its closest point. Surface runoff in the vicinity of the well occurs at a low rate because of the mostly forested conditions, but follows the land surface towards the Eagle River, located 900 m south of the well. A driller's log of the well, with casing data, is presented in Table 1. Two relatively shallow intervals in the well were screened and tested. The depth of the well was sounded during 1984 and 1985, and was found to be 13.4 m, suggesting that sediments had infiltrated into the well during test pumping, and had not been removed. The well was successfully slug-tested to assure its adequacy as a reliable observation well. The aquifer tapped by the Ravenwood well, called the Ravenwood aquifer by Munter (1984), has also been tapped by 246 TABLE 1. Log of Ravenwood Observation Well. Depth interval (m) 0.0-0.3 0.3-3.7 3.7-6.1 6.1-11.0 11.0-13 0 7 13.7-14.6 14.6-16.5 16.5-17.4 17.4-18.6 18.6-26.5 25.6-87.8 87.8 87.8-107.6 107.6-109.1 Depth interval (m) 0.0-13.1 13.1-14.6 14.6-16.2 16.2-17.7 17.7-18.3 18.3-26.8 26.8-29.9 29.9-87.8 Material description Organic topsoil Silty gravel-loose Gravelly silt Silty gra~zl with water, 2.2 X 10 1/s± Gravely silt, hard Sandy gra~Il with water, 2.2 X 10 1/s±(screened) Gravelly silt, hard Silty gra~jl with water . 8.8 X 10 1/s±(screened) Gravelly silt, hard Gravelly silt, hard_3 w/seepag~2 8.8 X 10 3.5 X 10 1/s Gravelly silt, hard to silty sand, some with water seepage Boulder, casing stopped driving Gravelly silt, hard Sandy gravel with water -won't stand open Casing information 15 em ID steel casing No.100 slotted screen (2. 54mm slot openings) 13cm ID steel casing No.100 slotted screen (2. 54mm slot openings) 13cm ID steel casing Steel casing, backfill~ with clean gravel Steel casing, backfill~ with cement grout Steel casing, backfillei with silty gravel Well completed March 29, 1983. other residential wells in the vicinity., The aquifer is relatively flat-lying, wit~l the thickness of the overlying confinin1i unit increasing to about 30 m to the north, toward the Eagle River valley margin. The Ravenwood aquifer is a few meters, or less, thick. A zone of ground- water discharge from the Ravenwood aquifer occurs about 300 m southeast of the Ravenwood well, where the land surface drops steeply to the river level and the Ravenwood aquifer is inferred to subcrop beneath the surficial soil mantle. The sediments overlying the Ravenwood aquifer are described as moderately well drained to moderately poorly drained, with low surface runoff by Zenone and others (1974). Numerous single-family domestic septic systems operate in the vicinity of the Ravenwood well without apparent difficulty, suggesting that the surficial diamicton deposits in the area have at least moderate permeability under thawed conditions. Potter Marsh Wells 1 and 2 The Potter Marsh wells 1 and 2 are located on artificial fill between the Old Seward Highway and Potter Marsh in south Anchorage (fig. 1). The geologic materi- als near the wells are of highly variable composition, from dominantly silty to dominantly gravelly fill, and are fairly level and well drained at an elevation about 10 m above the level of the marsh. The wells are about 2 m apart, about 5 m from the road ditch of the Old Seward Highway, and about 80 m from Potter Marsh. A small tributary to Potter Marsh passes within about 100 m of the well site, and is about 3-4 m lower in altitude than the land surface at the well site. The wells are located outside of a small alluvial fan associated with the stream, suggesting that streamflow ·does not directly affect water levels in the wells. The wells were drilled with an air rotary drilling rig, and logs are shown in Table 2 and Table 3. The abundance of dark grey silt in the sediments penetrated by well 1, and the location of the site near Turnagain Arm, suggest that most of the sediments at the site are of glaciomarine or glacio- lacustrine origin. 247 TABLE 2. Log of Po-tter Marsh Observation Well 1. Depth interval (m) o. 0-1.5 1.5-2.7 2.7-3.0 3.0-8.7 8.7-10.7 10.7-11.6 11.6-12.8 12.8-13.7 Material description Dark grey silty gravel fill Dark grey silt, damp Grey gravel Dark grey silt, moist, with clay, sand, and gravel Dark grey, silty, fine- grained sand, with coarse sand to gravel Light grey, silty, very fine-grained sand, with sand and occasional gravel Dark grey, silty, sandy gravel Dark grey, medium-to coarse- grained sand, with subangu- lar to angular gravel. Yield tested by aiE~ift method aE 2 1.3 X 10 - 2.2 X 10 1/s for 25 min. Cased with 15 em ID steel casing to 13.7 m below land surface. Well drilled and logged May 22, 1984. TABLE 3. Log of Potter Marsh Observation Well 2. Depth interval (m) o. 0-1.5 1. 5-2.4 2.4 2.4-3.5 Material description Dark brown gravelly sand fill Dark grey silty gravel fill Wood fragment Gravel and water, yield tested b~2 airlift meth£~ at 8.8 X 10 to 1.8 X 10 1/s for 30 min. Cased with 15cm ID steel casing to 3.0 m below land surface. Well drilled and logged May 22, 1984. vJater-level Monitoring Equipment Water-level data described in this paper were recorded with a microprocessor attached to a float and pulley system that senses water-level changes. The float and pulley system causes the shaft of a poten- tiometer to rotate, which is sensed by the microprocessor. The microprocessor con- verts the signals into digital data that are stored on an erasable solid state memory module. The module is periodically retrieved from the field, and the data are processed with DGGS computer facilities. Water-levels recorded by the micro- processor are converted to a common datum and verified for accuracy with wetted- steel tape measurements. WATER LEVEL AND CLIMATOLOGIC DATA Water-level data collected at the Potter Marsh and Ravenwood wells during water years 1984 and 1985 are shown in Figure 2. c: E LJ.J (l) 1-(.) <( co -$ ,_ ::J 0 Vl 1--o c:: :c ~ 1-~ a.. LJ.J 0 0 (l) aJ 2 3 4 5 · · · · · · · · · · Missing data ----Well dry RAVENWOOD WELL The data show that the largest water- level rises in the Ravenwood well occur during the March-April and August- September periods for water years 1984 and 1985. These periods correlate with sig- nificant episodes of snowmelt and rain- fall, respectively, in the Anchorage area. Less distinctive periods of rising water levels are evident from the Ravenwood well data for January 1984, and from the data from all three wells for January 1985. The magnitude of these water-level rises varies from about 20 to 60 percent of the range of the annual water-level fluctua- tion for each well. The water-level rises are anomalous episodes in the otherwise gradual decline of groundwater levels during the winter months. For more than two months prior to the January episodes of rising water level, temperatures in Anchorage were mostly sub- freezing (NOAA, 1983; NOAA, 1984b), indicating that surficial soils near the well sites were probably frozen. Selkregg (1974) shows that soils in Anchorage are POTTER MARSH WELL 2 POTTER MARSH WELL 1 / ltV I ·.; J FMAMJ JASONDJ FM J A S 1984 Water years 1984-1985 Figure 2. Water-level data from observation wells, water years 1984-1985. 248 frozen to an average depth of more than 0.6 m during January. Water-level data collected at the Ravenwood well during January, 1984, are shown in Figure 3, with climatological data obtained at the Anchorage Interna- tional Airport in west Anchorage (NOAA, 1984a). Subsequent to above-freezing temperatures on January 11 and 12, 1984, and 0. 53 em of precipitation (including 9.1 em of snow) on January 11, the water- levels in the Ravenwood well began to rise during the first few hours of January 13. The water level continued to rise through January 14, with above-freezing tempera- ~res that prevailed during this period, some or most of the precipitation recorded on January 11 and 14 must have fallen as rain, or melted soon after falling as snow, thus becoming available for percola- tion through the snowpack and infiltration into the ground. January 1985 had two distinct periods of above-freezing days accompanied by pre- cipitation events and rising water levels in the Ravenwood and Potter Marsh wells 1 and 2 (Figure 4). The initial rise of water levels in the Ravenwood well and Potter Marsh well 2 occurred on January 10, the warmest day of a 6-day thaw on which 0. 20 em of precipitation (with no measureable snow) fell (NOAA, 1985). Water-level rises that also occurred during late January were during or shortly following above-freezing temperatures and measureable amounts of precipitation. Observed water levels in Potter Marsh well 1 show a subdued, yet distinguishable pattern of fluctuation similar to those observed in the Ravenwood well and Potter Marsh well 2 (Figure 2, Figure 1). DISCUSSION Water-levels in observation wells can respond to a variety of stimuli, both natural and man-induced. Pressure phenom- ena such as air entrapment, tides, external loading, earthquakes, or baro- metric effects last, at most, a few days (Freeze and Cherry, 1979, pp. 230-233), in ('") 3 _4 a: E Daily precipitation w- 1-"' <t :;; S:'t: ::> 0 "' ..... ., c:-::t: (Q ..... - ~ s: w 0 oa; co E u 10 15 20 25 30 JANUARY 1984 10 --i m 0 s: "'0 m jj -10 l> --i c jj -20 m n -30 5 10 15 FEBRUARY 1984 Figure 3. Water-Level Data from the Ravenwood Well and Climatological Data from the Anchorage Weather Service Meteorological Office, Anchorage International Airport. 249 a: E w- 1-"' <t a; S:'t: ::::l 0 "' 1--c 0.5 C"l 3 1.0 :::1: ~ ~-­ a_ s: w 0 OQi Potter Marsh Well 1~--------- co -t 10 r'h $ -o 0 m :0 l> -t -10 c :0 m E -20 0 u ~ JANUARY 1985 FEBRUARY 1985 Figure 4. Water-Level Data from Observation Wells and Climatological Data from Anchorage Weather Service Meteorological Office, Anchorage International Airport. contrast with the elevated water levels that persisted for weeks at the observa- tion wells. No pumping, dewatering, or artificial recharge is known to have occurred near the observation wells that could have resulted in the observed water levels. Schneider (1961) has suggested that melting at the base of the seasonal frost layer within a few days of the onset of above-freezing air temperatures can result in a significant amount of ground- water recharge. In Anchorage, several periods of above-freezing temperatures that were devoid of precipitation clearly result in level or steadily declining water levels (Figure 3, Figure 4) • This suggests that either the above-freezing weather did not persist long enough to cause melting at the base of the frost layer or that the moisture content of the soils at the base of the frozen soil layer were too low to contribute signficantly to groundwater recharge. In either case, the 250 Anchorage data do not support the hypo- thesis that the initial surge of ground- water recharge is attributable to melting at the base of the seasonal frost layer. The data shown in Figures 2 through 4 exhibit strong correlations between above- freezing temperatures, precipitation, and rising water-levels during midwinter, a period usually characterized by ground- water recession conditions. Although the present study did not include detailed tracking of moisture and temperature conditions in soils, the elimination of other mechanisms leaves direct groundwater recharge resulting from surface water infiltration and perocolation as the most reasonable explanation of the observed water-level rises. The responsiveness of the water levels in the three monitoring wells to the climatic conditions is related to the thickness and type of strata that overlie the aquifers that are tapped. With about 3 m of unsaturated gravelly fill overlying the aquifer, Potter Marsh well 2 shows a response to precipitation events within a few hours to a day. Recharge probably occurs in close proximity to the Potter Marsh wells because of the local distri- bution of the artificial fill. The Ravenwood well responds to climat- ic conditions in a matter of a few hours to a day or two. This would seem to be relatively rapid, considering the thick- ness (about 6 m) of till in the area and the relatively low hydraulic conductivity of till. Freeze and Banner (1970) noted the occurrence of relatively rapid re- charge through till, and postulated that fracture flow may be a significant compon- ent of flow through till deposits. At the Ravenwood site, fractures may have been present in the unfrozen till above the Ravenwood aquifer, or in the frozen soils near the land surface or both. Potter Marsh well 1 responds sluggish- ly to climatic conditions because of the thickness and low permeability of sedi- ments penetrated by the well (Table 2). The fact that the water-level in the well responds at all, however, illustrates the importance of recharge to aquifers typically considered to be confined and relatively unaffected by short-term surficial hydrologic processes. Annual low water levels in the Raven- wood well occur during March of both years of record, immediately prior to major spring snowmelt episodes. The recharge events that occurred during January of 1984 and 1985 caused the elevation of the annual low water level in the Ravenwood well to be higher than it otherwise would have been because of the long imd gradual recession of the hydrograph from the re- charge events (Figure 2) . The elevation of the annual low water level is important because numerous residential wells tapping the Ravenwood aquifer have a low tolerance to water-level declines because of rela- tively low static water levels (Munter, 1984). The occurrence of mid-winter groundwater recharge could prevent diffi- culty for some Anchorage domestic well 251 owners with shallow wells and low static water levels. CONCLUSIONS Groundwater recharge is concluded to have occured through frozen soils at Anchorage, Alaska during mid-winter periods of above-freezing temperatures and precipitation, some of which occured as rain. Significant recharge is not evident during mid-winter periods of above- freezing temperatures devoid of precipitation. The primary factor governing ground- water recharge in areas where soils are at least moderately permeable and frozen may be the availability of significant quantities of water at the land surface. Because soils of at least moderate permeability are widespread, the major reason for the lack of recharge in most areas where soils are frozen during part of the year is probably the unavailability of water, rather than the reduced perme- ability of frozen soils. Mid-winter groundwater recharge at Anchorage may constitute a significant percentage of total annual groundwater recharge during some years. Inaccuracies may result if the effects of mid-winter recharge are ignored during consideration of seasonal water-level fluctuations for water-supply or waste-disposal purposes, or runoff forecasting. ACKNOWLEDGEMENTS The author thanks Larry L. Dearborn, Gary S. Anderson, and an anonymous reviewer for helpful reviews of the manuscript. LITERATURE CITED Freeze, R.A., and Banner, J., 1970. The Mechanism of Natural Ground-Water Recharge and Discharge 2. Laboratory Column Experiments and Field Measurements. Water Resour. Res. 6(1): 138-155. Freeze, R.A., and Cherry, J.A., 1979. Groundwater. Prentice-Hall, Inc., Englewood Cliffs, N.J., 604 pp. Kane, D.L. and Stein, J., 1983. Water Movement into Seasonally Frozen Soils. Water Resour. Res. 19(6): 1547-1557. Munter, J.A., 1984. Ground-water Occur- rence in Eagle River, Alaska. Alaska Dept. of Natural Resources, Di v. of Geological and Geophysical Surveys, Report of Investigations 84-21. 15 pp. NOAA, 1983. Climatological Data, Annual Summary, Alaska, 1983. U.S. Dept. of Commerce, National Oceanic and Atmospheric Admin. 69(13). NOAA, 1984a. Climatological Data, Annual Summary, Alaska, January, 1984. U.S. Dept. of Commerce, National Oceanic and Atmospheric Admin. 70(1). NOAA, 1984b. Climatological Data, Annual Summary, Alaska, 1984. U.S. Dept. of Commerce, National Oceanic and Atmospheric Admin. 70(13). NOAA, 1985. Climatological Data, Annual Summary, Alaska, January, 1985. U.S. Dept. of Commerce, National Oceanic and Atmospheric Admin. 71(1). Schmoll, H.R, Dobrovolny, E., and Gardner, C.A., 1980. Preliminary Map of the Middle Part of the Eagle River Valley, Municipality of Anchorage Alaska. U.S. Geological Survey Open-File Report 80-890. Schneider, R., 1961. Correlation of Ground-Water Levels and Air Tempera- tures in the Winter and Spring in Minnesota. U.S. Geological Survey Water Supply Paper 1539-D, 14 pp. Selkregg, L., 1974. Alaska Regional Pro- files, Southcentral Region. University of Alaska, Arctic Environmental Information and Data Center. 255 pp. 252 Zenone, C, Schmoll, H.R., and Dobrovolny, Ground Water for in the Eagle E., 1974. Geology and Land-Use Planning River-Chugiak Area, Geological Survey Open-File 74-57. 25 pp. Alaska. U.S. Report Zenone, C., and Anderson, G.S., 1978. Sum- mary Appraisals of the Nation's Ground-water Resources-Alaska. U.S. Geological Survey Professional Paper 813-P. 28 pp. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 RESIDENTIAL WELL DEVELOPMENT OF A LOW PERMEABILITY BEDROCK FLOW SYSTEM William A. Petrik 1 ~STRACT: Complaints of inadequate domestic well yields in the hillside area of Eagle River, Alaska, prompted a survey of existing water well data, measurement of water levels in private domestic wells, and interviews with homeowners regarding water use and water shortage problems. Water levels were measured in 59 domestic wells during August and September, 1985. Fifty-six of these wells were drilled into bedrock. The depths of the wells vary from 11 to 215 m (36 to 706 ft). Well logs show unconsolidated sediments above the McHugh Complex metamorphic rocks to vary in thickness from 0. 3 to 37 m (1 to 121 ft). Aquifers within the McHugh Complex rocks have virtually no primary permeability and low to very low secondary permeability. Reported well yields vary from 0.01 to 1.32 1/s (0.1 to 21.0 gpm). The median well yield is 0.13 1/s (2.0 gpm). Comparison of 50 original reported static water levels to 1985 measurements indicate current water levels are an average of 0. 3 m (1 ft) above originally reported values. This is not considered to be a significant change. No definitive areal trend of either rising or falling levels exists. Apparent water level changes exhibit an irregular areal distri- bution, are commonly many meteTs (tens of ft) in magnitude, and cannot be consist- ently correlated with any known factor. Domestic water use is highly variable and shortages have occurred, indicating that some wells are inadequate for domestic needs. (KEY TERMS: aquifer; primary permeability; secondary permeability; water levels.) INTRODUCTION The community of Eagle River is located 24 km (15 mi) from downtown Anchorage, Alaska. Eagle River has experienced much growth in the last decade. Increased concentrations of residences now occupy the moderately sloped hillsides surrounding the core of the town. On the hillside, most water is provided by private domestic water wells. Along the north hillside area closest to Eagle River, many reports of domestic water well levels declining and wells going dry have been received by the Alaska Department of Natural Resources (DNR). Beginning August 28, 1985, DNR's Division of Geological and Geophysical Surveys (DGGS) conducted a five week field survey of domestic water wells to determine the nature of these problems and their implications. This reconnaissance level study was centered in the Meadow Creek drainage area within the SE~ of Sec. 1, Tl4N, R2W, and Sec. 6, Tl4N, RlW, Seward Meridian (Figure 1). On a smaller scale, the study area is bounded by Eagle River Loop Road and Meadow Creek on the south, the concealed Border Ranges fault on the west (Updike, 1986), and Chugach State Park and limits of residential construction on the north and east (Figure 2). 1Hydrologist, Alaska Department of Natural Resources, Division of Geological and Geophysical Surveys, Water Resources Section, P.O. Box 772116, Eagle River, Alaska 99577. 253 149° 35' R2 W R I W 149° 25' 61° 21' -+~....,l"""":m-....-,""1"""7l,..,.~·==="'!"":"'"-=~"'!"":,.,-ll'l'l'l'.,.,..---,...,.,.,.,..,.,.nnm-r~-~.,._ 61° 21' T15N TJ4N so• R2W CONTOUR INTERVAL 100 FEET (30.5 METERS) RIW \ Bose from Anchorage (8-7), Alaska, I= 63,360 scale, U.S.G.S. topographic map, 1960. \SCALE \ \ \ \ \ 0 6 I 0 2 3 12 18 24 30 Miles 2 Miles Kilometers SCALE -----. ---0 10 20 30 40 50 Kilometers Figure 1. Location map of the study area. 254 TJ5N TI4N '"%j R2W RIW 1-'· OQ ~ 1'1 (1) N 0 Cl.l Hlrt ~ rn p... ro'<: n rt Ill 1-'-1'1 X489 0 (!) ::s Ill .. 1-' Ill 0 ::s n P...Pl rt Ill 1-'- 1'1 0 (1) ::s Ill rn s :u O"d HI < ~ 1-'- (1) rt 1'1 ::r '< 1-' 1-'0 N o n ( Ill U'l rt U'l '< 1-'- 1-'· 0 (1) ::s 1-'!Jl p... ( Ill ::s (1) p... 1-' I-''< rn 1-'-. (1) 1-' p... p... Ill rt Ill 0 R2W RIW Bose from Anchoraga(B-7) SW, Alaska, 1•25,000 scale, U.S.G.S. topographic mop, 1979. HI a (1) LOCATION OF WELLS AND REPORTED YIELDS SCALE 1=12,500 Ill rn ~ 1'1 (!) p... <I Gallon per Minute % 0 1000 2000 Feet CD ... 0 <.06 Liters per Second No yield data available 11:: 0 z 0 300 600 Meters ( (1) 1-' 1-' rn 1-' 1-3 Gallons per Minute Deepened, blasted or Ill @ ;;) Q .06-.19 Liters per Second Cl: CONTOUR INTERVAL 20 METERS (88 FEET) replacement well ... >3 Gallons per Minute (~~,, Area of very low Approximate Mean A A' • I I Line of Section > .19 Liters per Second ~~j yield wells Declination, 1979 1-'• ::s (1) rn Approximate location of the Border Ranges Fault (Updike, 1986) GEOLOGIC SETTING The mountains immediately surrounding this area are composed of McHugh Complex (Jurassic /Cretaceous) rocks (Zenone and others, 1974). The lithologies of this complex include weakly metamorphosed silt- stone, conglomeratic sandstone, graywacke, arkose, greenstone, metachert, and argil- lite (Magoon and others, 1976). Strikes in this area generally trend between N60E and N90E and dips range from 45N to 90N (Updike, 1986). Three joint alignments have been noted nearby in this rock com- plex a few miles up Eagle River valley. Joint strikes and dips are: N34W, verti- cal; N80E, 80NW; and N38E, 46S, respect- ively (Updike, 1986). Flow systems encountered within the McHugh Complex lithologies have virtually no primary permeability and low to very low secondary permeability. As reported on drillers' logs, water enters wells through fractured zones in bedrock. Most wells are reported to have two or more zones of water inflow (Figure 3). Within the study area, the McHugh Complex is overlain by glacial alluvium, younger glacial moraine deposits, and colluvium (Zenone and others, 1974). According to drillers' well logs, these surficial deposits vary in thickness from 0.3 to 37 m (1 to 121 ft) in the study area. The thickest surficial deposits occur over the eastern portion of the study area (Section AA' and BB', Figures 2 and 3). Calculated from drillers' well logs, the median depth through these unconsolidated deposits to the McHugh Complex is 7.3 m (24ft). The slope of the study area is domi- nantly moderate (15-25 percent) with minor steeper areas (25-45 percent). The study area is dominantly hummocky terrain with some smooth to rolling uplands, mountain ridges and valleys, and river bluffs along Headow Creek. Generally, stable slope conditions exist, except steeper areas and bluffs are possibly unstable, and poten- tial for landslides or rockfalls exists (Zenone and others, 1974). 256 DATA COLLECTION After initial reports of domestic water well problems were received by DNR, the Water Resources Section of DGGS selected specific wells for obtaining water level measurements. Prior to measurement, homeowners were questioned for information regarding water quantity or quality problems, length of residence in the home, number of residents in the home, household water consumption pat- terns, and knowledge of neighbors having well problems. Water levels were measured in 59 domestic water wells to the nearest three thousandths of a meter {hundredth of a foot). A 152m (500 ft) electric water level indicator (sounder) was utilized for measuring water levels in most wells, and a chalked steel tape was used for the rest. Homeowners were asked not to use their water while measurements were being made and readings were taken only if the well pump had not run within 30 minutes. Water level measurements were checked by taking a second reading five minutes after the initial reading. In some instances, the electrode of the sounder encountered rust and condensed water on the inside of the well casing, making it difficult to discern when contact with the water surface was made. To obtain a more reliable reading, a steel tape that was chalked ·for the first 6 m (20 ft) was used, and contact with the water was generally obvious. In an attempt to maintain consistent measure- ments, sometimes the steel tape was used to find the general water level first, then the sounder would be used for the final measurement. Driller-reported static water levels were sometimes lacking and often ambiguous because the methods of measurement were variable and undocumented. Some drillers estimated the height of water column in the well, and others measured the depth from the top of the well casing. Famil- iarity with individual drillers' methods allowed adjustment of the data to a common A Southwest 1500 1300 1100 900 Feet 200 Feel Legend 1 Domestic Water Wells 1-71 Well Identifier "' 1985 Measured Static Water Level Driller-Reported Zones of Water Inflow E::!j Glacial Deposits and /or Colluvium D Me Hugh Complex 0 "' .. Reported Well Yields <I Gallon per Minute 0 <.06 Liters per Second ~ 1-3 Gallons per Minute .06-.19 Liters per Second • >3 Gallons per Minute >.19 Liters per Second • • Static water level ?}L <404 ft.(l24m.) elevation. a' East 300 200 100 Meters Horizontal Scale 0 500 750 0 100 200 Vertical Exaggeration • 1.5 Datum is Mean Sea Level A' • Northeast 500 400 1000 Feet 300 Meters Figure 3. Generalized geologic sections showing well depths, yields, 1985 water levels, zones of water inflow, and overburden thickness. 257 datum. Accuracy of reported yield data is questionable for many of these wells because of varying methods by drillers to estimate yield. In addition, driller measured static water levels may have been taken prior to total recovery of a newly drilled well. Any resulting inaccuracy is most likely to occur in low yield wells. Drillers 1 well logs were the primary source of domestic water well information. Domestic water well logs were obtained from the U.S. Geological Survey and DGGS Water Resources Section offices, Munici- pality of Anchorage Department of Health and Human Services, Alaska Division of Land and Water Management, drilling companies, and directly from the homeowners. Most well logs document the common domestic water well information including drilling company, driller, landowner, legal description of property, date of drilling, depth of well, lithologies encountered and their respective depths, estimated yields, casing size, casing depth, static water levels, and well pumping test results. RESULTS Water Level Changes Of 59 water levels measured, 50 had original static water levels that were reported on their well logs by the driller. Comparison of original static water levels to 1985 measurements indicate no definitive trend of either rising or falling levels. Water levels lower than those reported by the driller were found in 22 wells. The declines ranged from 0.1 to 58 m (0. 3 to 189 ft) with a median decline of 4.6 m (15.2 ft). Water levels in 29 wells were found to be higher than the static water level reported by the respective driller. Increases range from 0 to 39m (127 ft), with a median increase of 4.3 m (14.1 ft). Using 49 wells, the median change was an increase of water level of 0.2 m (0.7 ft). Water level measurements made during this survey varied from 1.4 to 89 m (4.5 to 292 ft) below land surface (bls). One well showed a water level decline of at least 87 m (286 ft) when compared to the original 258 reported water level. The absolute decline was undeterminable because the 152 m (500 ft) limit of the well sounder was reached before the water level could be measured (Well 2-46, Figure 3). Using data from 49 wells where original data were available, measured water levels in the hillside area of Eagle River indicate an average reading 0.3 m (1 ft) higher than levels reported by drillers. However, no distinctive area of either increasing or decreasing static water levels was found. Fluctuations exhibit a highly irregular areal distribu- tion and appear to be independent of total well dep1:h or distance drilled into bedrock. Well Depths, Yields, and Flow Systems Water levels were measured in 59 domestic water wells of which 56 were drilled into bedrock. Total well depths vary from 11 to 215 m (36 to 706 ft) bls with a median of 84 m (277 ft). Water yield of different inflow zones in a well is reported cumulatively. Reported total well yields vary from 0.01 to 0. 76 1/s (0 .1 to 12.0 gpm). The median reported yield of 53 wells is 0.13 1/s (2.0 gpm). Although the data exhibit a wide distribu- tion, they suggest that the lowest yielding wells typically exhibit the greatest changes in water levels {Figure 4). The data also suggest that reported well yield decreases with the increasing distance that a well is drilled into bedrock (Figure 5). Existing hydrologic data for the Chugiak-Eagle River area indicate median yield for bedrock wells is 0.13 1/s (2.0 gpm) and 0.63 1/s (10.0 gpm) for aquifers in unconsolidated sediments (Johnson, 1979). A continuous groundwater flow system was undefinable in this study area after examination of available data including cross sections and drillers 1 well logs. Moreover, existing flow systems in this area are isolated, variable, and randomly occurring. Adjacent wells with only several meters (tens of ft) separating them show drastically different potentio- metric surfaces (e.g., wells 1-46 and 1.9000 50 40 30 Ci) 20 a: w 1-w ! 10 ..J w > w • ..J -6 a: w 1- c( ;: -10 ~ w CJ -20 z c( J: 0 -30 • -40 • -50 o. 03 o. 30 3.00 30.00 REPORTED WELL YIELD (GPM) Figure 4. Relationship of water level change with increasing well yield (well yield axes are psuedo-logarithmic scale). 2-46, Figure 3). This strongly suggests a lack of fracture connectivity in rocks in this immediate area. Further evidence for poor hydraulic connectivity is the dis- tinctly separate elevations of groundwater inflow in adjacent wells (Figure 3) and the fact that a few dry wells have been drilled. Water level Changes in Shallow Wells Figure 6 shows a plot of the change of water levels with respect to the distance the well was drilled into bedrock. Very similar results were found when plotting change of water level versus total well depth. Based on 1985 water level measure- ments, all wells less than 43 m (141 ft) deep exhibited no decreases in static 259 water levels when compared with driller reported levels. Of nine wells less than 43 m (141 ft) deep, three wells are cased to bedrock, two wells are cased into bed- rock, three bedrock wells are not cased into bedrock, and the remaining well has insufficient casing data. The increases varied from 0 to 3.9 m (12.7 ft) with the median and average increase of 1.4 m (4.5 ft). Yields of these wells varied from 0.1 to 1.3 1/s (2.0 to 21.0 gpm) with a median of 0.4 1/s (6.0 gpm). These wells did not show decreases because their reported yield is relatively high. In addition, above average precipitation during July and August, 1985, could cause an increase of water levels in shallow wells as a response to groundwater recharge. DISTANCE WELL DRILLED INTO BEDROCK (METERS) 0 20 40 60 80 100 120 140 160 180 200 I I I I I I I I I I I I I I I I too. o-r'----''---'--.J..-----L _ _.__.L._____J. _ _.__.L.__.J. _ _.__,L____..L _ __.L__L____.,L_....L._L----L-..L__L,r 6.3000 • s to. o-..... Q. • r 0.6300 (/) 0 .... ..... • _, 0 • .... ....1 • c !!:! _, • • • !.!! > • • • • • > ....1 ....1 _, w .... • • _, ~ • w ~ 0 • • • • • c w 1-• • w a: 1- 0 a: Q. t. o-• 0 w • -0.0630 0.. a: w • a: • • • • • • • • • • • • 0.1-Tr~~~rrTT~,~~~rrT~r~~~rr~·~~~~rTTTTr~~~~rTTr~~~~rTT,~~~~rJ,-0.0063 0 100 200 300 400 500 600 700 DISTANCE WELL DRILLED INTO BEDROCK (FEET) Figure 5. Inverse relationship between well yield and the depth the well was drilled into bedrock (well yield axes are psuedo-logarithmic scale). Well Age Characteristics The oldest well measured was drilled in June 1971 and the newest in June 1985. The average well age at the time of the survey was 6.75 years. No relationship is apparent between well age and change in water levels. Older wells do not have greater changes in water levels than younger wells. The hypothesis that newer wells are depleting available water supplies and dewatering existing wells is not supported by the data. In addition, no direct correlation was evident when comparing the long term precipitation data with changes in water levels. 260 Water Usage Households reporting low water quantities available from their wells had varying water use habits and estimated well yields. Of 53 homeowners inter- viewed, 18 have wells with reported yields of 0. 06 1/ s ( 1. 0 gpm) or less (Figures 2-6). Some of these households reported above average water consumption that included daily showers and laundry, car washing, and lawn watering, and did not have sufficient amounts of water for their needs. AYe rage water consumption in the lower elevation of Eagle River is approximately 455 liters (120 gallons) per person per day (Munter, 1984). DISTANCE WELL DRILLED INTO BEDROCK (METERS) 0 20 40 60 80 100 1"20 140 160 180 200 50 • 40 • • • • • 30 ..... "" en r-a: Ill 20 w w 1-IL w .., • • ...J # • 10 ~ ...... w • ...J > • • w • \ • • w ...J ~ ....._ -• . ...-. --·-...._ 0 > • • • w I[ • ...J w • r-• • a: c( -10 w 5: -4 • 1- c( ~ -20 :t w •• • w Cl (!' z z c( -30 c( I: :I: () -12 0 • -40 • -16 -50 • -20 -60 0 100 200 300 400 500 BOO 700 DISTANCE WELL DRILLED INTO BEDROCK (FEET) Figure 6. Change in water level plotted against distance well drilled into bedrock (shallow wells had no decrease in water levels). Three areas of very low yield water wells [less than or equal to an estimated 0.06 1/s (1.0 gpm)] are outlined in Figure 2 . Two of these areas are also charac- terized by reported below average water usage. Below average water usage is characterized by daily drinking and cooking needs, and washing dishes and showering on a limited non-daily basis. Two areas of low water use include resi- dences in Broadwater Heights Subdivision and Seidler Subdivision No. 1. The third area, characterized by average water usage, is in the center of the southwest quarter of Sec. 6. Attempts at Increasing Well Yield Some wells that have failed have subsequently been deepened, blasted, acidized, redrilled, or abandoned. After deepening, two of three very low yield 261 wells in the area exhibited significant yield increases. Yield in the first well increased from 0. 04 to 0.19 1/s (0. 6 to 3.0 gpm) as reported by the driller after it was deepened from 84 to 137 m (275 to 448 ft). The second well was deepened from 92 to 171 m (300 to 560 ft) and the reported yield increased from 0.01 to 0.02 1/s (0.1 to 0.3 gpm). The third well was deepened from 53 to 122 m (175 to 400 ft) and the reported yield increase was negligible. One other low yield well in the Meadow Creek area drainage was acidized and another blasted. In each case, reported yield was insignificant. CONCLUSIONS The average change in static water levels in the study area in the past 15 years is insignificant. Changes in water levels are large at many wells and tend to decrease with increasing yield. These large fluctuations of water levels in this area apparently are linked to the low yield of the wells. It appears that their water levels are affected most of the time by on-site domestic use. In addition, after 43 m (141 ft) depths, water level changes are independent of total well depth or distance well was drilled into bedrock and exhibit irregular areal dis- tribution throughout the area. Isolated complaints of declining water levels in the study area are legitimate. All of the problems of water acquisition in this area are a result of drawing from a low permeability bedrock aquifer. Three areas of declining water levels are characterized by unusually low-yield wells and varying water-use habits. Also, it has been determined that declining water levels are not a function of well age. Evidence indicates that water level changes are most affected by water usage in low-yield wells. The problem of insufficient water supply has mostly been alleviated by deepening existing wells, drilling new or supplemental wells, or implementing more conservative water-use habits. No unified continuous flow system in the study area has been definable based on existing data. All wells less than 43 m (141 ft) deep showed relatively stable water levels over a period of years. These shallow wells also are relatively high yielding and derive water from unconsolidated sediments as well as bedrock. An inverse relationship exists between well yield and distance well drilled into bedrock, suggesting that the option of drilling deeper to obtain water may not be as successful as drilling a new well at a different location. Too often, the reliability and accur- acy of historic water yield and water level data are questionable. Some of the original driller-reported water level readings and estimated yields at time of drilling are considered unreliable because very low permeability results in long periods (days) of water level readjustment after the stresses of drilling or yield 262 testing. Periodic rerneasurement of levels and yields would further assist definition of the response of the aquifer, to resi- dential pumping stress versus performance problems of specific wells. Baseline data collected by this study should, in time, help resolve long term water level trends, affects on the groundwater system, its relationship to domestic water usage, a~ lead to more confidence in the foregoing correlations. Acknowledgements The author would like to thank the residents of the study area for their courtesy and cooperation during acquisi- tion of field data. Special thanks go to Jim Munter and Larry Dearborn of DGGS for their technical editing, review, and comments while compiling this manuscript. LITERATURE CITED Johnson, P., 1979, Hydrogeologic data for the Eagle River-Chugiak area, Alaska: U.S. Geological Survey, Water Re- sources Investigations 79-59, 17 p. Magoon, L.B., Adkison, W.L., and Egbert, R.M., 1976, Map showing geology, wildcat wells, Tertiary plant fossil localities, K-AR dates, and petroleum operations, Cook Inlet Area, Alaska: U.S. Geological Survey, Miscellaneous Investigations Series, Map I-1019 (Sheet 1 of 3). Munter, J.A., 1984, Ground-water occur- rence in Eagle River, Alaska: Alaska Division of Geological and Geophysical Surveys, Report of Investigation 84-21, 15 p. Updike, R.U., 1986, Division of Geological and Geophysical Survey Engineering Geology Section, oral communication. Zenone, C., Schmoll, H.R., and Dobrovolny E., 1974, Geology and ground water for land-use planning in the Eagle River- Chugiak area, Alaska: U.S. Geological Survey Open-File Report 74-57, 25 p. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 HYDROLOGIC MONITORING OF SUBSURFACE FLOW AND GROUNDWATER RECHARGE IN A MOUNTAIN WATERSHED Michael E. Campana and Richard L. Boonel ABSTRACT: The mountainous regions of the Great Bas in are major water sources for the drier, lower-elevation regions. These mountain watersheds, which receive much of their precipitation as snow, serve not only as surface water catchments but also as source areas for much of the ground- water recharge received by the valley-fill aquifers in the intermontane basins. An understanding of the hydrologic processes in the mountain watersheds, particularly those occurring during winter, is requi- site before hydrogeologists can understand and quantify recharge to the valley-fill groundwater reservoirs. A small-scale study conducted in a Sierra Nevada water- shed just east of Lake Tahoe utilized four major groups of instruments to delineate infiltration, recharge and subsurface flow contributions to streamflow. Foremost among the groups was an automated tensi- ometer-transducer system. Tensiometers were linked to a central scanning valve via hydraulic tubing, which was buried to protect against freezing. The investiga- tion indicated that deep percolation/ groundwater recharge occurred beneath the snowpack during the winter months but could not determine hew much of it discharged locally and how much flowed off-site, perhaps as p<?crt of a deeper system. The experience gained during the study has greatly aided continuing studies in the same area. (KEY TERMS: groundwater recharge; moun- tain watershed hydrogeology; snowpack; infiltration.) INTRODUCTION The mountainous regions of the Great Basin are important sources of water for the dry, lower-elevation regions. The mountain watersheds, which receive sub- stantially more precipitation than their low elevation counterparts, serve as catchments for the intermontane basins and source areas for the few large perennial streams in the Great Basin. However, their importance with respect to the groundwater resources of the region is often overlooked. Just as they serve as source areas for surface water, so do they function as catchments for much of the groundwater recharge received by the basin-fill alluvial aquifers. And, since much of the precipitation received by the mountainous regions is in the form of snow, an understanding of the interrela- tionships among the snowpack, snowmelt and subsurface flow in mountain watersheds is requisite before hydrogeologists can hope to quantify mountain-block recharge on a large scale. In addition, a knowledge of the different recharge and subsurface flow pathways is necessary to understand sub- surface and surface water quality in moun- tain watersheds. In the Eastern Sierra Nevada, where the study was conducted, an important aspect with regard to subsurface flow paths concerns the disposition of acid rain derived from the large metropol- itan areas to the west. The Eastern Sierra Nevada also serve as a major catchment for the western Great Basin !Respectively, Water Resources Center, Desert Research Institute, P.O. Box 60220, Reno, NV 89506 and Department of Geological Sciences, Mackay School of Mines, University of Nevada, Reno, NV 89557; and U.S. Environmental Protection Agency, 345 Courtland St. N.E., Atlanta, GA 30365. 263 in Nevada. We believe that this study represents the first attempt to identify the year-round subsurface flow regime in an Eastern Sierra Nevada mountain water- shed and as such, is important in guiding future efforts along similar lines. A hillslope in a small watershed was instrumented with a variety of instrument groups. The hydrologic regime of the hillslope was monitored from summer through winter with emphasis on delin- eating the hydrologic processes occurring during winter. Particular attention was paid to the following: 1) identifying cold-season groundwater recharge; 2) determining contributions of subsurface flow to streamflow during the winter; and 3) ga1n1ng a qualititative understanding of the subsurface flow processes operating at the site. Acquisition of experience in the 1nstallation, use and maintenance of the instrumentation, especially during the winter, was also a major objective of the study. DESCRIPTION OF STUDY AREA The study area is located on a hill- slope having a northeast aspect along Clear Creek in Carson City, Nevada (Figures 1 and 2). The site is just east of Lake Tahoe approximately 9.5 km west of downtown Carson City, Nevada, in the Toiyabe National Forest. The experimental site (Figure 2) lies at an elevation of about 2000 m in the north-trending Carson Range of the Sierra Nevada and is under- lain by Cretaceous granodiorite. Ande- s1t1c and dacitic dikes and plugs are abundant throughout the Carson Range and evidence of this is seen in a contact zone in the northwest portion of the site. The average annual precipitation is about 7 5 cm/y; approximately 80% of this total occurs as snow. More than 50% of the total precipitation falls during January, February and March. Summer maxi- mum temperatures can be as high as 37° C; minimums can be as low as -9° C. During the winter, high temperatures can reach 2l°C; lows can approach -35°C. Soils in the area are part of the Toiyabe-Rock outcrop complex and the Corbett-Toiyabe association. The Toiyabe soil is shallow and well-drained, formed in residuum from the granodiorite. The 264 soil has a surface layer of gray, stony loamy coarse sand approxim!ltely 10 em thick. Below this is a layer of gray loamy coarse sand, gravelly coarse sand, and sand approximately 18 em thick. Weathered granodiorit ic bedrock occurs at a depth of 28 em. The Corbett soil is moderately deep and well-drained, formed in colluvium derived from weathered gran- odioritic bedrock. The surface layer of dark grayish-brown stony loamy coarse sand is about 20 em thick. The next layer consists of about 80 em of light gray gravelly and cobbly loamy coarse sand. Weathered granodioritic bedrock is found at a depth of 100 em. Location of I IJ---t+Lake Tahoe I and vicinity., I I I 0 6 10 miles ...._ _ _,___.....J 0 16.1 km CARSON CITY STUDY AREA Figure 1. Location of Study Area. Native vegetation of the area is characterized by Jeffrey pine (Pinus jeffreyi), lodgepole pine (~ murrayana), white fir (Abies concolor) and willow (Salix sp.) with an understory of antelope bitterbrush (Purshia tridentata), sagebrush (Artemesia tridentata), green manzanita (Arctostaphylos patula), tobacco brush (Ceanothus velutinus), horsetail (Equisetum arvense) and a variety of grasses. F1 ' \ IU 0 0 0 .... CQ (I) (I) • T2 eT1 CONTOUR INTERVAL= 12 m 0 15 30m I I I , -,. \ \" \ "" , ___ ..,. P6 • ,_,-,. , / SPRING ZONE ~ SPRING G = PRECIP. GAGE D = SCANNING VALVE F = FLUME X = SOIL TEMP. STATION •P = PIEZOMETER NEST = TENSIOMETER PLOT --- Figure 2. Map of Experimental Site. Within the study area, a spring zone discharges into Clear Creek, a steeply incised perennial stream which has its headwaters at Snow Valley Peak (2 ,827 m). Clear Creek flows eastward into Eagle Valley where it enters the Carson River. The Clear Creek watershed has an area of 39 kln 2 and delivers an average annual water yield of 4.8 ·10-3 km 3 into Eagle Valley. INSTRUMENTATION Four major groups of instruments were used at the site. The installation, oper- ation and performance of the equipment, particularly during the winter, were a major objective of the study in addition to the purely hydrologic considerations. Since this study represented an early attempt at year-round surface/ subsurface hydrologic monitoring in the Sierra Nevada, future studies (ongoing at the present time) would benefit from the m1s- takes and experience of the authors. 265 Piezometer nests provided data on vertical hydraulic gradients in the saturated zone and changes in groundwater levels. Flumes (H-type) were placed on the perennial stream to determine streamflow gains/ losses through the experimental site. Vertically-distributed psychrometers in the soil provided temperature gradients. The last major instrument group con- sisted of 5 tensiometer nests linked to a central 24-port hydraulic scanning valve by 1.6 mm OD flexible vinyl tubing (Figures 3 and 4). Trenches at least 45 em deep were dug for tubing installation to reduce the possibility of freezing during the winter months. A total of 22 tensiometers and two reference reservoirs were used in the soil moisture monitoring system. The outlet of the scanning valve was connected to a differential pressure transducer, whose output was connected to a strip chart recorder. Lead-acid batteries were used for power. The scanning valve-transducer assembly was installed 1n a wooden box, insulated and buried about 50 em deep. The batteries and recorder were housed in an enclosure placed above the average snow level. In this manner, the entire tensiometric system was completely automated. The system used is similar to ones described by Anderson and Burt (1977) and Williams (1978); more detailed information can be found in Boone (1983). Monitoring was initiated in July 1982 and extended through March 1983. POROUS CERAMIC CUP CAP CLAMP TO FLUID SCANNING SWITCH EPOXY RESIN GLUE CLEAR, RIGID PLASTIC TUBE DIAGONAL CUT END Figure 3. Tensiometer Construction and Arrangement (after Williams, 1978). RESULTS AND DISCUSSION TemperatUPe Data Temperature data, collected at two different sites, indicated gradual decreases in soil temperature throughout the study period. At no time, however, except for a few days in November, did 266 soil temperatures fall below freezing. Frozen soil was not observed at the snow/soil interface after the first snow· fall in early November. The insulatin~ effect of the snow (which averaged over one meter deep at the end of the study) was the major reason for the unfrozen soil. A slush layer was observed at the snow/soil interface, but it had meta· morphosed to a granular, icy layer by mid-December. At station 1, the tempera- ture gradient decreased from 3.9°C/m in October to 1.6°C/m in March; at the other station, the decrease was from 2.6°C/m to 2.3°C/m. CHART RECORDER PRESSURE TRANSDUCER REFERENCE FREE WATER LEVEL TENSIOMETER PVC CAP Figure 4. Automatic Tensiometer -Transducer System (after Williams, 1978). The limited temperature with the tensiometric data, the soil heat flux may have data, coupled suggest that been suffici- ent to melt the base of the snowpack 1n some areas and permit infiltration into the soil. The icy layer referred to earlier may have been discontinuous or not entirely impermeable. Tensiometric Data Tensiometric data from plots 3 and 4, which are representative of the entire site, and precipitation data (expressed as liquid water equivalent) are shown in Figures 5 and 6. Tensiometry established that from July to about early November, hydraulic gradients in the soil alternated between upward (response to evapotranspir- ation) and downward (response to precipi- tation/infiltration events). During this tme, the tensiometer-transducer system performed well, although the system occa- sionally had to be shut down to purge it of air. This period essentially served as a "shakedown" period to ensure that the system worked well. In November, matric potential at all soil depths decreased slightly from the near-zero or near-saturated conditions which resulted from the October storms. From November 1982 through February 1983 matric potential values remained fairly constant between -2.0 to -8.0 cent ibars with a downward hydraulic gradient. How- ever, a slight increase in matric poten- tial for most depths beginning around November 19th to 20th did occur due to the precipitation events of November 18th and 19th (6. 9 em total) and November 21st and 22nd (at least 1.3 em total). Additional pulses of deep percolation may have occurred as a result of the storm (rain- on-snow) events of December 20th to the 21st (2. 9 em total) and January 24th to the 26th (1.6 em total), but the data from plots 3 and 4 do not indicate this; these pulses are more evident from the piezomet- ric data. Conditions at plot 4 were generally similar to those at plot 3 throughout most of the study period. The maintenance of a downward hydrau- lic gradient in the soil in the vicinity of tensiometer plots 3 and 4 from November through February indicates that deep per- colation (potential groundwater recharge) occurred through this period. This down- ward movement of water occurred under near steady-state conditions which were occa- 267 sionally disrupted by additional pulses of moisture movement due to precipitation events. In addition, soil moisture move- ment occurred at relatively high values of matric potential, indicating flow under "slightly" unsaturated conditions. It should be noted that the term "slightly" is subjective, since the lack of data did not permit the determination of the degree of saturation of the soil during this period of downward flux. Although mois- ture characteristic curves for the site soils were unavailable, comparison with those for soils of similar texture indi- cated that degrees of saturation were around 0.90 and that the soil hydraulic conductivity represented nearly-saturated conditions. Piezometric Data Unlike the tensiometric data, hydrau- lic head data from the piezometer nests were temporally discrete and obtained manually. This proved difficult at times during the winter because of deep snow conditions. Unfortunately, piezometric data in close proximity to the main tensi- ometer plots (3 and 4) were unavailable. Piezometer nest 1, located on the banks of Clear Creek, indicated horizontal flow throughout the study period; piezometer nest 6 indicated a slight upward gradient throughout the study. For illustrative purposes, data from nests 2, 3 and 4 are most useful and are shown in Figure 7. Piezometer nest 2 showed an upward gradi- ent, while nests 3 and 4 indicated down- ward gradients. In each nest, the datum for hydraulic head 1s the bottom of the deepest piezometer. Each of the nests in Figure 7 shows a fairly quick response to storm events during the late summer and fall. Piezom- eter nest 4 indicates the most change, probably because of its close proximity and direct hydraulic connection to Clear Creek. Increases in water levels in nest 4 probably reflect a rise in the water table within sediments of the stream channel, similar to a buildup of the zone of saturation along a stream bank as des- cribed by Chorley (1978). Piezometer nest 4 is located in that part of the channel occupied by Clear Creek during high dis- charge. Nest 4 also shows a rapid N en co -.s::. ...... E (,) -z Q .... < !:: e: 0 w a: Q. -U) ... al :2 ... c:: Q) (,) -..J ~ 1-z w 1- 0 Q. 2 a: .... < ::iE 0.5 2.20 1 2 2 1.0~------------------------------------------------------------------------------------------ -40 -30 -20 -10 0 30 1982 . : ~~-. ,.. ,_.,-.... ~ '/,~,Y.'· ..... J''. I 'J\ I • • ,, ... \ ., . ., ·, 60 90 ' ' \ ----=-· . . . . . ...... . 120 150 TIME (DAYS) 0.6 m depth 0.9 m depth 1.2 m depth 1.5 m depth ·~---------------­.................. ----------:~,, ,,---· ·-·-·-·' . . ..... '.' 180 210 Figure 5. Site Precipitation and Tensiometric Data for Plot 3. .. .. 1983 -0.0 .c ..... E 0 ~ 1 r I ., I II I II I -z 0 t= 0.5 -< r- 0:: 0 w a: a. '~ '~ If 2.20 1.32 2.87 1.0 --60 rn / 0.6 m depth N .. en as a:> :2 -50 -.;" 1.2 m depth c Cl) "' "' 0 -40 "' -' ...J I ' ~ ... ,, ' r--30 ~ ........ " ' z \ ' w \ ' r--20 \ ' 0 \ ' a. ' Q -10 --- a: r-< 0 -------------- ::E JUL AUG SEPT OCT NOV DEC JAN FEB- 30 60 90 120 150 180 210 1982 TIME (DAYS) 1983 Figure 6. Site Precipitation and Tensiometric Data for Plot 4. ..... E ..... Q -< w J: 0 ::::i ~ -< a: Q > J: e ..... Q -< w J: 2 ...I ~ -< a: Q > J: ..... E ..... Q -< w J: 0 ::::i ~ -< a: Q > J: ·"· 1.8 ,.... .......... .,.,.,..·-·'" '--·-·-·......._. . ......_..-·-·-·--.--·- /. ·-. ..., J '-------.,.-----._-- -.......... \ ------·-....... '-/ --- 1.5 SEPT ---f.-OCT --t--NOV --+-DEC -+---JAN-+-FEB --i 10 20 30 40 50 60 70 80 90 100110120 130140150 160170180 190 1982 Time {days) 1983 1.8,--------------------------------------------------------, 1.2 A ,.,..----r--J-" "-......._. ......... ..,... _____ _ J I· ,........... . \ -·--·--·-·-. -· --__./ ,... ....... ..,; . . .......... / ·-·-· / / _ ......... 0.9 SEPT-+--OCT -+--NOV ~ DEC --t-JAN ---t--FEB--; 0.6~~--~~--~~--~-+-~-+--~-+--~-+--~~--~~--~ ....... 10 20 30 40 50 60 70 80 90 100 110 120130140 150 160 170 180 190 1982 Time {days) 1983 1.2 0.9 SEPT--+--OCT-+--NOV-+--DEC ---+--JAN -+--FEB ---t 10 20 30 40 ao eo 10 ao 110 100 110 120 1so 140 1ao 1eo 110 180 190 11182 Time {daya) 1983 / 0.8 m depth 1.2 m depth ...... ....... ...... 1.8 m depth ........ ........ ....... NEST #2 / 0.8 m depth 1.2 m depth ...... ...... 1.8 m depth ....... ...... . ...... ....... NEST #3 0.8 m depth / 1.2 m depth ....... ...... ....... ...... 1.5 m depth ............... ...... NEST #4 Figure 7. Hydraulic Head Data for Piezometer Nests 2, 3 and 4. 270 response to the aforementioned ra1n-on- snow events of late December, 1982 and late January, 1983. Nests 3 and 6 also showed overall water level rises from the beginning to the end of the study period. This indicates a rise in the saturated zone of the spring zone, possibly due to increased recharge associated with an , areal shift in recharge/discharge areas. It is also possible that the increasing 'heads could have been the result of later- ' al unsaturated flow from upslope. During piezometer installation, a confining clay layer approximately 5 to 10 ern thick was found about 2.4 m below the surface near piezometer nest 2. Ground- water existed above and below this layer. ~ similar clay layer was found at approx- imately 0. 75 m below the surface near the top of the spring zone (approximately 15 m southeast of tensiometer plot 3). Assuming horizontal continuity, this clay layer could extend below the tensiometer plots at a depth of approximately 5. 5 to 7.6 m and may play a major role in ground- water transmission in this area. This clay layer results in a groundwater system s~ilar to an unconfined aquifer overlying a confined aquifer and was probably very significant in the development of this spring zone. The SJ:-ring zone is a slump- type structure, the top of the zone being the edge or mainscarp of the slump. These types of structures have been observed in other spring areas in the Clear Creek ntershed. A perched water table atop the clay layer, coupled with the steep inc is ion of Clear Creek, may have resulted in slope oversteepening and slumping. It is possible that groundwater flows through the unconfined system below the tensiometer plots and contributes to streamflow. Water detected in the ten- foot piezometer of nest 8 around early February seems to confirm such 'a conclu- sion. The water table had risen approxi- mately 0.08 m above the bottom of this piezometer by early March. The rapid response in piezometer water levels to storms could be a combination of recharge directly onto the spring zone and contri- bution of groundwater from this flow system. A medium for direct hydraulic connection between recharge and this groundwater system could be the thinner soils found on topographic spurs farther upslope. 271 Surface Water Data Streamflow data were collected from two H-type flumes installed on Clear Creek about 160 m apart. Errors in discharge determination resulted from sedimentation in the flume, which changed the approach velocity and altered the stage-discharge relationship. Other sources of error were present, although sedimentation was undoubtedly the dominant one. The flume errors could account for the discharge differences between the two flumes, so the streamflow data were inconclusive and of little use. RECHARGE IMPLICATIONS The tensiometric data from plots 3 and 4 indicated that deep percolation (potential groundwater recharge) occurred beneath the snowpack during the winter. The hydraulic gradients in the soil at these plots were directed downward during the period from approximately November to March. Furthermore, with the exception of a few pulses of percolation from precipi- tation events, the downward gradient at each site was approximately constant, as was the rnatric potential at a given depth. These data indicate that deep percolation was occurring at this time, as was ground- water recharge. No piezometric data or deep (greater than a depth of 1.5 m) matric potential data are available from the immediate vicinity of plots 3 and 4, so recharge to the groundwater reservoir was not actually observed. However, the soil hydraulic data suggest recharge occurred, because soil-moisture movement was occurring under approximate steady- state conditions, which indicates that the soil was serving primarily as a "trans- mission zone". Frozen soil was not observed during the study period, so water movement through the soil was possible. Although the slush layer at the snow/ground inter- face gradually metamorphosed to a gran- ular, icy layer by mid-December, a trans- formation that would seem to preclude infiltration from the snowpack, dye studies suggest that such ice layers are not entirely impermeable but possess vari- ous permeabilities which force water to take a more circuitous route to the soil (Langham, cited in Nale and Gray, 1981, p. 398). Fracturing in the ice layers would also permit the movement of water into the soil. CONCLUSIONS The major conclusion of the study is that deep percolation, and thus, ground- water recharge, occurred beneath the snow- pack during the winter months; further- more, the rain-on-snow events provided additional pulses of infiltration superim- posed on the steady downward deep percola- tion. Because this observation is mostly qualitative, more investigation is needed into winter-season recharge. This phenom- enon is often overlooked in hydrologic studies/budgets in the Sierra Nevada, the Great Basin and perhaps other regions as well. The stuay could not determine how much recharge eventually discharged into Clear Creek and how much flowed off-site, perhaps as part of a deeper subsurface flow system. The automatic tensiometer-transducer system exhibited rapid response, and al- though some problems, such as tensiometer tailures, did occur, future studies can build on our experiences. A major con- straint on the use of the system is the trenching required in order to bury the hydraulic tubing below the frost line. Trenching was labor-intensive and time- consum1ng since much of it had to be done manually. As suggested by others (McKim, et al., 1976; 1980), this could possibly be circumvented by the use of an ethylene glycol (antifreeze) water mixture, although more research is needed to assess the effects, if any, of such a mixture on the soil mat ric potentials. The use of 1.6 mm OD vinyl tubing placed limits on the areal extent of the system because of the frictional head losses. The authors 1 experience leads them to recommend the use of 3.~l nun OD nylon tubing in similar studies; cost constraints prevented its use in the present study. Studies along the lines reported here are continuing, because this research spanned less than one year and sampled a winter that pro- duced one of the heaviest snowfalls on record. The instrumentation at the site has been expanded and monitoring continues. 272 ACKNOWLEDGEMENTS Many people assisted 1n this study and deserve thanks; however, without the help of C .M. Skau none of this would have been possible. Karla Cosens and Christim Stetter provided invaluable, excellent- quality assistance throughout the prepara- tion of this manuscript. The Toiyabe National Forest of the U.S. Department of Agriculture allowed access to its land for the study; the Carson Ranger District of the U.S. Forest Service was most helpful. Financial support was provided by the Office of Water Research and Technology, U.S. Department of the Interior (Grant ifl4-34-0001-1244); the U.S. Geological Survey 1 s Great Bas in Regional Aquifer Systems Analysis (RASA) Program; and the Desert Research Institute, University of Nevada System. REFERENCES Anderson, M.G. and T.P. Burt, 1977. Auto- matic Monitoring of Soil Moisture Con- ditions in a Hillslope Spur and Hollow. Jour. of Hydrology, 33:27-36. Boone, R.L., 1983. Groundwater Recharge and Subsurface Flow Processes on a Hillslope in the Clear Creek Watershed, Eastern Sierra Nevada. Unpublished M.S. Thesis, University of Nevada-Reno, 164 pp. Chorley, R.J., 1978. The Hillslope Hydro- logical Cycle. In: Hills lope Hydrol- ogy, M.J. Kirkby (ed.). John Wiley & Sons, pp. 1-42. Male, D.H. and D.M. Gray, 1981. Snowcover Ablation and Runoff, In: Handbook of Snow, D.M. Gray and D.H. Male, (eds.). Pergamon Press, pp. 360-436. McKim, H.L., R.L. Berg, R.W. McGraw, R.T. Atkins and J. Ingersoll, 1976. Devel- opment of a Remote-Reading Tensiometer/ Transducer System for Use 1n Sub- freezing Temperatures. In: Proceedings of the Second Conference on Soil Water Problems in Cold Regions. American Geophysical Union, Washington, D.C., pp. 31-45. McKim, H.L., J.E. Walsh and D.N. Arion, 1980. Review of Techniques for Heasuring Soil Moisture in situ. U.S. Army Corps of Engineers Special Report 80-31, Cold Regions Research and Engin- eering Laboratory, Hanover, NH, 17 pp. Williams, T .H .L., 19 78. An Automatic Scanning and Recording Tensiometer System. Jour. of Hydrology, 39:175-183. 273 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 DISCHARGE u;mER AN ICE COVER Henry S. Santeford and George R. Alger 1 ABSTRACT: In 1983, the authors proposP.d a theorE:tically bused mod,:l, augmented with a conVI"ntional resistance equation (~anning 1 S equation), to relat •2 th e• stage to discharge on ice covered riv•:=rs. A cooperativ(~ field studv was established with the U.S. Geological Survey and U.S. Army, Cold Regions Research and Engine~ring Labor.;tory to test the proposed model. HH~ results of this first wint2r season indicate that the average d i ffer .:>i1Ce between thP. predicted and measured discharge values was within the 1 i!nit of error of th!'? discharge measurem~nt (± 3 pero?nt). The model v1as also applied to unpublished historic records o-f' the USGS for the sarne s i tA.s. .1\ l though there are traditionally only 2 to 4 data points rer winter season, the results were the same as those reported for the more extensive studv. (KEY T~WijS: ·ice:> cover; stage-dischargo2 relationships.) I NTRODIJCTI ON Many hydrologists vi@W the collection af discharge data as a routine type of operation. For th~~ open water condition, a stage vs. discharge relationships is developed for the chosen gage location. Then a simple monit•Jring of stage can be used to develop a continuous record of di schct rge. Through periodic checks of actual discharge measurements and appropriate quality control, reliable ~cords can be obtained. Howev~r. wh~n an ice cover forms on an open chann~l , the stage vs. discharge relationship which had r.xist~d for th .; open water condition is no longer valid. The presence of the ice cov~r altPrS t he size and shape of the rigid boundary in contact with the moving wut '~r. This, i:l turn, causes the energv slope to sV~epPn. In addition the icP cov ·::r which is generally assumed to be floatin~~' h«s a buoyant displac.~~n~'nt causin9 a further 111creas ~~ it1 stage. These two factors incre~s~d resistance and buoyJnt displacement -- oft~~n Wolrk togeth2r t'J produce a s ~ag~ associati:d with ice c.JVI?.rt::d conditions which is l a rg er than the stJgP which would have existed at the sam2 disch~rge for the open vJatar condition. This increas •~ in stag2 is often t '::rm ed 11 iu· indu c Ad b.:Jckvnt·~r 11 or m e r~:l'/ 11 back,)'·' i;;:>l~11 • Tfw ·,u:~'· t :1 ~ y;!a rs n numb.::r ')T tech niqut-:':; 'J,l l'.? b ~~t>n d :;v ~~·loped in an atte~~t to r e late t~e ooen water and ice covered sta~~ vs. discharg~ relationships. All of these ~arlv techniques suffer fr'crr th2 sa.::J<:~ g,:r:P.ra 1 problems. First, none of the older techniqu e s considers the buoyant displace1nent of the ice. Consequently, the buoyant displacement of the ice (distance from equival~nt free water surface to underside of the ice cover) is not generJlly measured at th!? gagr.: location. Secondly, none of the early techniques can be shown to be theoretically sound, i.e. they are merely based on regression analysis or curve fitting. And thirdly, all of the 1 Both, Department of Civil Engineering, Michigan Technological University, Houghton, :1ichigan 49931. 275 early techni~ues assu~e that th~ same functional r~lationship exist from the timl: of t:w first Jppea1Aance oF ice at fr:~eze-up through the breakup period. A siinple anal~rsis of the flow regime clearly indicates that at freeze-up and 1gain at breakup, the flow must be both unsteady and non-uniform as the transition is m(l.de bet1,'/een the open ·.~ilter and ice covered conditions (Sant~ford and 1\lger, 1983). However, onr.e th2 trJnsition is complete, a st1blo2 neriod exists during 'Hhich there is a fixed relationship between the open i'ln.tor and ice covered condi ti IJns. The authors have ter;n this 11 the p·~ri od of st1ble ic'= co.1trol 11 (S0.nteford and Alger, 1984a, 1984~. 1984d). \l:=tn/ 0 f the i"at:-.r surv•=v 'lgenci es in the northern hemi s nhere re 1 y on the i:1terpretative discharge technique to 2stirni1te '.AJi nV~r di scharqe records. Actual discharge measure~ents are made ~t routi~e time intArvals, often once per month. Th>?. a~tual d·ischarge measurement is mad2 at a convenient location, not nocessarily at the gage cross-section. Consequently, the inhrmation r.ecessary to Cl1rrect for the buoy.'lnt displac~mr:nt of th~ ice at the ga9e is not always availJble from -t:h2 field data. The winter discharge record is then constl~ucted b,'/ sketching a hydrograph bet1~2en the plotted disch:1rge values obtai ned at thA time of Vv~ routine (monthly) measu1~e~w~nts. The hv:Jl'ologist is aid~-.cJ in his sketching bv the stRqe record, local air temperatur·~ records, precipitation data, and discharge :n~asure'Tlents from nearbv strGams. The procedure is very time consuming .1nd highly dep2nd•~nt upon the skills of the hydrologist. In 1933, the authors proposed 1n ani! l.!tti cal t•?chni que i"hi ch waul d pro vi de d direct conversion from winter stage records tn discharge. Th~ modA.l was based on a th·=oretic.1l analysis of open channel flow auq;n.;nted bv 3 conv~ntional t'·::sistctnce -'"quation (:1anning•s .::qua ti on). Us hg unpub l i :;hed r2cords Jf the U.S. Genlogical Survey encouraaing r~sults w~r= obt~ined for data from sev.~ral seL~ctd >trea;ns in :'iichigan•s Upper Peninsula (Santefard a~d Alger, 1934b; Alger and Santehrd, 1984). 93s2d on th~se pr~liminary findings a 276 detailed study was initiated with the aid of the U.S. Geological Survey and the U.S. 1.\rmy Cold Regions Research and Engineering Laboratory. Three ~aging sites 'tier~ chost:n f8r use in the initial study (Fi']ure 1). At each site the frequ2ncy of discharg~ measurement was i ncl~eas,:>d from once pP.r month to as many as 3 per week. Aarm weather conrli ti ons :1ampered the data collection efforts at the Red C~dar site. Thus, only the t\'IO northern stations ·.~!ill be discussed in th~ summary (see Santeford, .A.l ger and Stark, 1986). Sturgeon River near Sidnaw (04040500) Red Cedar River near Williamston (04111379) Figure 1. Map of Michigan showing location of study areas. THE PROPOSED ~ODEL Earlier works of the authors have shown th:1t for the sam:~ discharge, the ratio between the ooen water mean hydraulic depth (D ) and the ice cov~red meon h;tdraulic dep1h (D.) is a constant t~rmed the ice adjustmrni factor (IAF). D /D. = constant = IAF 0 1 (1) The magnitude of the IAF will be site specific depending upon the location of the gaging site related to the control sections and the geometry of the gaging section. Using a cross-section obtained at the gage location and the open water rating curve it is a simple matter to construct functi ona 1 re 1 ati onships between stage, discharge, area, top width, and mean hydraulic depth. These relationships can be presented either as mathematic functions or in tabular form as shown in Table 1 for the Sturgeon River near Nahma ,Junction. During the ice covered period the stage record is influenced by the buoyant displacement of the ice as well as th~ increase in energy s 1 ope resu 1 t i ng from the added resistance offered by the ice cover. Thus, before the stage record can be used to predict discharge, it must first be corrected for buoyant displacement. Since most natural rivers do not have regular prismatic cross-sections and further s i nee the dens i tv of the ice cover is not a-constant. the term 11 float qepth 11 will be introduced to account for the buoyant displacement. In the development of equation 1, a key element was the under ice area available to the flnw. For the given under ice area there is a stage which would exist if the buoyant displacement of tile ice was not present. That va 1 ue of stage is termed GH.. If the actual recorded stage is termJd GHw, it fa 11 ows that GH -GH. = float depth. w 1 S i nee GH. is the more c6nvenient relationship as desired term, it is to express th~ GH. = GH -float depth 1 w ( 2) Float depth can be defined as the correction which when applied to the actual stage reading yields a corr~cted stage with the appropriate area available to the flow. When an actual discharge measurement is made on an ice covered stream, the under ice Flow area ry_nd the actual stage Table 1. Ice correction data table for Sturgeon River near Nahma Junction. Mean Hydraulic Stage Discharge Area Top Width Depth (ft) (ft3/s) ( ft 2 ) (ft) (ft) 4.10 86.0 99.1 60.3 1.64 4.09 84.4 98.4 60.2 1.63 4.08 82.8 97.8 60.2 1.63 4.07 81.2 97.2 60.2 1.62 4.06 79.6 96.6 60.1 1. 61 4.05 78.0 96.0 60.1 1.60 4.04 76.4 95.4 60.1 1.59 4.03 74.8 94.8 60.0 1.58 4.02 73.2 94.2 60.0 1.57 4.01 71.6 93.6 59.9 1.56 4.00 70.0 93.0 59.9 1.55 277 (GH ) are both known quantities. E t vJ • 1 t . h . I h . n r~r1ng a rP. a 1ons 1p suc1 as s own 111 Table 1 with the under ice area, one can readilv obtain SH. and thus the float depth. The floa~ d2pth can also be obtained throuah field measurements of the ice cover. -The procedure consists of measuring the distance from the free \vil ter surface to the under side of t.he ice cover at numerous location along the gage cross-section. From thes2 data one can compute the cross-sectional area occupied by the ice. For the given stage the tota 1 wetted area is known and the area available to the flow can be found as the residual. If the float depth is a known quantity or can be estimated for periods between measurements, then the adjusted stage record can be used to give a direct indication of discharge similar to what is done for the open water condi ti rm. However, it is often more convP.nient to use only one rating curve for summer and winter condition and apply a correction for the ice covered condition. The ice adjustment factor (IAF) is that correction factor. The IAF is also readily obtained from field data. Here a simple illustrative example is used to de~onstrate the procedure. On February 11, 1935, the USGS measured discharge at the Sturgeon River near Nahma Junction and reported the following data: Discharge = 83 cfs Stag-2 (GH, ) = 4.92 ft Float Dep!h = 0.89 ft Therefore: GHi = GHw -FD = 4.92 -0.89 = 4.03 ft From Table 1 at a stage of 4.03 ft, the mean hydraulic depth is 1.58 ft. Since this value is for an ice covered condition, it represents D.. Returning to Table 1 with discharge 1equal to the measured value of 83 cfs, the mean hydraulic depth that waul d have existed for open water condition, D , is found to be 1.63 ft. 0 Then: IAF = 00 /Di = 1.63/1.58 = 1.03 278 If conditions are stable, tf-Je IAF should be a constant for an_v given gage location. Furthemor2, the adjusted stage vs. dischurge relationship for the ice cov2red condition should correspond to the open water relationship. Because 0f ti"Je genera 1 requirements used in selecting favorable gagin9 sites, most gages which exhibit an ice effr:ct on the stage vs. discharge rAlationship also experience a period of sta~le ice control. However, there are exceptions. As mentioned previously, freeze-up and breakup will produce unstable conditions at all gage location. In addition, the nature of the control section and the resulting water surface profile can also result in unstable conditions. For example: (a) a gage located along an M-2 profile upstream of a rapids or riffle which remained free of ice; and (b) a gage located in an ice covered cha~nel a short distance downstream from the outlet of a 1 ake. In both cases there ~.,i 11 be a family of rating curv~s depending upon the buoyant displacement of the ice. Space limitation do not allow for il full discussion of the various water surface profile which can exist for ice covereu condition. .L\ full discussion of ice effects on water surface profile and thus stable vs. unstable condition is presented in: Stark, 1986; Santeford, Alger and Stark, 1986; and S3nteford, 1986. RESULTS As mentioned previously, during thf~ 1984/85 winter season the frequency of winter discharge measurement was increased. The actual num~er of measurements varied with each of the test sites. The measured stage vs. discharge data are shown in Figure 2 for the Nahma Junction site. The relationship could be described as a 11 Shot gun pattern 11 • The stage values were then adjusted for both float depth and increased resistance (i.e. IAF) and replotted in Figure 3. It will be noted that very close agreement is obtained with the open water rating curve. The IAF for each measurement was computed. The mean va 1 ue was 1. 03 with a standard deviation Qf 0.01. Using the mean value 1-w w u. 8 z 7 w (!J < 6 1- (/) c w ~ 5 (/) < w :E 4 0 0 0 ""' t" Open-water ra mg curve 100 300 DISCHARGE, IN CUBIC FEET PER SECOND Figure 2. Semi-log plot of stage and discharge data collected during the 1984-85 winter season for the Stur- geon River near Nahma Junction. for the IAF and a 1 inear interpolation of float depth between successive readings, the hydrograph of Figure 4 was somputer generated from the stage hydrograph. The actua 1 measured discharges are also shown. Previous studies by the authors had used the unpublished records of 'JSGS for the same station. UsinCJ the raw field ~t) from the monthly ~easurernent for the ice covered condition for the period 1970-75, the mean va 1 ue of the IAF was computed to be 1.0~ with a standard deviation of 0.02. This is basically the same as that found for· the more ~xtensive study during the 1984/35 winter season. ,1\t the tim~ of this writing onlv a portion of the data for the 1985/36 \•linter season was available. Yere again, the IA.F has a value between 1.02 and 1. 03. Of the 3 sites used in this study, the Sidna\>J site received the most intensive study. Results si'llilar to those shown for Nahma Junction were a 1 so obtained for the data fro~ the Sidnaw 279 1-r w w L1. z u.i 6 (.!) < 1- C/) c 5 w 1- C/) :::J ci 4 < Open-water rating curve~ ~ 50 100 300 DISCHARGE, IN CUBIC FEET PER SECOND Figure ~· Adjusted stage vs. discharge relatlons for the 1984-85 winter for the Sturgeon River near Nahma Junction. site. The ice adjustment factor was found to be 1.00 (mean value). As part of the analysis of the Sidnaw data it was assumed that field measurement (i.e. float depth and discharge) were made on January 4 and again on January 29. A linear interpolation of flJat depth was used for all intervening days. Using the "ctual stage record and the int<?rpoLlted float depth the model 1tJas used to prdict the discharge for the intervening days. These results are shown in Table 2. Also, sh0wn in Table 2 are the actual measured discharges on 9 different occasions during the 26 day time period. The average error between predicted and :neasured d i scha rJe was 2. 9 percent. Under the best of fi~ld condition, it is generally assumed that a discharge measurement wi 11 have an error of ± 3 percent. Thus, the average error between the predicted and measured values of discharge was within the limits of error of the discharge measurements. 800 700 600 500 c / Based on open-water rating z 400 0 (.) w (J) 0: w a.. .... w w LL (.) m ::::> (,) z 300 200 on proposed model w " 1 0 0 0: < 90 :r: (,) 80 (J) c 70 60 5 10 15 20 25 30 JANUARY o Measured 5 1 0 15 20 25 5 1 0 15 20 FEBRUARY MARCH Figure 4. Hydrographs for the 1984-85 winter for the Studreon River near Nahma Junction SU~MARY AND CONCLUSIONS In 1983, the authors proposed a theoretically based model to relate stage on an ice covered river to discharge. l4hen the condition at the gage can be classified as 11 a period of stable ice control 11 a simple relationship exists between the open water and ice covered rating curves based on mean hydraulic depth. However, the stage records for the ice covered period are influenced by a buoyant displacement resulting from the floating ice cover. In order to develop a rating curve for the ice covered 280 period, the stage records must first be corrected for the buoyant displacement. In order to test the model, frequent discharge measurements were made on two different rivers in Michigan 1 s Upper Peninsula. In both cases the results show that the average difference between predicted and measured values was within the expected limits of error associated with a discharge measurement (± 3%). A review of unpublished historic records of the U.S. Geological Survey for the period 1970-75 showed that the same relationships existed for the historical record as were Table 2. Comparison between predicted and measured discharge'during January, 1985 for the primary gage at Sidnaw, Michigan. Date Stage Discha~ge (ft3/s) Error (ft) Predicted Measured ( ft 3 /s) Percent January, 1985 4 4.23 66.1 65.9 0.2 0.3 5 4.19 61.4 6 4.18 60.3 7 4.26 69.4 8 4.27 70.6 66.7 3.9 5.7 9 4.23 65.7 10 4.19 61.0 11 4.16 57.0 12 4.13 54.7 57.2 2.5 4.5 13 4.11 52.8 14 4.10 51.7 15 4.12 53.6 52.9 0.7 0.1 16 4.12 53.6 17 4.11 52.5 18 4.13 54.3 19 4.15 56.2 53.6 2.6 4.7 20 4.11 52.2 21 4.08 49.3 22 4.08 49.3 50.7 1.5 3.0 23 4.09 50.1 24 4.13 53.8 25 4.13 53.8 26 4.14 54.6 55.4 0.8 1.4 27 4.12 52.6 28 4.12 52.6 29 4.11 51.5 49.7 1.8 3.6 Average Error: 1. 75 2.9 a Predicted discharge is based on an assumption that discharge and float depth were measured only on January 4 and 29. was used for all intervening days. found in this study. Thus, although the data base for any given winter season may only contain a few actual discharge :neasurements the results confi rrn thJt a consistent relationship exists between the appropriate open water and ico:: covet'ed rating curves. 281 A linear interpolation of float depth ACKNOi~LEDGf-1ENT The authors greatfully acknmvl edge the cooperation and support of the U.S. Geological Survey and the U.S. Ar:ny Cold Regions Research and Engine2ring Laboratory. REFERENCES Alger, G. R. and H. S. Santeford, 1984, 11 1.\ Procedur.~ for Calculating River Flow RJ te Under an Ice Cnve1~, 11 ?roceedings: IAHR Ice Sy.nposium, Hamburg, Sermany, ~'\ugust 27-31, Vol. 1, p. 389-398, Discussion: Vol. III, p. 443-450. Santeford, ~. S., 1986, 11 Stage, Discharge and Ice,11 Proceedings: Sixth Northern Research Basins Symposium/Workshop, Houghton, ~I. January 26-30, 1986 (In > 1 . .,_ . ) puD 1ca"1on . Santeford, ~. S. and G. R. Alger, 1983, 11 Effects of An Ice Cover -A Conceptua 1 !'iOde 1,11 Proceedings: Frontiers in Hydraulic Engineering, ASCE Hydraulics Division Specialty Conference, Cambridge, MA, August, p. 24?.-247. Santeford, ~. S. and G. R. Alger, 1984a, 11 The Hydraulics of River Ice -A Su:nmary,11 Proceedings: Fifth Northern Research Basins S~nposium, Vierumaki, Finland, ~arch 19-23, p. 3.25-3.56. Santeford, H. S. and G. R. Alger, 1984b, "Predicting Flowrat::~s in an Ic0 Covered Stream," Proceedings: Third Int2rnational Specialty Conference on Cold Regions Engineering, ASCE/CSCE, Edmonton, Alberta, A~ril 4-6, Vol. III, P. 1031-1043. Santeford, ~. S. and G. R. Alger, 1984c, 11 rl.Jdraulics of Breakup,11 Proceedings: Workshop on Hydraulics of River Ice, Fredericton, New Brunswick, June 20-21, p. 95-112. 282 Santeford, H. S. and G. R. Alger, 1984d, 11 H_vdraul i cs of FrE~eze-up, 11 Proceedings: Water for RPsource Development, ASCE Hvdraulics Division Specialty Conference, Coeur d 1 Alene, Idaho, August 14-17, p. 574-57<3. Santeford, H. S., S. R. ~~lger and .J • .4. Stark, 1986, 11 lce in S:.reams --Its Formation and Effects en Flow,11 U.S. Geological Survey, Water Resources Investigation Report (in publication). Stark, J. A., 1986, Streamflow Jurinq a Period Stable Ice Control, A Maste~·s Thesis, Michigan Technological University, Houghto:-~, ~~ichigan. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 HYDROLOGY OF TWO SUBARCTIC WATERSHEDS Robert E. Gieck Jr. and Douglas L. Kane 1 ABSTRACT: A water balance can provide information needed for sound water resource management. Two interior Alaskan watersheds located on Ester Dome west of Fairbanks, Alaska were studied from May 1982 to June 1984 to assess possible groundwater recharge. Numerous hydrologic measurements were made, but basically run- off and precipitation were measured and evapotranspiration was calculated using climatic variables. Generally, light showery precipitation and high evapotran- spiration rates preclude or severly limit groundwater recharge during the summer months. Only during periods of substan- tial rainfall is groundwater recharge possible. During snowmelt every year, the groundwater recharge potential is quite high because of water stored in a snowpack that has accumulated over a 6 month period and evapotranspiration demands are minimal. (key terms: water balance; recharge; runoff; evapotranspiration; snowmelt) INTRODUCTION For much of interior Alaska's history, the most serious water resourc& problem has been to supply sufficient water for economic exploitation of its mineral re- sources. Recently, as rural lands have been developed for more diverse uses, the adequate and equitable supply of water for many types of use has become a serious concern. The Ester Dome area west of Fairbanks is an excellent example of such an area. Here traditional mining opera- tions, growing residential population, potential industrial development, agri- culture, and wildlife must compete for the limited water resources available. Water balances were calculated for two basins on Ester Dome to provide information about the area's water resources. A water balance is a valuable tool for managers and engineers who must make decisions about the area's water resources. Current problems can be solved, and future problems avoided by understanding the interrelationships of climate, runoff, groundwater storage, and land use. Of particular importance is groundwater recharge. Groundwater is the primary source of domestic water in the area; placer miners and wildlife rely on streamflow and surface storage. STUDY AREA Ester Dome is located within the Yukon- Tanana Upland of interior Alaska, at latitude 64° 53' north and longitude 148° 02' west. The area lies north of the George Parks Highway and approximately 11 Km west of Fairbanks within the Fairbanks Mining District. While much of the area remains undeveloped, several portions have been extensively modified by human activities. Vegetation is typical of the Yukon- Tanana Uplands. The well-drained, south- facing slopes support forests of Paper Birch (Betula papyrifera), White Spruce 1Graduate Research Assistant and Associate Professor Respectively, Institute of Northern Engineering, University of Alaska, Fairbanks, Alaska 99775-1760. 283 (Picea lauca), and Quaking Aspen (Populus tremuloides • The relatively sparse undergrowth consists of shrubs and forbs. The valley bottoms have shallow slopes of poorly drained soils, and the north-facing slopes support forests of Black Spruce (Picea mariana), with occasional Paper Birch, Green Alder (Alnus crispus), willow (Salix spp.), and Larch (Larix larecena). The undergrowth consists of a thick mat of mosses, lichens, tussock grasses, and shrubs. The geology of the Ester Dome area was described by Forbes (1982). The bedrock is primarily crystalline schists of the Yukon-Tanana metamorphic complex. The mineral soils of the area are composed of layers of micaceous loess originating from the glacial outwash plains of the Tanana Valley to the south. Unconsolidated soil thickness varies from 5 em to over 55 m. Permafrost is usually absent on the south-facing slopes. Discontinuous permafrost is encountered over much of the valley floor and on north-facing slopes. Well logs of the area show permafrost begins as shallow as 60 em and can extend to depths beyond 45 m. Massive ground ice occurs in the area (Pewe, 1982). Ester Dome lies within an area of continental climate characterized by warm summers and winters of severe cold. The average annual temperature at Fairbanks International Airport (133m msl) is -3.3 oc, with an average precipitation of 24.7 em and an average annual snowfall depth of 169 em. Two watersheds were studied on Ester Dome. Ester Creek is a 26.2 square-Km basin (21 percent permafrost) with a mean elevation based on area of 378 m. Happy Creek is a 25.0 square-Km basin {57 percent permafrost) with a mean elevation based on area of 216 m. Two 73.4 square-m runoff plots were observed during snowmelt. These permafrost-free plots \'lere located in the Goldstream Valley about 8 Km northeast of Ester Dome at an elevation of 290 m. METHODS AND MATERIALS The water balance equation for a basin and specific time period can be written as P - I ± F ± E -RO ± SS ± GW = 0 (1) 284 where P = precipitation I = interception losses F = lateral flux of water across soil boundaries E = evaporation, condensation, or actual evapotranspiration RO = runoff SS = change in soil moisture storage GW = change in groundwater storage. Equation 1 can be simplified by selecting the boundaries of the soil volume to define a watershed, thus eliminating the lateral soil water flux (F) term. Precipitation (P), soil moisture (SS), interception (I) and evaporation (E) were combined into a single surplus precipitation term (SP). This term represents precipitation input after accounting for evaporation losses and changes in soil moisture. The total evaporation demand or potential evapo- transpiration (PET) is controlled by the amount of energy available for evapor- ation. The interception term is included in PET since basically all interception evaporates in our case. When PET demand exceeds precipitation, stored soil moisture is used to fulfill the demand. Soil storage was established at 10.2 em (Thornthwaite and Mather, 1955). When soil moisture storage is exhausted, actual evapotranspiration (AET) is less than PET. Precipitation in excess of PET is returned to soil moisture storage until the full 10.2 em capacity is reached. The remaining precipitation is surplus precipitation (SP). Considering these simplifications equation 1 reduces to SP -RO ± GW = 0. (2) Potential evapotranspiration (PET) values were estimated from evaporation pan measurements. Pan data were adjusted using a 0.7 pan coefficient that was determined from a detailed mass balance for the upper one meter of soil calcula- tion using equation 1. Soil moisture data from the runoff plots for June and July 1984 were used to obtain this pan coeffi- cient. When pan evaporation data were not available, potential evapotranspiration was estimated using the Thornthwaite method (Dunne and Leopold, 1978). Local water balances {equation 2) were calcu- lated monthly excluding the winter months. Evaporation from the snowpack or conden- sation during snowmelt was estimated using an equation suggested by the U.S. Army Corps of Engineers (1956). This method was used for the 1983 data from start of snowmelt {15 April) until the disappear- ance of snow at the lowest elevations (28 April). Due to missing pan evaporation data in 1984, this method was used from 26 April until 17 May. The pan coefficient was used to adjust measured pan evapora- tion after these dates. Runoff volumes were calculated by multiplying the mean discharge of a time period by the length of the time period. Discharge measurement sites were estab- lished on Happy Creek and on Ester Creek. Discharge measurements were taken using standard cup-type current meters. Measurements of discharge were made one to three times daily during snowmelt or heavy rainfall runoff events. Discharge was measured each week during mid-summer and fall. Additional snowmelt runoff data were collected at the runoff plots in Goldstream Valley. Snowpack water equivalents were deter- mined at snow courses designated by elevation along the existing road network. Snowpack water equivalents were obtained using an Adirondak snow sampler. The reported amounts of snow water equivalent are averages obtained for eight to ten measurements at each course. In the late-spring of 1982, eight meteorological sites were established in the Ester Dome area to collect precipita- tion data. Summer precipitation data were collected using both standard 20.3 em dipstick and tipping bucket raingages. These sites were: Coutts (2), Ester Dome Road (3), Gedney (4), Rice (6), Stone (7), Swainbank (8), Ester Dome Summit (10), and Nugget Creek (13) (bracketed numbers show site location in Figure 2). The following year 4 additional sites were added: Quartz (11), Lasonsky (12), Willow Creek (14), and St. Patrick Creek Road (15). Additional precipitation data were obtain- ed from the U.S. Department of Commerce, National Weather Service, Alaska Climato- logical Summaries. Annual precipitation was obtained by adding the summer precipitation at each site to the maximum snowpack water equivalent prior to ablation. Air temper- ature and relative humidity were measured 285 at the Ester Dome Summit site (10) using a recording hygrothermograph. Evaporation data were obtained at the the Ester Dome Summit (10) using a standard 122 em diameter by 25.4 em deep evaporation pan, stilling well and hook gage. Additional evaporation data were obtained from the University of Alaska-Fairbanks, Agricul- tural Experiment Station (1). Linear regressiones were developed by comparing short term meteorlological data collected on site to data obtained at nearby stations with longer periods of record. Regression lines were calculated using the least squares method. Signifi- cance of fit was determined by t-test at the 95 percent confidence level. Soil tensions were obtained at the Stone (7) and Gedney (4) sites weekly using tensiometers installed at regular depths (20, 40, and 60 em) during the summer of 1982. Summer 1984 soil moisture data were obtained in the upper one meter of soil from a long-term study taking place at the Goldstream Valley runoff plots. These data were collected using time domain reflectometry (Stein and Kane, 1983). Three unpumped wells were monitored on Ester Dome: the Swainbank site (8), and the upper and lower wells at the St. Joe American Mine on Henderson Road. The depth to the piezometric surface in the wells was measured using both an acoustic well probe and a well tape. The wells were monitored during the 1981-1982 water year by Northern Testing Laboratories Inc. using a well tape. RESULTS The water balances in Ester Creek water- shed, Happy Creek watershed and the runoff plots for the 1983 and 1984 snowmelt periods (calculated using equation 2) are given in Table 1. The distribution of snowmelt is shown on a percentage basis in Figure 1. Precipitation, evaporation and soil storage amounts end at the disappear- ance of the snowpack, while runoff quanti- ties include the recession portion of the snowmelt hydrograph. Monthly water balances were calculated for each basin. Precipitation and actual evapotranspiration were estimated for the mean watershed elevations to give average basin conditions. Runoff was calculated Component (em) Ester Creek Happy Creek Plot 1 83 84 83 84 8384 Plot 2 ~84 Snow Water Equiv. {+) 15.5 12.7 13.7 10.7 12.4 9.9 14.7 9.9 Soil Deficit (-) 1.0 1.5 5.6 3.6 3.6 2.5 3.6 2.5 Runoff (-) 3.8 2.8 3.6 3.0 3.0 2.5 4.8 1.5 Actual Evaporation (-) 1.5 1.8 1.5 1.8 1.5 1.8 1.5 1.8 Snowme 1t Peri ad Groundwater Recharge 9.2 6.6 3.0 2.3 4.3 3.1 4.8 4.1 Table 1. Snowmelt period water balances (15 April to 25 May 1984). from streamflow measurements. Table 2 presents the 1983 water balance for the two watersheds. Total annual precipitation is plotted in Figure 2. Isohyets have been superimposed over elevation contours to show the dis- tribution of precipitation in the area. An average pan coefficient of 0.7 was obtained from mass balance calculations using soil moisture data. Detailed soil moisture measurements at the runoff plots allowed us to make this determination. Regression analyses show that precipita- tion increased as elevation increased. Snowfall increased 14 percent per 100m increase in elevation. Rainfall increased 11 percent per 100 m increase in eleva- tion. The air temperature lapse rate was -0.83 °C per 100 m increase in elevation. a) Ester Creek watershed. NET Component ANNUAL (em) APR MAY JUN JUL AUG SEP OCT-MAR (em) p 15.5 0.5 1.9 5.6 10.9 3.0 0.0 (37.4) PET 1.6 9.2 9.9 8.4 3.3 1.9 0.0 (34.3) P-PET 13.9 -8.7 -8.0 -2.8 7.6 1.1 0.0 (3 .1) SSi 9.2 10.2 1.4 0.0 0.0 7.6 8.7 SSf 10.2 1.4 0.0 0.0 7.6 8.7 8.7 ~ss 1.0 -8.7 -1.4 0.0 7.6 1.1 0.0 ( -0.5) AET 1.6 9.2 3.3 5.6 3.3 1.9 0.0 (24.9) SP 12.9 0.0 0.0 0.0 0.0 0.0 0.0 12.9 RO 0.8 2.8 0.6 0.4 1.1 2.0 3.3 11.0 GW I2.T -2.8 -=o:o -D.4 -IT -2.0 -D ~ b) Happy Creek watershed. NET Component ANNUAL (em) APR MAY JUN JUL AUG SEP OCT-MAR (em) p 13.7 0.4 1.7 4.9 9.6 2.6 0.0 (32.9) PET 1.5 9.7 10.4 6.9 3.3 2.2 0.0 (34.0) P-PET 12.2 -9.3 -8.7 -2.0 6.3 0.4 0.0 (-1.1) SSi 4.9 10.2 0.9 0.0 0.0 6.3 6.7 SSf 10.2 0.9 0.0 0.0 6.3 6. 7 6. 7 ~ss 5.3 -9.3 -0.9 0.0 6.3 0.4 0.0 (1.8) AET 1.5 9.7 2.6 4.9 3.3 2.2 0.0 (24.2) SP 6.9 0.0 0.0 0.0 0.0 0.0 0.0 6.9 RO 2.0 1.6 0.3 0.03 0.03 0.5 0.03 4.5 w-4.9 -TI -o.J -m -m -O:S -m 2:l SSi -initial soil moisture and SSf = final soil moisture. Table 2. Monthly waterbalances Apri 1 1983 to March 1984. 286 DISTRIBUTION OF SNOWMELT EVAPORATION SOIL MOlSTURE ~~- ESTER CREEK EVAPORA nON SOIL WOISllJRE ,f:r\ ~\Q RECHARGE Jell: PLOT 1 EVAPORAnON SOIL WOIS1\JRE ~~ 52X ESTER CREEK EVAPORATION SOIL WOISllJRE -Ed_ PLOT 1 1983 1984 SOIL WOISllJRE ~m· RU 28lll: ~CHAAGE 22X HAPPY CREEK EVAPORAnON SOIL WOJSTURE --·~ 33V0RECHAAGE 33lll: PLOT 2 SOIL WOISllJRE EVAPORAn:@33lll: 17lll: RU UECHAAGE 21111. 21lll: HAPPY CREEK EVAPORA~:ep=-WOJSTURE RUNOFF 0 15lll: RECHARGE 41lll: PLOT 2 Figure 1. Distribution of 1983 and 1984 snowmelt. Monthly evapotranspiration decreased by 3.5 percent per 100 m increase in elevation. DISCUSSION Winter Period Potential groundwater recharge is un- likely during the winter in interior and northern Alaska. Unless significant melting of the snowpack occurs, which is Annual Precipitation Oct. '82-Sept. '83 Figure 2. Site locations and precipitation distribution. isohyets and heavy solid lines are watershed boundaries. Heavy dashed lines are quite rare, all precipitation remains frozen and stored at the surface as snow or ice. Snow accumulation generally begins in late September or early October, depending upon elevation. Due to cold air temperatures and very low insolation, snowmelt does not usually begin until mid-April. By late April or early May, sufficient energy is provided by rising air temperatures and increased insolation for melt to begin. Without snowmelt inputs, winter is a period of net loss of stored groundwater. Groundwater storage changes through the runoff component as baseflow. The gradual drop of water tables in observed wells is evidence for the lack of winter ground- 287 water recharge and loss of groundwater storage as baseflow. The magnitude of winter baseflow is generally unknown, since ice-covered streams and aufeis accumulations prevent accurate stream gaging. However, a reasonable recession constant was estimated and applied to baseflow at freeze-up, so a crude estimate of winter baseflow was made for each basin. The water balance equation for the winter months is very simple. All terms in the equation estimating groundwater recharge are zero except runoff (RO), which is equivalent to the baseflow. Snowmelt Period Significant infiltration can occur during snowmelt despite the extensive seasonal frost (Kane and Stein, 1983). Steady and gradual melt of accumulated winter precipitation, while evaporative demands are low, provides soil moisture recharge. Soils saturate rapidly and infiltration rates, although relatively low, allow water to enter the soil. Some of the sno~melt infiltration may be used later to satisfy evapotranspiration demand. The Ester Creek watershed more effi- ciently retained snowmelt infiltration in 1983 and 1984. At least 50 percent of the snowpack became recharge in the Ester Creek watershed, while less than 22 per- cent of the snow water equivalent became recharge in the Happy Creek watershed. Several factors contributed to the dispar- ity in recharge effectiveness. The higher mean elevation of Ester Creek watershed gives it a relatively higher snowpack water equivalent and a lower soil moisture deficit. Also, permafrost in the Happy Creek watershed reduced the area contributing to groundwater recharge, and increased the proportion of runoff. Runoff dominates water balance losses during this period. Peak annual stream discharge usually occurs during the snowmelt period in interior and northern Alaska. Happy Creek snowmelt peak dis- charges occurred earlier, were higher, and receded earlier. More snowmelt runoff occurred in Happy Creek watershed, despite its lower terrain and gentler slopes. Geologic mapping of the area and well logs show that permafrost underlays much of Happy Creek lowland. Aerial photo- graphy of Ester Dome was used to estimate the areal distribution of permafrost within each basin (NASA, 1978). Vegeta- tion type was used as an indicator of permafrost-prone areas. The percentage of the basins with permafrost were: Ester Creek watershed 21 percent and Happy Creek watershed 57 percent. Much of the infiltration in the Happy Creek basin may result in recharge to suprapermafrost water, with recharge not reaching the deeper subpermafrost groundwater. In comparison, Ester Creek watershed is dominated by well-drained, south-facing, permafrost-free soils. 288 Well data indicate that at least some of the snowmelt infiltration reaches the water table. The wells, one near the summmit of Ester Dome and two along a ridge dividing the two basins, showed a rise in the late spring snowmelt season. Summer-fall Period Summer-fall hydrologic season was a time of net water loss. The water balance during this period was dominated by high potential evapotranspiration. Summer precipitation tended to be light and showery, providing scattered rainfall followed by dry periods. Fall precipita- tion, on the other hand, can be heavy and sustained. August has the highest average monthly precipitation in interior Alaska. However, the average August precipitation is lower than the average maximum snow water equivalent. Evapotranspiration demand exceeded precipitation throughout the summer months at lower elevations in 1982 and 1983. Early summer evapotran- spiration was nearly equal to the 1982 and less than the 1983 precipitation in the upper elevations. Long hours of daylight, warm air temperatures and low relative humidity maintained high demand for soil moisture. By mid-August, the evapotranspiration demand drops rapidly. Deciduous vegeta- tion generally loses its leaves and stops transpiring in late August, and actual evapotranspiration may be less than the potential. If large precipitation events occur late in the period, as happened in August 1983, groundwater recharge is possible. Without significantly above- normal precipitation events, evapotran- spirative losses probably still exceed inputs in the lower portions of the basins. Kane and Stein (1984) found that large September precipitation events, if they occur close to freeze-up, retard recharge the following spring, thereby offsetting any potential gains in ground- water storage. Basically, ice in the soil pores by reduces the infiltration the following spring. Long term average data and the regress- ion equations were used to estimate long term average conditions at all elevations on Ester Dome. Early summer (May and June) had greater PET loses than precipit- ation inputs at all elevations. July was a transition month, precipitation just exceeded PET at the highest elevations. August and September, however, had PET less than precipitation over most of the area. PET exceeded precipitation only at the lowest elevations. The soil moisture gains in August and September do not tot- ally replace the soil moisture lost in May, June and July. Both the average (4.7 em) and highest recorded summer precipita- tion (15.7 em) (both for August) are lower, respectively, than the average (10.7 em) and highest recorded (24.9 em) 1 April snow water equivalent. Compared to the April relationship, which includes the average March 31 snow water equivalent (USDA Soil Conservation Service, 1984) and average April precipitation, it becomes obvious that (at least on a long-term basis) the snowmelt period (usually April) is the most significant recharge period -- even when 20 to 30 percent runoff losses are taken into account. The calculated 0.7 pan coefficient is lower than the 0.81 coefficient for well-watered rapeseed at Delta Junction, Alaska calculated by Braley (1980). He concluded that all summer precipitation was used to satisfy evapotranspiration demand at his low elevation sites. Our plots, where the 0.7 pan coefficient was determined are located in a mixed stand of Aspen, Birch and White Spruce on a well-drained, south-facing slope. The Thornthwaite method was used to estimate monthly potential evapotranspira- tion when pan data were unavailable in the early summer. Thornthwaite•s method tends to underestimate the amount of potential evapotranspiration in early summer and overestimate fall potential evapotranspir- ation. Patrie and Black (1968) found a similar disparity at the Agricultural Experiment Station using long-term average data. Soils became steadily dryer throughout 1982, a year of steady precipitation input. It is unfortunate that soil moisture data were not obtained for the summer of 1983, since extremely light precipitation occurred early in the summer followed by heavy fall precipitation. However, suction lysimeters used to obtain soil water for a study of the groundwater geochemistry of Ester Dome (McCrum, 1985), indicated relatively dry soils with tensions greater than 50 em of water 289 throughout the summer. After the late August precipitation event, the suction lysimeters easily withdrew soil water, and tensions were less than 50 em. Therefore, relatively wet soils were present in late August and September after heavy precipi- tation. Some recharge could have occurred. A major flaw in treating the soil as a storage reservoir is that recharge can occur when soils are not saturated. The rate of recharge in unsaturated soils is dependent on soil moisture content and pressure gradients. Dry Fairbanks silt loams (less than 20 percent moisture by volume) have hydraulic conductivities less than 1*E-4 em per second (Kane et al., 1978). So without high pressure gradi- ents, recharge is small. The greatest potential for summer recharge occurs when a large input of water causes large pore pressure gradients. This occurred in late-August 1983. Over 13 em of precipitation fell at the highest elevations (above 600 m) from 20 August to 1 September 1983. Potential evapotran- spiration was only 1.3 em, leaving a pre- cipitation excess of about 11.7 em. The Swainbank well (640 m elevation) response was dramatic during this rainy period, rising over 46 m. The lower elevations received substantially less precipitation and higher potential evaporation demand. Less than 10 em of precipitation fell on the lowest areas (below 250 m) from 20 August to 30 September. The estimated potential evapotranspiration was about 3.1 em during this period. Table 2 shows that at the mean elevation of both watersheds (Happy Creek 216 m and Ester Creek 377 m) the soil had been completely depleted of stored moisture prior to this precipita- tion. Neither basins• average input of precip- itation fully replenished the soil mois- ture deficit. Due to the high pressure gradients between the upper and lower soils, some recharge must have occurred. Some of this summer recharge remained in soil storage until after freeze-up and some became runoff. Considering these losses, the summer recharge must have been less than the snowmelt recharge. The 1983 water balance suggests that significant recharge in the summer-fall period is unlikely, especially at lower elevations. High potential evaporation demand early in the summer depletes stored soil moisture. Heavy precipitation is required to overcome evapotranspiration demand and replenish depleted soil moist- ure if recharge is to occur. Runoff losses are relatively small during the summer months. Baseflow is often considered to be an estimate of groundwater recharge in a steady-state system. The 1983 annual baseflow esti- mates were: Ester Creek 8.9 em and Happy Creek 1.8 em. Baseflow nearly equaled the 9.1 em of estimated snowmelt groundwater recharge (Table 1) in Ester Creek water- shed. The Happy Creek baseflow is much lower than the 3.0 em of estimated snow- melt recharge, perhaps because of the large area of permafrost within the basin. CONCLUSIONS On and around Ester Dome (an area west of Fairbanks, Alaska) mining, and both residential and potential industrial development must compete for a limited water resource. Recent concern about the just and adequate distribution of water in the area has illuminated the need for more complete water resource information. Areas in the Yukon-Tanana upland adjacent to Fairbanks, such as Ester Dome, are of particular interest as development occurs because of limited groundwater storage in fractured schist aquifers. There are several basic questions which should be answered before sound water management decisions can be made in devel- oping areas. How much water is there? How and when does it get there? How and when does it leave? A water balance can partially answer these questions. Two watersheds on Ester Dome (Ester Creek and Happy Creek) were instrumented to collect data necessary for water balance calcula- tions. Data collection began in June 1982 and ended in June 1984. Water balances were calculated, solving for groundwater recharge. An annual water balance was made from a synthesis of 1983 water balances. Conclusions for the seasonal periods and annual synthesis were: 1) The snowmelt period was the primary time for significant groundwater re- charge. This was a time of net gain in the area's water resources. The water 290 balance during this period was dominated by high runoff. Steady input of melt from the winter's accumulation of snow and low evaporation rates provided significant groundwater recharge, despite the reduced infiltration rates of the frozen soils caused by seasonal frost. 2) The summer-fall period was dominated by high rates of potential evapotranspir- ation. This period was one of net loss to the area's water resources. Light precip- itation and high rates of evapotranspira- tion kept early and midsummer groundwater recharge very low. Groundwater recharge may occur in the late summer and fall as potential evapotranspiration rates de- cline, given sufficient precipitation input. Areas of higher elevation, where evapotranspiration is lower and precipita- tion greater, are more likely to have sig- nificant potential groundwater recharge. 3) The winter period is dominated by baseflow. All precipitation inputs are temporarily stored at the surface as ice or snow. Without snowmelt inputs the winter period was one of continual and significant water loss in the watersheds studied. 4) The upper elevations (above 600 m) of the Yukon-Tanana Uplands may receive twice the precipitation observed by the National Weather Service at the International Air- port. The precipitation increases per 100 m increase in elevation were: snow pack water equivalent 14 percent and rainfall +11 percent. Monthly potential evapotran- spiration decreased by 1.7 to 3.5 percent per 100 m increase in elevation. The average environmental lapse rate was 0.83 oc per 100m. The Swainbank well rose in response to snowmelt and following concentrated rainfall in the fall. Receding water tables were observed in the winter and early to mid-summer. REFERENCES Braley, W. A. 1980. Estimates of evapotranspiration from barley and rapeseed in interior Alaska. M.S. Thesis. University of Alaska, Fairbanks, Ak. Dunne, T., and L. B. Leopold. 1978. Water in environmental planning. W.H. Freeman and Co. Forbes, R. B. 1982. Bedrock geol.ogy and petrology of the Fairbanks Mining District, Alaska. Alaska Department of Natural Resources, Division of Geological and Geophysical Surveys. Open-file Report 169. Kane, D. L., R. D. Seifert, and G. S. Taylor. 1978. Hydrologic properties of subarctic organic soils. University of Alaska, Institute of Water Resources Report 88. Kane, D. L., and J. Stein. 1983. Water movement into seasonally frozen soils. Water Resources Research. 19{6):1547- 1557. Kane, D. L., and J. Stein. 1984. Plot measurements of snowmelt for varying soil moisture conditions. Geophysica. 20(2):123-136. McCrum, M. A. 1985. A chemical mass balance of the Ester Creek and Happy Creek watersheds on Ester Dome, Alaska. M.S. Thesis. University of Alaska, Fairbanks, Ak. NASA. 1978. Aerial Photo of Ester Dome. 23-317, JSC 386 Jul., 78 Alaska Cir. 60. Patrie, J. H., and P. E. Black. 1968. Potential evapotranspiration and climate in Alaska by Thornthwaite•s classification. USDA, Forest Service, Pacific Northwest Forest and Range Experiment Station, Juneau, Alaska. Forest Research Paper PNW-71. Pewe, T. L. 1982. Geologic Hazards of the Fairbanks Area, Alaska. State of Alaska Department of Natural Resources, Division of Geological and Geophysical Surveys. Special Report 15. Stein, J., and D. L. Kane. 1983. Monitoring the unfrozen water content of soil and snow using time domain reflectometry. Water Resources Research. 19{6):1573-1584. Thornthwaite, C. W., and J. R. Mather. 1955. The water budget and it's use in irrigation. The Yearbook of Agriculture, Water. U.S. Department of Agriculture. U.S. Government Printing Office, Washington, D.C. pp. 346-358. 291 US Army, Corps of Engineers. 1956. Snow hydrology, summary report of the snow investigations. U.S. Government Printing Office, Washington, D.C. USDA Soil Conservation Service. 1984. Snow surveys and water supply outlook for Alaska as of April 1, 1984. Portland, Ore. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 THE WATER BALANCE OF THE UPPER KOLYMA BASIN V. K. Panfilova 1 ABSTRACT: The water balance must be estimated for the economic development of cold regions. The sparse hydrometeor- ological network in these regions requires indirect methods to estimate the water balance. The author used runoff and its coefficients for the Upper Kolyma Basin (99,400 square km). Instead of evaporation data for this area, the author used the balance method with due account of the orientation and wetting of the main tributaries in the Upper Kolyma Basin. The water balance components differ with region and with different landscape zones. (KEY TERMS: water balance; precipitation; runoff; evaporation; landscape zone.) INTRODUCTION The Upper Kolyma valley is one of the most developed regions of the North- Eastern part of the USSR. The further development in the mountainous areas of this region demands a more profound research of natural resources, especially of water resources. The hydrometeor- ological service concentrated in the Kolyma valley and the middle mountain areas (above 600 m). This network does not provide for adequate data c.oncerning the different water economy accounts. This paper describes the elements and structures of water balance distributions in the different high-altitude zones of the region. The work is of great practical importance because there is no mutual estimation of the water balance elements for this region. REGIONAL PHYSIOGRAPHICAL FEATURES The water balance depends on some natural features of the region. The Upper Kolyma Basin is the middle mountain land with average height of 927 m and the maximum height up to 2,200 rn. The annual precipitation varied from 200 to 400 mm. About 60-70 percent of annual precipitation falls during the warm season (May-September). The rigorous climatic conditions promote the total spread of permafrost. By September, the upper permafrost limit is: 60-70 em below the surface with moss vegetation; 120-130 em below the surface with trees and shrubs on loam; and 150 em below the surface of the rudaceous rocks and alpine tundra. Permafrost provides a natural impermeable layer that hinders water exchange between surface water and that of subpermafrost zone. This water exchange is possible only in limits of valley taliks of the rivers with catchment areas more than 7,000 square km. During the cold season (October-April), other rivers are frozen up to the bottom. The rivers of the Upper Kolyma Basin are fed by rain and snowmelt; up to 95 percent of its runoff occurs during the May-September period. During the cold season, aufeis is formed from supraperma- frost water; 5-10 percent of annual runoff is stored as aufeis. The volume of growth of the aufeis depends on talus area in some river basins (Kuznetsov and Nasybulin, 1970). 1 Problems Research Laboratory for Northern Development, Geography Faculty, Moscow State University, Moscow, 119899, USSR. 293 Areas of talus are widespread in the region. These areas exhibit a larger infiltration capacity, 1 esser evaporation and good conditions for water condensation. Up to 30 percent of Upper Kolyma Basin is covered by talus. River runoff is formed in the specific landscape zones of the region: alpine-tundra zone (above 1,500 m); elfin wood-tundra zone (1,500-1,000 m); larch open woodlands zone (1,000-600 m); larch taiga zone (600-400 m); willow-chosenia forests, larch open woodlands ("mari") and grass moor along the flood plains. DATA AND METHODS Observations in the study area were conducted by the Kolyma Water Balance Station on catchments from 0.30 to 21.2 square km and four catchments from 100 to 1,750 square km. The observation results depict the hydrological cycle (Kuznetsov, et al. 1969). A number of scientists attempted to account for the normal annual precipitation and runoff for the different observation periods throughout the Upper Kolyma Basin (Garzman and Ryabchikova, 1972). The author used the following equation for the water balance: P = R + E where P, R and E are the normal annual precipitation, runoff and evaporation, respectively. This approach allows linking the water balance elements of the Upper Kolyma Basin. Precipitation. The corrected data of 53 meteorological stations provided normal annual precipitation accounts. Most stations are located near the Kolyma mainstream. On the uplands--at the western, south-western and southern parts of the Kolyma Basin--the precipitation observations were not carried out. The meteorological stations are located at 600-800 m on the western part of the basin, and 200-400 m on the eastern one. Only one station, with representative observations is located at 1,220 m. The author used the method of "call ecti ve regional analogy" to determine the depth of rainfall at the different altitudes. This method is based on the regional 294 dependence of the precipitation from the sea-level elevation. The met-stations grouping made on the area data for right- and left-bank, and western and eastern tributaries of the Kolyma River revealed that in three main areas, P = f(A) (where A is the met-station's sea-level elevation): I. The western part, left bank of the Kolyma-Ayan-Yuryakh (14,140.2 square km) and Berelekh basins (9,829.9 square km); II. The western part, right bank of the Kolyma-Kulu (15,654 square km), Tenke (4,450 square km) and Detrin (6,626.6 square km) basins. This area is divided into two s~bareas: IIa) Kulu Basin and lib) Tenke and Detrin basins. III. The eastern part, left-bank tributaries basins--Debin (5,444 square km), Tascan (11,911.8 square km) and right bank basins--Bokhapcha (13,600.0 square km), Orotukan (2,340.9 square km), Srednekan (1,728.5 square km). Normal annual precipitation in all of these basins is somewhat greater than in western ones. For these areas, the relationship of P = f(A) for the regions are as follows: 3.27 (I), 0.60 (IIa), 0.47 (IIb), 0.29 (III). The existing met-stations network does not allow for reliable extrapolation in the "precipitation-elevation" relationship. We used the indirect method for precipitation accounting--on the basis of runoff and its resultant coefficients. Runoff. The mean weighted elevation of the studied basins are 600-800 m (the eastern part) and 900-1,200 m (the west). Consequently, the precipitation data are sufficient for the 400-800 m elevation zones, but the runoff data are represen- tative for the 900-1,200 m zones. For R = f(A), we used the runoff data of 125 rivers. All observations were reduced to annual series. These rivers were grouped on the basis of catchment area: 35 percent of rivers have the catchment area less than 10 square km, 50 percent have less than 50 square km, 60 percent have less than 100 square km, and 78 percent have less than 500 square km (in increasing summation). In accordance with the position of the catchment of the mean weighted elevation, there are three main groups of rivers: 38 percent of the river•s catchments are located in the 600-900 m elevation range; 52 percent in 900-1,200 m range and 10 percent in 1,200-1,400 m range. The runoff data from 125 rivers reveal its distribution for each main tributary of the Kolyma. We have obtained representative curves for R = f(A) for the Berelekh, Ayan-Yuryakh, Debin Tascan, Kulu, Detrin and Srednekhan basins. There are no direct data for the Tenke, Bokhapcha and Orotukan basins, so we used the analogy between the runoff distribution and the elevation of the neighbouring basins. The combined analysis of P = f(A) and R = f(A) confirmed the existence of the regional boundaries for regions I, IIa, lib, differing in the precipitation data. In the eastern region, differences in the character of the vertical runoff distribu- tion were found. We divided this region into four subregions: Ilia) Tascan Basin; IIIb) Debin Basin; IIIc) Bokhapcha Basin; IIId) Orotukan and Srednekan basins. The long combined "precipitation- runoff .. observations were carried out on the Kolyma water balance station in the Kulu Basin. We used these data to deter- mine the runoff coefficients for different elevational landscape zones. Nasybulin (1976) demonstrated that the hydrological conditions of the water balance station are representative throughout the Upper Kolyma Basin. This result has enabled us to use the regional runoff coefficient (Ila) for the other regional accountings. These calculations show the precipitation and runoff distributions depend on an area•s elevation. We have tabulated the accounting depths of precipitation and runoff using regional curves of P = f(A). Since no evaporation data are available for the Upper Kolyma Basin, we determi~ed the amount of evaporation in mm as the preci- pitation-runoff difference for the vertical zones. On the basis of values of P, R, E, we then determined a (the runoff coefficient) and s (the evaporation coefficient) for the same vertical zones of each subregion. See the accounting examples in Table 1. The calculations were used to map the water balance of the Upper Kolyma. With these water balance maps and the landscape map planimetry made it possible to 295 TABLE 1. The elements of water balance distribution and structure in subregion II I d. Elev Precip Run Evap A(m) P(mm) R(mm) E(mm) 400 500 150 350 0.30 0.70 600 595 260 335 0.44 0.56 800 640 370 270 0.58 0.42 1000 675 450 225 0.67 0.33 1200 705 520 185 0.74 0.26 1400 730 570 165 0.78 0.22 1500 750 590 160 0.80 0.20 determine the mean weighted values of the water balance structure for different landscape zones (Table 2). RESULTS Both precipitation and runoff in Regions II and III increased with the elevation. In Region I, runoff increased with the elevation, but the precipitation decreased down to 1,100 m above sea level. This event is typical for the enclosed valleys like the Berelekh Basin. Besides, Region I is located in the western and northwestern parts of the Upper Kolyma Basin and is the farthest area away from the eastern and the northeastern humid flows. Higher than 1,100 m, the pluvio- metric coefficient has positive values, which is confirmed by R = f(A). This phenomenon may be explained by the transport of humid air masses to high elevations. The precipitation and runoff values in the Upper Kolyma Basin increase eastward and to the right-bank areas. The Orotukan and Srednekan basins receive the most precipitation. In the western part of the Upper Kolyma Basin, the evaporation is less than in the eastern parts because talus and scree cover most of the small river basins in the west. DISCUSSION This study provides the water balance TABLE 2. Water balance for different landscape zones. Region, subregion Landscape (percent) Water Balance elements (mm) Precip Run Evap 1 2 3 4 5 p R E a 8 I 20.2 2.4 42.5 12.0 22.9 330 180 150 0.55 0.45 21.8 6.9 33.5 7.1 30.8 375 200 175 0.53 0.47 I! a 36.0 1.0 33.7 8.9 20.5 375 220 155 0.59 0.41 lib 31.6 0.5 29.6 10.9 27.3 410 240 170 0.59 0.41 37.7 0.2 31.8 12.7 24.8 425 255 165 0.60 0.40 II! a 16.2 4.1 28.4 17.3 34.0 500 210 290 0.42 0.58 IIIb 24.2 4.6 36.3 8.0 27.0 490. 210 280 0.43 0.57 II Ic 33.3 2.2 34.3 10.3 20.0 545 265 280 0.48 0.52 IIId 26.2 3.7 40.4 15.1 14.5 570 290 280 0.51 0.49 31.9 5.3 28.0 10.2 24.2 560 280 280 0.50 0.50 Upper Kolyma Basin 27.2 2.7 32.5 11.6 26.1 433 233 200 0.54 0.46 1 -alpine-tundra; 2 -elfin wood tundra; 3 -larch open woodlands; 4 -larch taiga; open woodlands (11 mari 11 ) and grass moor of flood 5 -willow chosenia forests, larch plains. for the Upper Kolyma Basin and the distribution of water in the landscape zones of the basin. These results are interesting for practical purposes, but the results include all errors and inaccuracies stemming from the water balance method used for estimating evaporation. Determined as the remainder term of the equation, the evaporation value includes all precipitation and runoff errors and other parameter effects of the water balance. In the Upper Kolyma Basin, the many talus slopes condense some amount of moisture. If this amount is ignored, this leads to understating the water supply term in the water balance. Besides that, the orographic differentia- tion of water balance elements has been determined for the main tributaries of the Kolyma only. It proved impossible to take into consideration the slope orientation. These shortcomings may be eliminated by more detailed water balance investigations. 296 REFERENCES Garzman, I.N., T.N. Ryabchikova, 1972. On Mean Annual Runoff and Precipitation Distribution on the Upper Kolyma Territory and the Northern Coastal Zone of the Okhotsk Sea. Transac- tions of the Far Eastern Hydro- Meteorological Institute. 39:5-17 [in Russian]. Kuznetsov, A.S., Sh.S. Nasybulin and A.I. Ipat'eva, 1969. The First Results of the Upper Kolyma Basin Water Balance Investigations. Transactions of Magadan Hydro-Meteorological Observatory. 2:21-37 [in Russian]. Kuznetsov, A.S., and Sh.S. Nasybulin, 1970. The Runoff Formation Peculiarities on the on the Rivers of the Upper Kolyma Basin. Transactions of Magadan Hydro-~1eteorological Observatory. 3:52-65 [in Russian]. Nasybulin, Sh.S., 1976. The Kolyma Water Balance Station's Runoff Data Repre- sentativity for the Upper Kolyma Territory. 11 The Natural Resources of the Soviet North-East 11 • Vladivostok. [in Russian]. !ULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 WATER BALANCE AND RUNOFF ANALYSIS AT A SMALL WATERSHED DURING THE SNOW-MELTING SEASON H. Motoyama, D. Kobayashi and K. Kojima 1 ABSTRACT: The amount of snow water equivalent, the snowmelting rate and the stream discharge were measured at a small watershed (11.2km ) during four snowmelting seasons. Water balance and runoff analysis were made on the basis of hydrological observations. Some im- portant characteristics of the water balance were found as follows; (1) 90 % of the snowmelt in the watershed ran off in both heavy and light snow seasons; (2) the amount of evaporation was estimated to be less than 1% of the total amount of snowmelt; (3) in the middle of the snowmelting season, the several days accumulated snowmelt amount equaled the snowmelt runoff without a change of groundwater storage in the watershed. (4) The simulation of daily snowmelt runoff was successful by using the same runoff model for three years. (KEY TERMS: snowmelt; water balance; snowmelt runoff; runoff tank model.) INTRODUCTION It is well recognized that snowcover constitutes a vital part of water re- sources; therefore, accurate forecasting of the amount of runoff water ·is indis- pensable for making efficient use of meltwater and forestalling damage due to snowmelt flooding in snowy regions. The total amount of snowcover just before the snowmelting season in experi- mental watersheds has been measured and forecasting of snowmelt runoff has been attempted by many investigators (Leaf, 1971, Rango, 1983, Sugawara et al., 1984). A prerequisite for forecasting of snovnnelt runoff is to clarify the water balance of a watershed, but there are still many unknown factors at present. In a few papers, water balance is dis- cussed on the basis of the observed snowmelt amount in the watershed and the stream discharge (Ono and Kawaguchi, 1974). The present study was conducted for the following purposes: (1) to clarify the water balance of a watershed during the snowmelting season on the basis of observed hydrological data, (2) to simu- late the daily snowmelt runoff. INSTRUMENTS The observations were carried out at an experimental watershed located in the northern part of Hokkaido, Japan (44 23'N, 142 17'E, drainage area 11.2km , elevation range 280-630m a.s.l., aver- age slope angle 11.0 , shown in Fig. 1) during the period from April to May in the years 1981, 82, 83, 84. 11nstitute of Low Temperature Science, Hokkaido University, Sapporo, 060, Japan. 297 Experimental Figure tal watershed. the measuring points snow water equivalent and the snowmelting rate. BH station for meteorological observations. P4 water gauge station. l) Snow Water Equivalent Observations in the Watershed Total volume of snow cover in the watershed was obtained in mid-April, before the melting season, every year. The snow water equivalents (HW) at about twenty sampling points in the watershed (Fig. 1) were measured. Some ten meas- urements of the depth made with a snow sonde gave an average snow depth (HS) while three or four measurements of the average snow density made with a snow sampler gave the average density ( at a point. The snow water equivalent is calculated as follows, HW = HS ( 1) 2) Observations of Snowmelt Amount in the Watershed The snowmelt amounts were also meas- ured at about twenty points in the wa- tershed (Fig. 1). The height of snow depth (HS) and snow density ( ) between 0-5 and 5-lOcm below the snow surface were measured temporarily, so that the snowmelt amount between the measuring period was calculated as follows, m = fl(HS) ( 2) 298 where a(HS) means the height of lowering snow depth during the measuring period. 3) Observations of Melting Rate at the Lower Station (BH) Melting rates were continuously ob- served by four different methods at BH station: (a) snow stake method, (b) lysimeter method, (c) snow pit observa- tions, (d) heat balance calculations. Details were described as follows. (a) Snow stake method: Height of snow depth (HS) and snow density between 0-5 and 5-lOcm below the snow surface ( ) were obssrved at 0900 and 1700 every day, so that the snowmelt amount (m) could be calculated by the equation m= (HS) , where (HS) means the height of the lowering snow depth during the period of 0900-1700 or 1700-0900. (b) Lysimeter method: Amounts of runoff water from the snowpack were measured using two types of shallow lysimeters, one buried in the ground , 90 x 90cm area, and the other buried in the snow, 80 x 80cm area. Collected meltwater was measured by tipping bucket rain gauge. (c) Snow pit observations: Snow pit observations were done once or twice a month. The snow density and water con- tent profiles were obtained. (d) Heat balance calculations: Neglect- ing the conductive heat flux through the snow cover during the snow-melting sea- son, the heat energy balance equation at the snow surface is written as follows, QT = QR + QA + QE, ( 3) where QR is the net radiation, QA is the sensible heat flux and QE is the latent heat flux. The values of QT usually become negative values at night when the meltwater of the surface layer refreezes and the snow temperature decreases below 0 C, but this negative values of QT are compensated by the positive values of QT the next morning. Each component is obtained as follows, QR: direct meas- urement by net pyrradiometer, QA and QE: calculation by bulk formulae obtained for this region (Ishikawa et al. 1982) QA=0.0096(Tj-To)VI MJ/m2h and QE=0.02(E,-Eo)V MJ/m:lh, where TI(°C), E (mb), and Vr (m/s) are air temperature, vapour pressure and wind speed at 1m height above snow surface, and T ( C) and E (mb) are surface snow temperature and surface vapour pressure, respec- tively. 4) Stream Discharge Stream discharge at the outlet of the watershed (P4, Fig. 1) was measured with a self-recording stage recorder. 5) Other Observations Precipitation was measured by tip- ping bucket rain gauge at BH station. WATER BALANCE DURING THE SNOWMELTING SEASON The water balance was calculated from the beginning day of the snow survey to the day of complete disappearance of snow from the watershed. 1) Water Balance Equation The water balance for the snowmelting season at the watershed is written as: tM + ~R = LQ + ~E + LET + Asd, (4) where M is the total amount of snow- melt, which equals the total amount of snow water equivalent at the beginning day of the water balance, LR is the total precipitation, [Q is the total amount of runoff depth, AE is the mass transfer at the snow surface (evapora- tion +, condensation-), ET· is the total evapotranspiration in the water- shed, ~Sd is the change of groundwater storage (gain+, loss-). 2) Each Component The snow water equivalents (HW) ob- tained at about twenty points on the watershed are plotted against the alti- tude (z) in Fig. 2. The linear relation- 299 1982 11983 11984 150 I em-water "' ;:.-.----• 11-13 Apr. 1982 ~.----• 0 ----'i- 150 "'~~ -"i 50 150 50 50 300 400 500 600 Altitude (m a.s.l.) Figure 2. Altitude distribution of snow water equivalents in 1982, 83, 84. The linear relationship between the altitude and snow water equivalent was classified by four sub-watersheds, ex- cept in 1982. (open circle: west, open triangle: north, cross: east, solid square: lower region). ship between them can be classified by four sub-watersheds ( West, North, East and Lower regions) except in 1982, a heavy snow season. The altitude dis- tribution of the sub-watershed area A(z) Figure 3. Altitude distribution of sub-watershed area. Square of oblique line indicates the area of 0.5km (W: west, N: north, E: east, L: lower re- gion). Table 1. Water precipitation, ET: estimated balance (I:M + l:R =l:Q +6E +YET+ 6Sd, where M: snowmelt, R: Q: runoff, E: evaporation or condensation at the snow surface, evapotranspiration, sd: estimated groundwater storage) !:.M rR 13 Apr.-15 Jun. 1982 106 14 13 Apr.-30 May 1983 60 13 14 Apr.-5 Jun. 1984 67 3 is shown in Fig. 3. The total snow water equivalent can be calculated as follows, HW = fHw(z)A(z)dz ( 5) The total snow water equivalent in watershed (including rain water) 1.34x10 m -water (120cm-water for the was areal mean) in 1982, a heavy snow season, but it was 7.6x10 m -water (68cm-water) in 1983 and 7.8xl0 m -water (70cm-water) in ~~0 ~7:1' rf.. ] fil ·. .. .. ·. :.0 I ·· ... I o-~~o Apnl lo4ay Jun• 1982 IO m-'soc' (b) o.l'---t;;-"-~""'""'-~"-4<~;..::...:~~~Sf.I.J:A~1 ,.-0 ....l June Figure 4. Separations of snowmelt run- off from the runoff hydrograph, indi- cated by the areas of oblique line part. fQ llO 68 53 _6E I_ET hSd LQ/ ([M+LR) 1 2 7 0.92 0 2 3 0.93 -1 2 16 0.76 (em-water) 1984,which were seasons of little snow (Table 1). The values of R, E and ET were as- sumed to b~ the same in all places. The amount of R was observed at BH station continuously. The value of E can be calculated by the empirical formula using the meteorological data at BH station, E=QE/Ls, where QE is the latent heat flux and Ls is the heat of sublima- tion. The amount of ET was assumed to be 1mm/day (Arai, 1980) after the day of disappearance of snow at BH station. Total amount of snowmelt runoff was calculated by two methods. In one method that the runoff amount was accumulated during the snowmelting season, and in the other the snowmelt runoff was sepa- rated from the hydrograph, which was indicated the area of oblique line part in Fig. 4. 3) Results The values of water balance compo- nents obtained are represented in Tables 1 and 2. Results shows that 80 -100% of the snowcover in the watershed ran off in the three years with heavy and little snow (Fig. 5). one characteristic Table 2. Runoff coefficient of the snowcover in the watershed. 81, 82, Q' are shown in Fig. 4. Q: snowmelt runoff. I: snowmelt(M) + rain(R). 81 82 Q' Q=Q'+S2-S1 I=M+R Q/I 1982 1 2 llO ll1(1) 120 0.93 1983 3 5 62 64(2) 68 0.94 1984 1 4 53 56(3) 70 0.80 (em-water) ( 1) 13 Apr.-15 Jun. ( 2) 13 Apr.-17 May ( 3) 14 Apr.-4 Jun. 300 100 50 1(1982 Q(1982) 1(198~)_____ 1(1984) ~-----;::-_-:_ .... ~ -~:~:::=cit,~;;)-~r--- / / / 30 10 20 30 April May Q(1984) 10 20 Jun• Figure 5. Accumulated runoff depth (Q) and accumulated precipitation in addi- tion to areal mean snow water equivalent on 13 April (I). (Solid line: 1982, thick broken line: 1983, thin broken line: 1984). of the water balance during the snowmel- ting season was that the amount of water loss from the watershed (evapotranspira- tion etc.) was less than that in other seasons. The total amount of mass transfer at the snow surface was found to be less than 1% of the input (amount of snowmelt + rain) to the watershed during the snowmelting season. The loss values of snowmelt energy by evaporation during the daytime were usually compensated for by the gain values achieved by condensa- tion during the night. CHARACTERISTICS OF WATER BALANCE DURING THE MIDDLE OF THE SNOWMELTING SEASON The water balance at the watershed is written as, 300 W1(578) 1982 h40SHI RI (o.)WEST RIDGE 300 1982 MOSHIRI ( b)EA.ST RIDGE E1(565) o 10 15 ro 2s Jo s April Figure 1982. cate 6. Variations of snow depth in The figures in parentheses indi- the altitude. Observational sites are shown in Fig. 1. rM + L:R =I:Q +.!lE +Asd +Asd 1 , (6) where LM, plained in change of snowpack. t,R, I:Q, /j,E and ASd are ex- section 3.1. ~Sd 1 is the liquid water storage in a The amount of snowmelt in a watershed was obtained by the snow stake method (m=A(HS)•fl, Eq. 2) along with the altitu- dinal distribution of watershed area (Fig. 3). Examples of time variation of snow depth are shown in Fig. 6. The altitudinal difference between the top and the bottom of the watershed was only 300m, but the difference of snow depth became more than 1m. The values of water balance compo- nents obtained are represented in Table Table 3. Water balance during the middle of snowmelting season. (LM + rR = LQ + ~E +~Sd + ~Sd 1 , where M: snowmelt, R: rain, Q: runoff, E: evaporation or condensation at the snow surface, Asd: change of groundwater storage, 11 Sd I: change of liquid water storage in a snowpack) I:M LR I:Q /1E /:lSd+/j,sd I I:Q/ (I_M+I:R) 24 Apr.-29 Apr. 1982 12.1 0.3 9.6 0.0 2.8 0.77 29 Apr.-10 May 1982 39.4 2.6 40.8 0.3 0.9 0.97 16 Apr.-19 Apr. 1983 11.2 0.3 11.4 -0.2 0.3 0.99 (em-water) 301 3. In the middle of the snow-melting season, melting rates were 20-40mm- water/day, the accumulated snowmelt wa- ter equaled the accumulated snowmelt runoff for several days, without a change of groundwater storage in the watershed. RUNOFF ANALYSIS OF DAILY AMOUNT We devised a runoff model for the snowmelting season in 1982 and tested the propriety of this model during the same season in 1981 and 1983. 1) Determination of the Daily Amount of Snowmelt The daily snowmelt amounts obtained by four different methods at BH station showed different values from each other (over 10-20% differences, Fig. 7). The 50 -40 Gl i3o ~:: ,,-~"'-\_---w---'ti'-- \ i~ -~~:,./ 15 20 25 April 30 5 10 May 1982 Figure 7. Daily snowmelt obtained by four different methods. (solid line: snow stake method, broken line:lysimeter method, horizontal thick solid line: snow pit observations, dotted line: heat balance calculations). horizontal profiles of snow structure or ice layer were varied from one observa- tional place to another, in spite of the smallness of the experimental site. But the snowmelt amounts that accumulated for several days did not vary from each other. We obtained accurate daily snow- melt values by comparing the results 302 obtained from different Input water for the runoff is decided on the basis of the methods. analysis snowmelt amounts at BH station. Usually the snow- melt amount at BH station was nearly equal to the areal mean value of the watershed. But in 1982, the snowmelting rate of the upper region was larger than that of the lower region for a long period, so that the input water was 1.1 times as large as the snowmelting at BH station from 12 April to 12 May in 1982 (Fig. 8). mm-day1 40 20 30 10 20 May --observed runoff snowmelt • rain I t\ Figure 8. Areal mean daily snowmelt (added daily precipitation, broken line) and daily runoff depth (solid line) in 1982. 2) Determination of the Runoff Model Sugawara's "tank model" was used here to calculate the runoff during the snow- melting season (Sugawara et al., 1984). A series of two storage type tanks was adopted, in which the upper tank had three outlets, two on the side for runoff and one on the bottom for infiltration, and the lower tank consisted of two outlets, one on the side and the other on the bottom (Fig. 9). The daily input water (snowmelt + rain) supplied to the TANK MODEL l [Q2sf l 90mm ~04~ 10.01 Figure 9. Derived tank model. Coeffi- cients of tank's outlets are in units of day upper tank every day and runoff water discharged from the outlets on the side. The distinguished attenuation constants of the recession curves in the runoff hydrograph were applicable to the runoff coefficients of tank's outlets. The runoff Q(t) at time t is expressed as, ( 7) where Qo is the runoff at time 0, 1/~ is a time constant (relaxation time). Other parameters of the tank structure were decided by the trial and error method until the calculated discharge fit well the observed one for the case of 1982. 3) Remnant Area of Snowcover and the Evapotranspiration in the Late Snow- melting Season After the snowcover disappeared in the lower region, the remnant area of snowcover decreased. The decreasing rate of snowcover area can be modeled as shown in Fig. 10, which is based on the characteristics of recession curves of the runoff hydrograph and temporary snow line observations. The evapotranspira- tion was assumed to be lmm/d after the day of disappearance of snow at BH sta- tion. o.o 0 5 10 15 Days Elapsed Figure 10. Remnant area of snowcover in the watershed (abscissa: elapsed time after the day of snowcover disappeared in the lower region, ordinate: ratio of remnant area of snowcover, as, to water- shed area, A). 4) Results The observed and calculated hydro- graphs in 1982 are plotted in Fig. 11, where the solid line indicates the 303 mm·day-1 --Observed --~-·Calculated 0~~~~~3~0--~10~~2~0~~~1~0~2~0==~ May June 1982 :=;40 0 c ti20 mm-day1 mm-day-1 10 20 April --Observed ----Calculated t y ,._ 'v 30 10 20 30 May 1981 30 10 20 30 May 1983 Figure 11. Daily runoff depth in 1982 and 81,83. (solid line: observed runoff depth, broken line: calculated runoff depth by using tank model in Fig. 9). observed hydrograph and the broken line is the calculated one. Without changing the tank structure, the simulations of daily snowmelt runoff were successful in both 1981 and 1983. We considered an analogy between the tank structure and the runoff process as follows: The upper tank has two runoff outlets and the lower tank has one. Runoff coefficients of these outlets were decided by the recession curves of the hydrograph at various ranges of runoff amounts. It seemed that the runoff fro1u the lower tank corresponded to the groundwater flow from the deep level, and the runoff from the upper tank was the groundwater flow from the shallow level. CONCLUDING REMARKS We measured the amount of snow water equivalent, the snowmelting rate and the stream discharge at a watershed during four snowmelting seasons. Water balance and runoff analysis were made on the basis of observed hydrological data. The following results were obtained. 1. 90 % of the snowcover in the watershed ran off in the three years with heavy and light snow. 2. The total amount of mass transfer at the snow-surface was estimated to be less than 1% of the input (amount of snowmelt + rain) to the watershed during the snowmelting season. 3. In the middle of the snowmelting season, melting rates were 20 -40 mm- water/day, the accumulated snowmelt wa- ter equaled the accumulated snowmelt runoff for several days, without a change of groundwater storage in the watershed. 4. A tank model was used to calculate the runoff process. The simulation of daily snowmelt runoff was successful for the three year period without changing the tank structure. ACKNOWLEDGMENTS The authors are grateful to the staff members of the Uryu Experimental Forest of Hokkaido University for their help in these observations, and they also wish to thank Dr. N. Ishikawa of the Insti- tute of Low Temperature Science for his helpful advice and useful suggestions. REFERENCES Arai, T., 1976, Nippon no mizu. Sansei- do, Tokyo (in Japanese). Ishikawa, N., S. Kobayashi and K. Kojima, 1982, Measurement of sensible heat flux in the snow-melting season I. Low Temperature Science, Ser.A, 41: 109-116 (in Japanese with English summary). 304 Leaf, c. F., 1971, Areal snow cover and disposition of snowmelt runoff in central Colorado. USDA For. Serv. Res. Pap. RM-66, Rocky Mt. For. and Range Exp. Sta., Fort Collins, Colo. 19pp. Ono, S and R. Kawaguchi, 1974, Syoryuiki ni okeru choseturyo suitei to 'sinrin no eikyo' no kaiseki. Ringyo Sikenjo Touhoku Siba. Nenpou: 114-119 (in Japanese). Rango, A., 1983, Application of a simple snowmelt-runoff model to large river basins. 51st Annual Meeting Western Snow Conference, Vancouver, Washing- ton: 89-99. Sugawara, M., I. Watanabe, E. Ozaki and Y. Katsuyama, 1984, Tank model with snow component. Research Notes of the National Research Center for Disaster Prevention, 65: 1-293. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 ESTIMATIONS OF SNOWMELTING RATE IN A SMALL EXPERIMENTAL SITE N. Ishikawa, H. Motoyama and K. Kojima1 ABSTRACT: Estimations of hourly and daily snowmelt were achieved by using the heat balance method and bulk mete- orological parameters at a small experi- mental site. The results were compared with actual snowmelt and runoff amounts. Calculated snowmelt by use of the heat balance method explained variations of actual snowmelt for short-and long-term periods. Hourly variations of snowmelt could not be predicted by using simple formulae. However, daily variations could be explained by the daily mean and maximum air temperatures, in which lower limit of snowmelt occurrence was ob- tained. During the intensive snowmelt season the predominant percolation velo- cities of snowmelt water were measured to be 0.25-0.45 em/min, which implies that a large lag time of 3-4 hours was required for melting water to reach the bottom of a 60-80 em snow cover. The lag time depended on the melt intensity of the surface and the depth of snow. (KEY TERMS: heat balance; degree-day factor; discharge; percolation veloci- ty.) INTRODUCTION The estimation of snowmelt amount in a whole watershed is of increasing im- portance in forecasting snowmelt floods, as well as providing for the efficient use of snow as a water resource in a mountainous region. It is logistically difficult to obtain snowmelt in a wa- tershed of wide area, therefore many estimating techniques have been devel- oped (Motoyama, et al., 1983, Granger and Male, 1978, Price, et al., 1976). Net radiation at a melting snow surface plays the most important role on the snowmelt process (Kojima, et al., 1971), while the sensible heat and the latent heat are smaller in magnitude than the radiation except in very warm and wet conditions (Takahashi, et al., 1981). However, for the sake of sim- plicity many empirical formulae using only air temperature have been proposed for estimation of snowmelt (Yamaguchi, 1971, Oura, et al., 1969, Ishii, 1959). Appropriate empirical equations have still not been established (Kojima, et al., 1983). In this paper the relation- ships between snowmelt and bulk mete- orological parameters have been exam- ined. The estimated snowmelt was com- pared with actual snowmelt and discharge at the small experimental site. These methods were tested in advance of appli- cation to the whole watershed. INSTRUMENTATION Measurements of surface snowmelt, runoff and meteorological parameters were carried out at the small experimen- tal site (about 400 ml area) during the intensive period of snowmelt from 1980 to 1985. The site is located at the 1rnstitute of Low Temperature Science, Hokkaido University, N-19, W-8, Sapporo, 060, Japan. 305 lowest part of Bifukazawa River water- shed (about 11.2 km~ in area) in Uryu- River in northern part of Hokkaido, Japan (45°N, 140°E). The topography of the area is surrounded by mountain ridges with an altitude difference of about 250 m. The general winter climate is characterized by very cold air tempe- rature (the minimum air temperature is below -35 °C) and well developed snow pack (the maximum snow depth is over 200 em). The snowmelt season lasts from the middle of April to the middle of May. 1) Measurements of surface snowmelt In order to obtain the short term (hourly) variations of snowmelt the following parameters were measured every one or two hours a day during the snow- melting season in 1985: Surface depres- sion (4h), liquid water contents (W) and snow density (f). The liquid water con- tent of snow was measured by using the calorimeters designed by Akitaya (Akitaya, 1978). Snow stake readings were done twice a day (09h, 17h) to determine the long term (daily) varia- tions of snowmelt during six different snowmelt seasons from 1980 to 1985. 2) Measurement of heat balance components Micro-meteorological observations were carried out at the site to derive the heat balance components at the snow surface. Observations were taken as follow: dry and wet bulb air temperature (ventilated platinum resistance thermo- meters), wind speed(an ultra sonic ane- mometer, three-cup anemometers and a wind vane), dew point tempera- ture(Lithium chloride dew points me- ters), sensible heat flux (ultra sonic anemo-thermometers), global and re- flected solar radiation (pyranometers), net radiation (a net radiometer), sur- face temperature (an infrared thermome- ter) and conductive heat flux in snow (heat flow transducers). All meteoro- logical data were measured at 20 sec intervals and averaged for 5 minutes by a data logger. 3) Measurement of discharge from snow cover. Discharge be obtained of the snowmelt water can by a snow lysimeter 306 (Herrmann 1978). In this investigation two shallow lysimeters (size was 90 em x 90 em) were used, one of which was buried just below the ground surface, and the other in the snow. Meltwater was collected in rain gauges and re- corded automatically. 4) Measurement of the percolation velocity of meltwater through the snowcover. Observation of the rate of downward movement of meltwater through snow cover was made by using a dye-stuff (eosine powder). A small amount of the dye was spread on the melting snow surface, which dissolved instantaneously and per- colated into the snow cover. Even if the dyed portion of the snow surface was limited, it was covered with a certain amount of snow for eliminating albedo changes. Measurement of the length of dyed depth every 0.5-1 hour gave the percolation velocity of meltwater. OBSERVATIONAL RESULTS 1) The relationship of snowmelt with heat balance The amount of water content in snow W), snow density <f), and surface de- pression ( A h) were measured at one or two hour intervals. The density of snow just below the snow surface (0-5 ern) gave the nearly constant value of 450 kg/ml during the observation period. l<g/ml ;-""'~ .,. ~i i f i i i 20 200 H--y;--t·-~--~ : 0 :o 09h 18h Apr.18 l<g/m3 ~·'· ~~!' !/:! !: 2ob~·hwt-1-t""i o l ! i 09h Apr. 22 ,~ho Figure 1. .,. ft...---.r--,::--.,_,__,_, 4uu ' ~ j 20 I L. ... L _____ i···] w 200 : w : 10 Oi 09h Apr. 19 400'' ,, :--,:-r-1. ..,.......~: ~~ r: . 201.,.-.rw··H-~~ : : 1 ! : 09h Apr.23 18h 20 w 20o·~·-···H·vn 10 09h ""· 20 Variations of liquid water content and densities of snow cover. (18-25, April, 1985, Moshiri) The liquid water content near the sur- face changed 9-15 % according to the time and weather condition (Fig.1). Using these values, surface snowmelt (M) was obtained directly ( M = ( 1-W)4h·f ) , and hourly snowmelt (~Mh ) at different weather conditions are shown in Fig. 2. Figure stake. AMh mm/hr mm 3.0 I; AM 12h 15h Apr. 18 2. Hourly snowmelt measured by (18-25, April, 1985, Moshir~) Hereafter the method by stake measure- ment is called the direct method. The maximum value of snowmelt occurred near noon and ranged from 3.3 rnrn/hr on a clear day (18 Apr.) to 1.5 rnrn/hr on a cloudy day (23 Apr.). The total daily snowmelt (g6Mh) was about 22 mm on the clear day and 9 rnrn on the cloudy day. 1.0 Q5 OR 20 10 O.QoL19uh.....J.-,-1.2h ............... """,s,...h..J--~,sh 0.009hi--L-J--'--'-....L,~2hJ....L__L..l'::!,s::-'"h...L.-L.""--';-;::,ah Apr. 18 Apr. 23 Figure 3. Hourly variation of each heat balance component. QR: net radiation amount QA: sensible heat flux QE: latent heat flux of condensation or evaporation of water vapor (18-25, April, 1985, Moshiri) 307 Heat balance components were obtained during the same intervals when the stake measurement were carried out. Their time variation appears in Fig. 3. For the procedure radiation was measured by a net radiometer and sensible heat flux was measured thermometer. by an ultra sonic anemo- Latent heat flux for con- densation or evaporation of water vapor was estimated by the empirical equation which was established for the experimen- tal site before (Motoyama, et al., 1983). The sign of each component means that the positive and the negative show the flux towards and away from the snow surface, respectively. The main compo- nent of heat balance is due to net ra- diation (QR) (80-95 % during the inten- sive snowmelting season). Next in im- portance is the sensible heat flux (QA), while the latent heat flux of condensa- tion or evaporation (Qe) is small in magnitude and does not contribute much to the snowmelt as shown in figure 3. A previous investigation (Kojima, 1979) showed that the conductive heat flux in snow cover and the heat from rainfall were usually very small in the season, therefore snowmelt (M) can by estimated by the following heat balance equation: M = QR + QA + QE, where the positive sign of M shows snowmelt, and the negative sign indi- cates refreezing of meltwater in the snowcover. Figure 4 shows the compari- son of hourly snowmelt obtained by stake measurement and the heat balance method. Their agreement is fairly good, but during the morning of clear days, values from stakes are lower than from the heat balance determination. With clear weather conditions, radiative cooling occurs at night, and the liquid water content near the surface tends to re- freeze. Thus the so called ice crust layer forms. Even if heat balance be- comes positive in the early morning, surface snowmelt does not occur before the crust layer warms up to o•c. Neg- lecting the conductive heat flux in eq.(1) might be one of causes of this discrepancy. While the heat balance method is pointed out as a suitable technique for estimating snowmelt, it is too compli- cated to use. Therefore, we consider a practical technique by using empirical formulae requiring only bulk meteoro- logical parameters. 2.0 MJ/m?hr 2.0 MJ/m?hr H 1.0 H Apr. 18 17h QO 12h Apr. 20 17h 2.0 MJ/m?hr 2.0 MJ/m?hr 1.0 ~ 1.0 . . H . ~ ·-.............. ··s· ~. ........ ............. O.Oil__.L_-,!12.,--h _L___L---'-----'-----,-;!17h O.OIL---'---~12,.-h _L_----'----'----'----;;!17h Apr: 22 Apr. 23 Figure 4. Comparison estimated by heat balance by stake method. (18-25, Moshiri). H: heat balance method S: stake method of snowmelt and obtained April, 1985, 2) The relationship of snowmelt with bulk meteorological parameters i) Hourly snowmelt Many investigators reported empirical formulae requiring only air temperature, but the most suitable formula, or melting coefficient termed "degree day factor" has not been established yet. In this investigation hourly values of snowmelt and meteorological parameters were accumulated during daily snow- melting period of a day. Figure 5 shows relationships between accumulated values of hourly snowmelt and bulk meteorologi- cal parameters (hourly mean air tempera- ture (Ta) and hourly net radiation), which appeared during each day. The relationships between snowmelt and net radiation were expressed by empirical equations with high correlation coeffi- cients. As mentioned before net radia- tion plays the most important role of the heat balance during the snowmelt season at the experimental site. The empirical constants (degree-hour factor) obtained on each day are similar for the entire observation period. From this figure it can be seen that the estima- tion of such short term snowmelt (hour- 308 ly) might be calculated by using only net radiation amounts. Correlation be- tween accumulated values of hourly snow- melt and hourly air temperature are high, however linear regression lines change with individual days. Therefore, it is not be possible to estimate hourly snowmelt by using only air temperature. 1.S L'AM em em Apr. 1.0 _:• • 18 ·22 1.0 X 19 • 23 X X • l. '20 -25 as -X . 0.5 •' . !X ' . A ·c·h L:'Ta 0.00 40 80 Figure 5. Relations of accumulated hourly snowmelt with bulk meteorological parameters. a) net radiation b) hourly mean air temperature (18-25, April, 1985, Moshiri) ii) Daily snowmelt Figure 6 shows the relationship of daily snowmelt (AM) with the daily mean (Td) and the daily maximum (Tmax) air temperatures, respectively. The re- gression lines are expressed by the following equations: AM= -4.1 + 3.0 Tmax A M= 4 • 1 • ( 3 • 0 + Td AM mm/d 40 20 Tmax. •c o~~-L~~~~~~~~ 0 5 10 15 Daily max. air temp. AM mm/d 40 .... 20 ... . •• •. • • ••• •11M = 4.1( 3.0 • Td) . . .. .. Td ·c 5 10 mean air temp. Figure 6. Relations of daily snowmelt with the air temperature. a) daily maximum air temperature (1980, Moshiri) b) daily mean air temperature (1980-1982, Moshiri) These equations illustrate that the daily maximum air temperature Tmax=l.4•c might be an approximate criterion of apparent snowmelt. Furthermore,, daily snowmelt occurs when the daily mean air temperature Td is higher than -3•c. iii) Estimation of long-term snowmelt Estimation of long-term snowmelt was done using bulk meteorological parame- ters. Unfortunately, no observational data of snowmelt exist, only depression of snow surface for long periods. Snow density near the surface was almost constant during the intensive snowmelt seasons. Therefore, depression of the surface were adopted instead of the snowmelt data. Figure 7 shows regres- sion curves between accumulated surface depression and bulk meteorological para- meters, where daily mean (Td) and the daily maximum (Tmax) of air temperature and daily amounts of net radiation (NR) were selected as bulk meteorological parameters. These values were accumu- lated for 10-14 days just before the day 100 mm (a) !100 mm (b) I C6h "Cday I 100! I 50 100 mm ll6h 50 MJ/m2 100 MJ/m2 100 Figure 7. Relationships of accumulated daily surface depression with accumu- lated bulk meteorological parameters (from 1983(III) to 1985(I)). a) accumulated daily mean air temperature b) accumulated daily maximum air temperature c) accumulated daily net radiation solid lines: means for three years 309 when snowcover melted away from the experimental site, summarizing three different years (1983-1985). In compari- son with short-term snowmelt, better relations appeared between bulk meteoro- logical parameters and snowmelt. The relationships between accumulated sur- face depression (J!Ah) and the net radia- tion for three years are shown by broken lines. The mean regression curve which is shown by the solid line can be expre- ssed by l!Ah = 0.93 l:NR + 0,98, r=0.990 It is clearly shown that regression curves obtained at different years fit- ted well with the mean. Linear rela- tions were obtained among accumulated amounts of surface depression (~Ah), daily mean air temperature (&Td), and daily maximum temperature (fTmax). It is noted that the empirical coefficients can change slightly each year. Mean regression curves for three years (solid lines) can be expressed by J::'Ah = 1. 38·£ Td + 1. 24 (t=O. 980) l:4h = 0.62·CTmax -2.51 (t=0.975) From these equations the amount of long- term snowmelt can be estimated by using bulk meteorological parameters. 3) Comparison of discharge with the surface snowmelt An accurate method is still not available for measuring snowmelt con- tinuously, therefore, discharges from a snow cover are used as an indicator of snowmelt in this paper. Figure 8 shows the variation of net radiation (NR) and discharge measured by a lysimeter (GL). Rainfall for every 30 minutes (P) are plotted on the same figure. Weather conditions during the observational pe- riod were mostly cloudy except April 22nd which was clear: Rain-fall on April 23rd was 0,5 mm, and on April 24 18 mm. Large discharge appeared during 24th of April in contrast to net radiation. Heat amounts of snowmelt due to rain- fall (Qp) can be estimated by the equation Qp= 70·Ta·P/~t (W/m 2 ), where Ta is the air temperature and P is the precipitation amount (mm) for At period (min). The mean air temperature was 2.7•c, and the maximum precipitation for 30 min was 2.3 mm at April 24th. It means that the heat of 14.5 W/mz re- leased from rainfall, which did not contribute much to snowmelt as mentioned before. Therefore, it can be said that most of discharge was due to the pre- cipitation and not sno~aelt. Large time differences are seen between the varia- tions of net radiation and discharge. The lag time between the initial raise is about 4-5 hours, but the peak time difference is about 1-1.5 hours. The lag time is considered as the trav- erse time of meltwater from the surface to the bottom of snowcover. The former depends on the depth of snow cover and the intensity of surface snowmelt (Kojima, 1984). NR400 Wtm2 200 NR 400 WJm2 200 Apr. 22 Apr, 24 p 2.01 mm 1.0 0 Apr. 23 Apr. 25 Figure 8. Hourly variations of net radiation (NR), discharge amounts (GL), rainfall (P). (22-25, April, 1985, Moshiri) Some investigators have described the ~ovement of snowmelt water through snow cover previously. In order to determine the percolation velocity of melt water, I·Jakahama, et al. ( 1968) measured time variations of liquid water content in a snowcover, and Fujino (1968) measured the time change of snow electro-con- ductivities after spreading some markers over the snow surface. Theoretical analyses of meltwater movement was done by Yoshida (1965) and Colbeck (1972). Wankiewicz (1979) reviewed the process of meltwater percolation in detail. From their works it can be said that the percolation velocity of snowmelt water depends on many factors: snow density, liquid water content, surface snowmelt rate and snow stratigraphy. In this investigation two kinds of dyes (eosine 310 and waterblue) were used to determine percolation velocity. After spreading small amounts of dye over the surface or inside the snowcover, the length of the colored depth was measured by snow pit observations after a specific time in- terval. Figure 9 shows the variation of percolation velocity of meltwater at specific depths (O.lm, 0.3m, 0.5m) and isopleths. Under clear conditions the percolation velocity at the upper part of the snowcover was much larger than in the lower part, and the maximum velocity appeared around noon. During cloudy conditions the velocity was almost con- stant at each depth. In this investiga- tion the predominant percolation veloci- ty of 0.25-0.45 em/min was obtained. This means that it takes 3-4 hours for the surface meltwater to reach the bot- tom of a snow cover of a 60-80 em depth (Fig. 8). Therefore, such a lag time effect should be corrected when com- paring the surface snowmelt and dis- charge, especially for a shorter time durations. Snowmelt obtained by three methods are compared with each other in Fig. 10. 1.0 an/min 1.0 em/min Z' -o-San -~ -•-30em ~ _.,_ 60em 6 0.5 0.5 ~ ~ .. -JI--·--fit--Jt- a. QO.___.___,...__.L.._....L.---'--.l__J 12h Time 16h uo~~~,2~h~~--~,~sh~ Time Apr. 19 ~:1 .,.o.62 ~~~~ 'h~s ~50 o~~o.s~s))"fl23 -o ~0.34 ~ -0.27--0.25~ 5, -an/min .12 OL-~~-L~--L_~_J 12h Time 16h em 50 Apr. 23 o.Jf ..... o.ss--o.3e -o.40 :::> Q37 Q30 ~ 0.30.D.3!V ~Q21~ =--010 "®? cm/~.1~--9.12 _:.-- 0'--~~~'--~-L~~ 12h Time 16h Figure 9. Percolation velocity of surface snowmelt. upper: time variation of the velocity at 5 em, 30 em and 60 em below the snow surface lower: isopleths of the velocity left: fine weather, right: overcast (April, 1985, Moshiri) The abscissa shows the estimated snow- melt by using the heat balance, and the ordinate shows the snowmelt from the stake reading (direct method), and dis- charge. These values are during 8 days just before complete snowmelt. accumulated the day of Accumulated daily snowmelt Estimated snowmelt by heat balance Comparison with actual Figure 10. snowmelt discharge. (18-25, April, of estimated snowmelt and 1985, Moshiri) The estimated snowmelt by heat balance method coincided well with actual snow- melt from the stake reading and dis- charge measurements. It shows that the discharge from the bottom of snowcover can be taken as the surface snowmelt at the intensive snowmelt periods if it takes for a longer period of time,such as 24 hours. CONCLUSION Surface snowmelt during intensive snowmelting periods was estimated by using the heat balance method and bulk meteorological parameters. These were compared with the actual snowmelt and discharge at the small experimental site. The results are as follows: 1) Estimated snowmelt by heat balance or a simple formula using net fitted well with the actual radiation snowmelt. The relationship between calculated and observed snowmelt, accumulated for 8 ~ys, could be represented by linear regression curve with a high correlation coefficient ( r=O. 990) • 2) Short term variations of snowmelt 311 could not be explained by simple formu- lae using air temperature only. How- ever, the daily snowmelt during inten- sive melt periods were explained by the daily mean and daily maximum air temper- atures, respectively. Furthermore, long term snowmelt, accumulated for 10-14 days up to the day of snow disappear- ance, correlated well with the accumu- lated daily air temperature. From the work above the approximate criterion of apparent daily snowmelt occurrence is 1.4•c for daily maximum air temperature, and -3"C for daily mean air temperature. 3) Hourly snowmelt estimated by the heat balance method and simple equations were compared with snowmelt discharge from the bottom of snow cover. The lag times ranged from 3 to 4 hours, but they de- pended on intensities of snowmelt, and the depth of snow cover. For long pe- riods such as 10-14 days good relations were found between estimated snowmelt and daily discharge, especially for accumulated ones. ACKNOWLEDGMENTS Authors would like to express their sincerely thanks to Moshiri Branch of Uryu Experimental Forest, Hokkaido Uni- versity for their logistic support to the investigations, and to Dr.K.Fujino, Institute of Low Temperature Science for his useful comments on the measurement of the percolation velocity of melt water. They also thank Misses A.Ohmori and N.Ochiai for the preparation of the manuscript. REFERENCES Akitaya, E., 1978. Measurements of Free Water Content of Wet Snow by Calori- metric Method. Low Temp. Sci. 36:103- 111. Colbeck, S.C., 1972. A theory of Water Percolation in Snow. J. Glacial. 11:369-385. Fujino, Down Cover. Granger, K., 1968. Measurement of Flow Speed of Melt Water in Snow Low Temp. Sci. 26: 87-100. R. J., and D. H. Male, 1978. Melting of a Prairie Snowpack. J. Appl. Meteor. 17: 1833-1842. Herrmann, A., 1978. A Recording Snow Lysimeter. J. Glacial. 20: 209-213. Ishii, Y., 1959. Studies on Snow Melting. Fundamental Research of Snow Cover. Hokkaido Electric Power Co. and Sapporo District Meteor. Observatory: 1-84. Kojima, R. s. K., D. Kobayashi, H. Aburakawa, Naruse, K. Ishimoto, N. Ishikawa, Takahashi, 1971. Studies of Snow Melt, Runoff, and Heat Balance in a Small Drainage Area in Moshiri, Hokkaido.II. Low Temp. Sci. 29: 159- 176. Kojima, K. 1979. Mechanism of Snow Melting and Heat Budget. Note Meteor. Res. 139: 1-33. Kojima, K., H. Motoyama, andY. Yamada, 1983. Estimation of Melting Rate of Snow by Simple Formulae Using only Air Temperature. Low Temp. Sci. 42: 101-110. Kojima, K., and H. Motoyama, 1983. Time Lag of Meltwater Percolation through a Snow Cover. Low Temp. Sci. 43: 181- 194. Motoyama, H., D. Kobayashi, and K. Kojima, 1983. Water Balance at a Small Watershed during the Snowmelt Season I. Low Temp. Sci. 42: 123-133. Oura, H., K. Kojima, D. Kobayashi, R. Naruse, and N. Ishikawa, 1969. A Study of Snow Melt in Ikutora. Low Temp. Sci. 27: 143-162. Price, A. J., T. Dunne, and S. C. Colbeck, 1976. Energy Balance and Runoff from a Subarctic Snowpack. CRREL Rept. 27: 1-29. Takahashi, S., A. Sato, and R. Naruse, 1981. A Study of Heat Balance on the Yukikabe Snow Patch in the Daisetsu Mountains. Seppyo, 43: 147-154. Wakahama, G., T. Nakamura, andY. Endo, 1968. Infiltration of Melt Water into Snow Cover II. Low Temp. Sci. 26: 53- 75. Wankiewicz, A., 1979. A Review of Water Movement in Snow. Proc. Modeling of Snow Cover Runoff. S.C.Colbeck and M.Ray (Editors). U.S.Army CRREL, Hanover, NH, 26-28. Sept. 1978: 222- 252. Yamaguchi, H., and S. Hasegawa, 1970. A Study on Snowmelt Runoff. Rep. Civil Eng. Res. Inst. 64: 1-174. 312 Yoshida, z., 1965. Infiltration of Melt Water in Snow Cover. Low Temp. Sci., 23: 1-16. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 A METHODOLOGY FOR ESTIMATING DESIGN PEAK FLOWS FOR YUKON TERRITORY J. Richard Janowicz 1 ABSTRACT: A sbnple methodology based on a regional approach was sought for estimating peak flows for the design of hydraulic structures in Yukon Territory. This procedure is needed for the design of projects with relatively short design lives such as those associated with placer mining operations as well as more conventional hydraulic design situations. Because of its varied use a sbnple, readily transferable method was desired. Single station flood frequency analyses were carried out on 90 hydrometric stations with at least six years of record using a two parameter lognormal theoret- ical probability distribution which is believed to be the most appropriate for sparse data regions. Sbnple linear regression relationships were developed between maxim.nn annual instantaneous dis- charge at selected return periods and drainage area for two hydrologic regions. An attenpt was made to improve on these relationships through the inclusion of a storage index factor which accounts for the relative location of lake or swamp storage within a basin as well as the size of the storage. The inclusion of this parameter in the developed relationships was statistically significant. Multiple linear regression equations were developed for the Interior and Western Mountains hydrologic regions and the Territory as a whole. (KEY TERMS: regional approach; hydraulic design; placer mining operation; two parameter lognormal theoretical probability distribution; sparse data region; storage index factor.) INTRODUCTION Functional relationships were developed for estimating peak flows in the Yukon Terri tory. This information is needed for the safe and economical design of hydraulic structures such as bridges, culverts, dams, diversion channels and road embankments. Methods which are traditionally used in more populated southern areas are not applicable in many areas of Yukon Territory due to the relatively small hydrometeorological data base and unique hydrologic regime types associated with subarctic and arctic environments. There are 94 hydrometric stations within the Terri tory or one station for every 5100 square kilometers. Most of these are relatively recent; however, with 47 of the total established within the last ten years. The mean record length is 12.4 years. Of the total number of hydrometric stations, 20 are partial record stations operated primarily on small streams. Hydrometric station locations are noted in Figure 1. Relatively few stations exist in the northern part of the Territory; however, the available data indicates a differing hydrologic response between northern and southern regions. This is presumably due to the presence of underlying permafrost which becomes more dominant with increasing latitude. The boundary between the zone of continuous and discontinuous permafrost coincides approximately with 67° latitude as noted in Figure 1. Hydrometric station density decreases significantly north of the Ogilvie Mountains which coincides approximately with 65° latitude. The majority of the peak flow design estimates required annually in Yukon Territory are needed for placer mining operations. Of the 300 Yukon placer operations most are on small streams with drainage areas less than 150 square kilometers. Since hydraulic structures associated with the placer 1water Resources Division, Northern Affairs Program, Indian and Northern Affairs Canada, Whitehorse, Yukon Territory, Y1A 3V1. 313 e INDIAN AND NORTHeRN AFFAIRS CANADA & WATER SURVEY OF CANADA --HYDROLOGIC RESPONSE ZONE BOUNDARY 100 100 200 KILOMEUES Figure 1. Yukon Territory existing hydrometric stations and hydrologic zones. have relatively short (two years), safety and the economic impact mining industry design lives considerations of failure are problems. less than other design The objectives of this study were to develop a simple methodology for estimating peak design flows for all Yukon watersheds. A simple criteria was desired since the placer mining industry would be a major user with design applications primarily associated with short design life projects. Use of this methodology would not be limited to placer operations; however, and a reasonable degree of confidence in the technique and subsequent design estimates is required. Because of its anticipated varied application a lower or upper limit to drainage basin slze was not used as a criteria in selecting hydrometric stations for analyses. The application of the developed relationship is intended to be used for all hydraulic design situations for at least preliminary design purposes. If safety or economic considerations merit, and given a sufficient amount of data, other techniques should be investigated. The developed methodology is expected to provide estimates of flows for ungaged basins with a reasonable degree of accuracy and confidence throughout most of Yukon Territory. SETTING Streamflow in Yukon Territory is generally characterized by peak flows in spring or summer as a result of snowmelt and glacial inputs. Smaller systems are susceptible to peak flows in response to rainfall inputs while intermediate size basins may experience secondary rainfall peaks in addition to the freshet. Streamflow decreases to a minimum in March or April though groundwater inputs generally maintain winter flows in all but the smallest streams. Most streams are generally ice covered from late October to April. Two primary hydrologic regimes have been identified in the Terri tory south of the Ogilvie Mountains. For discussion purposes these are referred to as the Interior and Western Mountains hydrologic regions. In the southwestern area of the Territory, streams which drain the St. Elias and Coast Mountains experience a rapid increase in streamflow discharge during the early summer in response to snowmelt at lower elevations. Flows increase in magnitude to a peak in later summer due to glacial melt at higher elevations. The St. Elias and Coast Mountain drainages may be considered separate subregions of the Western Mountains hydrologic region. Streams draining the St. Elias Mountains have shorter response time than the Coast Mountains; therefore, exhibit a flashy response due to temperature and precipitation inputs during the summer. The Coast Mountains have smaller glacierized areas and longer stream lengths which in combination with the natural regulation 314 provided by the extensive lake systems tend to remove this variability. Streams which drain the remainder of the southern region are characterized primarily by a rapid rise in discharge in the spring due to snowmelt inputs. Peaks generally occur in June, after which time summer rainfall maintains high flows for a few weeks. Small basins throughout the southern Terri tory may experience rainfall peaks, while intermediate basins may experience secondary peaks due to summer rainfall. The available data indicates an increase in peak flows with increasing latitude north of the Ogilvie Mountains. This is presumably due to the increasing presence of underlying permafrost as it becomes more dominant. The dominance becomes complete within the direct Arctic drainage north of the British Mountains as indica ted by the recorded unit discharge which is the highest in the Territory. DEVELOPMENT OF REGIONAL RELATIONSHIPS Single Station Analysis Single station flood frequency an!ilyses were carried out using all hydrometric data in Yukon Terri tory with at least six years of record. Sample data was obtained from Water Survey of Canada (WSC) and Indian and Northern Affairs Canada ( INAC) computer files current to 1983. In addition selected peripheral stations in British Columbia and Alaska were used to supplement the Yukon data. Peak flow events produced by glacier outburst floods and ice jams were not included in the analysis. Records were not screened for mixed populations and outliers. A computer program initially developed by Environment Canada (Condie, Nix and Boone, 1976) and revised in 1981 was used for the anlyses. The program utilizes the plotting formula: T = N + 0.5 m -0.24 where T is the return period, N is the sample size and m is the rank. The developed routine fits four theoretical probability distributions to the sample data using the method of maximum likelihood. The two parameter lognormal probability distribution was selected for the present study on the basis of previous work which included the development of a goodness of fit criteria based on classical tests (Janowicz, 1983). For application in sparse data regions it is believed that two parameter probability distributions are superior to those with three parameters due to the generally smaller data sets which make it difficult to fit the more complex distributions. Regionaliza tion Procedure Hydrologic regions as defined by Janowicz ( 1984) were used as the basis for the development of the regional relationships. In the former study sufficient data were available to statistically define two hydrologic regions only within the Terri tory. As discussed above it is known that the hydrologic response of northern areas varies considerably from southern regions. This is indicated by the few available hydrometric stations in the northern part of the Terri tory, though these are insufficient in number to provide a statistical definition. Janowicz (1984) used multivariate statistical analyses to define the hydrologic regions. A cluster analysis was initially used to define rudimentary statistical groups. Input parameters included a timing parameter and selected sample statistics of the annual mean daily flow series divided by drainage area. Only stations with a minimum of ten years of record were used. The analysis yielded two well defined groups with evidence of a third transi tiona! group. These essentially correspond in area to include the stations draining the Coast Mountains as the most significant group, with the St. Elias stations forming the weaker transitional group. The third group consisted of the remaining stations. 315 Multiple discriminant analysis was then used to amend boundary locations. It was determined th<:t t too few stations existed in the St. Elias region to yield a statistical distinction from the Coast Mountains. These were combined to form the Western Mountains region. Region boundaries were adjusted during subsequent analyses. Two statistically defined hydrologic zones were obtained with relatively high degrees of accuracy. The defined hydrologic regions are shown in Figure 1 • Simple Linear Regression Simple functional relationships between mean annual maximum instantaneous discharge at selected return periods, and drainage area were developed for the hydrologic regions discussed above. Of the physica 1 watershed parameters, drainage area is generally the most significant and can be readily extracted from topographic maps. The least squares simple linear regression analyses were carried out on a Hewlett Packard 9826A micro-computor using the HP statistical library and graphics package. Because of the inverse relationship between drainage area and unit flood events, a curilinear relationship was initially sought using a polynominal curve fitting routine. Best results were obtained; however, with a linear fit using logarithmically transformed data. Values of the standard error of estimate were calculated using logarithmic units and represent the relative goodness of fit provided by the developed relationships. It became apparent during the analyses that several of the outliers represented hydrometric stations with considerable upstream lake storage. Both data groups were screened for stations with significant upstream storage. These were used to form a third group of 13 stations. The analysis was repeated with 53 and 24 stations for the Interior and Western Mountains regions respectively. The analyses were successful yielding values for the coefficient of determination (r2 ) of 97, 94 and 86 percent for the Interior, Western Mountains and storage controlled responses respectively, with standard errors of estimates of 6.1, 8.1 and 11.4 percent respectively. Plots of mean annual maximum instantaneous discharge against drainage area, with 95 percent confidence limits, are presented in Figures 2 to 4 for the three response types. ;;; ::!! '::: ~ <( r u "' 0 10000 1000 100 10 10 100 1000 10000 DRAINAGE AREA (SQ. Km.) Figure 2. Mean annual maiimum instantaneous discharge and 95 percent confidence limits for the Interior hydrologic region. 10000 1000 100 10 316 10 100 1000 10000 DRAINAGE AREA (SQ. Km.) Fiaure 3. Mean annual maximum inatantaneoua diacharge and 96 percent confidence limits for the Weatern Mountain• hydrolo1ic region&. 10000 1000 ;;; 100 ::;; ::: w " " <t J: u 10 "' Q 10 100 1000 10000 DRAINAGE AREA (SQ. Km.l Figure 4. Mean annual maximum instantaneous discharge and 95 percent confidence limits for Hydrometric stations with significant upstream lake storage. These relationships provide a means of obtaining design peak flow estimates with a reasonable degree of confidence for ungaged basins throughout most of the Terri tory. The developed relationships are presented in Table 1. Table 1. Summary of Simple Linear Regression Equations. Y = aiAREAib DEPENDENT REGRESSION REGRESSION VARIABLE CONSTANT COEFFICIENT y a b MEAN 1 0.226 0.865 ANNUAL 2 0 617 0.707 FLOOD 3 0. 139 0 817 2-YEAR 1 0. 188 0.876 2 0.467 0. 736 3 0.091 0.862 5-YEAR 1 0 301 0.856 2 0.971 0.676 3 0.200 0.798 10-YEAR 1 0.385 0.850 2 1. 430 0.640 3 0.305 0.760 50-YEAR 1 0.629 0.811 2 2.670 0 594 3 0.633 0.704 1 -INTERIOR 2 -WESTERN MOUNTAINS 3 -LAKE CONTROL Storage Index Factor STANDARD ERROR OF ESTIMATE r2 IPERCENTI 0.97 6. 1 0.94 8 1 0.86 11.4 0.96 6.5 0.94 8.3 0.87 11.8 0.97 5.8 0.92 7.9 0.86 11.3 0.97 6.4 0 91 8.4 0.84 11.5 0.96 6.7 0.89 9.5 0.81 11.9 Since lake storage appeared have a considerable effect on flows within the study area, to peak this parameter was investigated further with the intention of improving the developed relationships. Surface storage within a watershed is provided by lakes, swamps and reservoirs in addition to the stream channel itself and depression storage. In northern permafrost areas the thick organic rna t which makes up the active layer also provides considerable storage. These storage systems attenuate peak flows by increasing the duration of the streamflow event. The relative degree of attenuation depends on the amount of storage and its location within the basin. Poulin ( 1971) found that a storage parameter which represented the location of the storage system with respect to the point of interest, as well as the size, provided a better index of attenuation than the amount of storage alone. The use of this index was investigated in an attempt to better define the relationships presented above. For the purpose of this study the index is defined as the relative amount of watershed drainage area, in percent, which is controlled by the upstream lake or swamp. The lake or swamp is required to have an area equal to or greater than one percent of the upstream area controlled by it. This parameter is defined in Figure 5. The storage index factor has a range in values from 0 to 100 percent for both hydrologic regions. The mean is 31.2 and 68.1 percent for the Interior and Western Mountains hydrologic regions respectively. The greater values within the Western Mountains region are a result of a program carried out in the 1960's to establish hydrometric sta tiona at lake outlets for the purpose of assessing hydroelectric potential in the area. In a study carried out jointly by Environment Newfoundland and Environment Canada ( 1984), the investigators found this parameter to be quite significant. With 17 independent parameters it was most often entered by the stepping routine after drainage area as the second most significant variable. The study indicated; however, that in a particular region, estimates diverged considerably when this parameter fell below 55 percent. 317 a>1°/o of A b >1°/o of B (A+Bl STOR = (A+B+Cl x 100 a> 1°/o of A b>1°/o of B STOR =~x 100 (A+B+C) Figure 5. Determination of storage Index factor (after Poulin, 1971). Multiple Linear Regression Further analyses were carried out using the storage index factor in an attempt to improve the developed relationships. The stepwise multiple linear regression analyses were carried out using a program from the Biomedical Program series (Dixon and Brown, 1979). The general form of the equation is: Y =a + b 1 x 1 ••• bnxn where Y is the dependent variable, a is a regression constant, x 1 , xn re- present the values of the independent variable, and b 1 , b are regression coefficients. The ~lue of the F- sta tis tic was set to allow the entry of independent parameters which were within the 99 percent level of significance. It was found that by logarthmically transforming the streamflow and drainage area data, which had values with a range of up to five orders of magnitude, a better fit was generally obtained. The storage index data was not transformed. The general form of the transformed equation then becomes: which in the present case becomes: Relationships were developed between mean annual maximum instantaneous discharge at selected return periods and the independent parameters of drainage area (AREA) and the storage index factor (STOR) at selected return periods. Relationships were developed for the two defined hydrologic regions as well as the Terri tory as a whole, for use in the region between the Ogilvie and British Mountains. The developed equations are presented in Table 2. The best results were achieved for the Interior region as indicated by values of the standard error of estimate and the coefficient of determination. Though the entry of the storage index factor accounts for a small portion of the variance, the difference is significant at the 99 percent level. The use of this parameter reduced the standard error of estimate from 11.3 to 9.3 percent for the mean annual flood. Similarily there was an increase in the coefficient of determination from 0.916 to 0.945. The development of an improved relationship for the Western Mountains hydrologic region using the storage index factor was more difficult than for the Interior region. It was discovered that the data representing intermediate sized basins draining the St. Elias Mountains detracted considerable from the general trend. By removing these four stations (White River, Kluane River, Alsek River and Kathleen River) the storage index factor was then readily entered into the relationship by the stepping function. These basins contain relatively large glaciarized 318 Table 2. Summary of Multiple Linear Regression Eq ua tiona. LnY = a + b11 n (AREAl + b2(STOR) DEPENDENT REGRESSION REGRESSION COEFFICIENT STANDARD ERROR VARIABLE CONSTANT y a b, MEAN 1 -2.013 0.947 ANNUAL 2 -0.172 0.574 FLOOD 3 -1.182 0.831 2-YEAR 1 -2.097 0.954 2 -0.364 0.584 3 -1.356 0.845 5-YEAR 1 -1.757 0.943 2 0. 193 0.562 3 -0.844 0.819 10-YEAR 1 -1.591 0.939 2 0.486 0.551 3 -0.583 0.807 20-YEAR 1 -1.442 0.935 2 0.796 0.524 3 -0.340 0.792 50-YEAR 1 -1.299 0.933 2 0.938 0.538 3 -0 . 163 0.790 100-YEAR 1 1.190 0.930 2 1.179 0.523 3 0.047 0.776 1 -INTERIOR 2 -WESTERN MOUNT A INS 3 -COMBINED areas and subsequently generate high peak flows. Addi tiona! support for the removal of these stations is provided by the observed distinction between the two subregions making up the Western Mountains region as noted during the regionalization procedure. With the four station removed the developed relationship is statistically OF ESTIMATE b2 r2 (PERCENT) -0.0120 0.945 9.3 0.0082 0.944 10.9 -0.0050 0.921 12.2 -0.0124 0.941 9.7 0.0090 0.947 11.0 -0.0048 0.920 12.5 -0.0121 0.947 9. 1 0.0072 0.939 11. 0 -0.0056 0.920 12. 1 -0.0120 0.947 9.0 0.0063 0.930 11.4** -0.0060 0.917 12.2 -0.0120 0.948 8.9 0.0067 0.909 12.7** -0.0061 0.909 12.6 -0.0118 0.946 9. 1 0.0047 0.901 13.0* -0.0068 0.909 12.5 -0.0116 0.943 9.3 0.0041 0.888 13.4* -0.0069 0.901 12.8 * -NOT SIGNIFICANT ** -SIGNIFICANT TO 95% significant at the 99 percent level. The entry of the storage index factor reduced the standard error of estimate from 12.6 to 10.9 percent for the mean annual flood. Similarily there was an increase in the coefficient of determination from 0.922 to 0.944. A relationship was developed for the Terri tory as a whole using the 319 combined data sets of 61 stations. The relationship is intended for application to areas north of the Ogilvie Mountains. Because of the limited data base north of the Ogilvie Mountains and the increasing dominance of permafrost with increasing latitude, care should be exercised in applying the developed equation within this region. It should not be used for estimating design flows north of the British Mountains within the direct Arctic drainage. The developed relationship is statistically significant at the 99 percent level. The entry of the storage index factor reduced the standard error from 12.8 to 12.3 percent for the mean annual flood. Similarily the coefficient of determination increased from 0.913 to 0.921. DISCUSSION Simple and multiple linear regression relationships were developed for estimating peak flows in most areas of Yukon Territory. The developed relationships are based on a single station flood frequency analysis which was carried out on 90 hydrometric stations with at least six years of record. The two parameter lognormal theoretical probability distribution was selected for use in this study since it is believed to be the most appropriate for sparse data regions. Relationships between annual maximum instantaneous discharge at selected return periods and drainage area were developed for two statistically defined hydrologic regions and a third group of hydrometric stations with relatively large areas of lake storage. Because of the considerable effect of lake storage in attenuating peak flows, it was believed that the addition of this parameter to the developed relationships would result in a significant improvement. The analyses were repeated using a storage index factor as a second independent variable. The storage index factor represents the location as well as the size of the storage feature within a watershed. The inclusion of the storage index factor parameter in the developed relationships was statistically significant. Multiple linear regression equations were developed for the Interior and Western Mountains hydrologic regions. A relationship using the combined data was developed for use outside these regions. The analyses were generally good as indica ted by the re la ti ve ly low standard error of estimates and the high coefficients of determination. The developed equations provide a simple method of estimating peak flows with a reasonable degree of accuracy for most areas of Yukon Territory. REFERENCES Condie, R., G. A. Nix, and L.G. Boone, 1976 (with revisions 1981). Flood Damage Reduction Program Flood Frequency Analysis. Water Planning and Management Branch, Inland Waters Directorate, Environment Canada, Ottawa, Ontario. Dixon, W .J. and M. B. Brown. 1979. Bio- medical Computer Programs p Sries, University of California Press, Berkeley, California. Janowicz, J.R. 1983. Peak Flow Re- gionaliza tion for Alaska. Un- published M.S. Thesis. Fairbanks, Alaska. Janowicz, J .R. 1984. Yukon River Basin Hydrometeorologic Data Network Assess- ment. Yukon River Basin Study Hydrology Report #2. Indian and Northern Affairs Canada. Whitehorse, Yukon. Newfoundland Environment and Environment Canada. 1984. Regional Flood Fre- quency Analysis for the Island of Newfoundland. St. John's, New- foundland. Poulin, R. Y. 1971. Flood Frequency 320 Analysis for Newfoundland Streams. Water Planning and Operations Branch, Department of the Environment. Ottawa, Ontario • JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 EFFECTS OF SEASONALLY FROZEN GROUND IN SNOWMELT MODELING Knut Sand and Douglas L. Kane 1 ABSTRACT: The Swedish HBV-3 runoff model was used for simulations of runoff during snowmelt periods in the Chena River Basin in interior Alaska. The model was calibrated using data from 1969-1974. The initial simulations showed a poor ability to simulate both snowmelt runoff and rainfall runoff events. Because frozen soils have much lower infiltration and storage capacities than unfrozen soils, the soil moisture routine in the model was modified to accept seasonally varying soil parameters. This modification substantially improved the runoff simulations. The model was tested with meteorological data from 1982-1985 using the modified soil moisture routine. The simulations in the test period were of approximately same prediction quality as the simulations in the calibration period, and provided good estimates of runoff generated by both snowmelt and rainfall. (KEY TERMS: runoff model; snowmelt; soil moisture; seasonal frost; frozen soil) INTRODUCTION High-latitude basins are characterized by winter precipitation storage in the form of snow on the ground and by seasonally frozen ground. The runoff regime is strongly influenced by these two factors. Application of a mathematical runoff model in these basins requires a model with separate accounting routines for snowmelt and soil moisture. The snowmelt and the soil moisture routines are closely linked together, and successful modeling of snowmelt runoff strongly depends upon correct response from both routines and upon a correct interaction between the routines. Physical characteristics--such as infiltration capacity and soil moisture storage capacity of most soils--change drastically when the soil undergoes freezing or thawing. Kane and Stein (1983) found for Fairbanks silt loam that the infiltration rate increased by a factor of 10-100 (depending on the soil ice-content) when going from frozen to unfrozen state. So far, this fact has not been seriously considered in the development of snowmelt runoff models. The objectives of this study were: 1. determine how well a mathematical runoff model with its original snowmelt and soil moisture routines simulates runoff during snowmelt in a basin characterized by a seasonal snowpack and extensive seasonal frost; and 2. modify the soil moisture routine of this model to hopefully improve the runoff simulations. 1Respectively, Research Engineer, Norwegian Hydrotechnical Laboratory, N-7034 Trondheim-NTH, Norway, and Associate Professor, Institute of Water Resources/ Engineering Experiment Station, University of Alaska-Fairbanks, Fairbanks, AK 99775-1760. 321 TEST BASIN We selected Chena River Basin (Figure 1) as test basin for this study because of the availability of hydrologic and meteorological data. The basin is located near Fairbanks in interior Alaska and has a total area of 5,125 square km. It typically has a seasonal snowcover lasting for about six months of the year and extensive seasonal frost every winter. Wind-deposited silt loam is the dominant near-surface soil in the area. Many of the valley bottoms and north- facing slopes are underlain by permafrost. Much of the watershed is forested with tree line at about 600 m (msl). Poorly drained soils exist on most permafrost sites, and well-drained soils occur in permafrost-free areas. Moss and organic material cover all mineral soils at varying depths; depths are generally greater for permafrost sites. RUNOFF MODEL We chose a distributed version of the HBV runoff model for our study (HBV-3). The model is used in many Scandinavian watersheds (Bergstrom, 1976), and it has also been applied to watersheds in Switzerland (Braun, 1985). We selected this model because it has a simple structure and requires relatively little data input. For most northern basins, availability of hydrological and meteorological data is a critical factor in model selection. The model •s structure is graphically illustrated in Figure 2. The soil moisture routine in the model has the following parameters: FC LP BETA - maximum soil moisture storage (mm) limit for potential evaporation (mm) empirical coefficient (no units) CHENA RIVER DRAINAGE BASIN N 0 10 20 30 km • • • Stream gaging station • Snow course Figure 1. Chena River drainage basin showing data collection sites. 322 1 The outflow from the soil moisture zone is achieved by the following free parameters: KO, K1, K2 -storage discharge parameters (no units) UZL 1 imit for fast drainage, upper zone (mm) PERC percolation capacity into lower zone (mm/day) Figure 2. precipitation snow routine rain, snowmelt r-'-----"1"---'---., evaporation soil moisture routine QO;KO·(SUZ-LUZ) Ql :KI·SUZ Schematic presentation of the HBV-3 model. INPUT DATA The HBV model requires daily values for precipitation and air temperature (daily mean), and monthly estimates of potential evaporation. An estimate of the initial conditions in the basin at the beginning of the simulation period is also necessary. These variables are: Snowpack depth (mm) Water content of snowpack (mm) Soil moisture storage (mm) Upper zone storage (mm) Lower zone storage (mm) 3 Transformed discharge 3 (m /s) Recorded discharge (m /s) We used precipitation and air temperature data from meteorological stations at Fairbanks Airport and Chena Hot Springs for 1969-1974 (NOAA, 1969-1974) and from 323 Fairbanks Airport, Monument Creek and Teuchet Creek for 1982-1985 (NOAA, 1982-1985; U.S. Army Corps of Engineers, 1985; and U.S. National Weather Service, 1985). Estimates of potential evaporation were made based on observations from University of Alaska, Agriculture Experiment Station (NOAA, 1969-1974). River discharge observations were obtained from gage No. 15493000 Chena River at Fairbanks (U.S. Geological Survey, Water Resources Division, 1969-1974 and 1982-1985). Snow cover depth and water content observations from snow courses within the basin were also used (Soil Conservation Service, 1969-1974 and 1982-1985). CRITERIA FOR PERFORMANCE OF MODEL The model uses a numerical criterion (Sutcliffe and Nash, 1970) to indicate model efficiency, or the agreement between the recorded and the simulated hydrograph. where F 2 -F2 R2 = _o-..-__ F 2 0 (1) R2 2 = efficiency of model F = sum of squares of residuals 0 between observed and computed discharges F2 = sum of squares of residuals between observed discharge and mean of observed discharge during the simulation period. The value of R2 will range from minus infinity to +1, where +1 represents complete agreement between the observed and the simulated hydrograph. In addition to the numerical criterion, we gained valuable knowledge from inspection of the plotted hydrographs. TABLE 1. Results from simulations (R 2-values). Year Calibration Test 1969 X 1970 X 1971 X 1972 X 1973 X 1974 X Average (6 years): 1982 X 1983 X 1984 X 1985 X (Average ( 4 years): DISCUSSION We calibrated the HBV model with available data from 1969-1974. Only the months April, May and June each year were simulated. To improve the simulations, we modified the model by introducing a seasonal variation in the parameters of the soil moisture routine (FC, LP, BETA and PERC). Finally, the modified version of the model was tested with data from 1982-1985. The modified model improved the simulations in the calibration period substantially. The simulations in the test period were also satisfactory. The R2 . . 1 · -on g1 na R2 -modified model model 0.14 0.80 0.62 0.68 0.48 0.89 0.52 0.90 0.04 0.83 R2 0.53 R2 0.88 = 0.39 = 0.83 0.80 0.77 0.74 R2 0. 77 = 0. 77 results of the modelling effort are shown in Table 1. Soil moisture parameter values that produced these results are shown in Table 2. When calibrating the original model, we got an underestimation every year of snowmelt runoff in the early part of the simulation period, and an overestimation of rainfall runoff in the late part of the period (see Figure 3). We assumed that this was because the soil moisture parameters are set as constants in the model, while they actually exhibit seasonal variation. In this case, it means that the model used too high an infiltration capacity when the ground was frozen, and too low an infiltration TABLE 2. Values of the soil moisture parameters in the original and the modified model. Parameter FC (mm) LP (mm) BETA (-) PERC (m/day) Original model 265 265 3 324 Modified model Frozen soil Unfrozen soil 190 190 8 0.4 320 320 2 2.5 CHENA RIVER AT FAIRBANKS April 1 -June 30 1972 350 300 --Measured Q --Simulated Q 250 ., :;:;.200 ..s 0150 100 50 0 20 30 10 20 30 May June Figure 3. Runoff simulation 1972' original model. capacity when the ground \'Jas thawed. We recalibrated the soil moisture routine and then only examined simulation of snowmelt runoff occurring early in the period. We recalibrated the soil moisture routine again, and then simulated the rainfall runoff events in the latter part of the period (see Figure 4). This gave two sets of soil moisture parameters, one set for frozen soil conditions and one set for unfrozen soil conditions. Figure 5 shows the hydrograph simulation when the soil moisture parameter sets for frozen and unfrozen soil are combined; parameters for frozen soil are used in the first part of the modeling period, and parameters for unfrozen soil are used in the last part of the period. The point in time when the soil moisture parameters should be changed from the frozen to unfrozen case were, for all the years in the calibration period, determined from inspection of hydrographs only. Our assumption is that we should change the soil moisture parameters when the seasonally frozen soil layer is partially thawed. We estimated this from: 1. the amount of snow on the ground and ice in the frozen soil at the beginning of the snowmelt season; and 325 CHENA RIVER AT FAIRBANKS April 1 -June 30 1972 350,--------------,-----~--- ~ "' 300 250 -;;;-200 E ~ 0150 100 50 Measured Q- Frozen Simulation - Q TTTT1TTTll I IIIII I I I lllliilillillllllilllll I llillill I I I II I li!TTIIIIIIIIIIIIII 10 20 30 10 20 30 10 20 30 April May June Figure 4. Runoff simulation 1972, with soil moisture parameters for both the frozen and unfrozen state. 2. the amount of incoming energy available for melting. Lacking observations of ice content in the frozen soil, we estimated the total amount of snow and ice from the following parameters: = precipitation previous September (mm) = accumulated degree-days of freezing during the previous winter (°C*days) = water equivalent of the snowpack at the start of snowmelt (mm) The amount of incoming energy available for melting can be expressed as degree-days of melting: = accumulated degree-days of melting from start of snowmelt (°C*days) Using data from the calibration period (six years), we found the following regression equation for the required input of energy for melting. DDm = -88.2 + 0.125*Sd + 0.139*Ps+ 0.053*DDf + 1.319*DDf/Sd (2) 1972 April 1 -June 30 350 r------'-----;--------------, 300 -Measured Q --Simulated Q 250 "' -;;;-zoo E 0 150 100 50 10 20 April April . ! 0 ~.zeo ~aoo f--------. .A ,.o 20 30 10 20 30 May June May May May May Figure 5. Modified runoff calibration, 1972. 326 which gave a correlation of r 2 = 0.95. Using just six years of data from the calibration period is a minimal amount of data to confirm the validity of Equation (2). However, we used Equation (2) for predicting the time to change the soil moisture parameters in the test runs (1982-1985). Figure 6 shows the simulation of the hydrograph for one of the years (1984) the modified model was tested for the Chena River basin. CONCLUSIONS We demonstrate that improvements in model simulations of snowmelt and rainfall runoff can be made by introducing seasonal variation into selected soil moisture parameters in the HBV model. The main reason for this is that substantially different hydraulic properties exist between frozen and unfrozen soils and in our case extensive seasonal frost is al~o present. For our watershed, fine grained soils are predominant and, therefore, higher moisture levels occur than for coarser soils. Further improvements in the modifications to the model could be made by having a gradual transition between the frozen and unfrozen conditions instead of the abrupt change we used. ACKNOl~LEDGMENTS Funds for this study were provided by U.S. Geological Survey. The Norwegian Hydrotechnical Laboratory and the Royal Norwegian Council for Industrial and Technical Research supported the time Mr. Knut Sand spent at University of Alaska-Fairbanks as a visiting research engineer. Dr. Sten Bergstrom, Swedish Meteorological and Hydrological provided us with a copy of the HBV-3 model and advice. U.S. Army Corps of Engineers and U.S. National Weather Service in Anchorage provided precipitation and air temperature data . April 1 -June 30 1984 350.--------------------, 300 -Measured Q --Simulated Q ~ w 250 ~200 E ~ 0150 100 50 1 I . t> -. ! 0 I .!! •• I"" Apl'"'ll .. ay .. ay Figure 6. Test runoff simulation, 1984. 327 LITERATURE CITED Bergstrom, S., 1976. Development and Application of a Conceptual Runoff Model for Scandinavian Catchments. Swedish Meteorological and Hydrological Institute, Norrkoping, Sweden, Report No RHO?. Braun, L.N., 1985. Simulation of Snowmelt-runoff in Lowland and Lower Alpine Regions of Switzerland. Geographisches Institut, Eidgenossische Technische Hochshule, Zurich, Switzerland, Heft 21. Kane, D.L., and J. Stein, 1983. Field Evidence of Groundwater Recharge in Interior Alaska. Proceedings, Permafrost: 4th International Conference. National Academy Press, Washington, DC. National Oceanic and Atmospheric Administration, U.S. Department of Commerce. Climatological Data, Alaska 1969-74. Vol. 55-60. National Oceanic and Atmospheric Administration, U.S. Department of Commerce. Local Climatological Data, Fairbanks, Alaska. International Airport, 1969-1974 and 1982-1985. Soil Conservation Service, U.S. Department of Agriculture. Snow surveys and water supply outlook for Alaska. 1969-1974 and 1982-1985. Sutcliffe, J.V., and J.E. Nash, 1970. River Flow Forecasting Through Conceptual Models. Part I - A Discussion of Principles. Journal of Hydrology, 10(1970), 282-290. U.S. Army Corps of Engineers. Precipitation and Air Temperature Data From Meteor-burst Stations in Chena River Basin. Unpublished data. U.S. Geological Survey, Water Resources Division. Water Resources Data, Alaska. Water years 1969-1974 and 1982-1985. U.S. National Weather Service. Precipitation and Air Temperature Data From Meteor-burst Stations in Chena River Basin. Unpublished data • JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 SOME ASPECTS OF GLACIER HYDROLOGY IN THE UPPER SUSITNA AND MACLAREN RIVER BASINS, ALASKA Theodore s. Clarke, Douglas Johnson and William D. Harrison 1 ABSTRACT: Proposed hydroelectric development on the Susltna River, Alaska, has raised Interest In the glacIers that form Its headwaters. Three separate aspects of the hydrology of these glaciers are addressed here. First, long-term glacier shrinkage, which releases water that Is not renewable In the norma I sense, appears to have produced on the order of 3-4% of the total Susltna River flow above the Gold Creek gauge site since stream gauging began. Second, the major glaciers of the basin are surge- type and have the potent I a I to produce, In a few ronths, up to 30 times the estimated annual sediment Input Into the proposed Watana Reservoir. The next surge of one of the glaciers, Susltna, Is predicted In the fIrst decade of the next century. ThIrd, wl nter precIpItatIon varIes by a factor of two among the glaciers, Maclaren Glacier receiving the most. (KEY TERMS: Glacier shrinkage, glacier surges, sediment supply, precipitation variations.) I NTROD U::T I ON This paper describes, In part, the results of a study of the g I ac I ers that head the Sus I tna and Maclaren rivers (Figures 1 and 2). It addresses three separate topics: ( 1) whether the glaciers have changed In volume since stream gauging began on the Sus I tna RIver, ( 2) If and when any of th.e g I ac I ers In the area may be expected to surge, and h~ surges might affect the Susltna River, and (3) how precipitation varies throughout th~ area. A previous paper provides glacier runoff and mass balances estimates (Clarke and others, 1985). Early phases of the work are descrIbed by R & M and HarrIson ( 1981 ) and R & M and HarrIson ( 1982) and summarIzed by HarrIson and others ( 1983). The mater I a I presented here shou I d be C:Jns I dered an u~ate to these three early papers. Glaciers cover about 790 square kilometers or 5,9% of the bas In area above the proposed Watana dam site, 5.2% of the area above the proposed Devil Canyon sIte, and 4. 9% of the area above the Sus I tna River gauge located at Gold Creek (Figure 1). Field measure~ents of precipitation, snow accumulation, Ice melt, glacier speed, and surface elevation were made on most of the major g I ac I ers In the bas In during 1981, 1982 and 1983. I. LONG-TERM GLACIER VOLUME CHANGE Long-term glacier volume change Is an Important part of any hydrologic feasibility or planning study because It may have a significant Impact on project- ed water SUI>Piy. In general, glaciers have decreas- ed In size during the last half century. Conse- quently, water to their basins has been supplied out of Ice storage. As the glaciers approach equlllb- rl um wIth the present c II mate, the amount <>f water from storage approaches zero. ThIs has I ed, In some Instances, to an overestimation of water supply (Bezlnge, 1979). It seems that before long-term water availability Is predicted from stream gauge records, smoothed trends of glacier release or storage of water over the per I od of record ,; holll d always be subtracted. This reduces the problem to a conventional on~ Jf I J:Jg-term prediction for !In unglaclerlzed basin, although, of course, even the conventional approach Is susceptible to errors caused by c II mate change. Mayo and Trabant ( 1986) present evidence that a definable climate change took place In the Alaska Range In the Gulkana Glacier region, starting about 1976. Yo I ume change est I mates for the Sus I tna bas In are based on measurements on an unnamed g I ac I er, commonly referred to as East Fork Glacier (Fig- ure 2), which makes up only 5% of the total glacler- lzed surface. Pre.vlous estimates of Its volume change over the period 1949 to 1980 were made from photogrammetrlc data (R & M and Harrison, 1981; Harrison and others, 1983). These estimates sug- gested an average change In thickness of -50 ± 18 m. If this were typical of the other glaciers, then 13% of the Susltna River flow at the Gold Creek gauge site would have been from glacier storage. Since this seems unreasonably large, two other methods for estimation of volume change were 1 Theodore s. Clarke, Douglas Johnson and William D. Harrison, Geophysical Institute, University of Alaska- Fairbanks, FaIrbanks, AI aska 99775-0800. 329 0'\ \ \ \ \ \ \ \ --- \ \ \ \ \ \ \ --....._ ---------- Figure 1. Location map. (From Acres American, 1982.) applied. The first used direct measurement of glacier surface altitude; the second used the runoff precipitation model of Tangborn (1980). Direct Measurement of Glacier Surface Elevation In 1982 surface el evatlons were measured at several points on East Fork Glacier as a check of those estimated photogrammetrlcal ly from 1980 photos In the earlier work. Elevations were measured with a he I I copter and Its altimeter. Measurement points were located either by Brunton co""ass bearings to map I dent If I ab I e features or by theodo II te and establIshed control points. The altimeter was ca I I brated per I od I ca I I y on rock poInts of known elevation. The results are shown In Table 1. The results agree with those from the 1980 photos except at the highest point. According to the altimeter data, this point has remained at roughly the same elevation since 1949 when the u.s. Geological Survey maps were made, but the data provided by the photogrammetrlc method show this point to have lost 40 m of elevation. This discrep- ancy might be explained by the fact that the 1980 330 aerial photographs of East Fork Glacier show almost no contrast In Its accumulation area. This makes It difficult to Identify the surface accurately In these smooth snowy areas. Also, one might expect the accumu I at I on area of a "norma I" (non surge-type) glacier In retreat to remain at roughly the same elevation because a decrease In annual balance over the surface of a glacier affects the volume of Ice transported by the glacier In a way that accumulates down-glacier. The change In volume of the glacier was obtain· ed by co111>arlng the altimetry data with elevations obtained from 1949 photos. Unlike the 1980 photos, the 1949 photos are of very good qual tty. The elevations obtained from these early photos agree wl th pub I I shed map e I evat Ions and are therefore probably accurate. In practice, the volume change was co111>uted by determl n I ng a thIckness change versus elevation relationship, multiplying It by the area per e I evat I on I nterva I determl ned from the nap, and finally, by Integrating over the elevation Interval spanned by the glacier. Taking the altimetry data as the more reliable, the average thickness change of East Fork Glacier comes out to -13 m water equivalent for the 1949 to 5 10 MILES 0 5 10 KILOMETERS I::::::==::::E;;;;;;;;;;;;;a if. VELOCITY POINT Figure 2. Glacier names, locations and drainage divides. Glacier center line velocity was measured where Indicated. The points on the figure were placed next to the glaciers tor clarity. (M:>dlfled from Harrison and others, 1983.> 1982 period, rather than the -50 m tor the 1949 to 1980 period estimated by the previous work. It this 13m of water equivalent loss Is again extrapolated over the remaining 95% of the Ice In the basin (with suitable caution) then, on the average, about 3 or 4% of the Susttna River flow at Gold Q-eek has been due to glacier recession as opposed to the 13% of the earlier estimates. This estimate has very large errors associated with It since It Is based on tour points on a glacier that makes up only 5% of the Ice In the basin. However, It does seem more reasonable considering that the glacier runoff over the 1981 to 1983 period, when the glaciers were In approximate equilibrium, totaled only about 13% of the flow at the Gold Creek gauge site (Clarke and others, 1985). Tang born Runott-Precl p ltatlon M:>de I Tangborn ( 1980) has suggested a model tor determining long-term historical glacier balances by co~arlson of adjacent glacterlzed and unglaclerlzed basins. The model works by determining differences In runoff that do not correspond to precipitation changes, and these differences are assumed to be 331 caused by changes In storage of water as glacier Ice. The annual precipitation In each basin Is determined by using a representative precipitation station and determining a coefficient that corrects for precipitation differences between the basins and the precipitation station. The sum of evaporation, transpiration and condensation, per unit area, Is assumed to be the same tor both basins. The coef- ficient can be determined It runoff from both basins and glacier volume change are known tor a period of at least 1 year and If a suitable precipitation station exists. The model was tested against published mass balances of nearby Gulkana Glacier tor the period from 1967 to 1977 (Meter and others, 1980). Six different precipitation stations and three different unglactertzed basins were checked for the best possible tit of the model. Phelan Q-eek was used as the glaclertzed runoff station since this drains Gulkana Glacier. The best correlation between ca I cuI a ted and measured ba I ance occurred when Talkeetna precipitation station was used with ~he unglaclertzed basin Sh lp Creek near Anchorage (r = 0.77). Further datal Is are given by Clarke (1986). Table 1. Comparison of photogrammetrlc data (R & M and Harrison, 1981; Harrison and others, 1983) to helicopter altimetry data on East Fork Glacier. The surface elevation changes tor the altimetry data are for the period from 1949 to 1982; the surface elevation changes for the photogrammetrlc data are for the period from 1949 to 1980. A loss of elevation Is Indicated by a negative sign. East Fork Glacier Location on Elevation Change Elevation Change Glacier Center Line Altimeter Photogrammetry {1949 Map Elevation) (m) (m) ( 1949 to 1080 1390 1590 2050 In applying the model to the Susltna basin, there was a cons lderab I e uncertaInty In what the actual balance was for the period from 1981 to 1983. The measurements, tor all Ice In the basin, came out to +0.06 m water equ Iva I ent when summed over the 3-year period, but the cumulative uncer- tainty tor the 3-year period was 0.6 m (Clarke, 1986). In Tangborn 1 s model this uncertainty plays a large role In the resulting change In glacier mass tor the period from 1950 to 1983. These dates were chosen because 1950 Is the t lrst year from which complete runoff data are available tor the Susltna River at Gold Creek. It It Is assumed that balance tor the period from 1981 to 1983 was +0.06 m, then the average loss from the glaciers above the Susltna River at Gold Creek gauge site for the period from 1950 to 1983 was -16 m water equlva lent. It the balance was +0.66 m, then the average loss comes out to -9 m, and It the ba I ance was -0.54 m, then a calculated balance of -22m water equivalent results. The results of the two methods of volume loss estimation are summarl zed In Table 2. They are uncertain, but not Inconsistent. They Imply that 3 to 4% of the water t I ow at Go I d Creek between 1949 and 1980 came from Ice storage. This amount Is -74 ± -43 ± -51 ± +16 ± (m) 1982) ( 1949 to 1980) 18 -67 ± 18 18 -32 ± 18 18 -78 ± 18 18 -40 ± 18 wIthIn the stream gaugIng error and wou I d therefore probably not be s lgn It I cant In terms of projected water supply. II. GLACIER SURGES The major g I ac I ers of the Sus I tna bas In are West Fork, Sus ltna, 11 East Fork", Mac I aren, and Eureka (Figure 2). All except East Fork are listed by Post ( 1969) as beIng surge-type. Surges are sudden episodes of rapid glacier speed triggered by some Internal Instability, during which Ice movement may be hundreds or thousands of meters wl thIn a few months. The effects on sedIment and water supp I y, particularly the former, may' be substantial. There are some descriptive reports of high sediment production during glacier surges (Uskov and Kvachev, 1979; Shcheglova and Chlzhov, 1981) and two direct measurements. Humphrey (1986) reported that the 1982-1983 surge of Variegated Glacier, Alaska, released as suspended sed lment the equIvalent of about 0.3 m of eroded rock from the bed of that glacier. Bjljrnsson (1979) reported an erosion rate of 0.014 m/yr from the surge of Bruarj~kul I Glacier Table 2. Summary of glacier shrinkage estimates by two different methods. % Total Time Area Glacier I zed ThIckness Method Span Covered Area Loss Error Altimetry 1949-East Fork 5 13 (m) large, 1982 Glacier see text Runoff 1950-all 100 16 +6, -7 Precipitation 1983 glaciers If model Model In bas In applicable 332 In Iceland. The two measurements differ by more than an order of magnitude, but both are extremely high when col!l>ared to sediment production In non- surge years. Although Variegated Glacier Is consid- erably smaller than Susltna Glacier, both are narrow valley glaciers underlain by faults. It Variegated Glacier Is representa~lve of the Susltna basin, then a surge of the 250 ~ Susltna Glacier could release as much as 200 x 10 kg of suspended sed lment Into the Sus!tna ~lver, assuming a rock density of 2. 7 x 10 kg/m • ThIs Is 30 times the estimated annual ~edlment lntl~x,3 Including bed load, of 6.8 x 10 kg (5.8 x 10 m ) Into the proposed Watana Reservoir (R & M, 1982). There Is little direct evidence about the effect of surges on water supply. However, there are three potential effects. First, there should be a temporary Increase In me It water because of the Increase In ablation area that accol!l>anles some surges. Second, the extreme crevasslng that occurs during a surge temporarl ly Increases effective surface area, and therefore ablation. Third, surges release stored water (Kamb and others, 1985), a I though It Is not c I ear whether thIs water comes from long-term storage or merely from the most recent summer season. Given these effects of surges on sediment and water supply, It seems worthwhile to review the past hI story of surges In the Sus I tna bas In, and what It ~y Imply tor the future, particularly since surges tend to be per I odIc (MeIer and Post, 1969). West Fork Glacier Is known to have surged sometime shortly before 1940 when Bradford Washburn photo- graphed It, Susltna Glacier underwent a strong surge between 1949 and 1954 (Post, 1960); photos that we recent I y exam I ned IndIcate that the surge was COI!l>lete by July, 1952. Maclaren Glacier underwent a weak surge or strong 11 pu lse" In 1971 (Mayo, 1978). Surface speed measurements on West Fork, Susltna, and East Fork glaciers Indicate flow regimes that reflect the surge behavior of the first two. For both of these g I ac I ers the rate of Ice flow from the accumulation area Is considerably less than the rate of snow accumulation there (Table 3). This Indicates a thickening of the accumulation area that wl II probably be terminated by another surge. The velocity data and details of how accumu- lation and outflow were calculated are given by Clarke (1986). Wast Fork Glacier The disequilibrium of West Fork Glacier evident In Table 3 Is consistent with Its past behavior. Oblique aerial photographs of the terminus, taken by Bradford Washburn In 1940, show It to be extremely broken up and chaotic (see Clarke, 1986). This Information, along with the looped moraine pattern, Is conclusive evidence that a surge took place. Post (written comm. to Steven WII bur, 1984) places the surge In 1937. Close Inspection of 1981 NASA color Infrared aerial photographs shows at least three successIve term Ina I moraInes, each of whIch was very II ke ly caused by a success Iva ly weaker surge. Unfortunately, the periodicity of the surges cannot be estimated quantitatively because little Table 3. Comparison of annual lee flow through several cross sections to the annual accumulation above the sections. The location of each cross section Is shown as a velocity point on Figure 2. Surface center line velocity Is assumed to be caused by 50% Internal deformation and 50% basal sliding. AI I quantities are given In water equivalents. The cross sections are slightly below the accumulation areas and are shown as velocity points on Figure 2. Average Annual Ice Flow May 1981- Glacier June 1983 Name <m 3 /yr x 10 6 > West Fork 54 t 21 Sus I tna, MaIn Branch 14 t 6 Sus ltna NW Tr I b. 36 t 14 Sus ltna Turkey Tr I b. 72 t 28 East Fork 31 ± 12 1981 98 t 33 50 t 19 21 t 15 89 t 15 Annual Accumulation Above the Cross Section (m 3 /yr x 106 > 1982 1983 82 t 33 113 t 33 34 t 19 71 ± 19 70 t 15 20 ± 13 25 ± 13 333 Volume Change Above Cross Section ( 1981-1983 average) (m 3 /yr x 10 6 ) +44 t 39 +38 ± 20 Information exists for West Fork Glacier prior to the Washburn photographs. tJofflt (1915) gives a brief description of the glacier as It was In 1913 but nothing to Indicate a surge had occurred recent- ly. If Its recurrence period Is similar to the 50 or so years for Susltna Glacier, discussed below, a surge may be expected fairly soon. Susltna Glacier Susltna Glacier, unlike West Fork, has a complex set of tributaries that were studied Indivi- dually, as summarized In Table 3. It can be seen that the main branch of Susltna Glacier Is trans- porting only a fraction of the accumulated snow down-glacier. This would Indicate that either this branch of the glacier Is the one causing the surges, or It Is at least a reservoir that depletes during a surge. Altimetry data collected In the accumulation area of Susltna Glacier also show this branch to be accumulating mass. A gain of 56± 18 m of elevation from 1956 to 1982 was meas urad by comparIng 1982 a ltlmetry data to 1956 map elevation data (Clarke, 19~6)~ This translates to a gain of 93 ± 30 x 10 m /yr, which Is reasonably consltfte_rt with the average rate of gain of 38 ± 20 x 10 m /yr tor the 1981 to 1983 period (Table 3). Examination of moraine patterns confirms that this basin did Indeed contribute a large quantity of Ice to the last surge. Figure 3 depicts the moraine patterns on Susltna Glacier before and after the early 1950's surge. Before the surge, Ice motion In the main trunk above Turkey trl butary appeared to be very smal I, with relatively vigorous flow from Turkey pinching It oft. After the surge, a large volume of Ice had clearly advanced from the basin of the main branch. A large volume of Ice appears to have come from Turkey trIbutary a I so, and Northwest trIbutary appears to have contributed very little Ice, If any, to the surge. These observations Indicate that flow and accumu I at I on In Northwest trIbutary were prob- ably In equilibrium before the surge, the meln branch was far out of equilibrium, and Turkey tributary was somewhere In between. There are two reasonably quantitative approach- es to determining Susltna Glacier's surge period. First, the lobe created In the moraines of the main glacle~ by Northwest tributary had an area of about 4.0 km In 1949. A surge of the main glacier took place about 1951, as already noted. By 198~ the new lobe had grown to an area of about 2.0 km (Figure 3). Assuming the surge occurred In 1951, and assuming the present glacier speeds to be similar to those In the past, a period of roughly 60 years Is Indicated. Second, close Inspection of the same lobe In 1949 aerial photographs shows about 47 oglves to have passed from Northwest tributary Into the main glacier trunk (see Clarke, 1986). Oglves, or Forbes bands, are known to form on an annua I basts (Nye, 1958). Again assuming the surge occur- red In 1951, a surge return per I od of 49 years Is Indicated. It could be argued that Northwest tributary surges Independently, but the slow growth 334 SHORTLY IE FORE IS 52 SURGE SHORTLY AFTER SURGE 1110 Main Br&nch Turkey Tr I b. NW Trlb. Figure 3. Evolution of moraine patterns on Susltna Glacier. Left and center diagrams are from Meter and Post (1959). Right dl agram Is sketched from National flero- nautlcs and Space Administration photo- graphs. (tJodlfled from Harrison and others, 1983. > of Its new lobe and the balance between accumulation and flow makes this seem unlikely (Table 3>. The next surge wou I d therefore be expected wIthIn the first decade of the next century. East Fork, Maclaren and Eureka Glaciers East Fork Glacier Is probably not a surge-type glacier, as suggested by the approximate balance In Table 3, and by evidence from the displacement of surface features that the speed has not changed much since 1949 <R & M and Harrison, 1982). Both Maclaren and Eureka glaciers are thought to be weak surge-type glaciers; they do not surge on the order of k II ometers II ke Sus f tna and West Fork. As noted previously, Maclaren Glacier under- went a "pulse" In 1971 (Mayo, 1978). No speed measurements were made on these glaciers. II I. PRECIPITATION VARIATIONS Another Interesting aspect of glacier hydrology In this basin Is the large difference In winter precipitation among the dl fferent glaciers. In the late winter of 1981, 1982 and 1983, snowpack thick- -;; w c: u~ z ~ <{ ·-...J§. <{ Q) cc ~ a:; w 3:: 1-"' z ... -Q) ;=~ E 2.0 1.0 2.0 1.0 0.0 2.0 1.0 1000 & MACLAREN c WEST FORK A SUSITNA 0 EAST FORK • TURKEY • NORTHWEST 1000 1000 ELEVATION (meters) 1500 1500 1500 2000 2000 2000 2500 MAY 1981 2500 MAY 1982 2500 MAY 1983 0.0~---------------------------------------------------------------------...J Figure 4. Winter accumulation versus elevation as determined from snow probe data. (Top figure Is modified from R & M and Harrison, 1981; mlddl~ figure Is from R & M and Harrison, 1982.> 335 ness was measured by probing at several points along the center line of each glacier, and snowpack density was measured at representative points on each glacier. The water equivalent thickness at each point Is plotted In Figure 4. These data are reasonably consistent wlth more accurate snow depths measured at a few sites where stakes were maintain- ed, Generally the winter precipitation gradients are the same from glacier to glacier, about 1,2 mm water equivalent per meter of elevation, but the a bso I ute amount of water varIes cons I dera b 1 y from glacier to glacier. Maclaren Glacier consistently received the most precipitation, and the two steep south-facing tributaries of Susltna Glacier consis- tent I y receIved the I east. An orographIc effect created by the Clearwater Mountains, which divide the tributary Maclaren River basin from the Susltna River basin, may direct moisture toward Maclaren Glacier and reduce precipitation In the Susltna basin to the west, It Is worthwhile to note that because Maclaren Glacier had a positive mass balance of nearly 0,3 m/yr and the others had generally negative balances, It produced less runoff over the study period even though It received considerably more preclpltat~n (Clarke and others, 1985), IV, DISCUSSION AND CONCLUSIONS An attempt has been made here to ( 1) determl ne whether the glaciers that head the Susltna and Maclaren rivers have changed In volume since stream gauging began on the Susltna River, (2) determine when these surge-type glaciers may surge again, and what the effects of surges are II ke I y to be, and (3) describe variation In winter precipitation throughout the area, The conclusions are as fol- lows: I, The elevation change due to glacier wastl!lg seems to be on the order of -10 to -15 m water equivalent tor the 1949 to 1983 period tor East Fork Glacier rather than the -50 m estimated by R & M and Harrison (1981) and Harrison and others (1983) tor the 1949 to 1980 period, ThIs amounts to 3 or 4% of the tot a I t I ow of the Susltna River at Gold Creek rather than 13%. This quantity seems more consistent with the tact that during 1981, 1982, and 1983, when the glaciers were In approximate equilibrium, the average runoff from the Susltna basin glaciers was about 13% of the total Susltna River flow at Gold Creek (Clarke and others, 1985), 2. West Fork and Susltna are surge-type gla- ciers, It sediment output during a surge of Susltna Glacier, tor example, Is simi tar to that of Variegated Glacier, a single surge may produce about 30 times the estimated average annual sediment Influx Into the proposed watana reservoIr. The rates of transport and d I sper- slon of such a large sediment Influx are unknown. A surge of Susltna seems likely 336 because about two-thirds of the snow accumulat- Ing In the basin of Its main branch Is not beIng transported out <Tab I e 3), and the accumulation area of this same branch has gained approximately 56 m of elevation since the last surge. It past history Is any Indica- tion, It appears that Susltna Glacier has a surge period of 50 to 60 years, which places the next surge sometime between the years 2000 and 2010. It Is also likely that West Fork Glacier will surge In the future, but no quantitatively determined period can be placed on It since no data are available tor the period prior to Its 1937(?) surge. 3, Accumulation varies considerably from glacier to glacier, with Maclaren Glacier receiving more winter precipitation than any of the other glaciers. Generally, the winter precipitation gradIents are the same throughout the bas Ins, about 1,2 ± 0,1 mm water equ Iva I ent/m e I eva· tlon, but each glacier's accumulation versus elevation curve Is shifted vertically with respect to the accumulation axis, The shift ranges over about 0,5 m water equivalent (Figure 4), REFERENCES Acres American Inc., 1982, Susltna hydroelectric project; feasibility report, Final draft re· port tor the Alaska Power Authority, Anchoraqe, Alaska, 8 vols. Bezlng, A,, 1979, Grande Dlxence et son hydrologle, Ia collection de donnees hydrologiques de base en Suisse, Association Suisse l'amenagement des eaux. Service Hydrologlque National, 19 pp. Bjornsson, H., 1979, Nine glaciers In Iceland, Jokul I 29:74-80, Clarke, T. s., 1986, Glacier runoff, balance and dynamics In the upper Susltna River basin, AI aska, M.S. ThesIs. Un Ivers tty of AI aska, Fairbanks, 98 pp. Clarke, T, s., D. Johnson and W, D. Harrison, 1985, Glacier runoff In the upper Susltna and Maclaren River basins, Alaska, In: L. P, Dwight, 1985, Resolving Alaska's Water Re· sources Conflicts. Proceedings. Alaska Sect I on, AmerIcan Water Resources As soc I a• tlon, Institute of Water Resources/ Engineer• lng Experiment Station, University of Alaska, Fairbanks. Report IWR 108, 212 pp. Harrison, w. D., B, T, Drage, s. Bredthauer, D. Johnson, c. Schoch and A. B, Follett, 1983, Reconnaissance of the glaciers of the Susltna basin In connection with proposed hydroelectric development. Anna Is of Glaclol• ogy 4:99-104, Humphrey, N. F,, 1985, Suspended sediment discharge from Variegated Glacier during Its surge and pre-surge phases of mot I on. A contrIbutIon to the workshop on Alaska hydrology, 10-13 April, 1985, In preparation. Kamb, B., c. F. Raymond, w. D. Harrison, H. Engelhardt, K. A. Echelmeyer, N. Humphrey, M. M. Brugman and T. Pfeffer, 1985. Glacier surge mecha11lsm: 1982-1983 surge of Variegated Glacier, Alaska. Science 227(4686):469-479. Mayo, L. R., 1978. Identification of unstable glaciers Intermediate between normal and surging glaciers. Academy of Science, USSR, Sect I on of G I ac I o I ogy, ProceedIngs of the International Workshop on Mechanism of Glacier Variations, Pub. 33, p. 47-55 and 133-135. Mayo, L. R. and D. c. Trabant, 1986. Recent growth of Gulkana Glacier, Alaska Range, and Its relation to glacier-fed river runoff. Short paper for u.s. Geological Survey Water Supply Series. In press. Meter, M. F. and A. s. Post, 1969. What are glacier surges? Canadian Journal of Earth Science 6:807-817. Meter, M. F., L. R. Mayo, D. c. Trabant and R. M. Krimmel, 1980. Comparison of mass balance and runoff at four glaciers In the United States, 1966 to 1977. Academy of Science, USSR, Section of Glaciology, Report of the Inter- national Symposium on the Computation and Prediction of Runoff from Glaciers and Glacial Areas, Pub. 38, p. 139-143 and 214-216. Moffit, F. H., 1915. The Broad Pass Region, Nye, Alaska. u.s. Geological Survey Bul I. 608. 80 pp. J. F •• glaciers. 1958. A theory of wave formation on lASH 47:139-154. 337 Post, A. s., 1960. The exceptional advances of the Muldrow, Black Rapids and Susltna glaciers. Journal of Geophysical Research 65:3703-3712. Post, A. s., 1969. Distribution of surging glaciers In western North America. Journal of Glaciol- ogy 8(53):229-240. R & M Consultants, 1982. Alaska Power Authority, Susltna Hydroelectric Project; Appendix B.S, Reservoir Sedimentation. Report for Acres American, Inc., Buffalo, NY, 49 pp. R & M Consultants and w. D. Harrison, 1981. Alaska Power Authority Susltna Hydroelectric Project; task 3 -hydrology; glacier studies. Report tor Acres American, Inc., Buffalo, NY, 30 pp. R & M Consultants and w. D. Harrison, 1982. Alaska Power 114Jthorlty Susltna Hydroelectric Project; task 3 -hydrology; glacier studies. Report for Acres American, Inc., Buffalo, NY, 22 pp. Shcheglova, o. P. and o. P. Chlzhov, 1981. Sediment transport from the g I ac I er zone, centra I Asia. Annals of Glaciology 2:103-108. Tangborn, w. v., 1980. Two models tor estimating climate-glacier relationships In the North Cascades, washington, USA, Journal of Glaciol- ogy 25(91) :3-21. Uskov, Ju. s. and v. 1. Kvachev, 1979. The dldal surging glacier. Data of Glaciological Studies, Chronicle, Discussion. Academy of Sciences of the USSR, Section of Glaciology of the Soviet Geophysical Committee and Institute of Geography, Pub. 36, p. 170-175. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 REGIONAL DISTRIBUTION OF STREAM ICINGS IN ALASKA K. G. Dean ABSTRACT. Stream-icing zones in mainland Alaska were mapped based on analysis of multi-date Landsat imagery. Mapped features include late winter overflow, residual ice-sheets and braided streams possibly susceptible to LCLngs. Results of the 1:250,000 scale mapping were generalized and the regional distribution of icings displayed. Almost all icings occur in or near upland or mountainous terrains. The Brooks Range and northeastern Alaska have the largest and greatest number of occurrences. The number and size of icings generally decrease to the south except in the vicinity of the Alaska Range. In northern Alaska many icings develop east of the Colville River but few to the west. This difference appears to be related to the availability of freshwater. In interior Alaska occurrences are numerous but small and are restricted to tributary stream channels. (Key Terms: Stream-icings, aufeis, naleds, Landsat, Alaska) INTRODUCTION Stream icings, also referred to as aufeis or naleds, are seasonal flood phenomena that develop in high latitude and alpine regions. Icings form when the hydrostatic pressure of water in aquifers or streams is sufficiently large to force water through its ice cover. These hydrostatic pressures largely result from restriction of water flow and the hydraulic gradient in stream channels and aquifers (generalized from Carey, 1973, and Harden et al., 1977). Factors of climate, hydrology, geology, permafrost, and topography influence icing occurrence and behavior (Carey, 1973). Icings occur during subfreezing temperatures and result in the flooding of stream channels and adjacent low-lying areas. The flooding can extend many kilometers beyond the source of the extruded groundwater. The flood waters do not generally drain but freeze in place. This process repeats itself many times during a winter, often causing the accumulation of extremely thick sheets of ice; for example, 5 m along the Echooka River (Sloan et al., 1976). Icings are a detrimental phenomena. They constitute hazards that can disrupt transportation and communication, hinder field exploration programs, and present difficult engineering problems for drilling operations, pipelines, roadways, buildings and other structures in arctic and subarctic regions. Icings may also indicate the presence of groundwater seeps or springs and perennially flowing water, which are potentially valuable resources. The intent of the study reported here was to map the location and extent of icings in Alaska and to discuss their distribution. PREVIOUS INVESTIGATIONS Many investigators have mapped or studied icing conditions in Alaska. Existing knowledge concerning occurrence, control and prevention of icings was summarized by Carey (1973). Large icings in Alaska and Canada have been mapped and studied by Dean (1984 a&b); Harden et al. (1977); Childers et al. (1975); Van Everdingen (1975); Leffingwell (1919); and Hall (1976; 1980). Icings along the trans-Alaska pipeline route were mapped by Sloan et al. (1976). Local icings have been studied by Slaughter (1982); Carlson (1979); Kane et al. (1973); Kane and Carlson (1977); and Corbin (1977); Stringer et al. (1985). METHODS AND PROCEDURES Large stream-icing zones in Alaska were mapped at a scale of 1:250,000 based on analysis of multi-date Landsat imagery. These features include late-winter overflows and residual ice sheets that range from 6 to 40,500 hectares (20 to 100,000 acres) in areal extent; braided streams that are susceptible to icings were also mapped (Figure 1). The icings were interpreted from Landsat MSS (multispectral scanner) data recorded during the 1972 to 1982 late-winter, spring, and summer seasons. The location and maximum extent of icings were mapped regardless of frequency and timing of occurrences. Most icings tended to recur at the same location each year, although the extent did vary. Selected images with minimum cloud cover and extensive icings were enlarged to 1:250,000 scale and used as primary data sources for each plate. Landsat images at 1:1,000,000-scale that were recorded on dates not selected for enlargement provided additional information. Visible-wavelength (band 5) data were used to map residual ice after spring thaw (Figure 2). These ice sheets appear white on band 5 images due to their high reflectance compared with surrounding vegetation, soil, or rock. Occasionally, the ice even had a higher reflectance than nearby snow cover. Landsat images recorded in near-infrared wavelengths (band 7) were used to map late-winter overflows (Figure 3). Active overflows appear dark gray to black due to absorption of these wavelengths ~ophysical Institute, University of Alaska, Fairbanks, Alaska, 99775-0800 339 '"' <._/ I I . I / . I( .. ~ )' -----_/--.:;- , .... ~ I. l -- l -1 ' -,_.. I ..... _----- L.--------·. r---~;;--. ·I I -A __]____. _ _,_j __ r-. I -~ I I I ----.l.-- Figure 1. Stream icings in the vicinity of Demarcation Bay. Solid areas indicate residual ice sheets; hatched areas indicate overflows, and open areas indicate braided streams possibly subject to stream icings. 340 IS SEA ICE - • Figure 2. Summer Landsat image (band 5) of the Demarcation Bay area. Residual ice sheets (A) are bright and coincide with overflows displayed in Figure 3. Numerous unlabeled ice sheets are present. -· Figure 3. Winter Landsat image (band 7) of the Demarcation Bay area. Stream overflows (B) are dark and coincide with ice sheets displayed in 'Figure 2. Several unlabeled overflows are present. 341 by water. The dark signature in the infrared wavelengths contrasts markedly with the highly reflective snow and ice. Very limited field observations were made during this investigation. Crosschecking of features observed on images, and discussions with individuals familiar with specific areas helped refine the maps. In some areas, cloud cover restricted the number of images available for interpretation. At a few locations, aerial photographs were used to verify occurrences. At specific sites, large icings previously mapped by Childers et al. (1975), Hall (1976; 1980), Sloan et al. (1976), and Harden et al. (1977), compare favorably with those mapped in this investigation. The presence of confusing spectral signatures in some areas may have resulted in misinterpretations; these areas are queried on the maps. Surface features that may be confused with or mask residual ice sheets include snow cover (usually at higher elevations), snow drifts along flood plains, river ice, gravel, and vegetation. Features that may be confused with late winter overflows include dense spruce forests; dense, leafless brush along streams; unfrozen stream channels; exposed clear ice; and exposed ground. Overflows were observed on late-winter images in areas where large residual ice sheets were mapped. Typically, the overflows are located in stream valleys (river or spring icings) and their extent is larger than the residual ice sheets. Icings are often associated with braided streams (Carey, 1973) and may cause braided-stream patterns to develop (Harden et al., 1977). Boundaries of flood plains thought to be susceptible to icings were delineated by braided patterns on topographic maps and Landsat images and by proximity to icing zones. Typically, streams that are susceptible to icings are finely rather than coarsely braided, and usually, meandering streams in mountainous regions with relatively short, braided segments are affected by icings in the braided reach. If springs are also present, icings are probably responsible for the braided patterns. In mountainous regions, bifurcated channels were also observed in icing areas. REGIONAL DISTRIBUTION Approximately 700 stream-icings were mapped. The locations of icings were plotted on a small-scale map of Alaska to display the regional distribution (Figure 4). The highest concentration and largest occurrences are associated with Alaska 1 s two major mountain systems, the Brooks Range and the Alaska Range. The number and size of icings decrease away from these two mountainous regions. The largest icing is in northeast Alaska and affects an area greater than 40,500 hectares (100,000 acres). Few icings were mapped in low-lying areas that are not in the immediate proximity of hills and mountains. In interior Alaska most of the icings are confined to small tributary stream valleys which contrasts with occurrences to the north where icings develop along primary and tributary streams. Generally, the number and size of icings steadily decrease to the south with those in the vicinity of the Alaska Range being the exception. Northern Alaska, which includes the Brooks Range and the Arctic Coastal Plains, has the greatest density of occurrences but the distribution differs east and west of the Colville River. East of the river, icings develop in most streams throughout the Brooks Range and on the coastal plain. Numerous 342 springs in the area are located at sites where 1cwgs form (Childers et al., 1975). Calcium carbonate deposits on some of these icings suggest that the water has flowed through calcareous bedrock at depth (Hall, 1980). The coastal dispersal patterns of anadromous fishes in the Beaufort Sea (Craig, 1984) indicate that there is sufficient water in streams where icings develop for fish to overwinter in this area (P.C. Craig, pers. comm., 1985). West of the Colville River few icings occur except on the south flank of the Brooks Range. Assuming that climatic factors and the distribution of permafrost do not vary significantly across northern Alaska, the paucity of icings is apparently related to geologic and hydrologic conditions. The absence of icings suggests that there are few freshwater springs, and/or that flow of water within streams and stream beds is not constricted sufficiently to produce the required hydrostatic pressures or that water does not flow in the streams during the winter. The dispersal patterns and species of anadromous fishes west of the Colville River are markedly different from those found to the east, which coincides with the difference in the distribution of icings. This may be related to possible differences in the number of freshwater sources and suggests that water beneath the winter ice-cover is lacking or not sufficient for overwintering fish. Thus, icings may indicate streams that do not freeze to the bottom at least in arctic regions. CONCLUSIONS Stream icings are common throughout Alaska and usually recur near the same locality annually. Their areal extent may vary each year. The highest frequency and greatest density of icings occur in the Brooks Range, the northeast coastal plain and foothills. The largest icings also develop in this area. Few icings occur in the western coastal plain and foothills. Generally, the number and size of icings steadily decrease southward, except in the Alaska Range, where there are numerous stream icings. In interior Alaska, icings are numerous but relatively small and are restricted to tributary stream valleys. Almost all large icings occur within or near upland or mountainous regions and often along finely braided streams. In northern Alaska many icings coincide with the location of known freshwater springs and hence icings may indicate the location of springs in other areas. The distribution of these northern icings also coincides with the dispersal patterns of anadromous fishes. This suggests that streams with icings are winter habitats of anadromous fishes. ACKNOWLEDGEMENTS The author wishes to thank Dr. W. J. Stringer, T.H. George and c. Helfferich for their helpful comments. The initial mapping of stream-icing zones was funded by the Alaska Division of Geologic and Geophysical Surveys. LITERATURE CITED Carey, K.L., 1973. Icings Developed from Surface Water and Ground Water. Cold Regions Research and Engineering Laboratory Monograph 111-D3, 68°~ 66"k_ I -- 164° /60" •• • • • • , l • • /56" /52" 144" 140" 136° \ I \ EXPLANATION 0 LARGE RESIDUAL ICE SHEET OBSERVED e RESIDUAL ICE SHEET OBSERVED WINTER SPRING OVERFLOW OBSERVED A BRAIDED STREAM POSSIBLY SUBJECT TO SEASONAL ICING AND FLOODING 152° 148" 144° Figure 4. Regional distribution of icings in Alaska. 343 70° 58" Hanover, NH, 67 pp. Carlson, R.F., 1979. A Theory of Aufeis arid Bed Erosion. In Canadian Hydrology Symposium 79, Cold Climate Hydrology. National Research Council of Canada, Ottawa, 197-205. Childers, J.M., Sloan, C.E., Meckel, J.P., and Nauman, J.W., 1975. Hydrologic Reconnaissance of the Eastern North Slope, Alaska. U.S. Geological Survey, Open-File Report 77-492, 65. Corbin, S.W., 1977. The Thermal Regime of a Stream in Central Alaska. M.S. thesis, University of Alaska, Fairbanks, 144 pp. Craig, P.C., 1984. Fish Use of Coastal Waters of the Alaskan Beaufort Sea: A Review, Transactions of the American Fisheries Society, 113:265-282. Dean, K.G., 1984a. Stream-icing Zones in Alaska. Alaska Division of Geological and Geophysical Surveys Report of Investigations 84-16, Fairbanks, AK, 20 pp. + 1 map. Dean, K.G., 1984b. Alaskan Stream Icings. The Northern En~ineer, 16(1):3-7. Hall, D.K., l9 6. Analysis of Some Hydrologic Variable on the North Slope of Alaska Using Passive Microwave, Visible and Near-infrared Imagery. In Proceedings of the American Society of Photogrammetry. Falls Church, VA, 344-361. Hall, D.K., 1980. Mineral Precipitation in North Slope River Icings. Arctic, 33:343-348. Harden, D., Barnes, P., and""Reiiilnitz, E., 1977. Distribution and Character of Naleds in Northeast Alaska. Arctic, 30:28-40. Kane, D.L., Carlson, R.F:"";8nd Bowers, C.E., 1973. roundwater Pore Pressures Adjacent to Subarctic Streams. In Proceedings for the North American Contrib., --2nd International Conference on Permafrost. National Academy of Science, Washington, D.C., 453-458. Kane, D. L., and Carlson, R. F., 1977. Analysis of Stream Aufeis Growth and Climatic Conditions. In Proceedings of the Third National Hydrotechnology Conference. Canadian Society for Civil Engineers, Quebec, 656-670. Leffingwell, E.K., 1919. The Canning River Region, Northern Alaska, US Geological Survey Prof. Paper 109, 251pp + maps. Slaughter, C.W., 1982. Occurrence of and Recurrence of Aufeis in an Upland Taiga Catchment. In Proceedings of the Fourth Canadian Permafrost Conference. March 1-6, 1981, Calgary, Alberta, Canada. National Research Council of Canada, Ottawa, 182-188. Sloan, C.E., Zenone, C., and Mayo, L., 1976. Icings Along the Trans-Alaska Pipeline Route. U.S. Geological Survey Prof. Paper 979, 31 pp. Stringer, W.J., T.H. George and R.M. Bell, 1985. An Aufeis Case Study. The Northern Engineer, 17:25-29. Van Everdingen, 1975. Use of ERTS-1 Imagery for Monitoring of Icings, N. Yukon and N.E. Alaska. In Research Program, Hydrology Research Division, Canadian Inland Waters Directorate, Water Resource Branch, Ottawa, Canada, Report Series No. 42, 75-87. 344 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 ESTIMATION OF GLACIER MELTWATER HYDROGRAPHS David Bjerklie and Robert Carlson 1 ABSTRACT: The melt component of daily hydrographs from glaciers was estimated using a multiple regression model. The model was developed for one basin using a period of record over which precipitation was not a significant contributor to the streamflow. The input variables were the average daily temperature and the log of the previous day's flow. The model was then applied to subsequent years of temperature data to test its applicability. The melt estimates from the model simulations were compared with other independent estimates. The results indicate that estimation of melt using the simple multiple regression model presented here is a useful method which gives annual melt volumes that are similar to the other estimates and allows the daily determination of melt flow. The capability of the model to predict daily melt flows should be evaluated more thoroughly because of the model's high degree of sensitivity to error in input data and to changes in model parameters, and because of the model's overly simple design. (Key Terms: melt; hydrograph; glacier) OBJECTIVE We investigated the use of a simplified technique to estimate melt flow hydrographs from glacierized basins. Daily temperature is often the only and the most detailed data available for many areas of Alaska. Multiple linear regression was used to evaluate mean daily temperature as an index to predict daily melt from the Phelan Creek basin, Alaska. This stream has been gaged by the USGS since 1966 and has included a systematic study of Gulkana glacier, which comprises 61% of the basin. INTRODUCTION There have been numerous efforts to model streamflow and melt from glacierized basins (Fountain and Tangborn, 1985). Modeling techniques have ranged from purely stochastic to purely deterministic, and have been concerned with time frames ranging from one day to many years. The major source of ablation energy 1Institute of Northern Engineering, Water Research Center, University of Alaska- Fairbanks, Fairbanks, Alaska 99775-1760. 345 is net radiation (Mayo and Pewe, 1963), however, in some cases temperature is more highly correlated with ablation than net radiation (Braithwaite, 1981). Modeling ablation with a simple linear temperature relationship of the form ablation = BO + B1(temperature) (1) where BO and B1 are regression coefficients can explain up to half of the variance and has been applied in areas where there is sparse glaciologic and hydrologic data (Braithwaite, 1981). Multiple regression analysis using the log of the previous days discharge, daily precipitation, daily temperature, daily incoming solar radiation and daily average vapor pressure as independent variables to predict daily streamflow, from a basin in Switzerland with 54% glacier area, explained up to 94% of the variance (Lang and Dayer, 1985). Lang and Dayer found that the previous days streamflow was a better predictor of streamflow than was the average daily temperature even though the basin area was only 34.5 square kilometers. This paper will examine the use of a statistical multiple regression model to predict melt flows from Gulkana Glacier. Daily streamflows and daily temperatures were available for the basin as well as ablation estimates from glacier stake observations, from the USGS. ALASKA Phelan Creek Average Basin ElevattOtt-17351W Percent Glaciated-611 Meteoroloqical Station Elevation-1J75oo latitude-63 deg 14 '"'" ~p>.-Figure I. 346 MODEL DEVELOPMENT The primary objective of this paper is to develop a means to estimate the melt component of the hydrograph. Periods of record were chosen for analysis during times of maximum melt contribution and minimum rainfall runoff contribution so that the hydrograph was dominated by the melt component. We assumed that the groundwater component was negligible in mountainous glacierized basins. The periods of record met the criterion that no more than 5% of the total flow over the period fell as precipitation. Under this criteria no spikes obviously associated with a rainfall runoff event were evident, and the volume of flow from rainfall runoff was minimal. In addition, no peaks were seen that did not also correspond to an increase in temperature. These data provided a means to independently analyze the melt component of the hydrograph. Regression analyses were run on streamflow and temperature data in order to determine the best prediction model. The regression model used was of the form y = BO + B1(X1) + B2(X2) .••• Bn(Xn) + e (2) where BO, 81 and 82 ••• Bn are regression coefficients and y is the dependent variable, X1 and X2 •.• Xn are independent variables, and e is the residual (Davis, 1986). We found that, similar to Lang and Dayer (1985), the log of the previous days streamflow (lnQ1) and the daily temperature (T) provided the best results (Table 1). The longest period of record was that for Phelan Creek in 1977. This period also showed the best r squared value and smallest standard error of the estimate. TABLE 1 Multiple Regression Results of Average Daily Temperature and the Natural Log of the Previous Days Flow vs the Natural Log of the Days Flow. Stream Period of Record Regression Equation R squared Phelan Cr. May 13-Sept 5 1977* lnQ•.146+.878lnQ1+.049T .987 Phelan Cr. May 15-Jun 30 1969* lnQ•.192+.855lnQ1+.041T .976 Phelan Cr. Jun 1-Jul 31 1976* lnQ•.355+.825lnQ1+.041T .906 lnQ•natural log of streamflow 1nQ1•natural log of previous days streamflow T•average daily temperature • For periods of record with total precipitation volume of SS or less of the volume of flow. The 1977 Phelan Creek data were divided into three independent data subsets consisting of the June, July and August data. These data subsets were analyzed using multiple linear regression to determine if there was any time dependent difference in the regression results (Table 2). The regression coefficients indicate that temperature played the greatest role as a predictor variable in August and the least in June, indicating that the temperature-melt,relationship changes through the summer. This agrees with known phenomena, i.e., as the summer progresses, more ice with a lower albedo is exposed and melt is enhanced. This also explains some of the variation among data sets seen in the initial regression results for Phelan Creek. Multiple Regression Results for Phelan Creek 1977 data Seperated into Monthly Subsets. Time Period Regression Equation R squared Correlation of lnQ vs 1nQ1, vs T May 13-Jun 30 lnQ•.185+.818lnQ1+.035T .991 .995 .675 Jul 1-Jul 31 lnQ•.240+.893lnQ1+.039T .940 .935 .444 Aug 1-Sept 5 lnQ•.466+.796lnQ1+.076T .934 . 927 .793 347 The phenomenon of snow and ice melt is more complex than is suggested by the simple model produced from the regression analyses here. However, the goodness of fit of these regression models seems to indicate that they would prove useful for determining the melt component from river basins with significant glacierized areas. To test this hypothesis, a standard regression model was adopted. The regression results from the Phelan Creek 1977 summer data was used as the model because the results were the best statistically and the data encompassed the entire melt season. QUANTITATIVE EVALUATION The regression model was used to predict sequences of melt flows for the ablation seasons (May 15 to September 15) for five years of Phelan Creek data, including the 1977 calibration data. The results are shown in Figure 2 along with temperature sequences, precipitation and actual streamflows. In general, it appears that the model predicted melt peaks that were also seen in the actual flow sequences (related to temperature peaks). The model did not predict peaks associated with precipitation events. However, precipitation in this basin can fall as snow or rain, and can exist as both for a given storm at different elevations in the basin. Thus it is not necessarily true that a peak in streamflow would be associated with a precipitation event. Numerous discrepencies are apparent by inspection between the predicted hydrographs and the actual hydrographs in terms of peak height and timing during periods of no precipitation and therefore melt only. Because of the interference of precipitation events, there is no sound statistical means to evaluate the predictive accuracy of the model. Thus, the only means to determine how well the model did was to compare the melt estimates with melt estimates determined using different means • Three independent ablation estimates were used to evaluate the model's PHELAN CREEK 1977 .. c ... 19 tJl ' tJl -M[AS .. EO STRLW'LOW "' '' ' ' . CALCI.UTED MELT FLOW ~ .. ... r u "' ::0 u :3: ~ 0 _j ... _./' y -· -I '" ~ ~ .. "' " D - ~RYS lMAY 13 TO SEPT 5) DAILY PRECIPITATION 1977 DAYS <MAY 13 TO SEPT 5> CRILY TENPEPRTURE 1377 10 G '-" ::!! ... !!I .... II: "' w 0.. r ... .... -· ... DAYS (MAY 13 TO SEPT 51 Figure 2. 341 u ill ' "' "' w .... ... r u iii ::0 u X 0 ..J ... 0.. u ~ u '-" w Q ... 15 .... 12 ... 0.. r ... .... •• u 17 11 • -I e -· -e PHELAN CREEK 1976 i~ f! ll ' ~ .. B .. DAYS <MAY 15 TO SEPT IS l DRILY PRECIPITATION 1976 DAYS <MAY IS TO SEPT lSI DAILY TEMPERATURE 1976 DAYS <MAY IS TO SEPT ISl r£: !d3S 01 Sf A~Wl SA~O . "' 1 < 61 3onll:!i!i3d~J31 AlHIO CSI ld3S 01 51 A~Wl SA~O 1<61 NOl1~1ldi~3dd ~li~O 1<61 ~33d~ N~l3Hd E- e •• •• •• •• ee1 6~£ ;tl ,., ("> ~ ... :l:l ... ~ i ~ CSI 1d35 01 Sl A~Wl 5A~O CS! 1d35 01 51 A~Wl 5k~O ><61 N011~1ldi~3dd Ali~O c S l l.d35 01 5 l A~W l SA~O ;._;H·'!'"'i· ""61 ~33i:!:Y . .. >lt!l3Hd ·- ,_ ... ,., ::t ., ,., "' 1-:D --< ~ ,., 0 1"1 "' ~ e •• ;g ,., '"' ~ •• ., ~ ... :l:l ... •• ~ i ••• ~ ee1 1- ., ,.. 0 :0: ;:; c "' ;:; ::t ,., II ... 1"1 "' Ill ' ., f!i ~ 01 c w Ill ' Ill "' w .... w %: u "' :::l ~ 3: 0 ...J L.. ~ u "' w 0 PHELAN CREEK 1869 " ,. ,. DAYS CMAY IS TO SEPT ISl DAILY PRECIPITATION 1969 '"" •• •• .. ,. DAYS CMAY IS TO SEPT !Sl DAILY TEMPERATURE 1969 ~ ~-----------------------n N OA1S lMAY IS TO SEPT !Sl Figure 2 (cont.). 350 predictions. The best available estimate was from the glacier stake estimate of total melt volume over the ablation season provided by the USGS (Mayo, personal communication). In addition to this for comparative purposes, a predictive ablation model based on average ablation season temperature (Mayo and Trabant, 1984) was used to estimate the melt volume over the ablation season. Baseflow volumes were also computed over the ablation season. The baseflow was assumed to represent water released from storage, which would include the melt component. Groundwater contribution was assumed to be negligible in this basin due to the mountainous terrain. The results of the estimates of melt season ablation by the four methods are shown in Table 3. The best approximation of the glacier stake estimate for all five years was provided by the multiple regression model (model simulation). However, the model estimates were fairly close to each other and to the glacier stake estimate, with the exception of the model presented by Mayo and Trabant (1984) for 1976 (alternative model). The baseflow estimates were consistently conservative. Stream Phelan Cr. 1969 May 15-Sept 15 Phelan Cr. 1971 May 15-Sept 15 Phelan Cr. 1974 May 15-Sept 15 Phelan Cr. 1976 May 15-Sept 15 Phelan Cr. 1977 May 13-Sept 5 TABLE 3 Melt Estimate Comparisons (meters). Model Simulation 1.77 1.80 1.76 2.27 1.86 Glacier Alternative Baseflow Stake Model 1.64 1.35* 0. 96 1.47 1.57* 1.24 2.00 1.91* 1.43 2.47 1.57* 1. 73 1.96 1.82* 1.58 * Estimated from USGS ablation model developed for Wolverine glacier (Mayo,1984). Seasonal Ablation•.19+.43Ta where Ta•avg. summer temp. R-squared•.65 SENSITIVITY ANALYSIS A series of simulations were conducted after changing values of the data input and the regression coefficients of the model in order to evaluate the sensitivity of the model to the model parameters and input data. The results of this analysis are shown in Table 4. It can be seen that minor changes in the data input as well as changes in the coefficients result in significant changes in the quantitative value of the flow simulations. The qualitative character of the simulation does not change, however, except when the Bl coefficient (associated with the lnQ1 variable) is altered. From this it can be concluded that, due to the sensitivity of the model to changes in model parameters and variables, it is probably not wise to apply the model developed in one basin to another, unless generalized information is all that is needed. It is possible that a more extensive study can define the variation in coefficients with climatic and physiographic variables in order to render the model more generally applicable. Results of Sensitivity Analysis of Simulations. Model form y=BO+B1(X1)+B2(X2). Model Used Peak Flow Minimum Flow lnQ=.146+.87BlnQ1+.049T* 1.5 .01 lnQ=.146+.BOOlnQ1+.06T 0.6 .01 lnQ=.300+.B7BlnQ1+.049T 5.0 .01 lnQ=O.OO+.B7BlnQ1+.049T 0.4 .01 lnQ=.146+.600lnQ1+.049T 0.1 .01 lnQ=.146+1.001nQ1+.049T 2.7EE13 .01 lnQ=.146+.87BlnQ1+.035T 0.7 .01 lnQ=.146+.B7BlnQ1+.060T 2.6 .01 lnQ=.146+.87BlnQ1+.049T 2.1 .01 +1 degree bias in T data lnQ=.146+.87BlnQ1+.049T 3.2 .01 +2 degree bias in T data lnQ=.146+.B7BlnQ1+.049T 1.0 .01 -1 degree bias in T data lnQ=.l46+.B7BlnQ1+.049T 0.7 .01 -2 degree bias in T data *Model used for simulations. 351 CONCLUSIONS Using temperature and the previous day's flow as predictors of a melt hydrograph seems to be a workable method for estimating the contribution of the glacierized portion of basins to streamflow. However, the model needs to be calibrated more accurately to better predict the total volume of melt from glacierized areas. Discrepancies in the timing and peaks of predicted melt flow indicate that the modeling approach used here is limited by the simplistic indexing of melt with temperature. In addition, due to the model •s high sensitivity to model parameters and the temperature input data, the model must be evaluated for individual cases, since each case can include data biases and differences in terrain and microclimatic conditions that could presumably effect the melt temperature relationship. An advantage of the use of the predictive model described in this paper is its capability to predict daily flow. This would enable the calculation of missing data and enable the continuous monitoring of the hydrologic conditions in glaciers. In addition, it might enable more detailed management of water from glacierized basins. REFERENCES Braithwaite, R.J., 1981. On Glacier Energy Balance, Ablation, and Air Temperature. Journal of Glaciology, Vol. 27, no. 97. 9 pp. Davis, J.C., 1986. Statistics and Data Analysis in Geology. 2nd Edition, John Wiley and Sons, New York. 646 pp. Fountain,A.G. and W.V. Tangborn, 1985. Overview of Contemporary Techniques. Pages 27-41 in G.J. Young. Techniques for Prediction of Runoff from Glacierized Areas. IAHS publication no. 149. Lang,H. and G. Dayer, 1985. Switzerland Case Study. Pages 45-59 in G.J. Young. Techniques for Prediction of Runoff from Glacierized Areas. IAHS publication no.149. Mayo,L. and T.L. Pewe, 1963. Ablation and Net Total Radiation, Gulkana Glacier, Alaska. Chapter 43 in Ice and Snow Properties, Processes and Applications. MIT Press, Cambridge, Mass. Mayo,L. and D.C. Trabant, 1984. Observed and Predi~ted Effects of Climate Change On Wolverine Glacier, Southern Alaska. Pages 114-123 in G.P. Juday and G. Weller. The Potential Effects of Carbon Dioxide-Induced Cimatic Changes in Alaska, Proceedings of a Conference. Univ. of Alaska, Fairbanks, Alaska. University of Alaska misc. publication 83-1. U.S: Geological Survey, Water Resources Div. 1969-1977. Water Resources Data for Alaska. Report AK-69-1 through AK-77-1. Young,G.J., 1985. Overview. Pages 3-23 in G.J. Young. Techniques for Prediction of Runoff from Glacierized Areas. IAHS Publication no.149. 352 PRECIPITATION -SNOW PACK -SOIL PROCESSES 353 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 SNOW SURFACE STRENGTH AND THE EFFICIENCY OF RELOCATION BY WIND 1 R. A. Schmidt ABSTRACT: An automated system developed by Martinelli and Ozment samples forces required to fracture grains from a snow surface. The device has sufficient sensitivity to measure differences in surface erodibility under saltating drift. Such measurements help explain differences in the efficiency with which wind transports snow. (KEY TERMS: snow; blowing snow; drifting; snow strength; transport efficiency.) INTRODUCTION Wind-drifting snow is part of the hydrologic cycle in cold regions of planet Earth and two aspects of the process particularly significant to hydrologists are: (1) the reduction in effective precipitation caused by evaporation during transport and (2) the nonuniform distribution of the snow that remains after redeposition. Methods to predict these effects often require estimates of snow transport rate as a function of wind speed, from which the duration of drifting is computed for a given depth of snowfall. The efficiency with which wind moves snow may be defined by the ratio of drift rate to wind energy. Experiments that measured drift rate and wind speed showed that after snowfall, transport efficiency increased as surfaces became hardened by the drift process (Schmidt, 1986). Drift efficiency seemed closely related to momentum lost by saltating particles on impact with the surface. The wind energy required to maintain saltation over a hard, old snow surface as opposed to a soft, new snow surface is analogous to the effort one exerts to keep a ball bouncing on a hard floor, in contrast to a foam pad. When a drift particle's impact fractures a new particle from the surface, kinetic energy is lost and the total momentum of new and old particles is less than when the particle rebounds without producing a fracture. Surface hardening increases the probability that a surface grain is bonded too strongly to be ejected by the impact of a saltating drift particle. The hardening of the surface under drifting snow occurs (1) by mechanical reduction of the size and complexity of the precipitation crystals, and (2) through cohesion and subsequent bond growth by sintering. Laboratory experiments on these processes allow us to make general statements about how the hardening will proceed in relation to time, temperature, humidity, and particle size. However, the nonhomogeneity of both snow particles and surface configurations produces a distribution of strengths in bonds attaching the most exposed grains that is more easily measured than predicted. A device useful for surface hardness observations should measure the same range of force particle impacts produce in saltation on a snow surface. Such forces may be estimated by assuming central, elastic collision of ice spheres (Schmidt, 1980). Impact velocities _yanging from threshold (near 2) to 10 m s or more may be expected, considering the wind speed near the top of a saltation trajectory. Sizes of saltating snow particles average near 0.2 mm diameter (Schmidt 1981). Forces up to 1 N seem possible (Figure 1), with the greatest number of impacts around 1Rocky Mountain Forest and Range Experiment Station, 240 West Prospect, Fort Collins, Colorado, 80526, U.S.A. 355 0.1 N. The corresponding range of a load cell (in grams force) is also indicated on the figure. !0 0 100 0=0 . Smm 60 40 (D 0 L 20 ~ " 10 6 L 6 $' 4 ~ 0 2 ~ "' I =:J w u 0 < CJ --' ~'-----'----'---'-2--"-_L3 --'-_j_---'--~s----'-~6-~7 ---'---,s~~g----'--------'1 &· 1 PARi!CLE VELOCITY <m s-l) Figure 1. Estimated impact forces of two ice spheres with diameter D in a central, elastic collision where one sphere is stationary and the other has the speed indicated on the abscissa (Schmidt, 1980). METHOD Several hardness gages measure the resistance to penetration of snow by a hard conical probe. These devices sample a much deeper layer than that interacting with saltating drift. Martinelli and Ozment (1984, 1985a, 198Sb) developed a method of rapidly sampling forces required to fracture individual grains from a snow surface in situ (Figure 2). A sensitive load cell measures the force on a solid plastic sphere 20 mm in diameter !.1 driven toward the surface at 0. 5 mm s by a stepper motor. Comparing output voltage from the load cell with its no-load value, the system detects contact of the probe and t~t surface with a sensitivity of 9.8•10 N (a touch of 0.1 gram force). Surface detection starts a probe-travel counter and arms a break-detector that changes state when a surface failure reduces resistance against_fhe probe by a specified amount ( 2. 94 • 10 N for these experiments). Probe travel and maximum force at failure are held for measurement and the probe is retracted while data are recorded on magnetic tape and printed for inspection by the system operator. 356 Each measurement takes approximately one minute. The probe may be repositioned within the support cylinder (Figure 2) to take· many samples around the circumference of a circle 8 em in diameter. An array of four thermistors measures surface temperature in the center of the sampling area. Approximately 100 measurements are possible before recharging batteries. The operator controls the instrument with a bar code reader that permits sampling under blizzard conditions without removing gloves or mittens. Data are transferred electronically to a desk-top computer for analysis. Figure 2. Photograph of the Martinelli-Ozment surface strength tester. During operation, the electronics are carried on a small sled. Cold room experiments in which a television camera viewed interactions of the probe with surface grains confirmed sufficient sensitivity of the instrument to detect fractures of individual grains from the surface (Martinelli and Ozment, 1985a). As those authors note, the probe contacts surface grains at points that give a range of moment arms around the fracture fulcrums, adding to the large variance of these measurements. The same is true of saltation impacts. RESULTS Martinelli and Ozment (1985b) provide extensive measurements on snow surfaces to confirm the useful range of the instrument. Their experiments suggest a linear relationship between force at failure and distance the probe traveled between surface detection and fracture. The constant of proportionality varies widely. During drifting in January and February, 1985 at Diamond Lake in southeastern Wyoming I sampled strengths on several pairs of surfaces, using the Martinelli-Ozment system. As an example, Figure 3 compares the distribution of surface strengths on (a) the upwind (eroding) face of a moving snow dune and (b) measurements near the top of the same dune, where deposition was occurring. Data consist of three replications of nine samples at each location. Figure 4 shows the relationship of probe travel and fracture force for the data of Figure 3. Both surface strengths and probe travels were smaller on the eroding face. CONCLUSIONS The device developed by Martinelli and Ozment provides both the sensitivity and speed to sample snow surface strength sufficiently that different surface conditions may be compared. For example, forces required. to fracture grains from the windward face of an eroding snow dune were less than those on a surface depositing immediately downwind of the dune crest. Such data will help explain the large variations in snow transport efficiency. ACKNOWLEDGEMENT The usefulness of the instrument results from many clever solutions of both mechanical and electronic design problems by Mr. Arnold Ozment, who expertly machined the system and also helped with field measurements. 357 >-u z w 01/15/85 Ca) UPW I NO FACE -2. 2 c 6 0. 16 27 SAMPLES w 0::: LL w > ....... 1-:s 0. 08 w 0::: >-u z w 0.00~~~~~~~~~~~~~~~~~~~ 0.00 0. 02 0. 04 0. 06 0. 08 0. 10 FORCE REQUIRED TO CAUSE FAILURE CN) 01/15/85 Cb) TOP, MOVING DUNE -3.6 c 6 0.16 27 SAMPLES w 0::: LL w > ....... 1-:s 0. 08 w 0::: 0.00~~~~~-L~~~~~~~~~~-L~~ 0. 00 0. 05 0. 10 0. 15 0. 20 0. 25 FORCE REQUIRED TO CAUSE FAILURE CN) Figure 3. Distributions of forces that caused fracture of surface grains or bonds for (a) upwind (eroding) face and (b) top of a moving snow dune where deposition was occurring. w "' ~-25 ,---..,--....,-----.----,--~-~---.--""Tl 01/15/85 + UPW I NO FACE CEROS IONJ 3 ~.20 o TOP OF DUNE CDEPOS IT I ONJ < u.. w <J) ::J c5 0. 15 D w "' ~ 0. 10 w "' w u "' D +o F CN) =A+B*L Crnrn) u.. ~-~5 " + + 0 + A=-~. ~~72 B= ~. 327~ r 2= ~. 7951 "'i + 0 L2 L4 L6 LB L = PROBE TRAVEL DISTANCE Cmml Figure 4. Forces that caused surface fracture were greater and the probe traveled farther on the deposition surface, for measurements in Figure 3. REFERENCES Martinelli, M. and A. D. Ozment, 1984. Controlling Snow Surface Measurements with a Handheld Calculator. Cold Regions Sci. and Technol. 9(3):277-281. Martinelli, M. and A. D. Ozment, 1985a. Laboratory Test of a Motorized Snow Surface Hardness Gage, Cold Regions Sci. and Technol. 10{2):133-140. Martinelli, M. and A. D. Ozment, 1985b. Some Strength Features of Natural Snow Surfaces that Affect Snow Drifting, Cold Regions Sci. and Technol. 11:267-283 Schmidt, R.A., 1980. Wind-speeds and Elastic Snow Transport. J. 26:453-467. Threshold Impact in Glaciol. Schmidt, R.A., 1981. Estimates of Threshold Windspeed from Particle Sizes in Blowing Snow. Cold Regions Sci. and Technol. 4:187-193. Schmidt, R.A., 1986. Transport Rate of Drifting Snow and the Mean Wind Speed Profile. Boundary-Layer Meteorol. 34(1986) 213-241. 358 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 WATER FLOW RATES, POROSITY, AND PERMEABILITY IN SNOWPACKS IN THE CENTRAL SIERRA NEVADA Bruce J. McGurk and Richard C. Kattelmann1 AlmRACf: The equivalent permeability of spring snowpacks near Soda Springs, Califomia was estin.ated fran snowpack ootflow data and calculated surface melt. The equivalent permeability results agree with reported values of intrinsic permeability and were f~ to decline !Yo o1er of magnitude (fran 5 x 10 to 5 x 10 m ) during spring sno.melt. Values of a permeability parameter also changed significantly over the melt seasons, and these values can be used in operational application of meltwater rooting procedures. Water drainage fran snowpacks was about three times as rapid in the early season than in the late season. The decreasing iqx>rtance of flow channels with time may be responsible for this change in drainage speed. (KEY 'llBIS: snowpack permeability, snmmelt, water nxwement through snow, snowpack. porosity.) Snow cover runoff is forecasted daily by several agencies using a variety of methods. :Many of the forecasting teclmiques currently used depend on eq>irical relations and perfonn adequately under typical or JOOderate weather and snow conditions. These procedures can produce serious errors, however, during Ull1lSUa1 or extreme conditions such as rain-on-snow events. Yet these are also the times w.ben the need for accurate prediction is greatest. Development and application of forecasting teclmiques based on physical measurements may offer the best rope of iq)roving predictioo. accuracy during these extreme events. Transmission of water through snow is 8IOOIDg the processes that need to be considered in a ca~prehensive JOOdel of snow cover runoff. This process is rost iqlortant during extreme events and in areas of deep snow and rapidly responsive streams (Anderson, 1979). The study of water umrement through snow based on poroos media flow theory was pioneered by Colbeck. in the 1970's (Colbeck, 1972). Salle koowledge of the snow's permeability is required before the porous media flow JOOdels can be adapted to a snow cover. Laboratory measurcmmts of permeability, however, are virtually iqlossible to perfonn accurately due to the triple-phase nature of snow as well as the physical changes that occur during the seasonal metam:>rphism of the snowpack. Nevertheless, snow permeability has been estimated in a variety of ways, and the reported values cover three orders of magnitude (Table 1). The nonha!x>geneity of snowpacks in the Sierra Nevada and mst other muntain ranges mandates that porosity and permeability be estimated in situ and throughout the melt season. Snowpack permeability can be estimated by a procedure (described later) that requires hourly data on both snow surface water input and basal outflow, water travel duration and distance, and snowpack density. When the procedure is used with isolated colUDilS of halngeneoos snow (Colbeck and Davidson, 1973; Denoth and Seidenbusch, 1978), an intrinsic or saturated permeability may be estimated. When the procedure is used with data fran a heterogenous, layered snowpack, the results could be termed a snowpack equivalent permeability rather than an intrinsic permeability, but this distinction is seldan made. At the end of the melt season, the permeability of snowpacks in California and v~l "Were found to be in the range of 1 to 4 x 10 m (Colbeck and Anderson, 1982). Tbe permeability of snow to air has been found to be strongly dependent on the metaJOOrphism of the snow and to change over time (Conway and Abrahamson, 1984). Intrinsic permeability is believed to be 1 Pacific Southwest Forest and Range Experiment Station, Forest Service, U.S. Department of Agriculture, P.O. Box 245, Berkeley, Califomia 94701. 359 determined mainly by grain size (Shimizu, 1970) and to change quickly during the early part of the melt season (Dmme et al., 1976). (h the basis of observations of extensive chatmeling in Sierra Nevada snowpacks (Gerdel, 1954, Kattelmann, this proceedings), we 'WOlldered if the gross permeability of the snowpack might change over time as the flow network is altered. The procedures developed by Colbeck (1972, 1978, 1979) to IIXldel the rooting of water vertically through a snowpack require separate estimates of !'fffl':ability and porosity or of a parameter k /1/J • which is the quotient of the cube root of int~insic permeability and the effective porosity (proportioo. of the pore volume available for water D¥>Vement). This parameter can be derived fran the procedure Colbeck and Anderson (1982) used and, as a single value. characterizes an :inportant property of the snow. Colbeck (1978) suggested that typical values be assigned to this parameter for use in operational forecasting and that they woo.ld change as the snowpack mtures. This paper describes a study that, in part, atteopted to find sane representative values of this parameter to allow the applicatioo. of both si.Dple lag equations (Colbeck, 1978) and cooprehensive rooting programs (Tacker and Colbeck, 1977). The study area, the U.S.D.A. Forest Service's TABLE 1. Snow porosity and permeability values by snow description. Snow Description Porosity Perme!~il~ty Method (x10 ,m ) Source Loose new dry snow Fine grain old dry Medium grain old dry Coarse grain old dry 0.86 .59 snow .68 snow .67 Newly fallen snow Fine grain snow Large grain snow Glass beads 1.0 mm Not reported Disaggregated wet <1 mm Disaggregated wet 1 mm Disaggregated wet >1 mm Disaggregated wet 3 mm .85 .7 .6 .37 .49-.58 .44-.50 .48-.49 .52 3 Disaggregated wet .65g/cm3 .32 Disaggregated wet .62g/cm .36 Not reported Midwinter 0.5-1 mm Early spring 1 mm Late spring >1 mm Spring snowpack Spring snowpack Shallow snowpack 3 .48g/cm3 .50g/cm3 .41g/cm ----------------- .64-.84 .60-.65 .39-.51 .47 .45 .55 4 2 6 10 4 0.06-8 3-16 .9 .6 .06-.7 .9-3 1-10 25 .1 .3 10-30 2-3 3 1 2.6 2 3 Permeameter Permeameter Not reported Shimizu (1970) Bader et al (1954) Kuroiwa (1968) de Vries [cited by Wankiewicz (1979)] Price (1977) Calculated from Denoth and Selden- column drainage busch (1978) Calculated from Denoth et_!l. (1979) column drainage Calculated from Colbeck and column drainage Davidson (1973) Not reported Bengtsson (1981) Permeameter Calculated from snowpack input and outflow rates 1 Kattelmann (1981) Colbeck and Anderson (1982) 1 Kattelmann, R.C., 1981. Hydrology of Compacted Snow. Unpublished MS Thesis, University of California, Berkeley, California. 91 pp. 360 Central Sierra Snow laboratory (CSSL), Soda Springs, Califomia, acCUIIIllates deep seasonal snowpacks. It is in the forested snowpack zone on the west slope of the Sierra Nevada at 2100 m and is Sllbject to a Dllritime climatic influence. Between stonns, wann, ~ weather often leads to the foxmation of surface melt-freeze crusts, which becaue buried by subsequent snowfall to produce a highly stratified snowpack. Occasional midwinter rain or radiation-induced melt is often sufficient to produce snowpack. outflow in any 10011th (Smith, 1974). Rain or meltwater percolating downward fran the surface becaues concentrated into distinct flow channels at the layer interfaces (Colbeck, 1979). The resulting snowpack and its intemal flow network are highly variable spatially and continue to change u spring snowmelt progresses. Solar radiaueters and other meteorological equipnent at the study site provided hourly average data reqnired by a radiation budget DMXlel to estimate hourly surface melt flux. Study intervals were selected fran 1984 and 1985 on the basis of data availabilty and weather. Because melt is easiest to model during clear weather, the 14 mtervals selected were mostly sunny with 0 teoperatures above 0 C. To match the melt outflow pattem, the modeled 24-hour intervals started at 0000 hours •2 .Meltwater outflow was collected by several 2 m meltpans resting m the soil surface and pltmbed into tipping buckets. Each pan was filled with gravel to allow rapid lateral flow, and the travel time fran pan to tipping bucket was approximately 2 minutes. Because outflow fran the pans varied scmewhat in timing and quantity, an overall average time/outflow depth hydrograph was calculated. An isotopic profiling snow density gauge was used to measure average snowpack density and daily depth (Kattelmann et al., 1983). The theory of porous media flow has been applied to liquid water flow through snowpack, to derive permeability and porosity (Colbeck and Andersoo, 1982: Colbeck and Davidsoo, 1973; Price, 1977). Because the snow's unsaturated permeability DllSt be known for the application of usual methods for describing flow, typical methods of estimating permeability cannot be used. However, because snow is a large-grained material, gravity forces override tensioo. gradients, so capillary flow can be ignored (Colbeck, 1972) • As surface melt and percolatioo occur, the vertical rate of mvement for a melt flux value u is given by (Dwm.e et al. , 1976) 361 dzl (1/n) = ; (e..~) u (n-1/n) dt u e (1) where n is a power law exponent (often assumed to be 3) that relates snow water saturation and intrinsic saturated permeability to relative or unsaturated permeability k = kS n (2) w e and effective porosity is a function of total porosity and irreducible water saturation fj = fj(1 -s . ) e W1 (3) If a meltwater flux u is estimated for the snow surface, and the elapsed time is known until an equivalent flux is released fran ~ base of the snowpack, a ratio parameter k /fj can be estimated with equation 1. To thise end, 14 melt/ontflow events were selected from CSSL records for 1984 and 1985. Before the events were plotted, night freeze crust water equivalents were subtracted from the oroo and 0900 hours of the calculated melt. The hourly melt values were then adjusted so that the total daily melt equalled the daily observed outflow. The effect of this adjustment was tested with one study interval and found to change travel times unifo:nnly and by ooly a minor amount. Travel times for selected flux values were calculated by plotting the melt and outflow fluxes oo a single graph and estimating the delay between melt and outflow for equivalent flux values (Figure 1). Recession:-li.Db wave speeds 11ere calculated by dividing snowpack depth by travel times for decreasing late aftemooo. fluxes between 1 and 4 11111/h (Colbeck and Andersoo, 1982; Dwm.e et al., 1976), and log-log plots of flux versus wave speeds were prepared (Figure 2). U,ast-sqnares fitting of a line to the points produced slopes of the lines for carparisoo with the theoretical value of 0.67, the flux exponent in equation 1 when n is equal to the integer value 3. Porosity was calculated fran snow density, liquid water, and irreducible water saturation using equatioo 2 and p s-pi ~=--- p s -p. w 'W 1 (4) Irreducible water saturation equals water cootent divided by porosity and is assumed to be 1).07 (Colbeck and Andersoo, 1982). By inserting a wave ~~ and a matching flux value into equatioo 1, the k /~ ratio was obtained. The saturated perme~ility was then estimated by mltiplying the flow parameter by the estimated effective porosity .,--... 0::: :r: '-.... :::::2: :::::2: -.........- X =:> _J LL.. 3:: 0 _J LL.. ~ =:> 0 0::: 0 ~ w :::::2: .,--... 0::: :r: '-.... :::::2: :::::2: -.........- X =:> _J LL.. 3:: 0 _J LL.. 1-=:> 0 0::: 0 ~ w :::::2: 10 9 8 7 6 5 4 3 2 08 10 9 8 7 6 5 4 3 2 08 4-27-85 DEPTH= 0.5 M 10 12 14 16 18 20 22 24 2 4 6 TIME OF DAY (HRS) 3-24-84 DEPTH= 2.2 M 10 12 14 16 18 20 22 24 2 TIME OF DAY (HRS) Figure 1. Melt and ontfiow flux for two days at the Central Sierra Snow Laboratory, Soda Springs, California. At A. a flux of 2 11111/hr has a travel time of over 3 boors through ~ck 0.5 m in depth, and the wave speed is 5.2 x 10 m/s. 6 1/3 k = ( ! . ~ ) 3 (5) est p e t e es 8 8 Average porosities and permeabilities were an n value of 3 in equation 1 (Figure 2), but higher n values might be mre suitable. ~ slopes are based on flux values of 1.11 x 10 m/s (4 nm/h) or less. Slopes ranged fran 0.63 to 1.02 and had a mean value of 0. 79 and a standard deviation (n = 14) of 0.11. No seasooal trend was apparent (Table 2), and even on successive days values were variable mi. occasiooally quite different. When flux rates in excess of 4 11111/h were included in the least squares fitting, the calculated slope value increased. calculated for each of the 14 events. Slope The slopes of the lines in the log-log plots of flux and wave speed are indicative of the suitability of 362 While the portions of the curves that were analyzed were all monotonically decreasing (see Figure 1 for saDples), one event (4-15-84) produced a slope value in excess of 1.0 (Table 2). Values greater than 1.0 indicate increasing flux rates ml 16 4 ~r-----r---.-----.---~~--~~~~ ~-I - co ,.__ co -e-~--~~--7-----~--~~··~~~~~ 10 1 1.5 2 3 4 5 6 7 8 9 1 0 10"8 10"7 MELTWAVE FLUX (M/S) Figure 2. Calculated meltwave flux values and observed wavespeeds for 4-27-85. Line has theoretical slope of 0.67, While fitted value is 0.70. iDply the formation of a shock wave, Which was clearly not the case. For this reasoo as well as for carparability with other literature values, a slope of 0.67 was used in equation 1 for the derivation of fd and k. The actual slope values were tested, buf they were not used because they caused the effective porosity and the penneability/porosity ratio to deviate uarkedly fran expected and literature values. <kJce flux rates dropped below approximately 4 um/h, a divergence between melt and outflow curves was observed, and ooly those points fran the diverging curves were used in the calculatioo of slopes, porosity, and penneability. Steep slopes, often in excess of 1.0, resulted fran including fluxes greater than 4 DID/h. Plots of melt and outflow flux often showed a constant delay (parallel lines) fran the peak values down to about 5 DID/h (Figure 2, 4-27-85) • This behavior fails to confonn with theory that higher flow speeds result fran higher flux values and higher water saturatioo values. The 4-27-85 ~t 's outflow flux decreased fran 7 DID/h (1.9 x 10 m/s) to 3.5 DID/h, yet the calculated_.rve speed c~ insignificantly fran 8.2 x 10 to 8.0 x 10 m/s. This situatioo was seen in over a quarter of the events analyzed. TABLE 2. Depth, density, effective porosity, flow parameter, permeability, and derived slopes for study intervals at the Central Sierra Snow Laboratory, Soda Springs, California. Date Depth (m) Density 3 kg/m k -9 2 (x10 ,m ) Derived Slope ---------------------------------------·--- 3-23-84 3-24-84 4-14-84 4-15-84 4-22-84 5-11-84 5-12-84 3-31-85 4-5-85 4-6-85 4-13-85 4-14-85 4-27-85 4-28-85 2.2 2.2 1.9 1.9 1.8 .9 .8 2.1 1.6 1.5 1.0 .9 .5 .4 420 420 420 430 420 440 440 340 380 380 390 390 410 410 0.55 .55 .55 .54 .55 .52 .52 .63 .59 .59 .58 .58 .56 .56 363 2.6 2.7 2.5 2.5 2.3 1.7 1.5 4.1 2.6 2.9 2.7 2.2 1.5 1.2 3 3 3 3 2 0.8 .5 18 4 9 4 2 .6 .3 0 81 .64 0 83 1.02 .82 .84 .79 .67 .75 .76 .95 .86 .70 .63 The value of the exponent n has been extensively debated because it is crucial in converting saturated pemeabilty to lDlSaturated pemeability. Because the slopes of the lines should be 0. 67 if the value of n in equation 2 is 3, these slopes fail to support the choice of n as being equal to 3 (Table 2 and Figure 3). Colbeck and Anderson (1982) found an n of 3.3 but concluded that, due to the ease of calculations with the integer value and the marginal i.q)rovement in fit, the value of 3 was preferable. Literature values of n range fran 1.4 to 4.6 (Table 3). Values of n for sand sauples n >3.5 2.8 4-5 1.4-4.6 Table 3. Literature values of n for water flow through snow. Sanple Source Spring snow Jordan, 1983 Glacial firn Anbach et al., 1981 Derived Morris and Godfrey, 1979 MetaJIX)rphosing snow Denoth et al., 1979 varied fran 2.5 to 7 and were even greater for fine textured soils (Mualem, 1978) • Our mean slope of 0. 79 can be used to estimate an n value of 4.8. Effective Porosity Because effective porosity is inversely related to snow density when the irreducible water saturation is held constant, effective porosity declined fran 0.65 3to 0.52 as density increased fran 330 to 440 kg/m during the spring melt and densification of the snowpack. Accurate estimation of effective porosity is dependent on accurate estimations of density and irreducible water saturation. Average density of the snowpack was measured as accurately as possible with an isotopic density gauge 1 to 6 m fran the snO\I.IIIelt lysimeters. In the range of densities encountered in the study, meJsured densities should have been within +20 kg/m of the actual density. This level of uncertainty affects calculated effective porosity to the extent of :!9.02. While irreducible water saturation was assumed to be 0.07 (Colbeck and Anderson, 1982), it is almost certainly between 0.04 and 0.10. Values of 0.02 and 0.08 have been considered the limits of irreducible water saturation (Anbach et al., 1981). Thus, uncertainty in the calculated effective porosity due to uncertainty in the irreducible water saturatim is less than -+0.03. Because of the uncertainties due to density·--;.,asurement and the value of irreducible 364 'VI ......... ~ Q h:l a.. ~ 3' 10·3 .... :2 ~ ., .... .... ..... .., Figure 3. Wavespeeds fran fluxes between 0.5 and 4 um/hr. Line has slope of 0.67, While fitted slope is 0.79. water saturation, the seasonal changes in the estimated porosity values should be viewed only as trends. In turn, uncertainty in the calculated effective porosity leads to uncertainty in the pemeability. Calculated pemeability was found to be fairly sensitive to effective norosity, decreasing fran 5.8 x 10-9 to 4.5 x 10-9 m2 as effective porosity c~ed fran 0.60 to 0.55 (a density change of SO kg/m ) • Therefore, uncertainties about density and irreducible water saturation contribute an uncerta~ty2 to calculated pemeability of about ±1.5x10 m in the range of values encountered. Pemeability The equivalent snowpac~ ~ability values ranged fran 0.3 to 18 x 10 m , and agree well with lll)St of the reported values (Table 1) • The pemeability values showed a strong seasonal trend in both years: pemeability decreased as the snowpack transmitted water over time and became mre dense (Table 2). The trend was 1mifom in 1984, with the values changing almost an order of magnitude fran March to April. Colbeck and Anderson (1982) also found pemeability decreased with increases in density. In 1985, little surface melt occurred 1mtil the end of March, and the high value on 3-31-85 may indicate that little water flowed through the snowpack before this time. Further, low peak flux values (less than 3.5 um/h) may have cxmprani.sed the calculatiooal teclmique s<:lne't\bt. 'lbe snowpack melted out earlier in 1985 than in 1984, but the end-of-melt permeability values were similar. Tbe ratio parameter, k113 td decreased "'e' during both melt seasons, supportmg Colbeck's (1978) hypothesis that this parameter characterizes SDOWpack properties and changes with time. Furt~f3 the values in both seasons were close to 2.8 m when regular daily melt outflows wer~13 first observed, and fell to approx:i.Dately 1.5 m when the snow depth neared 0.5 m. The differences in this parameter within the pairs of data is of a DBgnitude ccmparable to that found by using 1954 <B. lysimeter data (Colbeck and Andersoo., 1982). 'lbe teclmique euployed herein is siuply not precise enough to eliminate variations of this magnitude, and additiooal seasons will be required to coo.finn the observed seasooal trend. Wave Speed Wave speeds (time between a given flux at the surface and an e~l value of outflow, divided into the snowpack depth) were calculated for fluxes between 1 and and 4 rrm/hr. 01 a few of the days, peak flux was less than 4 rrm/hr, and oo.ly two wave speeds were calculated. In 1984, wave ~eds were approximately 6, 10, and 18 x 10 m/s at respective fluxes of 1, 2, and 4 rrm/hr in March and declined to about one-half of these values for the same fluxes by May. Wave speeds in March were saaewhat higher in 1985 than in 1984 but dropped to essentially __Jhe same May values as in 1984 (3, 5, and 9 x 10 m/s at fluxes of 1, 2, and 4 rrm/hr, respectively) • These May 1985 wave speeds were about oo.e-third of their March counterparts. Jo~ (1983) found average wave speeds of 3 to 10 x 10 m/s at fluxes of 1 to 3 rrm/hr during the late SliOIIIDel t season. Tbe decline in wave speed over the course of the melt seasons was opposite what might be expected fran decreases in porosity aloo.e. Wave speed is inversely related to effective porosity (Colbeck, 1978) so that as porosity declines in the melt season, wave speed shoold increase. Based on Colbeck's eqoatioo for wave speed at a given flux, f"!9a f~ux of 3.3 rrm/hr and a permeability of 0.6 x 10 m , wave speed increased insignificantly fran 23 cm/hr to 26.5 cm/hr for a decline in the effective porosity of 0.65 to 0.55 (Price, 1977). In 1985, as effective porosity dropped fran 0.64 to 0.56, we observed a decrease in wave speed fran 68 cm/hr to 22 cm/hr at a flux of 3 rrm/hr. We eypothesize that the observed decrease in wave speeds is due to the transition fran a channel-dcminated flow system in the early season to II01'e widespread matrix flow in the later part of the 365 melt period. If the same surface flux is concentrated in channels of small areal proportion instead of distributed over most of the snow, then flow would be more efficient and faster in the channels than throughout the snow volume. The channels 'WOUld tend to consist of larger diameter grains (and tlms larger pore spaces) and 'WOUld have higher water saturations than the entire mass of snow. Both of these conditions favor increased flow efficiency. Equivalent permeabilities of spring snowpacks ~ a2 Sierra Nevada site declined fran 5 to 10 x 10 m -ft ~ onset of cootinued melt to about 0.5 x 10 m before snow 1/Nover disappeared. Similarly, the ~~Yf k /-e decreased f!Jfl ~t 2.8 X 10 m to abOUt 1.5 X 10 m over the two melt seasons. A value of this parameter, interpolated fran the results presented here, could be used in meltwater routing procedures, as suggested by Colbeck (1978) • The exponent n, relating intrinsic permeability and effective saturatioo. to unsaturated permeability, was found to be higher than the camnnl.y used value of 3 and to be strongly dependent en the flux range chosen for analysis. As expected, the effective porosity decreased during both seasons, but by only a small 81001Dlt. The results presented here support Colbeck and Andersoo.'s (1982) conclusioo that deep Sierra Nevada snowpacks drain more slowly than reported by Bengtsson (1981) • However, they seem to drain two to three times more rapidly in the early part of the melt season than they do near the end of the season. We hypothesize that as flow becanes more unif01m throoghout the snowpack with aging, flow is less efficient and a greater volume of snow drains at a slower rate than when channels daninate the flow. 1lms, channeling, in additioo. to layering and densificatioo., DllSt be coosidered in defining an equivalent permeability for the snow cover. dz/dtl k k 'W u wave speed or rate of propagation of a given value of u. 2 saturated permeability, m • unsaturated2or relative permeability to water, m • exponent, often assumed to equal 3. effective saturatioo.. water saturation as fraction of pore volume. S . irreducible water saturation, W1 g gravity. 9.8 m/s. t time. s. 3 2 u volume flux of water. m /m -s. z depth. m. 3 Pi ice density. 0.917 Mggn • p water density. 1 M§/m • P 8 snow density. Mg/m • ~ ~sity. fie effective porosity. -4 2 J.l viscosity. 10.1 x 10 ~/m pg/J.l flow factor. 5.47 x 10 /mrs. Aubach. W •• M. Blumthaler. and P. Kirchlechner. 1981. Application of the Gravity Flow Theory to the Percolation of Melt Water through Fim. Journal of Glaciology 27(95):67-75. Anderson. E.A.. 1979. Streamflow SilJillation Models for Use on Snow Covered Watersheds. In: Proceedings Modeling of Snow Cover Runoff. S.C. Colbeck and M. Ray (Editors). U.S. Army Cold Regions Research and Fngineering Laboratory (Cima). Hanover, New Hanpshire. pp. 336-350. Bader. H., R. Haefeli. E. Bacher. J. Neher, 0. Eckel, and C. Thams, 1954. Snow and Its Metauorphism. U.S. Army Corps of Fngineers, SIPRE Translation 14. 303 pp. Bengtsson, L., 1981. Snov.melt Generated Rlmoff fran &!all Areas as a Daily Transient Process. Geophysica 17(1-2) :1~122. Colbeck, S.C., 1972. A Theory of Water Percolation in Snow. Journal of Glaciology 11(63):369-385. Colbeck, S.C., 1978. The Physical Aspects of Water Flow through Snow. Advances in Hydroscience 11:165-207. Colbeck, S.C., 1979. Water Flow through Heterogeneous Snow. Cold Regions Science and Technology 1:31-45. Colbeck, S.C. and E.A. Anderson, 1982. The Permeability of a .Melting Snow Cover. Water Resources Research 18(4) :904-908. Colbeck, S.C. and G. Davidson, 1973. Water Percolation through lbJI:>geneoo.s Snow. In: The Role of Snow and Ice in Hydrology: Proceedings of the Banff Syuposium, ll'mn-IAHS. pp. 242-256. Conway, H. and J. Abrahamson, 1984. Air Permeability as a Textural Indicator of Snow. Journal of Glaciology 30(6):328-333. Denoth, A. and W. Seidenbusch, 1978. A .Method for the Determination of the Hydraulic Conductivity of Snow. zeitschrift for Gletscherknnde und Glazialgeologie 14(2):~213. Denoth, A., W. Seidenbusch, M. Bl\llllthaler, and P. Kirchlechner, 1979. Sane Experimental Data on Water Percolation through llcm:>geneou.s Snow. In: Proceedings Modeling Snow Cover Rlmoff, S.C. Colbeck and M. Ray (Editors). U.S. Army ~. 366 Hanover, New Hanpshire. pp. 253-356. Donne, T., A.G. Price, and S.C. Colbeck, 1976. The Generaticn of Rlmoff fran Subarctic Snowpacks. Water Resources Research 12(4) :677-685. Gerdel, R.W., 1954. The Transmissicn of Water through Snow. Transactions American Geophysical Unicn 35: 475-485. Jordan, P., 1983. Meltwater Movement in a Deep Snowpack 1. Field Observations. Water Resources Research 19(4) :971-978. Kattelmann, R.C., 1986. Measurements of Snow layer Water Retenticn, Proceedings AWRA Cold Regioos Hydrology Syuposium. (accepted) Kattelmann, R.C., B.J. McGtrk. N.H. Berg. J.A. Bergman, J .A. Baldwin, and M.A. Hannaford, 1983. The Isotope Profiling Snow Gage: Twenty Years of Experience. In: Proceedings of the 51st Westem Snow Conference. pp. 1-8. Knroiwa, D., 1968. Liquid Permeability of Snow. !Am Publication 79. pp. 3~391. Morris, E.M. and J. Godfrey, 1979. The Fnropean Hydrological System Snow Routine, In: Proceedings Modeling of Snow Cover Rlmoff, S.C. Colbeck and M. Ray (Editors). U.S. Army amiL, Hanover, New Haupshire. pp. 269-278. Mualem, Y.. 1978. Hydraulic Cooductivity of Unsaturated Porous Media: Generalized Macroscopic Approach. Water Resources Research 14(2) :325-334. Price, A. G •• 1977. Snownelt Runoff Processes in a Subarctic Area. McGill University Sub-Arctic Research Paper 29. 106 pp. Shimizu, H •• 1970. Air Permeability of Deposited Snow. low Teq>erature Science, Series A 22:1-32. Smith, J .L •• 1974. Hydrology of Wam Snowpacks and their Effects upon Water Delivery ••• Sane New Concepts. In: Advanced Concepts in the Study of Snow and Ice Resources. H.S. Santeford and J.L. Smith (&litors). Naticnal .Academy of Sciences, Washington, D.C. pp. 76-89. Thcker, W.B., Ill and S.C. Colbeck, 1977. A Calputer Routing of l.bsaturated Flow through Snow. u.s. Army ama Special Report 77-lo, Hanover. New Haupshire. 39 pp. Wankiewicz, A., 1979. A Review of Water Movement in Snow. In: Proceedings Modeling Snow Cover Runoff. S.C. Colbeck and M. Ray (Editors). U.S. Army rnREL, Hanover, New Haupshire. pp. 222-252, JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 IN SITU ELECTRICAL MEASUREMENTS OF SNOW WETNESS IN A DEEP SNOWPACK IN THE SIERRA NEVADA SNOW ZONE OF CALIFORNIA 1 James A. Bergman ABSmACr: A twin-disc sensor that measures changes in snow capacitance has been developed for the detemination of in situ snow wetness. In its current fonn the sensor is capable of measuring water in transit fran rain-on-snow and spring melt, surface diumal melt water flux, and the gradual wetting of snow as the winter seasm progresses. Wetness increases of up to 2 percent by volume and subsequent snowpack drainage has been measured during and after rain-on-snow. Surface diumal melt water flux ranging fran 0.5 to 2 percent volume wetness has been measured consistently by sensors near the snow surface. All buried sensors indicated an average seasonal progressive snow wetting of about 2.5 percent volume wetness. A sensor was calibrated in snow by spraying incremental additions of water on a sensor-snow bl~. A good linear response was indicated by an r of 0.998. Based oo snow calibration values, sensor volume wetness C<qlared favorably with freezing calorimetry volume wetness. After 35 COOJ>arative measurements, wetness fran the sensors averaged 2. 6 percent while calorimetry volume wetness averaged 2 percent. (KEY '!BifS: capacitance; snow wetness; snow; free water; Sierra Nevada.) The threat of damaging floods fran excessive snowpack outflow caused by rain-on-snow and excessive spring melt camined with ever increasing demands on the variable Sierra Nevada snow zone water supply, have created a need for in-place uxmitoring of the mre elusive snow-water relationships. Snow wetness (free water, liquid water) and water flow routing through snow are two of the mst difficult snowpack properties to measure. During rain-on-snow and spring melt, the speed at which these two COOJ>OD.ents increase may indicate the severity of snowpack outflow and when it may occur. 1be met8D¥>rphic state of a snowpack can significantly affect the quantity and timing of snowpack outflow. Internal snowpack structure and layer developnent are a potential hinderance to water flow through snow (Berg, 1982). 1be structure of layers in the early winter snowpack may not be as caJJ>licated as the structure that develops by early spring. By mid-to-late season there may be as many as 10 to 1S major stonn layers and associated ice lenses, creating a highly structured snowpack that is uore likely to retard rain and meltwater flow. Flow channel developnent may reduce the time it takes for rain and meltwater to reach the soil (Kattelmann, 1985) • Rain-on-snow in the Sierra Nevada snow zone occurs JOOStly in Deceuber and Janoary, before snowpack structure becaoes very COOJ>licated. Flow channels produced by rainwater flowing in this portion of the snowpack may remve sane of its retarding effect on subsequent water flow. 'Ibis inconsistent snowpack water flow and the seasonal variation in snow acC1l1111lation create a need for methods that measure water in transit and suspended liquid water within the snowpack. Most measurements of snow wetness, including water in transit (surficial meltwater, rainwater) euploy saupling methods that destroy the snowpack. 'lbey include heating and freezing calorimetry, dilutim of an acid or dye, centrifugal extraction, and portable plate and cam shaped capacitors (resonators) of various sizes. During the saupling procedure, the extraction of the snow sauple interrupts the natural water flow paths within the snowpack. 'Ibis COOJ>ranises the integrity of the snowpack and excludes that portim fran further free-water analysis. 1be DBjority of these methods are excellent for point wetness measurements and sensor calibration but they cannot be used to continoally track in situ snow wetness and water in transit throughout the winter. During the past decade, methods have been developed to measure in situ snow wetness in essentially undisturbed snow. Linlor (1900) ~logist, Pacific Southwest Forest and Range Experiment Station, Central Sierra Snow Laboratory, P.O. Box 66, Soda Springs, California 95728. 367 developed an electranagnetic system for measuring snow wetness based on the attenuation of microwaves in the 4 to 12 Wz frequency range. Over these frequencies, the attenuation of microwaves by snow varies proportionally with the 8100Ullt of liquid water associated with the ice. A system of horizontally opposed transmitter-receiver pairs was installed at different heights above the soil surface before snowfall to track the uovement of free water with the sensor array. Although this method showed pranise, it was COIJ)lex, inconsistent and data reduction was difficult. Linlor also tried to determine gross snow wetness by placing a transmitter on the soil surface and measuring the vertical microwave attenuation by the snowpack. The main problem encountered with this method was the deflection/reflection of the microwave beam by intervening ice lenses. The results were inconsistent and the measurements were difficult to obtain. Snow wetness has been studied by measuring changes in the electrical potential or capacitance of snow (Ani>ach and Denoth 1972; Denoth, et al. 1984; and Linlor and &nith 1974). Snow capacitance was measured by inserting closely spaced parallel plates of various sizes into the snow. Radio frequencies used varied between 1 and 20 Mlz. At frequencies below 10 MHz, grain size and shape canbined with the quantity of liquid water present to influence the capacitance reading. At frequencies above 10 Mlz, the effect of grain size and shape was effectively eliminated. The parallel plate sensors "WOrk well when inserted into the snow, but because they are destructive, they do not appear to be suitable for in situ wetness measurements. This paper describes the developnent of a permanently mounted electrical instrument with widely spaced discs that measures in situ snow wetness by m:mitoring changes in snow capacitance. In theory, the instrument was modeled after hand held plate resonators. Dt:y snow, as a conductor, has a specific capacitance for a given induced charge. The ability of snow to carry an induced charge is determined by its wetness. Therefore, given a specific charge, the potential of snow to carry this induced charge should be directly related to the amount of liquid water in it. Capacitance may be stated as the ratio of a charge on a positive plate of a condenser to the difference of potential fran a corresponding negative plate. lbus, snow wetness can be determined by measuring capacitance between t'WO closely spaced parallel plates or by measuring the 368 field capacitance surrounding thin discs. The electrical field flux surrounding a thin disc is the basis for this sensor design. The field extends outward from the flat outer portion of each disc and then loops back toward the negative side of the sensor, in this case the mast (Figure 1). Field flux fran the inner portion of the discs has a DDre direct path to the negative side of the sensor. Significant field flux occurs within approximately 3 radii of each disc and tlms covers a larger volume of snow than the measurement field of parallel plate capacitors. A twi.n-1l!sc sensor was designed to measure a large (0.04 m ) volmte of snow f1i1e withstanding snow loads of up to 0.15 Kg/ em • The discs are Inner coaxial cable conductor Figure 1. Three Dimensional Electrical Field Surrounding a Twin-i>isc Sensor. fabricated fran T6 aluminum, 3 mn thick and 10 an in diameter. An acrylic spacer, 2 an thick and 10 an long, attaches and electrically isolates each disc an opposite sides of a T4 alumi.mln channel mast (Figure 1). The discs may be considered as the positive side of the sensor and the mast the negative side. During the 1983-84 winter, eight sensors were munted 61 an apart an a 5 meter high mast, beginning at 30 an above the soil surface. During the 1984-85 winter, each sensor was IOOOD.ted oo. its own mast and three measurement depths were duplicated. For the 1985-86 winter, three of the sensors had their 10 an discs replaced with discs 15 an in diameter to increase the resolutioo. of the sensor. The capacitance field surrOWJding each twin-disc sensor is measured by a Hewlett-Packard, Model 4342A, Q meter. (Trade names and coomercial products are mentioned only for infonnatian. No endorsement by the U.S. Department of Agriculture is inplied.) It is a stable solid state voltmeter which is connected across an internal variable capacitor (portioo. of the tuned circuit), to measure the reactive voltage in terms of the circuit Q. The coil portion of the tuned circuit is connected externally and represents the unknown to be measured. By inserting low :iq)edance in series with the coil and high inpedance in parallel with the capacitor, the parameters of the unknown ca1p00.ents (twin-disc sensor) can be measured in terms of their effect an circuit Q and resonant frequency in units of picofarads (pf) • Accuracy of the Q meter capacitance circuit is .:!:.().1 pf. To mdntmdze distortion of the capacitance readings, the effect of the coaxial transmission line m signal stability was detennined. Ideally, the signal will not be distorted if the standing wave ratio (SWR) of the transmission line remains bel<7of 0.25. Standing wave ratios of 0.25 to 0.50 my cause a slight distortion of the capacitance readings. During winter 1983-84 a wavelength of 555 meters, which corresponds to a frequency of 5.4 KHz, was used to measure snow capacitance. The SWR of the transmission lines. whose length varied fran 8.3 to 12.6 meters, was 0.015 to 0.023. The frequency of 10.7 mz, which corresponds to a wavelength of 30 meters, was used in winter 1984-85 and is being used for 1985-86. At this frequency, ~'s for the signal transmissioo. lines vary fran 0.27 to 0.42 indicating a possibility for distortion of the capacitance readings. Distortion caused by signal line SWR's above 0.25 shonl.d remain relatively constant and be directional. If this is the case, then capacitance readings fran all sensors will be slightly distorted in cne direction oo.ly, causing the measurement peak 369 to shift in the directioo. of the distortion. This should not significantly affect sensor response to changes in wetness. Capacitance readings may also be influenced by the size and shape of snow particles. Because snow metamorphosis is continual, its effect may vary and be amidirectianal. And because the directioo. in which the measurement peak shifts is not known, recorded sensor response to changes in wetness is sanewhat unreliable. For this reason, frequencies above 10 Hlz are used for measuring snow capacitance. Since all sensors are made fran the same materials, the calibration results fran one twin disc sensor is used to represent all sensors. In February 1984, me sensor attached to the center of a 76 an portion of the mast material, was inserted into a cubic meter box of kiln dry fine sand. Measured amounts of water in increments equal to 1. 7 percent by volume, were added to the sand until a maxinun detectable wetness of about 7 percent was reached. Due to scale ltmdtations, the Q meter could not accurately detect changes beyond 7 percent volume wetness in sand. After each incremental additioo., the admixture was stirred to distribute the misture evenly. At each wetness level capacitance measurements were cootinually made until sensor variation and instrument drift were accounted for (Table 1) • A plot of the sand wetness calibration values iqpicated a good linear fit (Figure 2) • An It of 0. 822 denotes the relationship of sand wetness to capacitance change. Due to Q meter detection ltmdts, this correlation could not be continued past 7 percent volume wetness. The average capacitance change for each 1 percent volume wetness increase in sand is 12.1 pf. During February 1985 the sensor that was calibrated in sand was tested for response to liquid ~ter in snow. After completely saturating a 0.027 m snow block, a mdninum capacitance reading was obtained. Capacitance readings were then obtained after a 45 minute drainage period and again after overnight drainage. Response was quite abrupt when a large quantity of water was applied (Table 2). After 45 minutes, capacitance values were close to those at the beginning of the test. In early March 1985, the same sensor was calibrat~ in a 46 an by 46 an by 31 an volume (0.064 m ) of mderately dry snow. Water was applied with a small adjustable hand sprayer in increments equivalent to 0.5 percent volume wetness or 320 ml. After each application of water. capacitance values were ooserved until a mdninum TAIIE 1. Sensor Calibration, in Sand, at a Frequency of 5.4 KHz. Wetness Percent Volume No. of Meas. Total Replicate Mean Capacitance Meas. (pf) Oumge (pf) 0 1.73 3.46 5.19 6.92 355 365 r 5 6 6 27 61 Increasing sand wetness c '0 • " 375 385 ~ 395 u ... .e ~ 405 ~ u • g. 415 u '0 iii "' 425 435 I I 455 "' I I I / I I ,J I I I 1. 73 I I I I I / ./ / I 444.6 422.5 395.8 373.6 361.5 .. / / / / I / / I / ----Perfect fit I I / I. • Sand wetness values 3.46 5.19 6.92 Sand volume wetness (%) Figure 2. During Sensor Calibration in Sand, Capacitance Change and Sand Volume Wetness were Closely Correlated. TAIIE 2. Sensor Response Test, in Snow, at a Frequency of 10. 7 1\llz. 0 22.1 48.8 71.0 83.1 Snow Measured Total Difference Wetness Capacitance Capacitance Between (pf) Cbange (pf) Measurements Air 181.6 0 0 Mod. Dry Snow 160.2 21.4 21.4 Saturated Snow 91.5 90.1 68.7 Drain 45 Min. 145.1 36.5 -53.6 Drain Overnight, Sub-freezing 167.0 14.6 -21.9 reading was obtained. Applicatioo. ceased after snow wetness reached 3.75 percent by volume. The sensor-snow block was measured after draining for 45 minutes and after draining overnight at 370 Difference Standard Standard Between Deviation Error of Meas. s MeanS- X X 0 0.075 0.033 22.1 0.373 0.152 26.7 1.377 0.562 22.2 6.226 1.198 12.1 6.583 0.843 subfreezing air te~~peratures. 'Ihe relationship between wetness and .yapacitance change showed excellent linearity (lt'=0.994) (Figure 3, Table 3). 1he capacitance reading after the sensor-snow block had drained for 45 minutes indicated that the snow retained about 1 percent by volume of the applied water. But, overnight drainage at subfreezing tetll)eratures reduced snow wetness to below the initial level at mich snow calibration began. Snow drainage can be rapid (Tables 2 and 3), 1he capacitance values were very close after draining for 45 minutes and overnight. Air and overnight drainage capacitance values, in both tests, were al.toost identical, indicating good repeatability of the sensor-snow measurement system. 'Ihe average capacitance change for each 1 percent volume wetness increase in snow was 3. 75 pf, 135 140 145 . il ~ 150 .§ .e ~ 155 ~ g,l60 ~ 170 175 '180 . --~--,.. -- ~---· -----_.,- --~ .... --r Increasing snow wetness ----Perfect fit • Snow wetness values 0 0.25 0.75 1.25 1.75 2.25 2.75 3.25 3.75 Snow volume wetness ('7o) Figure 3. During Sensor Calibratioo. in Snow, Capacitance Oumge and Snow Volume Wetness were Directly related. TAliE 3. Sensor Calibration, in Snow, at a Frequency of 10.7 mz. Wetness Measured Total Difference Percent Capacitance Capacitance Between Voll.UIIe Change Change Measurements (pf) (pf) Air 181.7 0 0 lht. Dry Snow 153.1 28.6 28.6 0.25 152.7 29.0 0.4 0.75 150.7 31.0 2.0 1.25 149.0 32.7 1.7 1.75 147.1 34.6 1.9 2.25 145.6 36.1 1.5 2.75 144.0 37.7 1.6 3.25 141.4 40.3 2.6 3.75 139.0 42.7 2.4 Drain 45 Min. 148.2 33.5 ~.2 Drain Overnight, Subfreezing 165.6 16.1 -17.4 Results of the sand and snow calibrations show that sensor response to 'Wetness change is DJJCh greater in sand than in snow. Sensor response is influenced by the dielectric constant of the measured mterial. The dielectric constant of sand (quartz) is about 4.2 while that for snow (ice) is about 88. This may cause a difference in the electrical absorptivity of the two materials when mixed with water. Sand and snow have dissimilar water retention properties, and surface tension characteristics. This may cause oore water to be retained by sand for a given vol1lllle, resulting in greater sensor response. Because snow is the measured medium, wetness values fran the calibratioo in snow will be used. Capacitance is being measured at the USDA Forest Service's Central Sierra Snow laboratory (CSSL), located at 2100 m eleva tim in the north central Sierra Nevada of California (Figure 4). The influence of maritime air masses at this location cause winters of long duration and deep snowpacks. Snow ca~prises about 85 percent of the total yearly precipitation. SnowstolDIS are usually large, intense and of low 'Wetness. Midwinter rain-on-snow nonually occurs at least once per season. The ... Castle Peak Central Sierra Snow Lab (2100 ml 0 10 20 30 Km "' •"' CALIFORNIA "' •"' "'"'"'"' • •"' • • • Figure 4. The Central Sierra Snow Laboratory is I..Dcated at 2100 m Elevation, in the North Central Sierra Nevada,Near Soda Springs, California 371 ... 0. 2 km seasonal snowpack. begins to acCUIIIllate in Noveui>er, with lllljor acClmlllation fran January tbrongh March. An average peak snow depth of 305 em and snow water equivalent of 90 em is reached by early April. ImErS AN> DISCVSSI<N 1983-84 Snow Season Eight sensors were installed on a 5 m high mast. Initial data indicated that the buried sensors responded to a wetness increase, fran water in transit, during a 3.2 em rain-on-snow event in late Deceuber 1983. Three sensors located at 30 em, 90 em, and 150 em above the soil surface showed an increase in wetness (capacitance decrease) during rainfall (Figure 5) • 'Ibe sensor nearest the snow surface decreased by 8.9 pf, which is equivalent to a 2.4 percent volume wetness increase when based on snow calibration values. 'Ibe sensor nearest the soil surface decreased by 2.4 pf, indicating a liquid water increase of 0.65 percent volume wetness. And, the sensor buried midway in the snowpack showed a 6.1 pf decrease, for a 1.6 percent increase in volume wetness. Mter rain ceased, the sensors regained about bwrthirds of the lost capacitance indicating that sane snowpack drainage had occurred. Decreasing wetness fran the snow surface downward may be due to the establishment of water flow paths (Kattelmann, 1985) • 'Ibe greater lllliiDer of surficial flow channels caused by the mechanical 1.8 1.5 1.2 0.9 0.6 0.3 ~ 0 ~ 1.4 ~ 1.2 ~ 0.9 ~ 0.6 0.3 0.8 0.6 0.3 90 (M SEI<;(R RAIN 10 15 20 r.t::c:nmE:R, 1983 Figure 5. Three .lllried Sensors responded to a 3.2 em Rai.n-<kl-Snow event in Deceuber 1983. 372 action of rainfall may be considerably reduced by the time the rainwater reaches the lower portions of the snowpack. As the smaller flow pathways cooi>ine and fOJ:ID larger channels, the distance between them increases causing liquid water to beccme more difficult to detect because of a greater amomtt of dry snow between the wet pathways. 'Iberefore, more than one sensor at each level may be necessary to IID!litor suspended and flowing water accurately. Freezing calorimetry measurements were made during May to check sensor operation (Table 4). Sensor wetness was detennined by subtracting SIIOII' capacitance fran capacitance in air, assumi.ng that the capacitance of totally frozen dry snow is close to that of air. Based on snow calibration values, data indicate good correlation between avera~ wetness fran all sensors and average wetness fran freezing calorimetry. These results need to be put in the proper context: snow sauples used in freezing calorimetry were obtained fran an area of snow 10m away fran the sensors. and freezing calorimetry wetness may vary as DllCh as .:!:. 0. 5 percent by volume because of user error and equipnent limitations (Bergman, 1978). 'Ibe above factors are likely to cause most of the variation between individual ccq>arative measurements. One of the basic problems of mounting a large number of sensors on a single mast is weight and snn copping of snow around it. Doe to reflected solar radiation, large melt cups formed around the mast midway through the winter. 'Ibe melt cup my have affected the flow of water around the sensor mast by acting as a large flow path, causing ovenneasurement. 'Ibe melt cup probably reduced the magnitude of measured diurnal meltwater flux because the sensor became exposed while it was below the snow surface. 'Ibe height of the mast canbined with the weight of the sensors caused considerable swaying in high winds. This tended to cause air gaps around the sensor nearest the snow surface, resulting in lower wetness readings at that level. 'Ibe temperature of the Q-meter was found to affect sensor capacitance values. Colder Q meter temperatures resulted in lower capacitance readings and vice versa. To c~ate for this effect, capacitance values fran a twin-disc sensor always in air were averaged. Capacitance values fran all other sensors were adjusted according to the deviation of the sensor in air fran its average over the measurement season. 1984-85 Snow Season To eliminate the influence of sun cupping and swaying of the mast, each sensor was IOOIJllted on top TAII..E 4. ~ison of Sensor Wetness and Wetness fran Freezing Calorimetry. ~te Height Sensor wetness Above Sand Snow (em) · Percent VoltDDe Calorimetry Wetness Percent VoltDDe -Measured at a frequency of 5.4 KHz.- 4/20/84 30 0.6 1.9 1.4 7t 90 o. 7 2.2 0.4 7t 210 0.5 1.5 2.4 4/23/84 30 0.8 2.1 0.4 7t 90 0.8 2. 7 2.1 4/27/84 30 0.6 2.1 0.9 7f 90 o. 7 2.3 2.5 4/30/84 30 o. 7 2.1 3.1 7f 90 0.8 2.4 1.3 S/00./84 30 0. 7 2.1 1.0 7f 90 o. 7 2.3 o. 7 S/04/84 30 ·o. 1 2.3 o.9 7f 90 o. 7 2.3 o. 7 S/10/84 30 0.8 2.4 2. 7 S/11/84 30 0.8 2.5 3. 7 S/16/84 30 .JWL _b.L _LL Partial Ave. 0. 7 2.3 1. 7 -Measured at a frequency of 10.7 :r.t:Iz.- 4/18/85 30 0.9 3.0 2.5 7t 30 1.0 3.1 2.5 7t 90 0.8 2. 7 2.0 7f 90 1.2 3. 7 2 •. 0 7f 30 0.9 3.0 3.6 7f 30 1.0 3.1 3.6 7f 90 0.9 2. 7 3.3 4/19/85 30 0.9 2.9 3.1 7f 30 1.0 3.1 3.1 7f 90 0.8 2.5 2.9 4/22/85 30 0.9 3.0 1.8 7f 30 1.0 3.1 1.8 7f 90 o. 7 2.3 2.2 4/26/85 30 0.9 2.9 0.5 7f 30 _LQ_ _b!_ ....Q.,1_ Partial Ave. 0.9 2.9 2.3 Overall Average 0.8 2.6 2.0 of its own mast. ~e buried. the mast stopped swaying and the sensor was not exposed Wltil SllOIIIIIelt. To better represent the snowpack and to check sensor response and repeatability. duplicate measurements were made at 30. 60. and 90 em above the soil surface (Table 4). &cavation of a sensor near the snow surface indicated that sensor-snow contact was excellent. All disc. spacer. and mast surfaces showed canplete contact with the snow. Even the snow directly below the 2 an thick spacer mintained good contact. Apparently. the lateral pressure fran the surrotmding snowpack exerted 373 enough force to close the gap that fonned during settling. lbroughou.t the winter. all of the buried sensors indicated a slow increase in wetness up to 1.5 percent by voltDDe (Figure 6). As the snow 3.0 2.4 1.8 1.2 (.5 4.0 3.5 ~ 3.0 ~ 2.4 ~ 18 11.2 0.6 4.5 4.0 3.5 3.0 2.4 18 12 0.6 smsat-150 Cll soorn sm>al-90 CM SWDI IEC. 1984 JAN. 1985 Figure 6. Snow VoltDDe Wetness Gradually Increased During the 1984-85 winter. surface neared each sensor. it showed a diurnal liquid water flux of up to 2 percent by volume (Figure 7) • This wetness flux is attributed to meltwater infiltration and drainage during springtime freeze/thaw and is corroborated by basal ootflow data collected fran six melt pans situated 4.1 3.8 3.5 3.2 2.9 sm9:ll-150 C>£ scum 2.6 2.3 I'' 17 ~14 ;u 0.8 0.6 3.2 2.9 2.6 2.3 2.0 17 APRIL, 1985 3.6 3.3 3.0 2.7 2.4 2.1 1.8 2.7 2.4~Slll<;ffi-90C>II<Iml 2.1 18 1.5 APRIL, 1985 Figure 7. Four '1\ri.n-Disc Sensors Responded to Surface Diurnal Meltwater Flux. beneath the snow, on the soil surface. During April, wetness fran sensors and wetness fran freezing calorimetry were coopared. Data indicate good sensor response when snow calibration values are used (Table 4), but sensors appear to undemeasure when sand calibration values are used. Wetness values were higher in 1985. This may be due to several factors other than an actual increase in liquid water. Because the measurement frequency was changed fran 5.4 KHz to 10.7 Hlz, the influence of the size and shape of the ice particles was eliminated, possibly increasing measured wetness. Individual sensor/mast mom1ting eliminated mast sway and sun cupping fran solar re-radiation, allowing for better sensor-snow contact and tbos higher wetness readings. 1985-86 Snow Season The sensor and JOOD.i.toring system were refined for this winter season. 'lbree of the sensors retain the original 1D-cm diameter discs While three others, at matching levels, have 15 c:m discs. The larger discs should increase the resolution of the sensor and respond more quickly to changes in wetness. A chart recorder is being used to continuously monitor the sensor nearest the snow surface, to track diurnal variations in wetness and to measure wetness during unattended periods. Data collected so far this season indicate a gradual increase in snow wetness (Figure 8). Wetness increase appears to be about 1 percent greater fran the two sensors with the 15 c:m diameter discs. This implies that larger diameter discs have a higher degree of response to the presence of SE:i&lt-90otSWlli-15 O!:DISC 18 12 Slls:R-30 OlllRrn-10 01 DISC ~-30C>Isamt--1501DlSC 10 15 20 ,. 5 r::F.Cr18Dt,l985 JNUt.lt'i,l986 Figure 8. Snow Volume Wetness Gradually Increases During the 1985-86 Winter. 374 liquid water. The actual wetness values may not be as meaningful until a calibration of the sensor with the 15 c:m discs is made. The magnitude of wetness fluctuation caused by diurnal freeze/thaw did not seem to be affected by the difference in the size of the sensor discs (Figure 9). Wetness at the sensor with the 15 an discs, located 150 c:m above the soil surface, did not seem to fluctuate to any greater extent than it did at the sensor with the 10 c:m discs. In fact, wetness at the 10 c:m disc sensor varied rore. High snow surface variability may be the cause for tbe equal response fran the different size sensors, because the sensors are separated horizontally by about 2m. 1.5 1.2 0.9 0.6 1.2 0.9 0.6 0.3 S&S:II.-150 CM tumi-10 C1 DIR£IER DISCS • s 6 1 s 9 w u u u a u u u ~ u w n ~ IE»t!Dl, 1985 Figure 9. Surface Liquid Water Flux Measured by Different Size Sensors. Two rain-on-snow events have occurred this season. A 1.9-c:m rainfall on December 28, 1985 caused a slight response fran the sensor with the 1S c:m discs located at 90 c:m above the soil surface and a rore marked response fran the sensors located at the 30 c:m level (Figure 10). The detection of increasing wetness and the indication of drainage after the rainstoml were greatest in the deeply buried sensors and indicates that water flow was greater in these areas. Water flow in the upper snowpack was more variable. <kle sensor respooded, while the other did not respond at all. 'Ihis strengthens the conviction that early season water flow in snow is extremely variable, spotty and may be restricted to flow channels. A 4.5-c:m rainfall on Jatmary S, 1986 caused just the opposite response. Both of the upper level sensors showed a large increase in wetness. lbre liquid water was indicated by the sensor with the 1S c:m discs. The response of the more deeply buried sensors was variable but to a DDCh lesser degree than the shallow sensors. The rore intense rainfall probably created rore flow paths in the upper half of the snowpack. As the rainwater infiltrated downward in fewer but larger flow channels, liquid water detection became mre difficult. At both levels, the sensors with the 1S em discs indicated the mst liquid water. 3.9 3.7 3.5 J.J 3.0 2.7 2.4-:::===-- J.o+---- 2.8 2.9 2.6 2.41.----- 2.:1-f---- 2.~-- 4.5 <H RAIN 1.9 (}of RAIN SENSCR-90 CM SOO'IH-15 CM DISC SEme:R-90 CM tnml-10 CM DISC SENSCR-30 01 SWIH-15 <H DISC Figure 10. Sensor Response to 'I'M> Rai.n-fu-Soow Events During the 1985-86 winter. Data fran 3 years indicate that buried twin-disc sensors respond to gradual changes in snow wetness and sharp rises in liquid water fran rain-on-snow and springtime diurnal freeze/thaw. Snowpack drainage was indicated by the sensors after each rain-on-snow event. Wetness determined by individual sensors carpared favorably with that detemined by freezing calorimetry, especially when average values were carpared. During calibration, cme twin disc tmit with 10 em diameter discs showed a good linear response to incremental changes in snow wetness. The sensors are capable of consistent measurements and can indicate lateral variability in ~ck liquid water. The calibration of the sensor with the 15 em discs is scheduled for April 1986. Wetness measured by buried sensors will be carpared with wetness values obtained by one of the mre recent methods, probably acid dilution. If received in time, wetness measurements will be made by A. Denoth' s new 375 hand held resonator and carpared with twin-disc sensor wetness. More field work is needed to find the DBXiDun practical disc diameter and the best available recording equipnent. A cooplete remte system is planned for installation in the near future. Ambach, W •• and A. Denoth. 1972. Studies on the Dielectric Properties of Snow. Zeitschrift fur Gleitscherkonde und Glazialgeologic 8(1):113-123. Berg, N. H., 1982. Layer and Crust Developnent in a Central Sierra Nevada Snowpack: Sane Preliminary <llservations. In: Proceedings of the Western Snow Conference, Reno. Nevada, pp. 180-183. Bergman, J. A •• 1978. A Method for the Determination of Liquid Phase Water in a Snowpack Using Toluene Freezing Calorimetry. Masters Thesis. University of Nevada (Reno). Denoth, A •• A. Foglar, P. Weiland, C • .Matzler, H. Aebischer, M. Tiuri, and A. Sihuola, 1984. A Calparative Study of Instrmnents for Measuring the Liquid Water Content of Snow. Journal of Applied Physics 56(7) :2154-2160. Kattelmann, R. c .• 1985. <llservations of .Macropores in Snow. Annals of Glaciology 6:272-273. Linlor, W. I., 1980. Pe:mrl.ttivity and Attenuation of Wet Snow Between 4 and 12 GHz. Journal of Applied Physics 51(2):2811-2816. Linlor. W. I.. and J. L. Smith, 1974. Electronic Measurements of Snow Sauple Wetness. In: Advanced Concepts and Teclmiqu.es in the Study of Snow and Ice Resources. National Academy of Sciences. Washington, DC. pp. 729-739. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 MEASUREMENTS OF SNOW LAYER WATER RETENTION Richard Kattalmann 1 1\&STAJ\Cl: The water holdirog capacity of a snowpack determines the initial runoff response to snowmelt nr rain-on-snow. :=arl y streamflow generated from these events has been difficult to forecast due to limited understanding of snowpack water retention. To identify the likely range of water holding capacity of newly dP.posited snow layer·s, water was artifici<~lly appliEd to small plots, and w<>ter retention was determined by a change-in-density tachni~ue. All of the average water retention valuas from fresh snow on level ground ranged from 0 to 5 percent by volu,.,e. About two-thirds of these values were 2 percent or less. Concurrent applications of dyed water suggested that relatively little ( <20 percent) of the snow volume contllJc:ted ~Vater. Other observations indicated that. cold content was not fully satisfied before water left a given snow layer and that losses from water· ·input continued long after snc~pack outflow started as the wett~~ VPlU~P incrAARArl. (KEY Ttf-!MS: Snowpack water store££!, sno~pack V'3ter r.10VP.ment, snowmelt runoff, rain-on-snow, snow hydro I ng:,.) INTROOUCTIIJN The capability of snow to retain rain and r,,,,[tw<Jter is a n.3jc>r uncertainty ir. st.r·rJamflow furecostin3. This liquid water holding capacity of ~now appears to hP. a highly variatle phenomenon ~pandinJ nn t.he development of the snowpack and its pr·CJper·ties at a given time. Estimates of water holding capacity range fr·cm 0 to r"ore th11n 10 percen7. by volume. Thu rt,i.ention capacity of a snowpack IVithin this ranue can influence st.rr:<<rofl.ow response to rain or meLt enormously. This initial ro~ronse is difficult tn forncast. fhe fundamental objactive of this investigation W3S to obt.:in somE· idea of the range of water rr.tP.ntion nf fresh snow under a particular application regime. This case study measured the approximate water holding capacity of 16 Layers over 3 years. Most of these layers storm were SLICCf!H&ively over several days as metmnorphism progrP.ssed. EARLY STUDIE:S A c;uotation from ~CJ.!I __ I1Y..Q.r~L9ll.:t. (u.s. Army Corps of Engineers, 196o: 314-315] surnmar·izes thr, basic principles rP.garding ~ater holding capacity: Liquid water entering a cold snowpack freezes within the pack, becomes part of it, and incresses the tempP.r·att•re within the snowrac:k. 0 Sr.nw at 0 G wi il impnund adrliti :mal water on crystal surfaces and in air 8paceu <•s hygroscopit; ;:.nd capillary water. S..rch 1\flt(<r (held 3gainst gravity) also becomes part of the snoi'Vpack and is ret a i r'P.d unt i!. the snow has mel ted. Pocl<e tE• o ,. r:F·I [ s of snow .,.hi ch arB cold and dry may exist in an otherwise wRt' sno~pack as the r£su!t of ice pl<:~no~ which have not yet disintegrated to allow the snow to becCime fully conditioned... Neither the retentivity or permaability of the sno~ io constant. Therefore, the transmission rat8B and wster storAQP. cApac:i ty of the snow vary with the eharaci:er or the st~ge of the metamc•rprisrr of t.r·e indiviltual sr.nw layers." The first significant work on the water holding capacity of snow was rlnne by Gerdel (1915, 1948. and 1954) 6t the CAntral Sierra Snow Laboratory [CSSL]. tmong Gerdel's contributions was t.he distinction of surface tension {ndsorhed or hygroscopic water] c<tpacity from capillary capacity. He reasoned that. new low dr:!n£.-i ty srro~~: C(lllld ret.;;~ in water as surfacP. films but the voids in such snow v1ere too large to hoi d n1uch ''"'•t.n· by r.Rfd llarity. He argued t.hat high rP.t .• mtion ut r·tdn in fr·f'sh snow [e.g., Church. 1!341; Horton, 1941; Wisler snd 8rAter·, 1949) should t.p 1 Pacific Southwest Forest and Range Experiment Station, USDA Forest Service, P.O. Box 245, Berkeley, California 94701. 377 consiclP.red as e very short-term ph<=mornenon r·esltl tir•o from surface tension or, more likely, pending of water at the soil surface. Several other studi8s included measured or estim~ted values nf . water holding capacity, gen~:~·ally as a minor topic (deQ.uervain, 1!:173; •,\lank i ewi cz, 19791. Reported meusurements <'lnd estimates of snc•wpack l'lc:•t.Hr ret.er•t.ior-vury widely (Table 11. \'lankie11dcz [19791 li~ted the follolll'ing as factors possibly influencing water holding capac1~y: drainagt intervaL. snow teKturp 0 impurities, impeding layers, isolated dry cells, and attraction of moisture up to a refreezing surface. deQ.uP.rvain (18731 listed snow density, grain size, grain shape, spedfic: :;urface of snow !JI'Ains, specific nun•i:o;or· of gr·oin conte<cl.s, end pore v1icith as important influences. Snow with fine orains bolds more water t.han dof!s conr·sp. grsined :>now t'lnd "«n optimuon dr·y densit.y fur maximum storage obviously exists" (deQuervair, 1973: 209}. Only two major &tudies known to date dealt excl.usively with snm~op!'tck water holding capacity, Sulahri<o [197£1 measured water retention as the difference between dry density and wet density aftar artificially applying water to a snowpack and NEW SNOW New 0.5 New D-5(x=11 New D-12 Dry 1-10 35% density 2 Cold, arctic 4 New fine grain 4.7 20-25% density 5-6 32% density, new s.s Above interfaces 10 13% density 10 35% density 0 Wet 0-4 Isothermal arctic 0-8 Old 0-40 Wet 1-2 Wet 1-2 Wet 1-2 Beginning to thew 1.2 Wet 1.2-3.5 Wet 2-3 Wet 2-3 Wet 2-3 50% density 2-3 mm 2.5 Wet 3-4 Wet 4-5 Large g r • refrozen 4.7 45% density 6 Wet 6 41% density, 1mm 8 53% density, wet 10 Residual water content Residual water content Change in density Change in density Estimate Change in density Calculated from theory Residual water content Residual water content Residual water content Estimate OLD SNOW Inflow-outflow Change in density Change in density Change in density Residual water content Estimate Residual water content Inflow-outflow Residual water content Estimate Residual water content Residual water content Residual water content Residual water content Residual water content Calculated from theory Estimate Residual water content Residual water content Residual we ter content 378 Gerdel and Codd ( 1945) Ebaugh and DeWa L La (1977) Sulahria (1972) Smith end Halverson (1969) u.s. Army c.o.E. (1956) Marsh (1982) Colbeck (1975) 8engtsson ( 19 81 ) deQuervain ( 1 97 3 1 Wakehema (1975) Kuzmin (1948) Himmel (1951 1 Smith end Halverson (19691 Marsh (1982) Sulahrie (1972) Gerdel (1954) u.s. Army c.o.E. (1956) Leaf (1966) Price,!.! ll· (1979) Gerdel (1945) Boys r and Merrill [ 1954) Ambach and Howorka [19651 Lemmela (1973) deQuervain (1973) Ambach (19651 Bengtsson (1981) Colbeck (19751 Kuzmin (1948) Ambach (1963) deQuervain (19731 Wankiewicz (1976) allowin3 it to drain. Areas of 1 m2 in different locations in the Carson Ran:~e of the Sierra Nevada were irrigated on 13 occasions. Density, temperature, and grain size before irrigation and wet density after irrigation were measured at 5 em intervaLs throughout excllvated profi Las. Other indices were calculated frcr:~ meteorolooi.::al ~servations at the sites. Using multiple regression analysis, Sulahria developed a predictive equation for fresh snow and another for metamorphosed subfreezing snow. These equations had R2 values, for his data set, of .886 for fresh snoVI and .755 for old snow. The equations relate water retention to dry density and to several terms involving different temperature measurements. These temperature terms included the mean snow surface temperature, the calculated cold content, average storm temperature, average temperature gradient of the snow layer, end the time weighted temperature (degree-days] of the snow layer. Sut~hria observed pending of water on discontinuities and layer interfaces, flow along such layers on sloping terrain, and vertical channels conducting water within the snowpack. Fresh snow close to freezing and older sno"' that full at low temperatures, but subsequently warmed to near fre.ezing, had :naximum water retention capacity. Other important observations wore the occurrence of channels in homogeneous dry 3now ~fter water was applied; the small wetted volume of very cold snow after i rri!:ption; the presence of subfreezing dry zones efter water was observed to flaw out of the pack; the warming of these zones within 24 hours; and the predictability of stratigra;1hic discontinuities within the pack that 111ay retain large quantities of water. The other known study of water hclding capacity was a laboratory investigation of Liquid water transmission and retention in new snow (Eb:;.ugh and UeWalle, 1977a and 1377b]. Snow samples were irriget'!d with a rainfall simulator in a controlled ter~~rert.'ture labor-3tor·y. \1ater application rEite, ;nJN temperature, rain temperature, sample depth, ,,,,d tim'! distribution of outflow volume were recorded. neg rP.ss ion gquet ions were water holding capacity, Lag time, coeffi ci er.-ts. Initi at snow density predictor (of those monitored] for capacity. They estimated liquid developed for and routing . was the best water holdino ·11ate r holding capacity as 3 24 * [initial snor1 density (g/cm )] -2.91 and found that it varied frr:Jtn 0 to 4.95 percent by volume with an averagfl of 1.026 pe:rcent. In the past decade, significant progress was ~hieved in understanding how liquid water interacts with snow (C:Jlbeck, 1978; WMkiewicz, 1379; and Jord1in, 1983a}. \'iater flo'Ns through snow primarily 379 in response to gravity. The gravity theory of water percolation through snow (Colbeck 1972] uses Darcy's law for one dimensional unsaturated flow: ~ ~ -K (d /dz + 1} where the flux u (m/s) is proportinnal to the sum of the capillary pressure gr8dient [d /dz) and tho gravitational pressure gradient (1 m/m). The constant of proportionality (K} is the unsaturated hydraulic conductivity. Gravity drainage is predominant during steady or decreasing flo1t, and the capillary pressure gradient may be ignored -8 except at very low flux (10 m/s] or at some interfaces (Colbeck, 1974). Experiments demonstrated that a negligible capillary pressure gradient exists throughout the snowpack as a whole, but significant pressure gradients may occur at major pressure or textural discontinuities (Wankiewicz, 1978b]. In ca leu lations of flux and !;he speed of the flux wave, the water retention or "irreducible water content" (Colbeck, 1972} be>Gomes important in determining the proportion of the pore volume occupied by mobile (as opposed to retained) water. With greater volumes of water available for flow, the permeability inGreases rapidly [Colbeck, 1978a). Values of the irreducible water content are usually assumed (Dunne, et ll·• 1376; Jorden, 1983a). Modeling of water flow through snow based on the gravity flow theory has baen extended to include the effects of ice layers and flow channels (Colbeck, 1979; Marsh, 1982), increasing flow (Jordan, 1983b), and refreezing and water retention [Bengtsson, 1982). Inherent variability in snow water holding capacity may be high. The irreducible water content of uniform sand varied widely (Wankiewicz, 1979) based on data from Mualem (1978}. Additionally, water holding capacity should not be assumed to be a constant quantity. Tbe pre'3en;;e of liquid N~ter induces metamorphism of the snow grains (Wakahama, ·t9o8), which in tL•rn alters the ability of the snowpack to retain water against gravity. This grain graVJth and r•esultant incre.:tse in po1•e sizes reduce th"l omount of II! ate r th~t can be hP. ld i:Jy capillary forces ~nd also increase the permeability of the snow to water (Marsh, 1982]. Therefore, water that is initially held may slowly be released as the physical structure of the snow chanaes. STUDY AREA ANll CONi:llTIONS The study was conducted at the Central Siarra Snow Laboratory [CSSL), a facility of ~he Pacific south11est Forest and Rang3 Experiment Station ne.3 r Soda Springs, California. The CSSL is located near the crest of the Sierra Nevada at 2100 m whAre Interst~te Highway BQ crosses the ranJe near l)onner Pass. The area is subject to a maritin1e climatic influence and accumulates d£1ep seasonal snowpacKs. Annual precipitation consists mostly of snow and sverages about 150 em. Average maxi~tum seasonal snowpack water equivalent is about 110 em (Smith and Berg 1982]. Warm periods of several days to one or two 111eeks usually occur betwRen storms and a II. ow melt-freeze crusts to develop. These crusts become buried by subsequent snowfalls and result in stronuly layered, heterogeneous snowpacks. Wel'lll storms with high elevation freeaing Levels occasionally deliver rain to the snowpack, further alterin~ its structure and raising the snowpack 0 temperature to about 0 c. The study was conducted during the wintArs of 1983 through 1985. 1983 was one of the wettest yeetrs on record in the study area. 1984 and 1985 were drier than average with ;>rolonged interstor1" dry periods. In 1983, storm temperatu1·es r·anged 0 from -1 to -3 C, and snov1 densities at tho time of i l'rigation were mostly 20 to 30 percent. In 1984, the thrae storms studied occurred at temperatures of 0 -4, -7, and -10 C, and the der.si ties of the Layers ran3ed from 14 to 2ti percent. Snowfall contributing to the 1985 Layers occurred at -6 to -9°C. ALL snow densities in 1985 were Less than 20 percent at the time cf measurement. DE:FINlTIONS !n this paper, the terms "water holding capacity" and "water retention" are considered equivalent and are used to include both water frozen when it enters a cold snowpack (that satisfying the "cold content" or "heat deficiency"} and liquid water held against gravity (the "irreducible water content"). In this context, the questirm of interest is how much water will be ret01ined b)· -:;ne snowp~;~ci< for a given period of time, given some input of l iguid water. Water holding capacity by l (3 .3 l . vo ume em water;cm snow = water hold1ng capacity by weight,(~ water/g dry snow} x dry snow density (Sc snow/cr.? snow) I density of water (1 g ;j water/1 em water). METHODS The study approach was to apply water artificially over an are~J of snow, allow it to drain, and then compare the densities of samples collected in the wetted volume to densities of corresponding snow that was not irrigated. The difference in density, after correcting for settling in the presance of water, was attributed to wat.er retained by the snoN. This case study was Limited 380 to a Level area of Less clearing at the CSSL. i;ban 200 2 m in a forest Mei:'bLII'ements were made only in layers of snow deposited iJy a single storm. The study was concerned ~ith how much water could be held by snow with relatively consistent propRrties throughout its volume. Thus, intRrstorm bound~ry Layers had to be avoided as did storm Layers with obvious discontinuities (such as grain size changes caused by dramatic shifts in storm tempereture]. Storm letyers had to be of sufficient depth (minimum 5G em) at the time of the first water application to allow sequeni;ial measurement of the same storm layer over severaL days. The volumes of the sno\lr that were irrigated were completely undisturbed portions of the snowpacks except for removal of the surfece snow in some instances. Thus, the margins and base of the irrigated area were not physically isolated in any way from the remainder of the snowpack. rha application of water alone defined the area that waH altered. The extPn~ of this area was determined by visual inspection. Only the central portion of the wetted areas was sampled for density, and no san•ples were taken within 15 em of the surface perimeter of the irrigated arP.a. Although irrigation usually took place on the nat~ral surface, occ01sionally surface snow had to ba removed before irrigation to avoid the v1ater r·ou!;ini.J problems of surface crusts or to expose a Layer which had be"ln buried by subsequent snowfall. Snowfall in ~XGPSS of one-half meter usually stoppPd measurements in buried l~yers. The JTIBE\Suremerot sitP. v1as kept Level by filling all pits before storms. 8oth the water application and the density sampling procedures were revised a fe•~ ti1res during the study as pro:Jlems were noticed or improvemBnts incorporated (rnble 2). Thu!';, thA r·esults are not comparable between seasons to the degree that they are dependent en the measurement technique and sampling procedures. A series of pi l.ot tests were made in 1r::ac:. Those tests incLuded comparison of different density sampler volumes, selection of area and intensity for water application, and limited study of the effects of Layer ir.terfaces, drainage intervals, and compaction. The results of these tests provided thR busis for the initial design of the st~dy. In 1983, two plots of 'i m a1·ea were irrigated following snow storms of adequate depth with a hand-oparai;ed pump sprayer. Water tu a depth of 2 em was distributed as evenly as possible over the area in about 45 minutes, for an overall appLication rate of 2. 7 cm/hr. Instantane~us intensities were somewhat higher because the spray nozzle was in motion and l.ess than the entire area was being sprayed at any given instant. The snoVJ was irrigated in the late afternoon, covered with 3 sheets of 2 em-thick expanded styrene and thin plastic film and allowed to drain overniaht. A comparison area within a couple of meters was physically isolated from the wettP.d area by a metal sheet driven vertically into the snow to prevent any lateral flow of we te r. The comparison a rea s were covered in the r-;ame manner as the irrigated areas. After 14-18 hours of drainage, an ac~ess pit 11as excavated on the edge of each 1 m area. Cylindrical samplers were inserted horizontally to obtain up to 12 snow samples from the central 50 em x 70 em portion of the area. The samplers had an inside diameter of 10 em, were 24.7 em long, and 3 enclm~erl a volume of 1940 em • The snow at the ends of tr.o cyl1nders was cleanly cut with a thin metal blade, and the sample was removed for weighing. The average settlement of the wetted area [in excess of undisturbad settlement] was estimated to the nearest centimeter by measuring the vertical distance to the wetted surfllce from a straight edge held 01t a level as close as possible to the surface of the surrounding unwetted snow. In cases where a definite boundary (such as a buri,3d crust] existed within 50 em of the surface, several measurements were made of the depth to the buried layer and averaged. Settlement was assumed to have occurred in the top 20 em. The water retention by volume was considered tn be the difference between the mass of a unit volume of wetted snow and the mass of a unit volume of dry snow. In practice, the mean of the weights of the unwetted samples was subtracted from the mean of the weights of the samples of wetted snow. This difference was then divided by the sample volume to obtain the wa~er retention on a :rolumetric basis (grams [or em ] of water per em of dry snow). Alternatively, the same result is obtained as the difference between the mean wet and dry densities. This value ceo be divided by the dry snow density to obtain the water retention on a weight basis (grams of water retained per gram of dry snow), if desired. In order to correct for I'Jater induced settling, the mean of the weights of wet. snow or the mean wet snow dansity must be multiplied by a factor that accounts for the more concentrated mass in the given volume. That is, if no 8ettling occurred, less mass would be found in the same volume than is actually measured. This factor is computed as the quotient of the difference between depth of the layer in which the settling occurs (usually assumed to be 20 2 Storm Treated Control Plot Area (m ) Sampling Days Mean WHC Date Plots/Day Plots & Irrigation Method & Measured (percent) ---------·-----------------------tl~!.!:!.f!.!!!. _________ yE.l!!m.!!! _____ -------------------- 2-7-83 1 1 1 A c i 1 2-9-8:3 2 1 1 A c 3 2-13-83 2 1 A c 1 5 2-18-33 ~ 2 1 A c 3 2,3,4 " 3-2-83 2 2 1 A c 3 2,3,3 3-7-83 2 2 1 A c 3 1 • 2' 3 '5-·5-83 1 1 1 A c 1 1 2-12-84 2 2 1 A D 2 0,0 2-16-84 2 2 1 A D 3 1,2,:? 2-2'!-84 2 2 1 A D 4 1,1,2,2 1-28-65 2 20 B E 2 2,3 2-1-85 2 20 8 E 1 2 2-9-85 1 2 20 B E 3 1 • 5. 5 3-8-85 1 2 20 [I E 4 1 ,2,2,5 3-26-85 1 2 20 B E 4 2,3,4,5 1940 3 *A ::: pump spray~::r 3 B "'" hose w/ spray nozzle c = up to 12 samples of c:n 3 D -one 25,000 em sample E -up to 30 samples of 13,000 em 381 em] ;md the d&pth of settler.~er11; (to the nearest centimeter} divided by the depth of the layer ( (2U r.;inus set;tlinJ depth]/20}. Often, no settling was obs~>rvnd and a default fto~ctor of o.s5 (1 err in 20 em] was used. The w-1ter retention value is very sensitive to the settlemen' factor. For 1984, an automatic water appl icstor was designed and constructed based on drop forming needle-type rainfall simulators. In trials, the applicator failed to distribute water evenly and was abandoned. The pun,J spraye'r was used again in the same manner as in 1983. A single large sample was excavated and weighed in eE!ch of two appl icetion areas and in the unwetted snow. The sampler was a rigid metal square (~0 x 50 x 10 em} that proved to be easier to use than the small samplers of 1983. A metal sheet was first inserted into an exposed pit wall at the edge of the iPrigoted <31'88 at a depth greater than 1C em. Then, the sampler was inserted fr·om above until conti.'lct with the sheet was made on all ~ides. Next, a second metal sheet was inserted along the upper edges of the square cutter, and the surpaunding snow was excavated, aLLo~ing removal of 3 the sample volume. The 2!5,000 em volumes of snow wel'e weighed, ond water retention values were calculated in the same manner as in the previous year. 2 In 1985, a single 20 m area was irriaated bufore each measurement day. A spray nozzl.e was used that produced a relatively even distribution uf water· over the aprl ication area (2:5 percent more water at the center than near the edges). The nozzle was operated at household line pressure through a hose attached to a tl'ipod. Applications of 1-hour duration r·usulted in ai:lout 2-cm watel' depth over most of the area and up to 2.5 em in the central few square meters. Water was applied in the late afternoon and allowed to drain overnight. When pc•ssible, the same layer of snow was irrigated in new plots on successive days to determine change in water Petention over time. A metal sheet was inserted horizontally at approximately 2U em depth at the edge of the area to be wetted 0 before i rrigatiorr. A second plate was similarly inserted in snow that was not to be irrigated. The depths from the snow surface to the four corners of each plate were recorded before irr·igation and again before sampling the next day. The difference between the overnight settling depths of the irrigated snow and the natural snow was used to correct the water retention in a manner similar to that described for 1983. Generally, settlement in the dr·y snow was fr·om 0.5 to 1 em per 10 em of original depth, and settlement in the wetted snow was about twice that in the dry snow. After 12 to 14 hours of drainage, large volume density samples were removed from both the natural 382 and irrigated snovJ and then weil:)hml. Sections of flue pipe, 15 c101 in diameter· by 90 cr11 Lon:;, were inserted horizontally to obtain snow volumas of approximately 13,000 cm 3 • Gre3t caPe was required to keep the sa:uplers horizontal during insertion, The depths fro:n the surface to the top of the sampler at etich end wel'e measured. If these depths differed by moPe than 1 em, the sample was discarded. Sometimes, up to half of the potential. samples were discard8d to ensure that the same layer of snow was sampled throughout the area. Genel'ally, 12 samples were obtained in the natural snow end up to 30 in the wetted snow, depending on the number of samples uiscarded. Water holding capacity by volume was then computed in the same manner as in 1983. RESULTS In 1983, 13 sets (two areas of the same snow layer irrigated at the same ti01e} of measurements were obtained from seven storm layers i:lefope they became naturally wetted. In 11 of the 13 sets, the mean of the measured water retention valutH' was 3 percent or less by volume. Water retention was 4 or 5 percent in the other two sets. In the relatively dry winter of 1984, few stol'ms were suitable for· messurement. Only three storm layers were sampled, and 13 sets of measurements were obt.<dned. In 8 of th<~ 13 cBses, the mean of the measured water retention values was 1 percent or less. The mean of the ethel' 4 sets was 2 percent. On two occasions, measur·ements were obtained from topcgraphic depressions where water was bei ieved to have com:entrated. The snow at ~hese sites retained about 1Q percent water by volume. At only 1 meter directly upslope from these sites, the snow had retained less than 1 percent water by volume. In 19Bti, 14 sets of measurements wepe obtained frorn five si:or·rn layers. Measured water holding capacity by vof.ume was 1 of' 2 percent on 7 dc.ys, 3 per·cent on 2 days, and 4 or 5 percent on 5 days, Successive measurements of three of the layers over several days indicated increasing then decreasing water retention for the data observed. Water holding capacity was at a minimum (1 - 2 percentj 1 to 2 days after deposition, increased to a maximum (4 - 5 percentj after 2 to 3 days, and then decreased to nearly zero within 5 to 7 days, as the grains enlarged and midwinter melt percolated into the layers. In a fourth layer, water retention was already up to 5 percent when first measured 2 1/2 days after the end of the storm. In 1983 and 1984, the absolute diffePence between plots was generally 1 percent or less. In a few cases, the difference was 2 percent. Because at most three applications !IVer-e made at any one time, assessing retention vari ebi L i ty the natural property variability of or from possible separating probLems the water physical with the experimental technique was not possible. Further work on this subject would require much greeter attention to sampling design and a more uniform uater applic~tion system. Within the ir·r·iuo~~>d ;:rreas, the densities of smaLL-volume samples (up to 30 in number) varied greatly. Densities in a single irrigated volume often ranged from the dry snow density (usually 0.25 3 3 ~em or less) up r.o 0.60 g/cm , over distances of only a few centimeters. Coefficients of variation ranged from 20 to 140 percent. This wide variation in snow density was most apparent with smaller volume (500 -2000 cm 3 ) samples. Visual inspection of the samples revealed that the high density samples included a large proportion of wet, nearly slushy, snow of Large grain size (1 -2 mm). The low density sampLes either lacked entirely or included only a small proportion of obviously wetted snow. Concurrent applications of dyed water to other areas of snow during the water experiments permitted documentation retention of the prnportion of the snoYI volume that oecomes wetted. A preliminary evaluation of these data suggested that less than 20 percent of the snow volume became wetted within 12 hours of the water application. During the 3 years of this study, only two rain storms occurred when the snow was not already saturated. In one storm, 9 em of rain fall over a 15-hour period, on snow 1 to 10 days old in the surface meter. About S em of water was retained by the snowpack. OnLy 1 em of rain fell on fresh snow during the other storm and all of it was retained by the snoYifpl'lck. The near-surface snow layers had water holding capacities of 3 to 5 percent by volume during tha~.e events, as determinP.d by chan!Jes in Snthv density (corrected for settling) over the storm periods. Snow pits excavated after the events revealed thut much of the snow did not becGme wetted during the stLlrms. DISCUSSION ALL of the average water retention values obtained from fresh snow on lP.vel ground ranged from Q tn ti perc~nt by volume. About two-thirds of these values were 2 perc:Ant or less. It appears that laye-rs of fresh, Low-density snow sampled in this study were not capable of retaining more than 5% liquid "at.er by volume, 12-18 hours after high intensity application. In most cases, the layer water retention may have been less than 2 pflrr.ent. These values tenc! to be at the lower end of <:he range of vallHlS rapor·ted for fresh snow and closer to common values for old snow (fable 1). In a 383 practical sense, water retention of 2 to 5 percent by volume means that a 50-cm thick snow layer could retain only 1 to 2.5 em of rainfall. rlesults from the two rain events suggest that tha high intensity artificial application of water did not seriously bias the apparent water retention as compared with Low intensity rainf~ll. The high variability of density in the wetted volumes and observations of widely distributed flow conducting channels indicated that relatively little {less than 20 percent) of the snow volume became wetted within 24 hours of water application. lhus, the distribution of water and the wetted proportion of the total volume may have been more important to total water holding capacity than the weter retention of that part of the snow that actLHll.ly conducted water. If snow that became fully wetted retained 30 percent water by volume, then a layer 50 em thick that became 10 per·cent wetted could have held 1.5 em of water, and one that became 20 percent wetted could have held 3 em of water. The proportion of snow occupied by flow channels is being studied at CSSL. In the High Arr:t.ir;, flow fingers covered about 22 percent of horizontal snow sections (Marsh, 1982). Data collection was not structured in a way that allowed statistical evaluation. The only obvious natural influence on water holding capacity was wetting opportunity. As natural melt occurred, the meo~;:ured water retention declined to zero. Surface melt occurs at the study site on most clear days throughout the wiPter (Smith, 1974]. This input of water satisfies the water holding capacity end also increases density, grnin size, and temperature. Thus, water retention declined coincidentally with inr~r·P.ased aae, density, grain size, and temperature, if meltwater input wAs the fundamental cause of the decline. Snow layer age is the most readily determined factor and, in general, water retention Approached zero within a week if the layer had not been buried by subsequent snowfall. At elevations below that of the study site (2100 m), the water holding capacity would probably be satisfied even faster. Because only thick snow layers that were not buried by additional snuwfall could be measured, the range of snow conditions sampled was relatively small. A greater number of replicated plots under a wider span of conditions would help to batter understand the water retention process. 08SERVATIONS AND CONJEClUH=s !)rJring the study, a v~riety of qualitative observations were made opportunistically. While many of thasa observations lacked sufficient detail to support definitive statements, they provide some Lli recti on for future studies. They also provided the baEds for san•e ideas concerning various processes involved in snowpack wat~r storage. Slope angle appeared to be a critical factnr in water movement and storage in snow. With even f•l ight slopes {about fj%), a larJe proportion of the uppl ied 1vater was routed Leterally downslope along one or more nearly imperceptible layers in the sno~. Such flow reduced the wetted volume directly beneath the application area to less than the wetted volume in sample site5 that were on level ground. Hater flow apparentiy !Jecame more concentrated on slopes due to the opportunity for lateral downslope movement. This study did not include interlayt:~r boundaries. Water is known to be retained at layer interfaces ll'there texture changes abruptly between fine-grain snow and under·lying coarse-grain snow (Wakahama, 1968, 1975; Colbeck, 1973]. Concentration and storage of substantial quantities of water at such interfaces has been observed near the study area (Kattelmann, 1985]. Water storage at these layer boundaries and at the snow-soil interface may be mc•re important than water storage in the snow layers. There is historical evidence that snowpacks can occasionally store large quantities of liquid water. If the snow matrix has insufficient capacity to account for these episodes of moderated water release to streams, then the interfaces are the next place to look. A few periods of record from snowmelt lysimeters at CSSL indicate that water continues to dr·ain from a saturated snowpack for several days aftE'r surface water input has ceased. This slow drainage could be expactcd frcm thnories of d~clining flux due to declining liquid water content (Colbeck, 1972] and snow metamorphisn' in the presence of liquid water (Wakahama, 1968]. Therefore, it may be better to think of water "detention" by snow rather than water "r·etention", at least above the long-term "i rrec!u::ible r.ater content" (Colbeck, 1972]. Conversely, slow drainage from the wetted por•tion of the snow could be partially offset by increasing the wetted volume. Eventually, the entire snow mass comes into contact with liquid water, and some of this water will be stored around the newly wetted grains. The length of time for the entire snow matrix to become wetted following the development of the initial flow channels has not been directly measured. Results of a model study indicated that ,a deep isothermal snowpack becomes thoroughly wetted about 4 days after channels have penetrated the snow (Marsh, 1982]. Thus, losses from surface water input apparently can continue to occur after water has started to drain from a snowpack. 384 In a similar manner, liquid water continues to be stored in the snow matrix by freezing long after water bP.iJins to drain fr-om the snowpack. Following the irrigation of subfreezing snow, dry sections of snow at the pre-application temperature were found on several occasions during tha study. Snow 0 tempP.rature~:; as luw as -4 C have been recorded up to 48 hours after substantial snowpack cutflow had been recorded fr·om snowmelt lysimeters at CSSL in 1985. Isolated cP-lls of subfreezing snow in fr·eely draining snowpacks have also been noted by other investigators (e.g. 1 U.S. Army Corps of Engineers, 1556]. Cold content or heat deficit must be satisfied eventually, but these obRervations indicate that water can be released before the 0 entire snowpack warms to 0 C. COf\lCLUSlONS Measurements from 1i:i stor·m layers indicated that the total water retention capacity of fresh snow was almost al~ays 5 percent or less by volume and usually less than 2 percent for the study conditions. Relatively little (perhaps less than 20 percent] of the snow conducted water within 18 to 24 hours following application of water. A limited number of observations indicated that topographic slope may reduce t~e wetted volume and water holding capacity below these values found for snow on level ground. No consisttlnt relationships between water retention and other snow characteristics were suggested by the data. In develcping forecasting techniques for· snowpack water release, future research should be directed towards determining tho effects of slope, snowpack structure, channel development, and snow/soil interaetions. Time is a critical consideration in assessing snowpack water storage. The intensity of the water input to the snow surface probably affects the distribution of water and the development of channels through the snowpack. For example, less water is probably stored if 5 em of rain falls in an hour rather than if it falls uniforrroly over a day. This storage should be considered as temporary, Detention in tha Gnaw can distribute the rainfall or snowmelt over a longer period of time than in the absence of snow cover and thereby reduce the flood peak downstream. HoVIever, the ultimate vr:>iume of runoff may not be changed very much. Similarly, the heat deficit and irreducible water content of the whole snowpack must be satisfied, but not necessarily as a condition for snowpack water release. Some of the water input at the surface that ante rs dry snow can be frozen or held by capillary fc•rces at the same time as other portions of the water are flowing through previously wetted channels to soil and streams. ~NOWLEDGEMENT. This research was supported by the USOI Bureau of Reclamation Office of Atmospheric Resources Research and Calif. Air Resources Board. REFERENCES Ambach, w., 1963. Messungen des freian wassergahaltes in schneedecke. Expedition Glaciologique Internationals au Greenland 1957-60. 4{4):174-180. Ambach, w., 1965. Research on the energy balance and free water content of a winter snowpack during ablation (a translation). Unpublished manuscript on file at the Central Sierra Snow Lebo ra tory. Ambach, W. and F. Howorka, 1965. and free water content of International Association Sciences (IAHS) Publication Avalanche activity snow at Obergurg l. of Hydrological 69:65-72. Bengtsson, L., 1981. Snowmelt generated runoff from small areas as a daily transient process. Geophysics 17(1-2):109-122. Bengtsson, L., 1982. Percolation of meltwater through a snowpack. Cold Regions Science and Technology 6:73-78. Boyer, P. and P. Merrill, 1954. Storage effect of snow on the flood potential from rain felling on snow. u.s. Army/Weather Bureau Cooperative Snow Investigations Technical Bull. 17. 22 pp. Church, J.E., 1941. The melting of snow. Proceedings of the Central Snow Conference. Michigan State College. Vol. 1: 21-32. Colbeck, s.c., 1972. A theory of water percolation in snow. Journal of Glaciology 11(63):369-385. Colbeck, s.c., 1973. Effects of stratigraphic layers on water flow through snow. CAREL Research Report 311. Hanover, New Hampshire. 13 pp. Colbeck, S.C., 1974. The capillary effects on water percolation in homogenous snow. Journal of Glaciology 11(67):85-97. Colbeck, s.c., 1975. Analysis of hydrologic response to rein-on-snow. CAREL Research Paper 340. Hanover, New Hampshire. 13 pp. Colbeck, s.c., l978. The physical aspects of water flow through snow. Advances in Hydroscience 11: 165-205. Colbeck, S.C., 1979. Water flow through heterogeneous snow. Cold Regions Science and Technology 1:37-45. Colbeck, s.c., E.A. Anderson, v.c. Bissell, A.G. Crook, D.H. Male, c.w. Slaughter, end D.R. Wiesnet, 1979. Snow accumulation, distribution, melt and runoff. EOS, Transactions, American Geophysical Union 60(21 ):465-468. 385 deQuarvein, M.A., 1973. Snow structure, heat, and mass flux through snow. In: The Role of Snow and Ice in Hydrology-Proceedings of the Banff Symposium, UNESCD-WMD-IAHS, pp. 203-226. Dunne, T., A.G. Price, and S.C. Colbeck, 1976. The generation of runoff from subarctic snowpecks. Water Resources Research 12:677-685. Ebaugh, W.P. and D.R. DeWalle, 1977a. Laboratory investigation of rainfall retention and transmission in fresh snow. Institute for Research on Land and Water Resources. NTIS PB-288 703. Pennsylvania State University, University Perk, Pennsylvania. 62 pp. Ebaugh, W.P. and D.R. DeWalle, 1977b. Retention and trans•ission of liquid water in fresh snow. !.!!.: Proceedings, Second Conference on Hydrometeorology in Toronto (American Meteorological Society, Boston), pp. 255-260. Gerdel, R.W., 1945. The dynamics of liquid water in deep snowpacks. Transactions, American Geophysical Union 26(Part 1):83-90. Gerdel, R.W., 1948. Physical changes in snow cover leading to runoff, especially to floods. IAHS Publication 31 (General Assembly of Oslo), pp. 42-53. Gerdel, R.W., 1954. The transmission of water through snow. Transactions ,American Geophysical Union 35[3):475-485. Gerdal, R.W., and A.A. Codd, 1945. Snow studies at Soda Springs, California. Annual Report of Cooperative Snow Investigations of the u.s. Weather Bureau end University of Navada-Reno. Hin~~al, J.M.,, 1951. Lysimetar studies of rain-on- snow phena.ene. Cooperative Snow Investigations Research Note 4. Horton, R.E., 1941. The role of snow, ice, and frost in the hydrologic cycle. Proceedings of the Central Snow Conference. Michigan State College. Vol. 1:5-21. Jordan, P., 19B3a. Meltwater movement in a deep snowpack. 1. Field observations. water Resources Research 19[4): 971-978. Jordan, P., 1983b. Meltwater movement in a deep snowpeck. 2. Simulation modal. Water Resources Research 19[4): 979-985. Kettelmenn, R.c., 1985. Wet slab instability. In: Proceedings of the International Snow Science Workshop. Aepen, Colorado pp. 102-108. Kuzmin, P.P., 1948. Snowmelt research end calculation. Proceedings of the State Hydrological Institute. Issue 7(61). Cited in Kovzel, A.G., 1969. A method for the computation of water yield from snow during the snowmelt period. Floods and their computation. IASH Publication 85(2): 568-606. LaChapelle, E.R., 1969. Field Guida to Snow Crystals. University of Washington Press, Seattle, Wast inJton. Leaf, C.F., 1966. Free water content of snow pack in alpine areas. Proceedings of the 34th Western Snow Conference. pp. 17-24. Lemmela, R. 1973. Measurements of evaporation- r:ondP.nsation HnLi melting from a snow cover. In: The Role of Sncv1 and Ice in Hydrology: ProceBdings of the Banff Symposium, UN£.t:iG0-\'114u-INl3 pp. !:!70-•_;Jg. ~ale, D.H. ~nd D.M. Gray, l98l. Snowcover ablatioh and runoff. In: Handbook cf Snow. Po rgf.lnon Press Canada, Ltd. pp. 360-436. Marsh, D., 1982. Ripening processes and meltwetar move~ent in arctic snowpacks. Ph.D. ThH&io. McMaster University. 179 pp. Muslem, Y., 1978. Hydraulic conductivity of porous JTtedia; genP.ralized m9crcscopic approach. Water Resources Research ~4(2};325-334. Perla, R.I. and E.R. LaChapelle, 19~4. Dilution r.tethod for measuriPI] Procbodings of the Confereo~e. pp. 80-85. liquid 52nd water in snow. western Snow Perla, R.I. ~nd M. Martinelli, Handbook, USDA Forest Handbook 489. 254 pp. Jr., 1876. J\v<.• i anche Service Agriculture Price, A.G., L.K. Hendrie, and 1. Uunne, 1979. Controls on the production of snowmelt runoff. In: Proceedings ~iodating of Snow Lover Runoff. S.G. Colbl'lck and ~1. Ray, [Edit<Jr's). u.s. Army CHREL, Hanover, l!ew Hampshire. pp. 2b7-268. Smith, J. L., 1974. Hydroloay of warm snowpAcks and their effects upon water delivery ••• some nPI!t concepts. ~; Advanced Cnncepts and Techniques in thE! Study of Snow and Ice Resource"• H. S. Sant.eford <md J. L. Smith (Editors]. pp. 76-89. Nation a I. Academy of Sciences, !i<~shington, U.C Smith, J.L. and N.H. Berg, 1982. Historical snowpack chara~teristics at the Central Sierra Snow Laboratory, a representative Sierra Nevada location. The Sierra Ecology r'rojoct, Vol. 3, Pari; 2. USOI-Bureau of REiclamatian, Sierra Sooperative Pi lot Project. l!-4 pp. Smith, J.L. and H.G. Halvorson, l969. Hydrology of sn0\1 profiles obtained with the profiling sronw gage. Proceedings of the 37th ~/estern Snow Conference, pp. 4l-48. 3ulshria, M.~., l972. Prediction of water retention capacity of high elevation snowpacks on the eaf;t side of the Sierra Nevflde. Unpublished Ph.u. dissertation. l~eno. ll5 pp. University of Nevada at U.S. Army Corps of Engineers, l9bo. Snow Hydrology. I~:.Jl:i PP. Wakahama, G., 1968. The metamorphism of ~et snow. 386 IAHS Publication 79 ! General Assetnb l y at Bono], pp. 8/U-378. Wakahama, G., 197~. ThB role of melt-water in densification pro~esses of snow and firn. International Symposium Cln Snow Mechanics, IAHS Publication 114. pp. 66-72. Wankiewicz. A., l97ti. water percolation within a det>p snowpack--Field investigations at a site an Mt. Seymour, British Columbia. UnpublishP.d Ph.D. di&sertation. University of British Columbia. l77 pp. Wankiewicz 0 A., 1~78a. snowpacks. 593-Ei!JU. Water prASSUre in ripe Resources Research 14[4}: Wankiewicz, A., 197~b. Hydraulic characteristics of snCIVI lysimeters. t->roceF.Jdinos of the 3bth Annual Eastern Sno\'1 Conference, Pll• 1UJ-115. Wankiewicz, A., l!:/9. A rE!view of water movement in sPow. In: Proceedinos, Hodel ing of Snov; Covl'r Hunoff. S.C. Colbeck and t-1. Ray (f:ditor:;j. U.S. Army Corps of Engineers, CRREL, Hanover, New Hampshire, piJ. 222-252. Wisler, C.O. and E.~. l:lrc,ter, 1849. Hydrology. John Wiley and Sons. pp. 136. Work, A.A., 1948. Snow-Layer density changes. Transactions GAophysic::~l Union 29[4]:525-546. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 PRECIPITATION MEASURED BY DUAL GAGES, WYOMING-SHIELDED GAGES, AND IN A FOREST OPENING 1 David L. Sturges ABSTRACT: Precipitation measured by dual gages and a Wyoming-shielded gage in exposed locations was compared to that measured in a nearby forest stand. The Wyoming shield provided adequate protec- tion from wind for rain, but not for snow. The Wyoming-shielded gage measured 37% to 53% of actual precipitation bet- ween November and March, when air tem- peratures were well below freezing and monthly wind speeds were at a maximum. Analysis of individual snowfall events indicated that undermeasurement was directly related to wind speed. The pre- cipitation deficit increased about 7% for each 1 meter per second increase in wind speed for storms with air temperatures below -2°C. Estimates of precipitation determined from dual gages were not statistically different (p < 0.05) from measurements in the forest stand in 8 of 12 months. (KEY TERMS: precipitation gages; pre- cipitation measurement; Wyoming shield; dual gages; precipitation under- measurement; snow.) INTRODUCTION Accurately measuring precipitation falling as snow is a difficult task. Turbulence in the windstream created by the gage causes snow particles to be deflected over or around the gage, so that precipitation is undermeasured. Various types of shields have been developed to reduce the deleterious effects of wind. The shield devised by Alter (193 7) and modified by Warnick (1953) is widely used in the United States. Precipitation is still undermeasured, however, if snowfall is accompanied by wind (Larson and Peck 1974). A shield specifically designed for gages operated in an exposed, windy environment was developed at the University of Wyoming (Rechard and Wei 1980). Use of the Wyoming shield at Barrow, Alaska tripled estimates of winter precipitation compared to measurements from unshielded gages (Benson 1982). Black (1954) earlier had noted that snow on the ground at Barrow contained two to four times as much water as indicated by measured precipitation. Hamon (1973) believed actual precipi- tation could be computed from data collected at shielded and unshielded gages through use of a calibration coefficient to account for the influence of wind, and he developed the following relationship: (1) where A is computed precipitation, U is precipitation at the unshielded gage, S is precipitation at the shielded gage, and B is the calibration coefficient with a value of 1.73. The use of data from shielded and unshielded gages to estimate precipitation is known as the dual gage procedure. Relatively few studies rigorously evaluated the performance of precipitation gages in exposed locations. Gages pro- tected by a Wyoming shield and a modified Alter shield measured 80% and 55%, respec- tively, of precipitation measured by a gage protected with a Nipher shield, a 1Research Forester, Rocky Mountain Forest and Range Experiment Station, Forest Service, USDA, 222 South 22nd Street, Laramie, Wyoming 82070. 387 significant undermeasurement (p < 0.05) compared to the Nipher-shielded gage. The study was conducted at the Regina, Saskatchewan, airport, over five winters (Jones 1984). Sturges (1984) compared precipitation measured by a Wyoming-shielded gage located on windswept rangeland with that measured in a nearby forest stand. Both gages measured similar precipitation for rainfall events. However, undercatch at the Wyoming-shielded gage was significant (p < 0.05) in months with snowfall. Between November and March, the Wyoming- shielded gage caught only 40% to 60% of precipitation measured in the forest opening. A linear relationship existed between wind speed and the under- measurement of precipitation when indivi- dual snowfall events were analyzed. Hanson et al. (1979) compared preci- pition computed by the dual gage method with that measured by a gage equipped with a Wyoming shield. Similar quantities of precipitation were measured by the two procedures. OBJECTIVE This study compared precipitation measured in a forest stand with that measured on adjacent windswept rangeland by dual gages and by a gage protected with a Wyoming shield. STUDY SITES Twin Groves The Twin Groves site lies on a gently sloping bench just off the northern flank of the Sierra Madre mountain range (41°2l'N, 107°10'W) 30 km west of Saratoga, Wyoming. Forest-protected gages were located in an 18-ha stand of lodge- pole pine within an opening 0.75 tree height in diameter (Figure 1). Trees were about 15 m tall. The gage protected by the Wyoming shield was placed 970 m upwind from the forest-protected gage and 360 m from the tree margin for the prevailing wind direc- tion. Rangeland dominated by sagebrush about 30 em tall surrounded the stand of trees. Elevations at forest-protected and 388 Wyoming-shielded gages were 2,465 and 2,445 ~ respectively. Foote Creek The Foote Creek site lies off the eastern flank of the Medicine Bow mountain range, 67 km northwest of Laramie (41°36'N, 106°15'W). The forest-protected gage was placed in a stand of aspen, willow, and alder bordering Foote Creek. Trees surrounding the gage were about 7 m tall and the gage was placed in an opening 0.5 tree height in diameter. The gage protected by the Wyoming shield and the dual gages were placed on shortgrass prairie rangeland. The Wyoming shield was 195 m from the forest-protected gage. Dual gages were 6.1 m apart and about 48 m from the Wyoming shield (Figure 1). Elevation at the study site was about 2,360 m. STUDY METHODS Precipitation was recorded con- tinuously by gages with an orifice 20.3 em in diameter. The chart drum, driven by a stepper motor controlled by a quartz crystal, rotated every 8 days. The orifi- ces of forest-protected gages and dual gages were 3 m above the ground while ori- fices of gages protected by the Wyoming shield were 2.3 m above the ground. The forest-protected gage at Foote Creek was shielded by a modified Alter shield as were gages at Twin Groves after November 1982. The shielded dual gage was equipped with a modified Alter shield which was further modified to constrain baffles so that their lower end pointed towards the precipitation gage at a 30° angle (Hamon 1973). Gages at Twin Groves were usually serviced monthly and gages at Foote Creek were serviced bimonthly. Interval preci- pitation indicated by the chart trace was corrected to match interval precipitation determined from weighing the reservoir bucket at the start and end of each ser- vice interval. Chart precipitation recorded in the forest opening at Twin Groves was corrected to match precipita- tion measured by the nonrecording gage for the service interval. Comparison of interval precipitation determined from the Figure 1. Forest-protected precipitation gages (standard and recording) at Twin Groves (left); precipitation gage protected by a Wyoming shield and the dual gages at Foote Creek (right). change in bucket weight and recorded on the chart, provided an internal check on data consistency. The reservoir of each gage was charged with antifreeze to pre- vent freezing of precipitation and with transformer oil to prevent evaporation. Wyoming shields were built to speci- fications of Rechard and Wei (1980). Two concentric rings of Canadian snow fence 1.2 m in height with a solid density bet- ween 40% and 50%, surrounded the precipi- tation gage. The outer ring was 6.1 m in diameter and the top of the fence was 2.6 m above the ground surface, while the inner ring was 3.05 m in diameter and 2.3 m above the ground surface. The outer and inner rings of snow fence inclined towards the precipitation gage at angles of 30 ° and 45 ° from vertical, respec- tively. Wind speed, wind direction, and air temperature were recorded by a mechanical weather station located near the Wyoming shield. Sensors were 2 m above the ground surface at Twin Groves and during the first year of study at Foote Creek. The sensor was then moved to a 3-m height when dual gages were installed. Independent time control marks were placed on the chart at 6-or 12-hour intervals by a 389 marking system that was accurate to about 2 minutes per month. DATA COMPARISONS AND ANALYSES Comparisons of precipitation measure- ments in the forest opening with those from the Wyoming-shielded gages were .based on a 9-year record period extending from November 1976 to October 1985 at Twin Groves, and a 4-year record period extending from November 1981 to October 1985 at Foote Creek. Data from dual gages were available from November 1982 to October 1985. Freese (1960) and Reynolds (1984) describe use of the chi-square statistic to evaluate the accuracy of a new esti- mating or analysis procedure compared to a standard procedure. This analysis was used to determine if precipitation measured by an exposed gage was signifi- cantly different from precipitation measured in the forest opening. Analyses were based on monthly gage catch ratios calculated by dividing monthly precipita- tion measured at the exposed gage by pre- cipitation measured in the forest opening. A ratio of 1.0 indicated the two gaging systems measured identical quantities of precipitation. Comparisons were restricted to months receiving at least 0.64 em of precipitation; periods of questionable data validity were omitted from analyses. The analysis procedure involved calculating a critical range about average monthly gage catch ratios for data from Wyoming-shielded gages at Twin Groves and at Foote Creek and for dual gages at Foote Creek, based on a probability level of o.os. Precipitation was significantly undermeasured when a ratio value of 1.0 was not included in the critical range about each mean monthly gage catch ratio. The correction for bias (average precipi- tation undermeasurement) described by Freese (1960) was applied if the average monthly gage catch ratio was less than 0.9. RESULTS Twin Groves and Foote Creek had simi- lar climatic characteristics (Table 1). Average annual precipitation was 56 em at Twin Groves and 59 em at Foote Creek, with 75% of the total falling as snow. Average annual temperatures were about 2°C. Wind speeds were at a maximum in winter months and averaged more than 6 meters per second (m/s) between November and March. Prevailing wind direction was from the southwest at Twin Groves and from the west at Foote Creek. Wyoming Shield Monthly gage catch ratios for the Wyoming-shielded gage exhibited a cyclic pattern through the year (Tables 2 and 3). The ratio was about 1.0 in summer months, decreased to about 0.4 in midwinter, and then progressively increased until summer. Precipitatation was significantly under- measured (p < 0.05) between October and May at Twin Groves and at Foote Creek (Figure 2). Monthly ratios ranged from 0.35 to 0.54 between November and March when air temperatures were well below freezing, and monthly wind speeds were at a maximum. The shield did provide ade- quate protection for rainfall events in June, July, and August because the ratio was about l.O. 390 1.2 0 1.0 ... 0.8 .r. CJ -; 0.6 CJ ~ 0.4 tU 0 0.2 1- - i- - • - f Twin Groves f Foote Creek ' ' ~~rtr ' =t-... ' ' 1 ' ' ~ ' ' ... ~ : I tr ~ IT ' T ~ ~ ~ ' tr ' ' ~ ' ' ' ' ' r!-' ' ' ' ' ... 0 _j_ 1 I _j_ 1 I I _1 _1 1 1 _j_ Oct N 0 J F M A M J J A Sep Figure 2. Average monthly gage catch ratios and associated critical ranges (0.05 probability level) for precipita- tion gages protected by a Wyoming shield at Twin Groves and Foote Creek. Dual Gage The average monthly gage catch ratio for precipitation determined from dual gage data was less than 1.0 throughout the year, but the critical range about mean monthly values included a ratio value of 1.0, except in October, November, May, and June (Table 3). Monthly analyses were based on 3 years of information and so results are not as representative of variability in yearly climatic characteristics as data from the Wyoming shield. The hypothesis that dual gages provide the same estimate of monthly precipitation as measured in the forest opening was accepted in 8 out of 12 months. Relationship of Windspeed to Gage Catch at the Wyoming-Shielded Gage Precipitation measurements at the Wyoming-shielded gage were both greater and smaller than precipitation measure- ments in the forest opening for indivi- dual rainfall events (Figures 3 and 4). Storm wind speed was not related to gage catch at either study site. In contrast to rainfall events, undermeasurement of precipitation falling TABLE 1. Average monthly climatic characteristics at the study sites and calculated precipitation for Wyoming-shielded gages and dual gages. Years Parameter of Oct. Nov. Dec. Jan. Feb. Mar. Apr. May Jun. Jul. Aug. record -------··- Twin Groves Temp (oC) 5 -0.4 -5.8 -10.2 -9.4 -9.2 -5.4 -1.2 3.3 8.9 13.8 14.0 Wind speed (m/s) Avg. monthly 5 4.8 6.5 7.8 6.2 7.2 5.9 5.1 4.4 3.9 3.2 3.2 Avg. storm * 5.9 5.8 7.6 4.5 7.8 6.8 5.9 5.2 5.6 3.2 4.4 Precipitation (em) Actual 10 5.6 4.9 5.2 4.9 3.9 6.5 5.6 6.6 2.4 4.0 3.0 Wyo shield *** 4.1 2.2 1. 8 2.7 1. 6 3.1 3.6 5.3 2.3 3.9 2.8 Foote Creek Temp (oq_ 4 2.9 -3.4 -7.5 -7.5 -6.6 -3.4 -0.3 8.3 11.8 17.6 18.0 Wind sEeed (m/s) Avg. monthly 4 5.8 7.4 8.6 8.4 8.0 7.3 6.1 5.5 4.6 4.0 4.1 Avg. storm ** 4.3 4.7 7.4 5.9 4.4 5.3 5.9 6.0 2.8 5.5 3.9 Precipitati~~cm~ Actual 8 5.0 5.6 5.1 4.6 3.1 7.1 6.4 9.7 2.4 4.2 3.0 Wyo shield *** 3.6 2.9 2.1 2.3 1.4 3.6 4.0 7.5 2.4 4.2 3.0 Dual gages *** 3.8 4.7 5.0 4.4 2.7 6.5 5.3 7.3 2.1 4.1 2.9 *From 7 to 24 individual storms were analyzed in a month. **From 5 to 17 individual storms were analyzed in a month. ***Actual precipitation measured in forest opening multiplied by average monthly gage catch ratio. 391 Sep. 8.2 4.3 5.4 3.6 3.1 9.8 5.2 4.5 3.3 2.6 2.5 TABLE 2. Mean monthly gage catch ratios (Precip. at Wyoming shield/precip. in forest opening) at Twin Groves, and critical range (0.05 probability level) associated with mean monthly gage catch ratio. Year Oct. Nov. Dec. Jan. Feb. Mar. 1976 ** 0.693 1977 0.799 0.630 .173 1.105 0.557 0.694 1978 1.067 .439 .409 * .318 .449 1979 .751 .409 .362 0.616 .509 .564 1980 .688 .623 .558 .523 .449 .427 1981 .817 .548 .296 .338 .402 .557 1982 .557 .411 .209 .586 .373 .347 1983 .863 .381 .253 .313 .270 .537 1984 .428 .224 .228 .354 .481 .439 1985 .602 .513 .270 .302 n (year) 9 8 9 8 9 9 Avg. .730 .458 .353 .543 .403 .480 Critical ~0.265*** ~0.189 +0.245 +0.353 +0.14 7 +0.170 range *Data missing. **Less than 0.64 em precipitation in service interval. ***Precipitation measured by Wyoming-shielded gage significantly range does not include a rato value of 1.0. as snow by the Wyoming-shielded gage was significantly related (p < 0.05) to storm wind speed when air temperatures were below -2°C (Figures 3 and 4). The data set included 79 events at Twin Groves; maximum wind speed was 11 m/s and minimum air temperature was -19°C. There were 54 events at Foote Creek; maximum wind speed was 11.3 m/s and minimum air temperature was -25°C. The gage protected by the lvyoming shield caught about half as much precipitation as measured by the forest- protected gage for winds 5 m/s, and less than a quarter as much precipitation for winds 10 m/s. The relationship between wind speed and the deficit in gage catch for storms with temperatures between +2°C and -2°C was weaker than for storms with tem- peratures below -2°C though correlation coefficients were statistically signifi- cant (Figures 3 and 4). Data sets included predominantly snowfall events. The deficit in measured precipitation at the Wyoming-shielded gage increased about 5% for each 1 m/s increase in wind speed. DISCUSSION AND CONCLUSIONS The study was based on comparisons of precipitation measured in a small forest opening to reduce the deleterious effects 392 Apr. May Jun. Jul. Aug. Sept. 0.914 * * * * * .663 0.740 0.943 1.104 0.961 o. 708 .589 .724 .791 1.000 .802 .788 .581 .780 ** O.Q78 .947 .993 .632 .814 .908 .882 1.087 1.000 .533 .761 1.120 1.010 ** .856 .676 .601 1.000 1.029 .821 .9!5 .731 .882 .951 .814 1.016 .887 ;552 1.075 .964 1.068 .927 .759 9 8 7 8 7 8 .652 .797 .954 .986 .937 .87! +0.164 +0.191 +0.142 ~0.127 +0.156 +0.!32 less than in forest opening if monthly ratio + critical of wind on gage performance, with precipi- tation measured in exposed locations. This approach was the one Hamon (1973) used when determining the calibration coefficient to relate precipitation measured by shielded and unshielded gages to actual precipitation. The relationship between the actual vertical flux of preci- pitation and that measured in the forest stand is, of course, unknown, as it is for all studies utilizing this approach. Average monthly gage catch ratios were utilized to develop a function to predict undermeasurement of precipitation at gages protected by a Wyoming shield through the year (Figure 5). Undermeasurement(%) 3.177 + 0.218X + 6.616X2 _ 5.258X 3 + 1.140X4 _ 6.578X 5 (2) 10 3 10 5 10 1 10 11 where X is the number of days past August 16. Though the independent variable is number of days past August 16, the expression implicitly defines the effect of air temperature in determining whether precipitation falls as rain or snow and the effect of wind on gage performance in months with snowfall. It is applicable to sites with climatic conditions reasonably w U:l w TABLE 3. Monthly gage catch ratios for the Wyoming-shleld.ed gage and for dual gages (Precip. at Wyoming shielrl or dual gage/precip. in forest opening) at Foote Creek, and tl1e critical range (0.05 pr:obabil.ity level) assocl.aled wlth mean monthly gage catch ratios for Wyoming shield and rlual g.qge data. Year 1981 1982 1983 1984 1985 n (years) Avg. Critical range Year 1981 1982 1983 1984 1985 n (years) Avg. Critical range Oct. Ratio Wyo shield 0.662 .896 .590 .693 4 o. 710 +0.159* Dual gage 0.838 .695 .761 3 0.765 0.081* Apr. Ratio Wyo Dual shield gage 0.577 .844 1.005 .564 0.853 .523 .647 4 3 0.627 0.835 +0.178 +0.204 Nov. Ratio Wyo shield 0.485 .567 .675 .340 4 0.517 Dual gage 0.812 .983 .705 3 0.833 +0.172 +0.159 May Ratio Wyo Dual shield gage 0~897 .700 0.804 .685 .709 .830 .760 4 3 o. 778 0.758 +0.125 +0.054 Dec. Ratlo Wyo Dual shield gage 0.452 .381 1.073 .434 1.106 .382 0.749 4 3 0.412 0.976 +0.044 +0.198 Jun. Ratio Wyo Dual shield gage 1.043 0.959 0.890 1.026 • 777 .926 .927 4 3 0.988 0.865 +0.063 +0.089 Jan. Ratlo Wyo shield 0.566 .529 .387 .490 4 0.493 Dual gage 1.015 0.891 .969 3 0.958 +0.094 +0.080 Jul. Ratio Wyo Dual shield gage 1.000 1.048 1.033 0.998 0.920 .990 0.995 4 3 1.009 0.983 +0.031 +0.061 Feb. Ratio \~yo shield 0.548 .422 .530 .344 4 0.461 Dual gage 0.639 1.405 .614 3 0.886 +0.116 +0.509 Aug. Ratio Wyo Dual shield gage 0.903 1.060 0.966 .950 .900 1.059 1.084 4 3 0.993 0.983 +0.088 +0.095 Mar. Ratio Wyo shield 0.477 .574 .594 .385 4 0.507 Dual gage 1.206 0.880 .677 3 0.921 +0.117 +0.282 Sep. Ratio Wyo Dual shield gage 1.000 0.895 .957 .600 .472 .607 .875 4 3 o. 776 .768 +0.247 +0.294 *Precipitation measured by Hyoming-shielded gage or dual gages is significantly less than in forest opening if monthly ratio + critical range does not include a ratio value of 1.0 ~ 100 0 .s::. 80 ~ (,) 60 ~ "C -IU G) (,) 40 + .. + .s::. G) 20 ~ + + Ill > 0 + Cl 0 + + 1: --++ + + ~ .-'~ + + E 0 20 ....... T> +2°C 0 .s::. 40 ~ >-(,) 3: -IU 60 ~ (,) .. G) 80 ~ "C I I I I I I 1: 100 ~ 2 4 6 8 10 12 14 Avg storm wind speed (m/s) ~ 100 0 ....... 80 Deficit= 5.85(wind speed(m/sll-4.39 .s::. r =0.52 (,) "C -60 IU G) (,) 40 .. .s::. G) Ill > 20 0 Cl 0 1: --~ " E 0 20 + 0 T = -2°C to +2°C .s::. 40 + + >-(,) 3: -.f IU 60 (,) + + + .. + t + + G) 80 "C 1: 100 ++ ~ ~100 0 80 Deficit= 7.63(wind speed(m/sll• 12.87 .s::. r = 0.82 (,) 60 "C -IU G) (,) 40 .. .s::. G) 20 Ill > Cl 0 0 1: --+ ~ ++ + E 0 20 + ....... T < -2°C 0 .s::. 40 >-(,) 3: -IU 60 (,) .. 80 G) "C 1: 100 ~ Figure 3. The relationship between wind speed and performance of the Twin Groves precipitation gage protected by a Wyoming shield. Precipitation events were stratified into three temperature regimes. Top--temperatures greater than +2°C. Middle--temperatures -2°C to +2°C. Bottom--temperatures less than -2°C. 394 ~100 0 .s::. 80 1- (,) 60 1-"C -IU G) (,) 40 -.. .s::. G) Ill > 20 -+ + 0 ± + Cl 0 1: --~ •• i'\-?ft. E 20 1-+ + 0 .s::. T > +2°C >-(,) 40 1- 3: -IU 60 1- (,) .. 80 -G) "C 1: 100 I I I I I ~ 2 4 6 8 10 12 14 Avg storm wind speed (m/s) ?;.100 Deficit= 5.17(wind speed(m/s))-0.65 .s::. 80 r= 0.71 (,) 60 "C -IU G) (,) 40 .. .s::. G) 20 Ill > 0 Cl 0 1: --~ 20 E 0 ....... T = -2°C to +2°C 0 .s::. >-40 3: (,) -IU 60 (,) .. G) 80 "C 1: 100 ~ ?;.100 80 Deficit= 5.92(wind speed(m/sll•16.41 .s::. r = 0.78 (,) 60 -"C IU G) (,) 40 .. .s::. G) Ill > 20 Cl 0 1: --0 ~ 20 E 0 ....... T < -2°C 0 .s::. >-(,) 40 3: -IU 60 + ++-~" (,) .. .. + + G) 80 "C 1: 100 ~ 2 4 6 8 10 12 14 Avg storm wind speed (m/s) Figure 4. The relationship between wind and performance of the Foote Creek pre- cipitation gage protected by a Wyoming shield. Precipitation events were stratified into three temperature regimes. Top--temperatures greater than +2°C. Middle--temperatures -2°C to +2°C. Bottom--temperatures less than -2°C. Undormom (%) = 3.177 • 0.218X • 6.616X2 -5.25BX3 • 1.140X4 -6.57BX5 ---;-T ---;or -;or --;orr- R2 = 0.115 -100 ! 80 60 40 20 Wind apeed v ov 0 40 v ~ v ~ ~~J ~ I 80 120 160 200 240 280 320 360 Number of days past August 16 I I I I I I I I I I I I I I I I I I I I I I I I I Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug - - Figure 5. Undermeasurement of precipita- tion through the year as a function of the number of days past August 16. Average monthly wind speed is also shown. similar to those in Table 1. Information about precipitation undermeasurement in relation to monthly wind speeds can be used to estimate undermeasurement in loca- tions where climatic conditions are appre- ciably different from monthly averages in this study (Table 1). The relationship is based on average monthly gage catch utios, so precipitation estimates for individual storms may be in considerable error if meteorological conditions during storms are not representative of average monthly conditions. From a theoretical standpoint, preci- pitation should be measured without error in the absence of wind regardless of gage location or type of shielding. Wind speed had no effect on the quantity of precipi- tation measured at the Wyoming-shielded gage when precipitation fell as rain (Figures 3 and 4), a finding that agrees with results based on monthly gage catch ratios. Intercept values for precipitation data collected at the Wyoming-shielded gage at storm temperatures between +Z°C and -Z°C were not significantly different from zero, in accordance with the theore- tical relationship (Figures 3 and 4). However, there was a significant (p < 0,05) precipitation deficit in the absence of wind at both study sites for storms with temperatures below -Z°C. Just 3 of 395 79 events at Twin Groves and 6 of 54 events at Foote Creek were accompanied by winds less than 2 m/s, and study data poorly defined the relationship between wind speed and precipitation under- measurement for low wind speeds. Correlation coefficients between wind speed and gage undercatch were not as strong for precipitation falling between +Z°C and -Z°C as for precipitation falling at lower temperatures. The presence of some rainfall events in this data set pro- bably contributed to the poorer rela- tionship, but at the same time tended to pull the regression line towards the ori- gin. Hanson et al. (1979) found that simi- lar quantities of precipitation were measured by a Wyoming-shielded gage and by dual gages for both rainfall and snowfall events. Precipitation calculated from dual gages at Foote Creek was usually not significantly different from actual preci- pitation, but gages protected by a Wyoming shield significantly undermeasured preci- pitation falling as snow. Thus, under study conditions, the dual gage approach was superior to use of a Wyoming shield to measure winter precipitation. As Hanson et al. (1979) reported, the legibility of the pen trace of a Wyoming- shielded gage was greatly superior to the pen trace of the shielded member of the dual gage setup. Wind-induced vibrations caused the pen of the shielded gage to "paint" a line up to 1 em wide, which obscured all detail of the pen trace. After 1 year of operation, support for the fixed windshield was separated from the post supporting the gage and the gage- support post was firmly guyed. These modifications vastly improved the quality of the chart trace and made it comparable to the trace of the Wyoming-shielded gage. The Wyoming shield provides adequate protection from wind if precipitation falls as rain. Summer precipitation at Foote Creek was nearly identical in exposed and forest-protected locations and monthly critical ranges were narrower than those for Twin Groves (Tables 2 and 3), a reflection of the close physical proximity of Foote Creek gages. Twin Groves gages were about 1 km apart, and much of the between-storm variation reflected random, small-scale differences in storm patterns. Precipitation received as snow measured in gages protected by a Wyoming shield should be considered a conservative estimate of actual precipitation. Undermeasurement can be a significant source of error if precipitation infor- mation is used as input data for hydrolo- gic models, or used as a design parameter for snow control engineering purposes. For example, precipitation measured at the Wyoming shield in months when snow reloca- tion is an important hydrologic process, averaged 47% of that measured in the forest opening. The severity of under- catch increases with increasing wind speed. Thus, undercatch would be more severe on the plains than in areas where wind is not such a dominant feature of the environment. ACKNOWLEDGEMENTS The Foote Creek study site is located on land belonging to the Bear Creek Cattle Company of McFadden, Wyoming; their interest and gracious cooperation with study efforts is appreciated. The Rocky Mountain Forest and Range Experiment Station laboratory at Laramie is main- tained in cooperation with the University of Wyoming. Station headquarters is at Fort Collins, in cooperation with Colorado State University. LITERATURE CITED Alter, J. C., 1937. Shielded Storage Precipitation Gages. Monthly Weather Review 65:262-265. Benson, C. S., 1982. Reassessment of Winter Precipitation on Alaska's Arctic Slope and Measurements on the Flux of Wind Blown Snow. Report UAG R-288. Geophysical Institute, Univ. of Alaska, Fairbanks. Black, R. F., 1954. Precipitation at Barrow, Alaska, Greater Than Recorded. Trans., Amer. Geophysical Union 35:203-206. Freese, F. 1960. Testing Accuracy. Forest Science 6:139-145. Hamon, W. R., 1973. Computing Actual Precipitation. In: Distribution of Precipitation in~ountainous Areas. Vol. I, P• 159-174. World Meteorological Organization No. 326. 396 Hanson, C. L., R. P. Morris, and D. L. Coon., 1979. A Note on the Dual-Gage and Wyoming Shield Precipitation Measurement System. Water Resources Research 15:956-960. Jones, K. H., 1984. A Comparison of Various Snow Gauges on the Canadian Prairies Over Five Winters. Environment Canada Atmospheric Environment Service, Scientific Services Regina, Saskachewan, Report No. CSS-R84-04. Larson, L. W. and E. L. Peck., 1974. Accuracy of Precipitation measurements for Hydrologic Modeling. Water Resources Research 10:857-863. Rechard P. A. and T. C. Wei., 1980. Performance Assessments of Precipitation Gages for Snow Measurement. Water Resources Series No, 76, Water Resources Research Institute, Univ. of Wyoming, Laramie. Reynolds, M. R. Jr., 1984. Estimating the Error in Model Predictions. Forest Science 30:454-469. Sturges, D. L. 1984. Comparison of Precipitation as Measured in Gages Protected by a Modified Alter Shield, Wyoming Shield, and Stand of Trees. In: Western Snow Conference Proceedings 1984. p. 57-67. Warnick, C. C., 1953. Experiments with Windshields for Precipitation Gages. Trans. Amer. Geophysical TJnion 34:379-388. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 THE MASS BALANCE OF SNOW COVER IN THE ACCUMULATION AND ABLATION PERIODS Esko Kuusisto ABSTRACT: The mass balance of snow cover is affected by solid and liquid precipitation, mass flux from the soil, the net amount. of snow transport by wind, evaporation (and sublimation) and water y1eld from the snow cover. The present status of knowledge of these components is reviewed. Even the measurement of their point values is still far from accurate, and areal estimates are very inadequate. -Gauge measure- ments normally underestimate solid precipitation at least by 20 %, but corrections based on in situ meteorological observations and gauge characteristics can lead to monthly values with an error of less than ± S %. Mass flux from the soil rather rarely occurs to such an extent that it would significantly affect the mass balance of the snow cover. Transport of snow by wind can essentially affect the areal distribution of snow cover even in partially forested basins. The combined evapor- ation from snow cover, interception storage and windborne snow can amount to several tens of millimeters per season. Water yield from snow cover is due to liquid precipitation and to three snowmelt components. Of these components, the basal snowmelt is a very inadequately known phenomenon which can cause considerable losses from the snow cover during the accumulation period. (KEY TERMS: snow cover, snow accumulation, evaporation, intercep- tion, snow transport.) INTRODUCTION Snow -Nature's white, porous reservoir. In the Northern Hemisphere, the average maximum extent of snow cover is on land about 49 x 106 km2 and at sea about 15 x 106 km 2 (Untersteiner, 1984 ). Thus one quarter of the hemisphere and one half of its land area is covered with snow in midwinter. The average maximum mass of the Northern Hemisphere's snow cover is about 9.5 x 101 5 kg, giving an average maximum water equivalent of 150 mm. Snow cover has a variety of influences. To the hydrologist, the most important is its role as a major source of streamflow. In Finland the spring runoff amounts to 40-60 % of the total annual runoff. Practically all annual flood peaks in northern Finland and 70-80 % in the south of the country are due to snowmelt. Snow-soil interactions are very important both hydrologi- cally and hydrochemically. The water immobilized in the snow Hydrological Office, Box 436, SF-00101 Helsinki, Finland 397 cover and the materials reaching the snow by wet and dry de- position are rapidly released during the ablation period. When meltwater penetrates the soil profile, its ion composition may change drastically. Therefore the study of the quality of groundwater and of the whole subsurface runoff is closely connected with snow accumulation and snowmelt (Soveri, 1985). Usually, the snow scientist is mainly interested in the water equivalent of the snow cover. He or she is glad if a reasonably accurate estimate of an areal mean of water equivalent can be obtained. However, estimates of the different components of the mass balance of the snow cover are also important. Simulation of the accumulation of snow cover for snowmelt-runoff models is crude without a knowledge of all the component processes. The areal distribution of the snow cover is deter- mined by the micro-, meso-and macroscale heterogeneity of different mass balance components. This distribution is be- coming increasingly important in snowmelt-runoff models, particularly if it is necessary to know the true origin of the water discharged during the spring flood. Even if successful forecasting of the magnitude of the flood is possible without this knowledge, the increasing use of hydrological models together with water quality models requires a thorough under- standing of runoff formation. In operative snowmelt-runoff models, two different ap- proaches are used to estimate the areal water equivalent at the beginning of the melting season. Either a snow accumulation model is used or the areal water equivalent is estimated on the basis of snow surveys. A discrepancy of several tens of milli- meters between these two results frequently occurs (Vehvilai- nen, 1986), and the model user finds it difficult to decide which value he should start with. Part of the discrepancy is due to poor knowledge of the dif- ferent processes affecting the mass balance of the snow cover in the accumulation model. It is of considerable significance to flood formation, which process has had the greatest influence. Losses from the snow cover to the underlying soil due to basal snowmelt are not losses from the point of view of flood forma- tion, whereas losses due to evaporation are. The economic value of the correct information on the type of losses in- volved may be considerable. THE MASS BALANCE EQUATION The mass balance of the snow cover can be expressed by the equation: (1) where: W = water equivalent of the snow cover, both in solid and liquid form P 5 = solid precipitation P1 = liquid precipitation Mg = mass flux from the soil D = net amount of snow transport E = evaporation Y = water yield from the snow cover All the components can be interpreted as accumulated values since the formation of the snow cover. The components P 5 and P1 are normally considered to include rime deposition and condensation, respectively, because these are inevitably in- cluded in measurements by precipitation gauges (WMO, 1965). The component D also includes that part of the intercepted snow which later falls down onto the snow surface. Evapor- ation includes sublimation; in fact all of this component is sub- limation in the case of dry snow and most of it even in the case of wet snow. The component Y is the sum of the following processes: -melting at the base of the snow cover -melting within the snow cover -melting at the snow surface -liquid precipitation into the snow cover Only the water produced by the first process is immediately released from the snow cover. The other components are re- leased with a delay and attenuation which depend on the re- tention of liquid water within the snow cover. SOLID PRECIPITATION It is well known that the aerodynamic error of precipitation gauges is much greater for solid than for liquid precipitation. Even gauges equipped with a wind shield often underestimate the amount of snowfall by 20-30 per cent. In very open sites with high wind speeds, unshielded gauges have been reported to catch only one third of the snowfall (Benson, 1982). The aerodynamic error also increases with decreasing tem- perature as a result of the change in structure of snow particles (Sevruk, 1982). If data on both these meteorological variables -wind speed and air temperature -are available, a rather re- liable estimate of the true amount of winter precipitation can be obtained for shielded gauges, except in very open terrain. For example in Finland the detailed studies carried out by the Finnish Meteorological Institute and the Hydrological Office have made us believe that the accuracy of monthly precipi- tation in winter is often better than ± 5 %. When exception- ally heavy snowstorms occur, the accuracy may be consider- ably lower. 398 Overestimation of solid precipitation may sometimes occur due to the catch of windborne snow by the gauge. Sevruk (1982) has presented corrections due to this phenomenon for the Tretjakov gauge as a function of wind speed and the dur- ation of the storm. For example, with a wind speed of 10m s-1 and duration of 12 h, the correction is 1.1 mm. LIQUID PRECIPITATION Liquid precipitation into the snow cover during the ac- cumulation period is a rather frequent phenomenon in many regions. It can usually be estimated rather accurately with or- dinary gauges. The percentage correction of gauged liquid pre- cipitation in Finland is slightly higher in winter than in sum- mer. This is due to a larger proportion of drizzle in winter. The liquid water retention capacity of snow cover in Fin- land is normally 5-6 % by volume at the beginning of the ab- lation period (Lemmelli, 1970). Although it is smaller in the accumulation period due to a lower snow density, liquid pre- cipitation is often retained and no water yield from the snow cover occurs. In Helsinki the liquid precipitation during a three month winter period (December 16 -March 16) averaged 31 mm in 1958-79. Only 8 mm of this amount was released (Kuusisto, 1984). During the ablation period, liquid precipitation may be an essential factor in flood formation. These rain-on-snow events can cause serious flooding e.g. in the Columbia River Basin (Harr, 1981). In Finland heavy falls of rain are not common during snowmelt; rainfall normally represents only 8-12 per cent of the total water yield. MASS FLUX FROM THE SOIL An upward water movement within the soil towards the freezing front is a phenomenon which has been known for sev- eral decades. In recent years, it has been studied e.g. by Mageau and Morgenstern (1980), and Kane and Stein (1983). This movement may extend through the soil-snow interface and add to the water equivalent of the snow cover. Santeford ( 1978) reported that in interior Alaska the aver- age magnitude of the vapor transport from below to the snow cover was about 30 mm of water during the winter season. This moisture originated mainly from the organic soil and moss layer. The thickness of this layer was reduced from 25 em by about one half as a result of the desiccation and compaction processes. The temperature gradient near the soil-snow interface must probably be rather large for the upward vapor transport to oc- cur. In addition, a significant moisture source must also be available. MELTING AT THE BASE OF THE SNOW COVER Heat flux from the soil or shortwave radiation penetrating through the snow can cause melting at the base of the snow cover. In Finland the latter can be significant only in the ab- lation season, but the former process may occur throughout the winter. In heavy clay soil below the depth of maximum soil frost penetration, the heat flux at Jokioinen in southern Finland averaged 2.8 W m-2 in November-April in 1964-70 (Fig. 1). It decreased steadily throughout the winter so that the average flux in April was only 50 per cent of that in November. Similar values have been reported elsewhere in southern Finland. The flux decreases towards the north due to the smaller storage of heat in the soil during the summer, in Lap- land the flux is typically around 1.0 W m-2. Thus the upward heat flux could theoretically cause considerable basal melting, in southern Finland over 20 mm/month and in Lapland 5-10 mm/month. The amount of basal snowmelt is determined by the energy balance at the base of the snow cover. Upward heat losses through the snow cover often exceed the heat flux from the soil, resulting in the formation of soil frost. The release of latent heat in the formation of soil frost es- sentially affects heat fluxes in the upper soil layers. Part of this energy is conducted downwards, leading to an increase in soil temperature below the freezing front, this phenomenon was already observed by Keranen ( 1920). The remainder is conduc- ted upwards, retarding the freezing process and also leading to melting of soil frost or snow, if the heat losses through the snow cover decrease. Studies by Solantie (1976) and Kuusisto (1984) can be used to derive areal estimates of basal snowmelt in Finland during the period December 16 -March 16. In the western coastal areas with a thin snow cover, basal snowmelt within this three month period is 10-15 mm. In southern Finland it is about 20 mm, in eastern Finland where the snow cover is thicker it amounts to 25-35 mm. Despite thick snow covers, the smaller heat flux from the soil and the lower air tempera- tures restrict the basal snowmelt to 10-25 mm in Lapland. Sr-------------------------------------, wm-2 ............ /Maximum 4r-: ••••••••••• , ........... .( . . Mean 1 3-x 3 .. ·············: :········· ··: I : •..•........ ,:: : : ........... , 0 .J . ............ ~ ........... ...._ __ -;. :r2-r .r··········· /··········= Minimum : ............ ,_ 1 . I I I I 0~----L-----~----~----~----~----~ Nov Dec Jon Feb Mar Figure 1. The average and extreme monthly heat fluxes below the depth of maximum soil frost penetration at jokioinen (61° N, 23° E) in 1964-70. Apr 399 At the base of a shallow snow cover, the heat flux is nor- mally directed downwards during the ablation season (Weller and Holmgren, 1974; Granger and Male, 1978; Ohmura, 1982). Part of the shortwave radiation penetrating through the snow cover then causes melting of the soil frost. SNOW TRANSPORT BY WIND » ... several kilometers long, up to 20m wide and 10m deep, they may contain from 20 to over 100 metric tons of water equivalent per meter of length along the bank.» This is how Benson ( 1982) described the snow drifts on the banks of the Meade River near Barrow, Alaska. On the Canadian Prairies, Steppuhn and Gray (1978) estimated the potential transport fluxes during the winter season to vary be- tween 2.6 and 22 tonnes/m. There are three modes of snow transport by wind: ground creep, saltation and turbulent diffusion. Generally, it is ac- cepted that most of the snow is transported by the two latter modes and mainly within a few centimeters above the snow surface (Gray and a!, 1978). If there is no crust on the snow surface, even a fairly light wind (3-5m s-1 ) may initiate snow transport. The formation of a crust by refreezing of meltwater inhibits transport, but for new snow above such crust it may again be very effective. The high water equivalents in snow drifts can cause serious errors in gamma snow surveys. Uryvaev and a! ( 1969) gave an example of a 10 % underestimate with drifted snow. Fritzsche and a! (1975) showed for an average water equivalent between 50 and 250 mm that underestimates up to 50% could occur if the ratio between the water equivalent of drifted and non- drifted areas became larger than 4. A correction to the results of gamma surveys can be applied, if ground information is available on the water equivalents of snow drifts (Cork and Loijens, 1980). EVAPORATION Total evaporation (Etot) during the accumulation and ab- lation periods consists of the following components: where: E.s = evaporation from snow cover Eg = evaporation from bare ground Et = transpiration Ei = evaporation of intercepted water or snow Eb = evaporation of windborne snow (2) Of these components, Es affects directly, Ei and Eb indirectly the mass balance of snow cover. A rather accurate method for the measurement of evapor- ation from snow cover is the use of evaporation pans. Differ- ent kinds of pans have been used (e.g., Croft 1944, Nyberg 1966, Lemmela 1970, Kaitera and Terasvirta 1972). The pans used by the Finnish Hydrological Office are lathe-turned from white plastic, they consist of double cylinders which are ad- justable according to variations in snow depth. Their surface area has been 500 cm2. The other components of Eq. (2) are much more inad- equately known than direct evaporation from snow cover. In the case of continuous snow cover and sub-zero tempera- tures, Eg and Et can obviously be neglegted. On the other hand, evaporation of intercepted or windborne snow can under certain conditions be much greater than evaporation from the snow cover itself. In Finland it has been estimated that the storage of intercepted snow in a coniferous forest may reach areally averaged values as high as 45 mm (Seppanen, 1959). At this level of maximum storage, the albedo of the forest is very high, thus reducing the net energy available for evaporation. Therefore the most favourable conditions for Ei may occur at intermediate values of interception storage. For example at an albedo of 0.40 the rate of evaporation from the interception storage could theoretically reach 1. 9 mm d-1 in southern Finland. With a canopy cover of 50 %, this would mean an areal value of almost 1.0 mm d-1. Snow particles transported by wind are subject to evapor- ation losses, which may be considerable. Tabler (1975) esti- mated that 3 7, 57 and 7 5 per cent of relocated snow in Wyoming evaporated over transport distances of 1500, 3000 and 4600 m, respectively. The availability of energy and the presence of a vapor pressure gradient largely determine the evaporation from snow. In midwinter, both these conditions are unfavourable in Finland (Fig. 2). In the whole country, the radiation balance is negative from November to March. However, days with a posi- tive radiation balance occur even in midwinter, and small amounts of evaporation can occur. 250.------------, MJrri2 .. g 150 .9 0 ..,. 100 c 0 :g :0 0 a: 0 :t! u i "0 c 0 :;:; 0 ... "' -~ NDJFMAM Figure 2. The monthly averages of the radiation balance and saturation deficit in southern Finland (solid line) and in northern Finland (dashed line) in November- April in the period 1971-80. In the ablation period, energy supply and saturation deficit are sufficient for evaporation. In that period, a good corre- lation exists between daytime evaporation from snow cover and the dewpoint temperature (Fig. 3). In an experimental field in southern Finland, this dependence took the form: 400 1.5r---------------, mm/12h c 0 :;:; 0 0.5 5 a. 0 > LJJ 0 • -10 -5 Dewpoint 0 Figure 3. The dependence of daytime snow evaporation on the dewpoint temperature during ablation period in southern Finland. E5 =-0.10 Td + 0.02 (3) where E 5 is expressed as mm/12 h and the dewpoint tempera· ture T d in °C. Table ( 1) summarizes the monthly estimates of different components of evaporation in southern Finland in January- April. The total evaporation during these months amounts to 39 mm in open field areas and 49 mm in forests. Direct and in· direct losses from snow cover are 18 mm in open areas and 36 mm in forests. Most of these losses occur during the ab· lation period. Depending on climatic and physiographic conditions, either evaporation or condensation may prevail during snowmelt (Table 2). For the whole melting season, the average mass ex· change due to evaporation-condensation rarely exceeds 1.0 mm d-1. In a single day, both evaporation and condensation may exceed 2.0 mm (e.g. de La Casiniere, 1974; Moore and Owens, 1984). INTERCEPTION Interception of solid and liquid precipitation by tree canopies often has a considerable indirect influence on the mass balance of the snow cover. In a coniferous forest, the in· terception storage formed by snow and rime can have an areal water equivalent of several tens of millimeters. The portion of snowfall intercepted can reach 20-30 % from the seasonal total (Braun, 1985). There are five processes by which the intercepted snow may leave the canopy (Miller, 1965): TABLE 1. Estimated values of different evaporation components in southern Finland in winter and spring months. Estimated value of the component (mm/month) Component Open field J F M A Es 0 6 4 Eg 0 0 2 18 Et 0 0 0 E· I 0 0 0 0 Eb 1 2 3 1 Etot 3 11 24 -falling of blowing of dry snow -sliding or falling of partly melted bodies of snow -dripping of flowing of meltwater -vapor flux from meltwater film -vapor flux from snow The last two processes result in the component Ei in Eq. (2). This component may be considerably greater than evapor- ation from snow cover. Many facets of tree crowns are perpen- dicular to the solar radiation, and therefore they obtain more shortwave energy than horizontal surfaces. The large surface- to-volume ratio of the intercepted snow and high wind speeds at the canopy level also enhance evaporation. Several methods have been applied in studying the removal mechanisms of intercepted snow. Goodell (1959) used weighing method, he found a 60 % loss in less than 3 hours, the whole loss being due to evaporation. Hoover and Leaf (1967) used time sequence photography. They concluded that mechanical removal predominated evaporation. Miller (1965) estimated that the removal of intercepted snow by sliding pro- ceeded at the rate of about 2 mm h-1 , while removal by evaporation seldom exceeded 1 mm d-1 . AREAL VARIABILITY OF MASS BALANCE COMPONENTS Three different scales can be distinguished in the areal dis- tribution of snow cover: 1. Microscale variability. This can be defined as the variation of snow properties over a homogeneous area. This area can be a section of an open field, located far enough from the edges of the surrounding forest. It can also be a forested area, consisting of trees of approximately equal age and spacing, or a slope with a constant angle in any terrain type. Characteristic linear distances of microscale variability range from a few centimeters up to 100m. 2. Mesoscale variability. This is mainly caused by variation of physiographic factors: terrain types, slopes, aspects, vari- ations in forest density etc. The characteristic linear dis- tance of mesoscale variability depends on the scale of varia- ation of physiographic factors. In Finland it usually ranges from a few tens of meters to several kilometers. Total 11 20 0 7 39 401 Forest J F M A Total 0 0 3 4 7 0 0 0 6 6 0 0 0 7 7 3 5 8 12 28 0 0 0 1 3 5 12 29 49 3. Macroscale variability. This depends mainly on the variation of climatological factors over a region. Typical scales in Fin- land are from a few kilometers (coastal effects) to hundreds of kilometers. These three types of variability have been discussed by sev- eral authors (e.g. Gray and a!. 1978, Gottschalk and Jutman 1979), but the definition of characteristic linear distances varies. -All components of the mass balance of snow cover vary in all these scales. Consequently the water equivalent of the snow cover has a considerable micro-, meso-and macrovariability, all of which have been rather extensively investigated. On the other hand, only a few studies have been made on the variability of the dif- ferent components, particularly in micro-and mesoscale. Let us consider a small, partially forested river basin in Fin- land. If it is not located on the coast, the macroscale varia- bility can be neglected. The following estimates of the effect of the variability of each component on the overall variability of the water equivalent in micro-and mesoscale can be made: microscale mesoscale Ps moderate small PI small small Mg small small D large large E small moderate y moderate moderate Even in a sheltered environment, the transport of snow by wind can often cause considerable areal variability of the water equivalent. In microscale this is manifested by ripples and ridges with a variety of forms, plus depressions around trees and other obstacles. In mesoscale the accumulation of drifted snow may occur e.g. in river beds and near the edges of an open field. Uneven radiative melting can cause considerable variations in water yield both in micro-and mesoscale. Solid precipi- tation is unevenly distributed in microscale in a coniferous forest, whereas its mesoscale distribution is rather even (if dif- ferences in altitude are small). Evaporation from intercepted snow in the forest or from snow transported by wind in the open can also have a moderate mesoscale effect on the distri- TABLE 2. The average daily values of snowmelt and snow evaporation in different climatic and physiographic conditions. Reference Site Elevation Observation Average Average (m) period melt evaporation mmd-1 mmd-1 Gold and Williams (1960) Open field (Canada), 45° N 100 March 1959 7 0.60 Treidl ( 1970) Open field (Michigan), 46° N January 23, 1969 15 -0.64 Dewalle and Meiman (1971) Forest opening (Colorado), 39° N 3 260 June 1968 50 0.18 de La Casiniere (1974) Open field in mountains (France), 46° N 3 550 July 1968 16 0.28 Open field in mountains (Spain), 41° N 1 860 April1970 10 0.49 Granger and Male (1978) Open field in prairies (Canada), 51° N Melting season 1974 8 0.09 Melting season 1975 5 0.17 Melting season 1976 3 0.05 Hendrie and Price (1978) Deciduous forest (Ontario), 46° N April 1978 10 0 Kuusisto (1978) Open field (Finland), 60° N 60 Melting seasons 1968-197 3 7 O.o3 Harstveit (1981) Open field in mountains (Norway), 60° N 435 April-May 1979-1980 12 0 -cloudy days 23 -0.70 -clear days 7 0.19 Braun and Zuidema (1982) Small basin, 23% forest (Switzerland), 47° N 800 Days with intense snowmelt, 23 -0.54 1977-1980 Eaton and Wendler (1982) Open field (Alaska), 65° N April1980 3 0.24 Kuusisto (1982) Open field (Finland), 61° N 104 Days with intense snowmelt, 14 -O.o3 1959-1978 Open field (Finland), 67° N 178 15 0.41 Moore and Owens (1984) Open field in mountains (New Zealand), 43° N 1450 Melting season 1982 31 -0.91 Vehvilainen (1986) Small basin, 82% forest (Finland), 64° N bution of the snow cover. In Finland the coefficient of variation of the areal water equivalent ranges in small basins typically from 0.2 to 0.5 dur- ing the period of maximum snow accumulation (Kuusisto, 1984). No quantitative estimates of the contribution of each mass balance component to these values have been made. RESEARCH AND DATA NEEDS It is obvious that fundamental measuring problems still exist in connection with all the components of the mass bal- ance of the snow cover. This is true even for point measure- ments, but overwhelming difficulties arise when it is attempted to estimate areal values. Some new methods of measurement will probably be devel- oped in the future. The attenuation of transmitted gamma radiation in the canopy at different elevations can be measured to study evaporation from the interception storage. A gamma transmitter placed at the soil-snow interface can perhaps give the water equivalent accurately enough to indicate the magni- tude of basal snowmelt or mass flux from the soil. The devel- opment of aerodynamic models together with observations on the structure of snow particles may improve the accuracy of 402 120 Melting seasons 1971-1981 5 0.08 gauged solid precipitation (Carlsson and Svensson, 1984). Accurate areal modelling of all the processes affecting the mass balance of the snow cover will, however, remain too complex for many years. Therefore the measurement of the water equivalent of snow will maintain its key role in snow ac· cumulation studies, and new, more accurate methods for these measurements should be developed. However, the practical value of the knowledge of the com· ponent processes will also increase. Such knowledge will help to diminish and interprete correctly the discrepancies between modelled and measured water equivalents of snow cover. This information is also needed in the study of the chemical com· position of the snow cover and in the forecasting of the effects of snowmelt on water quality. The need for the management of snow cover for different purposes (water supply, water quality, transportation, ava· lanche prevention) will also increase in some areas. This man· agement requires improved knowledge of some processes af· fecting the mass balance of the snow cover, particularly snow transport by wind. REFERENCES B~nson, C., 1982. Reassessment of winter prectp!tation on Alaska's Arctic Slope and measurements on the flux of wind blown snow. Univ. of Alaska, Inst. of Geophysics, Report R-288, 26 p. Braun, L., 1985. Simulation of snowmelt-runoff in lowland and lower alpine regions of Switzerland. ETH, Geographical Institute, Zurich, 168 p. Braun, L. and P. Zuidema, 1982. Modelling snowmelt during advection- melt situations is a small basin. Proc. Symp. Hydr. Research Basins, Bern, vol. 3, pp. 771-780. Carlsson, B. and U. Svensson, 1984. Wind correction factors for snow precipitation gauges. Fifth NRB Symposium, Vierumiiki, Finland, March 19-23, 1984, pp. 1.15-1.30. Cork, H. and H. Loijens, 1980. The effect of snow drifting on gamma snow survey results. Journal of Hydrology 48, pp. 41-51. Croft, A., 1944. Evaporation from snow. Am. Met. Soc. Bulletin 25, pp. 38-42. d~ La Casiniere, A., 1974. Heat exchange over a melting snow surface. Journal of Glaciology 13(67), pp. 55-72. D~walle, D. and J. Meiman, 1971. Energy exchange and late season snowmelt in a small opening in Colorado subalpine forest. Water Re- sources Research 7(1), pp. 184-188. Eaton, F. and G. Wendler, 1981. The heat balance during the snowmelt s~ason for a permafrost watershed in interior Alaska. Archives for Meteorology, Geophysics and Bioclimatology 31, pp. 19-33. Fritzsche, A. and E. Feimster, 1975. Snow water equivalent surveys of the Souris River Basin. EG & G Report EGG-1183-1668 (Prepared for U.S. National Weather Service). Gold, L. and G. Williams, 1960. Energy balance during the snowmelt period at an Ottawa site. IAHS Pub!. 51, pp. 288-294. Goodell, B., 1959. Management of forest stands in western United States to influence the flow of snow-fed rivers. IAHS Publication 48, pp. 49-58. Gottschalk, L. and T. Jutman, 1979. Statistical analysis of snow survey data. SMHI Rapporter RHO 20, 41 p. Granger, R. and D. Male, 1978. Melting of a prairie snowpack. Journal of Applied Meteorology 17(12), pp. 1833-1842. Gray, D. and a!., 1978. Snow accumulation and distribution. Modelling of Snow Cover Runoff, Hanover, New Hampshire, 26-29 September, 1978, pp. 3-33. Harr, R., 1981. Some characteristics and consequences of snowmelt during rainfall in western Oregon. Journal of Hydrology 53, pp. 277-304. Harstveit, K., 1981. Studier av sn0smelting i Dyrdalen 1979 og 1980. Norsk Hydrologisk komite, Rapport 9, 25 p. H~ndrie, L. and A. Price, 1978. Energy balance and snowmelt in a de- ciduous forest. Modelling of Snow Cover Runoff, Hanover, New Hampshire, 26-29 September, 1978, pp. 211-221. Hoover, M. and C. Leaf, 1967. Process and significance of interception in Colorado subalpine forest. Proc. Int. Symp. on Forest Hydrology, Pergamon Press, pp. 213-222. Kaitera, P. and H. Teriisvirta, 1972. Snow evaporation in southern and northern Finland, 1969-71. Aqua Fennica 2, pp. 11-19. Kane, D. and J. Stein, 1983. Water movement into seasonally frozen soils. Water Resources Research 19(6), pp. 1547-1557. 403 Keranen, J., 1920. Uber die Temperatur des Bodens und der Schnee- decke in Sodankylii. Helsinki, 200 p. Kuusisto, E., 1978. Optimal complexity of a point snowmelt model. Symposium on the Modelling of Snow Cover, Hanover, New Hampshire, 26-28 September, 1978, pp. 205-210. Kuusisto, E., 1982. Temperature index, energy balance-or something else? Symposium on the Use of Hydrological Research Basins in Planning, Sept. 21-23, 1982, Bern. Kuusisto, E., 1984. Snow accumulation and snowmelt in Finland. Pub!. of the Water Research Institute 55, Helsinki, 149 p. Lemmelii, R., 1970. On the snowmelt runoff and groundwater forma- tion on a glacifluvial esker in southern Finland. Univ. of Helsinki, Dept. of Geophysics, 119 p (in Finnish). Mageau, D. and N. Morgenstern, 1980. Observations on moisture mi- gration in frozen soils. Can. Geotechn. Journal 17, pp. 54-60. Miller, D., 1965. The heat and water budget of the Earth's surface. Ad- vances in Geophysics 11, pp. 175-302. Moore, R. and I. Owens, 1984. Controls of advective snowmelt in a maritime alpine basin. J. of Climate and Applied Meteorology 23, pp. 135-142. Nyberg, A., 1966. A study of the evaporation and condensation at a snow surface. Arkiv fOr Geofysik 4, Stockholm. Ohmura, A., 1982. Climate and energy balance on the arctic tundra. J. of Climatology 2, pp. 65-84. Seppiinen, M., 1959. On the quantity of snow lodging on branches in pine dominated forest during the time of snow destruction in Fin- land. IAHS Publication 48, pp. 64-68. Sevruk, B., 1982. Methods of correction for systematic error in point precipitation measurement for operational use. WMO, Operational Hydrology Report 21, Geneva, 91 p. Solantie, R., 1976. The computation of the water balance of Finland in the period 1931-60. Univ. of Helsinki, Dept. of Geophysics, 350 p (in Finnish). Soveri, J., 1985. Influence of meltwater on the amount and compo- sition of groundwater in quaternary deposits in Finland. Pub!. of the Water Research Institute 63, Helsinki, 92 p. Steppuhn, H. and D. Gray, 1978. Modification of a conceptual model to estimate the sublimation potential for wind-transported snow. Univ. of Saskatchewan, Div. of Hydrology, Saskatoon. Tabler, R., 1975. Estimating the transport of blowing snow. Univ. of Nebraska, Research Committee of Great Plains Agricultural Council, Publication 73, pp. 85-117. Treidl, R., 1970. A case study of warm air advection over a melting snow surface. Boundary-Layer Meteorology 1, pp. 155-168. Untersteiner, N., 1984. The cryosphere. In 'The Global Climate', ed. by J. Houghton, Cambridge Univ. Press, pp. 121-140. Uryvaev, V. and a!., 1969. Principal errors of airborne gamma surveying of snow cover. In 'Determination of the water equivalent of snow cover', ed. by. L. Vershinina and A. Dimaksyan, Jerusalem. Vehviliiinen, B., 1986. Modelling and forecasting snowmelt floods for operational purposes in Finland. Second Scientific Assembly of IAHS, Budapest, July 2-10., 1986. Weller, G. and B. Holmgren, 1974. The microclimates of the arctic tundra. Journal of Applied Meteorology 13, pp. 854-862. WMO, 1965. Guide to hydrometeorological practices. No. 168 TP 82, Geneva, 281 p. CHANNEL HYDRAULICS AND MORPHOLOGY 405 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 EROSION CONTROL FOR PLACER MINING Larry A. Rundquist and N. Elizabeth Bradle/ ABSTRACT: Techniques for the control of nonpoint-source sediment pollution are reviewed for application to placer mines in northern environments. Erosion-control techniques are divided into three categories: drainage control, site grading, and stockpile protection and location. Drainage control serves two primary functions: 1) to exclude un- polluted water from the mine site and 2) to direct water that has become polluted within the site through a treatment system prior to discharge to the receiving watc:rbody. Drainage-control techniques include stream channel diver- sions, bedrock drains, overland-flow diversion structures and settling-pond systems. Site grading serves to reduce the height, steepness, or length of erodible slopes, thereby reducing the potential for erosion of the slope. Site grading should be conducted concurrent with the mining operation to reduce erosion potential. Locating and pro- tecting stockpiles to minimize contact with watcrbodics also reduces nonpoint-sourcc sediment pollution. Stockpiles of vegetation, organic overburden, fine-grained inorganic overburden, oversized rocks, and fines removed from settling ponds may be retained for usc in site rehabilitation; stockpiles of these materials should be located away from erosive watcrbodies or protected with erosion-resistant materials. (KEY TERMS: erosion control; nonpoint-source pollution; placer mining; drainage control; stockpiling.) INTRODUCTION Placer mining has historically increased turbidity in downstream receiving waters. Much attention has been directed at studying and regulating the water quality of the point- source discharges from placer-mining opera- tions, with minimal study or control of non- point-source pollution. However, nonpoin t- source pollution may cause substantial water quality degradation. Recent advances have been made in the control of nonpoint-source pollu- tion from agriculture, silviculture, and urban construction. Measures proven to be effective in controlling sediment discharge resulting from these applications may be useful in reducing the erosion potential from mined areas. The term "nonpoint-source pollution" is used in this paper to include all stream turbidity, suspended sediment, and siltation (deposition of solids on the stream bed) resulting from soil erosion caused by human activity that cannot be traced to a single discharge point. Point-source pollution, in contrast, is highly localized with an origin that can be traced to an identifiable point where the polluted dis- charge reaches a body of water. For example, turbid effluent from the outlet of a settling pond is a point-source pollution, while sedi- ment washed off an exposed hillside is non- point-source pollution. A placer-mining operation typically removes upland and riparian vegetation and the soils overlying alluvial gravel deposits. Overburden is stripped to expose the gold-bearing gravels, which are processed to separate the placer deposits from the lighter alluvium. The mining process results in unvegetated slopes and steep piles of overburden and alluvium that contain fine particles, which are easily eroded during rainstorms and flood events. Tailing and overburden piles typically are located within the active floodplain and are susceptible to erosion by the stream during break-up and high flow events. Without proper erosion control, the fine particles will enter the local drain- age system. Unlike point-source pollution, the nonpoint-source pollution from exposed and unstabilized slopes may continue long after the operation of the site has ceased. This paper presents a summary of information obtained during a study to develop best manage- ment practices to minimize nonpoint-source pollution from placer mining. The best manage- ment practices are documented in a reference manual (Rundquist et al. 1986a) and a support- ing technical report (Rundquist et al. 1986b). The development of the best management prac- tices was based on a review of literature in the fields of mining engineering, drainage and erosion control, hydrology and hydraulics, aquatic and terrestrial biology, and aquatic- and terrestrial-habitat rehabilitation. The lSe . nsor Consultant and Staff Engineer, respectively, at Entrix, Inc., Business Park Blvd., Suite 6, Anchorage 99503 407 literature review focused on references with application to the control of nonpoint-source pollution in stream valleys in Alaska or other northern environments. The study was designed to provide guidance to resource-management agencies reviewing placer mining permit applications, with parti- cular emphasis placed on the control of non- point-source pollution and site rehabilitation. The study results also will assist placer miners in the development of mining plans that include design elements for nonpoint-source pollution control and site rehabilitation. If appropriate erosion-control procedures for surface-water drainage, grading, and stockpiling are outlined during the planning phase and implemented during site operation, nonpoint-source pollution can be substantially minimized. Appropriate measures most often include: development of drainage systems; proper diversion of stream channels; construc- tion, maintenance, and rehabilitation of an adequate settling-pond system; grading of tailings piles; and siting and protection of stockpiles. Site-specific conditions at a placer-mining site dictate which techniques are most effective at each site. The following sections of this paper present general techni- ques for drainage control, site grading, and stockpile placement for placer-mining opera- tions. DRAINAGE CONTROL A comprehensive erosion-control plan to limit nonpoint-source pollution should focus on drainage control. Drainage control has two objectives: to divert water around a mine site in order to minimize the amount of water that contacts erodible material within the site and to collect all turbid water within the site to allow for treatment prior to discharge to the stream. Prior to the initiation of mining activity, a drainage plan should be developed which delineates the surface-flow pattern around and through the site. This plan is useful in evaluating the need for, and location of, drainage-control structures. It can also be used when designing a plan to reestablish the drainage pattern during site rehabilitation. Channel Diversion Instream disturbances at a mining site should be avoided. If mining in a stream channel is necessary, the surface flow should be diverted and isolated from the mining activity prior to the initiation of mmmg within the existing stream channel. The 408 diversion channel should be sited, designed, constructed, and opera ted to a void excessive erosion or deposition, to contain floods within the range of acceptable site risks, and to meet fish-passage requirements if fish are present. An overflow channel near the diversion channel may be constructed to divert potential flood flows around the settling ponds and any other structures, such as the camp, to be protected. The overflow channel can function as an access road within the site except during flood flows in excess of the diversion channel design flood. The location for the diversion channel will depend on the physical characteristics of the site, the equipment available to construct the channel, and the proposed sequence of site operation. Preferably, the stream should be placed permanently into a newly constructed or rehabilitated stream channel. At sites where mining cuts are narrower than the entire flood- plain, a permanent stream channel could be constructed on the mined portion of the flood- plain prior to mmmg the other portion. Whenever possible, the stream should be divert- ed only once to minimize impacts to fish and wildlife. Design criteria for a diversion channel are presented in Rundquist et al. (1986a and 1986b) and Simons, Li, and Associ· ates (1982). Overland Flow Diversion Surface runoff from the valley walls should be intercepted and diverted around the site. Gullies or other topographical features which may concentrate runoff can provide significant amounts of flow during or after rainfall events. Overland flow can also result from melting permafrost, groundwater seepage, and snowmelt. A drainage system can divert most of the surface runoff around the site and retain the remaining sediment-laden water within the site. Upslope berms may be used to divert water around the site (Herricks et al. 1974, Becker and Mills 1972). In areas of permafrost ditches draining the intercepted flow should follow the contours closely to avoid degrada- tion. Overland flow may also be diverted by stockpiles of strippings and overburden pushed uphill onto the valley walls and out of the active mining areas (Figure 1). Overland flow interception structures formed in this manner should be protected from erosion. Drainage structures in areas of permafrost will likely transport significantly larger amounts of water than those in non-permafrost areas, and should be sized accordingly. Areas of permafrost exhibit rapid responses to prectpttation. In a study of similar sized streams with and without permafrost, Slaughter DITCH TO STitEAM Figure I. Using stockpiles to divert overland flow. et al. (I 983) found that the presence of permafrost resulted in increased amounts of overland flow. Due to the large amounts of flow and the potential for drainage structure failure caused by the degradation of perma- frost, additional maintenance of drainage structures may be required in areas of perma- frost; drainage structures should be inspected regularly. Groundwater Diversion Groundwater flow should also be diverted around the mining site. Combinations of both surface and subsurface drainage control will likely provide better and more economical results for erosion control (USSCS 1979). Bedrock drains located upstream of the excava- tion area may be used to divert groundwater before it seeps into the excavation area. While the bedrock drain should generally be only deep enough to prevent down valley seepage from flowing into the active mining area, the drain should be designed to have sufficient capacity for aufeis growth. The bedrock drain may be backfilled with cobbles and large gravels to act as a french drain (Sowers 1979). The water collected in an upstream bedrock drain can be routed around the mine site in a ditch which flows into the stream downvalley (Figure 2). A series of bedrock drains may be dug throughout the mmmg site to collect additional groundwater infiltration from the valley walls (Peele 1941 ). However, ground- water or surface water which contacts unstabi- lized slopes or enters the excavation area must be collected and treated with other turbid water within the site. Site Drainage Water in contact with surfaces disturbed by mining will likely become sediment laden and 409 Figure 2. Bedrock drain routing groundwater flow around mine site. turbid. Therefore, water that is not diverted around the site will require treatment prior to discharge. Overland flow resulting from precipitation or groundwater upwelling within the site should be routed into settling ponds. Settling-pond systems should be designed to retain all sediments originating on the site. Efforts to prevent water from becoming sediment laden appear to be more efficient than efforts to treat sediment-laden water (Vesilind and Peirce 1982). Concurrent with mining activity, or at least at the end of each mining season, the overland flow drainage pattern should be reestablished and stabilized. Surface-water control is the primary concern in the design of the drainage pattern. Any water that is detained or im- pounded within the site will reduce the rate and amount of surface runoff (Beasley 1972) and thus reduce erosion. Storage of water will occur when small depressions or terraces along the contours are constructed (Frank Moolin and Associates 1985). Site grading should reesta- blish a drainage pattern to limit sediment contributions to the system (Figure 3). __d--··~~ c1r:=J SECTION A·A SECTION 8-1 Figure 3. Grading should create a drainage pattern with small depressions. Settling Ponds Although settling ponds are essential to the control of point-source pollution, they are also an important aspect of nonpoint-source pollution control. Settling ponds are used to clarify water which has been used in mining activities such as sluicing or hydraulicking which produce point-source pollution. Settling ponds should also be used to clarify sediment- laden surface water and seepage from excavated or stripped areas which would otherwise contri- bute to nonpoint-source pollution. Surface water and seepage from the mining area must be controlled to limit nonpoint-source pollution; all sediment-laden water produced on the site should be routed through the settling pond system. However, the flow of unpolluted water into a settling pond should be minimized to increase the settling pond efficiency. SITE GRADING Grading tailing piles and steep, excavated slopes within the mmmg site will reduce erosion and encourage revegetation (Becker and Mills 1972). A grading plan should describe the sloping and contouring necessary to reesta- blish desired elevations within the site and to provide for an appropriate drainage pattern. The mining site should be graded to cause surface water from erodible areas to flow through the settling-pond system. Site grading should be done concurrent with mining activity, at the conclusion of each cut, or at least at the end of each mining season. Grading during mine operation will benefit seasonal erosion control; if tailings are properly placed and graded during mining, additional grading will not be necessary upon site closure. Slope Modifications Steep slopes and tailing piles should be graded as soon as possible. Surface erosion from slopes is greatest during the first year following disturbance (Cook and King 1983). Reducing the steepness of tailing piles will significantly reduce surface-water pollution (Vesilind and Peirce 1982). A slope of 3 hori- zontal to I vertical or less is recommended (Becker and Mills 1972, Dryden and Stein 1976). Large piles of stripped loess with thawing ice lenses should be graded to slopes of 3 to I when placed. Slopes that are steeper toward the top and shallower toward the bottom are preferred as they have been shown to have less sediment loss than ilat slopes or slopes that are steepest at the base (Haan and Barfield 1978, Meyer and Romkens 1976). 410 Grading steep slopes will also reduce potential slope failures which can release large amounts of sediments into nearby streams. Slope failures are generally avoided with 3 to 1 slopes (Schwab 1982). In addition, earth- moving equipment such as front end loaders and bulldozers are able to operate on slopes as steep as 3 to 1 (Johnson 1966). Permafrost material that will be removed after thawing should be graded to a shallow grade to mtntmtze erosion during thawing. Adjacent permafrost materials that will not be removed should not be graded; they should be left as a cut face for the duration of adjacent mmmg operations (Figure 4). This will minimize the area of permafrost degradation adjacent to the mine site by reducing the surface area of permafrost exposed to warm air, water, or solar radiation and minimizing disturbance to the insulating vegetation on the top of the cut bank. The length of time that the cut face is exposed should be minimized, especially in ice-rich permafrost terrain. Tailings should be pushed up against the cut face following the conclusion of mining at that location to insulate the cut face (Figure 4). Figure 4. Site grading during and after mmtng adjacent to permafrost terrain. Terracing and furrowing parallel to the contours should be incorporated in site grading, particularly outside the active floodplain, to reduce the effective length of the slope and thus decrease further erosion (Figure 5). Backsloping terraces encourage deposition of eroded soils (Beasley 1972). Small furrows (trenches) approximately 0.5 ft in depth and spaced closely together along the contours of the slope can effectively reduce erosion (Doyle 1976, Becker and Mills 1972, Cuskelly 1969). SIDE VIEW Figure 5. Terraces and furrows on side slopes. Contoured tailing piles and steep slopes should be roughened or scarified to reduce erosion by increasing water infiltration, minimizing slope failures, and encouraging natural revegetation (Rutherford and Meyer 1981, Troeh et al. 1980). Tailing piles should be reduced in height to minimize abrupt varia- tions in surface elevations. Timing of Grading Site grading during site operation most likely will be more cost effective and effi- cient than site grading at the end of each mining season since the equipment will already be in the immediate vicinity; grading at the completion of each cut rather than at the end of the mining season will discourage repeated handling of material. Frequent moving and handling material will increase the time and costs involved in grading. In addition, a lesser amount of material will be handled at one time. Tailing piles and overburden may be mixed and graded together unless required for use in site reba bili ta tion. Progressive site grading can realize the benefits of improved erosion control and vegetation reestablishment (Johnson 1966). STOCKPILE PLACEMENT Proper placement and protection of stock- piles will minimize erosion o'f the stockpiles, ensure maximum retention of materials for site rehabilitation, and mtntmtze distances that stockpiled materials must be moved during site rehabilitation. Stockpile placement and protection, if properly planned and executed, will help control erosion. General Recommendations Stockpile locations should be identified after consideration of drainage patterns and the eventual use of the material. Stockpiles should be located in areas where they can be easily protected and where rehandling will be minimized. Stockpiles should preferably be located where the results of the site assess- ment work indicate unproductive material. The amount and type of material required during site rehabilitation should be estimated to determine the quantities of each material to be stockpiled. Nonpoint-source pollution from stockpiles can be minimized through proper siting and containment procedures. The principal objec- tive is to select sites where the stockpiled material is protected from surface flow, including active stream channels and runoff 411 from rain or snow melt. Stockpiles should also be located and protected according to the acceptable risks at the site. Stockpiles should generally be located in elevated areas away from flood waters such as on the valley walls or in flat areas where flood waters often have shallow depths and low velocities. Stockpiles should not be located near an active channel or within the active floodplain where high flows could erode the sediments. Areas of concentrated surface runoff should also be avoided. Armoring the lower slopes of the stockpiles with cobbles, large rocks, or logs available on site is suggested if protection is needed (Joyce et al. 1980). An armoring layer of cobbles, large rocks, or logs may inhibit the erosion of fine sediments. Organic mater- ial and settling pond fines may have a high moisture content and may require containment berms around the perimeter of these stockpiles. Location of stockpiles should also minimize handling. Properly choosing washplant, screen- ing plant, grizzly, or stacker locations and the directions in which tailings are pushed can result in stockpiling in the proper location without expending additional equipment time for material redistribution. Recommendations for Specific Materials Materials stockpiled for use in site rehabi- litation include six types of material that can be removed from a work area. These are: I) trees (greater than 6 inches in diameter at breast height); 2) small trees, shrubs, and grasses with the top layer of organic soil; 3) shallow soils, often containing a mixture of fine organic and inorganic material; 4) large quantities of fine inorganic material, such as deep loess deposits; 5) oversize inorganic material such as large boulders; and 6) fines removed from settling ponds. If the top layer of organic soil is very thin, the upper six inches of the surface material should be removed and stockpiled with the shrubs and grasses. Depending on the area where the mine is located, one or more of these categories may not be present. Large trees require little protection since surface flow past a stockpile of trees will not substantially contribute to nonpoint-source pollution. Stockpiles of trees should be placed in an area where they will not interfere with the mining operation and be available to enhance stream and floodplain habitats during site rehabilitation after the site is closed. Large trees should be stockpiled separately from other types of material and not mixed with other organic material during temporary stor- age. In some cases, the trees can be used to construct log crib containment berms for smaller organic material. The most common forms of vegetation that are removed during a placer mining operation are the small trees and shrubs that occupy previ- ously undisturbed floodplain riparian zones. Proper stockpiling and protection of this material is important since it will serve as the stock for natural revegetation, and since the material is easily eroded. When feasible, the amount of time of exposure of stripped slopes should be minimized to reduce erosion. The small trees, shrubs, and grasses should be piled with the top layer of organic soil in a protected location. Placing these two types of materials together accomplishes several objec- tives. First, the woody material will provide some protection for the organic soil from wind and surface water erosion. Second, many species of shrubs and grasses will continue to grow within the stockpile. This will improve the survival rate of vegetation placed during site closure. Third, when the shrub layer and surface soil are removed together, disturbance of the root zones will be reduced, thus enhanc- ing survival of the vegetation. This material should be kept moist to maintain the viability of the woody slash. The shallow soils located beneath the top layer of organic soil and often consisting of fine organic material, sands, and perhaps some gravel and cobble should be piled separately from the trees and the shrub-organic soil piles. This material should be protected since it can be eroded easily. If necessary, large rocks can be used for protection along the side slopes of the piles. Containment berms may also be constructed around the stockpiles. At some locations deep deposits of loess, or fine inorganic soil, must be removed to expose placer-bearing gravels. In these situations the large volume of material that must be handled dictates that it be moved a short distance and only handled once. Typically, the most efficient handling of this material is to push it up the valley side slopes into fan- shaped, flat-topped mounds (Figure 1). At sites where large volumes of this material must be handled, final stockpile locations should provide maximum protection from wind and sur- face-water runoff which can easily erode this material. If organic rehabilitation material is limited at the site, portions of this loess material may be used both as a leveling layer and as a seedbed to enhance vegetative recovery. In situations where highly erodible stock- piled material cannot be protected using berms 412 and riprap as described, temporary seeding and fertilizing using annual grasses may be useful to help stabilize this material. SUMMARY If appropriate erosion-control procedures for surface-water drainage, grading, and stockpiling are outlined during the planning phase and implemented during site operation, nonpoint-source sediment pollution can be substantially minimized at a placer-mining site. Appropriate measures include the following: 1. Divert active channels around the site in a diversion channel if the active channel bed is to be mined; a one-time diversion into a final rehabilitated stream channel is recom- mended to minimize impacts to fish and wildlife. 2. Intercept and divert surface runoff around the site to prevent contact with erodible soils. 3. Intercept and divert groundwater around the mining activity using bedrock drains. 4. Route sediment-laden water within the mining site into settling ponds using proper grading. 5. Design settling ponds to accommodate all point-and nonpoint-sources of sediment- laden waters within the site in order to assist in the control of nonpoint-source pollution. 6. Grade the site concurrently with site operation. Steep slopes and tailing piles should be graded to reduce their height, slope steepness, or slope length to leave an undulating terrain. Roughening the surface reduces erosion potential and encourages natural revegetation. 7. Locate or protect stockpiles of materials to be used for rehabilitation to mmtmtze erosion potential while minimizing distances that materials must be moved during site rehabilitation. ACKNOWLEDGMENTS This paper was based on the Technical Report and Reference Manual produced during the Best Management Practices for Placer Mining Study funded by the Alaska Department of Fish and Game, Habitat Division in Fairbanks, Alaska. The authors wish to thank their colleagues in this study for their contribution. We also want to thank the reviewers of this paper, the Technical Report, and the Reference Manual. REFERENCES Beasley, R.P. 1972. Pollution Control. ersity Press. Iowa. Erosion and Sediment The Iowa State Univ- Becker, B.C. and T.R. Mills. 1972. Guidelines for Erosion and Sediment Control Planning and Implementation. Prepared for Office of Research and Monitoring U.S. Environmental Protection Agency. Washington, D.C. EP A-R2- 72-0 15. Cook, MJ. and J.G. King. 1983. Construction Cost and Erosion Control Effectiveness of Filter Windrows on Fill Slopes. Inter- mountain Forest and Range Experiment Station Research. Ogden, Utah. Note: INT -335. Cuskelly, S.L. 1969. Erosion-control Problems and Practices on National Forest Lands. Trans. Am. Soc. Agric. Eng. 12( I): 69-70, 85. Doyle, W.S. 1976. Strip Mining of Coal: Environmental Solutions. Noyes Data Corporation. Park Ridge, New Jersey. Dryden, R.L. and J.N. Stein. 1975. Guidelines for the Protection of the Fish Resources of the Northwest Territories During Highway Construction and Operation. Frank Moolin and Associates, Inc. 1985. Best Management Practices Manual on Erosion and Sedimentation Control. Prepared for the Alaska Power Authority. Haan, C.T. and B.J. Barfield. 1978. Hydrology and Sedimentology of Surface Mined Lands. University of Kentucky. Lexington, Kentucky. Herricks, E.E., J. Cairns, and V.O. Shanholtz. 1974. Preplanning Mining Operations to Reduce the Environmental Impact of in Use Drainage on Streams. In: Water Resources Problems Related to Mining. R.F. Hadley and D.T. Snow (eds). American Water Resources Association. Minnesota. Johnson, C. 1966. Practical Operating Proce- dures for Progressive Rehabilitation of Sand and Gravel Sites. National Sand and Gravel Association. 413 Joyce, MR., L.A. Rundquist, and L.L. Moulton. 1980. Gravel Removal Guidelines Manual for Arctic and Subarctic Floodplains. Prepared for U.S. Fish and Wildlife Service by Woodward-Clyde Consultants. Meyer, L.D. and MJ. Romkens. 1976. Erosion and Sediment Control on Reshaped Land. Proc. Third Federal Inter-Agency Sedimenta- tion Conference, Water Resources Council. Washington, D.C. Peele, Robert. 1941. Mining Engineers' Hand- book. John Wiley and Sons, Inc. New York. Rundquist, L.A., N.E. Bradley, J.E. Baldrige, P.O. Hampton, T.R. Jennings, and MR. Joyce. 1985a. Best Management Practices for Placer Mining. Reference Manual. Prepared for the Alaska Department of Fish and Game. Rundquist, L.A., N.E. Bradley, J.E. Baldrige, P.O. Hampton, T.R. Jennings, and MR. Joyce. 1985b. Best Management Practices for Placer Mining. Technical Report. Prepared for the Alaska Department of Fish and Game. Rutherford, C. and K. Meyer. 1981. Revegeta- tion on Gold Dredge Tailings, Nyac, Alaska. Prepared for BLM, Anchorage, Alaska. Simons, Li and Associates, Inc. 1982. Surface Mining Water Diversion Design Manual. Prepared for U.S. Department of the Interior Office of Surface Mining. Slaughter, C.W., J.W. Hilgert, and E.H. Culp. 1983. Summer Streamflow and Sediment Yield from Discontinuous-Permafrost Headwaters Catchments. Proc. Fourth Annual Conference on Permafrost 1983. Fairbanks, Alaska. Sowers, G.F. 1979. Soil Mechanics and Founda- tions: Geotechnical Engineering. MacMillan Publishing Co., Inc. New York. Troeh, F.R., J.A. Hobbs, and R.L. Donahue. 1980. Soil and Water Conservation for Productivity and Environmental Protection. Prentice-Hall, Inc. New Jersey. U.S. Soil Conservation Service. 1979. Drainage of Agricultural Land. Water Information Center, Inc. Port Washington, NY. Vesilind, P.A. and J.J. Peirce. 1982. Environ- mental Engineering. Ann Arbor Science. Michigan. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 RIVERBANK EROSION PROCESSES ON THE YUKON RIVER AT GALENA, ALASKA William S. Ashton and Stephen R. Bredthauer * ABSTRACT: Periodic measurements of riverbank recession on the Yukon River at Galena, Alaska have been made since 1946. Intensive studies of channel shape and riverbank erosion were conducted in 1959, 1984 and 1985. Erosion rates varied from 0.3 m/yr (1.0 ft/yr) at banks with devel- oped vegetative protection (peat or bank debris) to 10.8 m/yr (35.5 ft/yr) at steep banks with active thermoerosional niching. Comparison of channel profile measurements from June 1984 and June 1985 indicate that the thalweg did not significantly change location or elevation during a 10-year recurrence interval flood. (KEY TERMS: thermoerosional niching; subarctic; Alaska; Yukon River; erosion.) INTRODUCTION Galena, Alaska is located on the downstream end of an 11 km (7 mi) long meander of the Yukon River (Figure 1). Erosion control structures were first bstalled in the early 1960's by the U.S. Amy, Corps of Engineers, when bank erosion threatened to undermine a flood control dike protecting the airfield from ice jam floods. The upstream key at the end of the airstrip is a series of sheet pile cells. Bank protection extends 793 m (2,600 ft) downstream of the sheet pile cells, and consists of a variety of designs, including the use of rock riprap and of gravel-filled 55 gallon oil drums. Both designs have a sheet pile wall at the toe. Since 1946, there have been sixteen sets of riverbank erosion measurements of the Yukon River at Galena. During 1984 and 1985, detailed measurements were made of the channel geometry, bed material and riverbank erosion rates at sites studied since 1946. The following is a summary of riverbank erosion processes observed during 1984 and 1985, and of data collect- ed since 1946. STUDY AREA The area climate is continental subarctic, with mean annual temperatures of -4.7 degrees C (23. 6 degrees F) and mean annual precipitation of 329 mm (13.0 in). Snowfall averages 1500 mm (59.1 in) per year. Galena lies in a wide flood plain composed of fine to coarse grained deposits on a 11 km (7 mi) long bend of the Yukon River. The study area has four physiographic phases, each with distinct permafrost, drainage, vegetation, and engineering characteristics. Each of these phases is temporary (relative to geologic time) and somewhat evolutionary in nature due to permafrost, flooding, river meandering, and associated vegetative changes. As the phases become older, ice-rich permafrost becomes more prevalent (Weber and Pewe, 1970). The river is constrained by mountains for approximately 330 km (200 mi) upstream of Galena, flowing in a confined channel until reaching the Yukon-Koyukuk Lowlands, where the river meanders for about 67 km (40 mi) until again becoming confined by the mountains. The drainage area of the Yukon River at Galena is approximately 676,000 sq km (261,000 sq mi). Twenty-three years of streamflow data *Respectively, Hydrologist and Senior Civil Engineer, R&M Consultants, Inc., 5024 Cordova Street, Anchorage, Alaska 99503. 415 FIGURE 1. BOTTOH TOPOGRAPHY (IN FEET) AND RANGE LINES, GALENA, ALASKA, OCTOBER 1959. RANGES 0 THROUGH G WERE RESURVEYED JUNE 1984, AND RANGES 4 THROUGH 10 WERE RESURVEYED JUNE 1985. KALA SLOUGH ENTERS THE YUKON FROM THE LOWER LEFT. are available at Ruby, approximately 80 km (50 mi) upstream of Galena and with a drainage area of approximately 671,000 sq km (259 ,000 sq mi). Mean annual flow is 4,730 ems (166,900 cfs), with average daily velues ranging from a winter minimum of 480 ems (17,000 cfs) to a peak of 27,500 ems (970 ,000 cfs). As is typical of most subarctic rivers, the Yukon River has low flows in the winter, with flows sharply increasing at breakup in May. Peak flows normally occur in early June, steadily decreasing throughout the summer except for occasional late summer peaks. The bankfull stage is approximately 38 m (125 ft) at Galena. However, at water levels of 34 m (112 ft), water begins to flow into old channels and lake beds on the north bank. Ice jams occur frequently in the area, with the maximum recorded stage of 40.5 m (133 ft) occur- ring during an ice jam in 1971. Galena is in the zone of discontinous permafrost. The mean annual freezing index is approximately 3000 degree-days C. Mean monthly air temperatures range from -24.2 degrees C (-1.5 degrees F) in January to 15.3 degrees C (59.6 degrees F) in July. Water temperatures rise rapidly shortly after breakup, reaching 10 degrees C (50 degrees F) by early June and up to 17-20 degrees C (63-68 degrees F) in late July or early August. The temperatures 416 then steadily decline, dropping to freez- ing in mid to late October. Freezeup occurs in November. Annual maxi.mum observed ice thicknesses range from 830-1,400 mm (32-55 in). Breakup dates range from May 1 to May 28. BANK AND BED MATERIAL Data on bank and bed materials are available from the literature and from a drilling and bed material sampling program conducted in 1984 and 1985 (Weber and Pewe, 1970; R&M Consultants, 1985; Ashton, 1985). Generalized soils profiles near the new townsite (range 7) and upstream of the townsite (Range B) are shown in Figures 2 and 3, respectively. The upper bank material is primarily composed of silt with some sand. The material becomes coarser in the lower layers, grading into sand, gravelly sand, and sandy gravel. Surface bed materials collected from the river with a 152 mm (6-inch) pipe dredge reflect two distinct types of surface bed material from the north bank to the south bank. Material gathered near the north bank consists of a well-graded mixture of gravel from 25.4 to 50.8 mm (1-2 inch) gravel down to fine sand, with median diameters ranging from 0.7 to 10.5 mm. Material found farther I:SO- 120- 110- 100- 90- 80- 70- 00- 00- I 8 + . ' I l ~;· 8 nr,..· + ~ I I I I ~r 8 8 8 8 8 + + + !: :; ' ~ ' 0 .:·~:~. OI.TIIMC. ~11011 .UIIY.Y IIIOMUII.IIT (F•In ltL.tY·I8. FIGURE 2. GENERALIZED SOIL PROFILE AT RANGE 7, LOOKING UPSTREAH, NEAR THE Nm'l GALENA TO~'lNSITE (R&I·1 CONSULTANTS, 1985). 00- ,_ FIELD liTE DUCIIII•TtOM lO.Of'IIIYlltU.U (NI)TAIOtllttOLI!) IIIYfiiiC( ----------c---------------- ~~~~~~: ICI .AVI:L IIITIIIlloi'IIIIO WITM .ILWII:Ll'f IAMO. IAtiD 8 ! 8 ! FIGURE 3. GENERALIZED SOIL PROFILE AT RANGE B, LOOKING UPSTREAI-1 (R&H CONSULTANTS, 1985). 417 out in the channel consists of fairly uniform medium to fine sand. Sub-bed material collected in the main channel is primarily a nixture of fine gravel grading to fine sand. Permafrost was encountered to ele- vation 30.9 m (101.2 ft.) along the bank at Range 7, but was not encountered below the main channel down to an elevation of 5.7 m (18.8 ft). This is typical of pemafrost patterns on large, migrating northern rivers, and has been reported for the tiackenzie River by Smith and Hwang (1973). The large heat mass of the water in the summer maintains a large thaw bulb beneath the river channel. Significant degradation of the permafrost on the north bank is not evident because the material is eroded away before the bank has a chance to thaw. However, permafrost was found at Range B at elevation 21.9 m (72 ft.), only 2.4 m (8 ft.) below the streambed. With the rapid bank erosion in this area, the bed would have been exposed for 4-6 years, and therefore it would be anti'Cipated that the depth of thaw would have been greater. EROSIONAL PROCESSES General The erosion processes observed on the Yukon River near Galena are similar to those described for other large northern rivers, as summarized in Lawson (1983). Lawson describes the erosion of perenially frozen streambanks as being comprised of three zones: bluff, transitional, and bank. The bluff zone is above high water, the bank zone is below low water and the transitional zone is the area between the two. There is no frequency of occurrence associated with these water levels. Therefore, for a specific lo- cation each zone is identified by field observations. Observations During 1984 and 1985 For the Yukon River at Galena the bank zone is below elevation 97 ft (29.6 m) and the bluff zone is above elevation 120 ft (36. 6 m). The transition between the bank zone and the transitional zone corresponds with the change in bank mate- 418 rial from sand and gravelly sand to silt and sandy silt. The following discussion concentrates on the erosion processes in the transitional zone. During the breakup flood there is little or no bank erosion. Between Range 0 and Range 5 (Figure 1), an ice shelf along the north bank prevents moving ice from hitting the bank (except along the sheet pile cells at Range 4). Upstream of Range 5, ice is hitting the bank, but since the bank is frozen and water temper- atures are still near 0 degrees C (32 degrees F), little erosion or thermoerosional niching occurs. Water temperatures increase rapidly once ice is out of the river, reaching 10 degrees C (SO degrees F) less than 2 weeks after the 1984 breakup. Water levels are generally highest in June (except for ire jams), and the flow is against the frozen upper silt layer of the bank. The warm water rapidly thaws the frozen silt, which is carried away by the current and waves, forming therrnoerosional niches of up to 5-7 m (15-20 ft) in depth. The niching continues until the soil can no longer support the weight of the frozen cantilevered block, which then fails. Once the block fails, it thaws and the material is washed downstream. Depending on river stage and the bank material, the soil blocks provide the freshly exposed bank face limited pro- tection from niching due to waves and currents, and no protection against thermal niching. During June 1984 niching progressed 0.4 m (1.3 ft) into a freshly exposed frozen silt bank in 24 hours due to thermal action and small waves (less than 5 em high). The rate of erosion varies throughout the meander bend, from little or no erosion at Ranges 4 and G to maximum rates at Range B through D (Table 1). The major portion of the thermoerosional niching occurs during the high water after breakup and into June and early July. The rate of niching decreases as water levels and nearshore velocities decrease, and as the larger bank material is exposed. During 1985 the water level stayed higher for a longer period than in 1984, with conse- quently more erosion (Patrick, 1985). During June 1985, cross-section measurements and a thalweg profile t,rere made at a discharge of 21,300 ems (754,000 TABLE 1. EROSION RATES NEAR GALENA, ALASKA Average 1984-1985 Erosion 1984 Erosion m(ft)(1) Erosion m(ft)(1) Rate June July October 1984 1963-1983(2) Range to July to October Total to June 1985 m/yr(ft/yr) 4 0.0 o.o o.o 0.0 0.0 5 0.15(0.5) 0.5(1.5) 0.6(2.0) 1.2(4.0) 2.0(6.5) 6 1.4(4.5) 0.9(3.0) 2.3(7.5) 4.6(15.0) 3.7(12.0) 7 0.3(1.0) 2.7(9.0) 3. 0 (10. 0) 2.6(8.5) 4.1(13.5) 8 3.4(11.0) 0.9(3.0) 4.3(14.0) 7.6(25.0) 4.0(13.0) 9 0.6(2.0) 1.2(4.0) 1.8(6.0) (3) 4.4(14.5) 10 1. 7(5.5) 1.5(5.0) 3.2(10.5) 7.9(26.0) 4.7(15.5) llA 0.15(0.5) 0.15(0.5) 0.3(1.0) (4) 4.1(13.5) A 0.6(2.0) 1.8(6.0) 2.4(8.0) (4) 4.4(14.5) B 1.8(6.0) 5.9(19.5) 7.8(25.5) (3) 4.4(14.5) c 4.9(16.0) 0.8(2.5) 5.6(18.5) 10.8(35.5) 6.1(20.0) D 0.9(3.0) 5.9(19.5) 6.9(22.5) 3.5(11.5) 4.6(15.0) E 0.3(1.0) 0.0 0.3(1.0) 3.5(11.5) F 0.0 0.0 0.0 0.0 2.3(7.5) G 0.0 0.0 0.0 o.o 0.8(2.5) (1) Measured from survey monument on north bank. (2) Determined by comparison of aerial photographs from 1963 and 1983, R & M Consultants, 1985. (3) Nonument destroyed (4) Monument underwater From: Ashton, 1985. cfs) at Range 7, with a total flow of approximately 22,600 ems (800,000 cfs) including flow through Kala Slough. This flow is approximately the 8-10 year recurrence interval flood and reached an elevation of 120 ft (36.6 m). The high water levels provided an opportunity to compare the bed elevations and bluff erosion between Ranges 4 and 10 with that whi.ch occurred during 1984, when the maximum flow was 11,400 ems (403 ,000 cfs) and the maximum water surface elevation was 113 ft (34.4 m). At Ranges 5 through 10 the bluff and transitional zone (ele- vation 27.4 m (90 ft.) to top of bank) eroded 1. 2 to 7. 9 m (4 to 26 ft), while the channel shape below elevation 27.4 m (90 ft.) was approximately the same as in June 1984. While there was no general lowering of the bed evident at these sites during the June 1985 flood, it was appar- ent that the bluff zone was eroding at a more rapid rate than measured during 1984 (Figures 4 and 5 and Ta.ble 1). 419 Historic Erosion Rates At Range 7, the U.S. Army, Corps of Engineers has collected bank erosion rates since 1946 (Table 2). These measurements indicate an average annual erosion rate at Range 7 of 4.8 m/yr (15.6 ft/yr), ranging from 0.3 m/yr (1.0 ft/yr) in 1965 to 9.1 m/yr (30 ft/yr) in 1968. During our study we measured 3.0 m/yr (10.0 ft/yr) during 1984. Comparison of aerial photography obtained in 1963 and 1983 indicates the long term upper bank erosion ranging from 0.8-6.1 m/yr (2.5-?.0 ft/yr) (Table 1). During 1984 the upper bank erosion was observed to vary from 0.3-7.8 m (1-26 ft) for the same area. The areas with low erosion rates (Ranges llA and E) were located on an old slough channel filled with peat and at a bank protected by debris, respectively. Erosion in any individual year varies, dependi.ng on the water stage, wind ol=o N = .. ~ Prolllo Oct. 19&11 tnrcug~ ,o\ug. 1165 Ioken ~y 11>1 U.S. A1my c~rpo of £ngln&Gra, Ala<k a Dloulcl. Ju"'> 1184 and Jun3 191!5 Prollll laktn ~Y R•M ConoultanU, Inc. ~~ ----1~-- -1 i- DISTANCE F"OM 1959 BASE LINE (FEET) Pro!olo $1911984 P"'lllo A•; .. 85 Prol•lo Junol984 Profllo JUUI98!> FIGURE 4. BOTTOil PROFILES, RANGE 71! LOOKING UPSTREArt (R&H CONSULTANTS, 1985). u.S.AIIIfT ~-~1 --- 1 ! -+ -- 130 t-------- 1 - ---------t ----------------------------+----- --~---~-r--r--r,-rl ~ E GEND __ Prof.ile Oct. 1959 (COEl I . I ._._ Prof1le Jun. 1984 (R6M) , ---r----+-----+---------4 ! I I I J I ------------t-----------r----t----j--- r--r--_ --_ ++------'------11-----+---t--="''!reR LC CL 2 JVNC 19/!4-CLEV. 114.~ --~----~------1-----r---~----r---+----------- ~~-----4+-~--~~--~~---r--t ----------+------+ ----- I w•r.• , 1 v<L •ocr~,. .. -niv 1 •-• ~=1f-:::::;_.;::::=;~l ===--t ___ =::t~~~====t===t==,~-~:=-==-::;-~~-~-;:-;::.;::;- __ __j_ ___ - :.bl-00~~~:.~:0 ---,~------+~-------~•"'oo"-~~~--~----j.;!;;~-o:_L~-"600:--~~•+o\n-o~~~~--+'of.voo----~~,;f,. ;-oo~-~~"'vc'~~di,.j--oo• -~'-•sot6oo~~------~}-l-~--------,o+-oo·-_ -----2+z'iro-----.. ~ 3000 3400 3600 - 2600 2800 DISTANCE FROM 1959 BASE LINE (FEET) FIGURE 5. BOTTOI1 PROFILES, RANGE C, LOOKING UPSTREAl1 (R&H CONSULTANTS, 1985). TABLE 2. HISTORICAL EROSION RATES, RANGE 7 Period of Erosion Date Years Winter 1946 October 1959 12.8 September 1960 0.9 October 1961 1.1 October 1962 1.0 October 1963 1.0 October 1964 1.0 August 1965 0.8 July 1966 0.9 July 1967 1.0 February 1969 1.6 February 1974 5.0 June 1977 3.3 August 1978 1.2 September 1982 4.1 TOTALS 35.7 From: u.s. Army Corps of Engineers, anc1 amount of exposed bank silt. Rapid upper bank erosion is typically caused by a combination of high water and wind driven waves. Ott (1981) reports up to 13.7 m (45 ft) of erosion during a 3 day storm, and local residents report erosion of up to 36.6 m (120 ft) in one year (H. Strassburg, 1984). Comparison of long-term erosion rates with the depth of thermoerosional niches observed in 1984 indicates that on the average the upper bank goes through one cycle of thermoerosional niching and block slumping per year. The amount of slumping at any specific location is dependent on the localized bank stratigraphy. SUMMARY Observations of bank erosion process- es along the Yukon River near Galena, Alaska found erosion rates similar to those summarized in Lawson (1983). The rates varied depending on water stage, frequency of wind waves, bank stratigraphy, and location along the meander bend. During 1984, bluff re- cession varied from 0.3 m (1 ft) in areas with developed bank vegetative cover (peat Amount of Annual Rate Bank Eroded m(ft) m/yr(ft/yr) 70.1(230) 5.5(18) 4.3(14) 4.9(16) 6.4(21) 5.8(19) 2.1(7) 2.1(7) 4.9(16) 4.9(16) 5.8(19) 5.8(19) 0.3(1) 0.3(1) 5 .8(19) 6.4(21) 1.5(5) 1.5(5) 14.6(48) 9.1(30) 15.2(50) 3.0(10) 12.2(40) 3. 7(12) 9.1(30) 7.6(25) 18.3(60) 4.6(15) 171(560) 4.8(15.6) 1983. 422 soil or vegetative bank debris) up to 7.8 m (25.5 ft) in areas with steep banks a~ thermoerosional niching. Channel cross-sections taken in 1984 and 1985 indicated that the bed elevation remained at approximately the same ele- vation in the study area during a 10-year recurrence interval flood, likely due to the relatively large bed material found near the north bank. However, rapid bluff erosion occurred clue to the high water stages. ACKNOWLEDGMENTS Data used in this paper were collect- ed during an erosion control study and design for the Alaska Department of TransportAtion and Public Facilities. Special thanks to boat operators Dan Patrick and Harvey Strassburg for their local insights of the Yukon River. REFERENCES bhton, W.S. 1985. Galena field trip Jure 1-3, 1985. Memo to V. Gretzinger and S. Bredthauer. R Consultants, Inc. June 6, 1985. 14 pp. & }1 Lawson, D . E. frozen 83-29. 1983. Erosion of perenially streambanks. CRREL Report Hanover, New Hampshire, 22 pp. Ott Hater Engineers, 1981. Plan of study, Yukon River erosion control at Galena, Alaska. Prepared for U.S. Army, Corps of Engineers, Alaska District. Anchorage, Alaska, 25 pp. with appendices. Patrick, D. 1985. Personal communication. July 1985. R&M Consultants, Inc. 1985. Galena bank stabilization, Project No. K-83513. Final Report. Prepared for Alaska Department of Transportation and Public Facilities. 2 volumes. Anchorage, Alaska. Smith, N.W. and C.T. Hwang. 1973. Thermal disturbance due to channel shifting. In Proc. 2nd International Permafrost Conference. National Academy of Sciences. pp. 51-59. Strassburg, H. 1 984. Personal communica- tion. October 1984. U.S. Army, Corps of Engineers, 1983. Galena streambank protection, Galena, Alaska. Section 14 Reconnaissance Report. Anchorage, Alaska. Weber, F.R. and T.L. Pewe. 1970. Surficial and engineering geology of the central part of the Yukon- Koyukuk lowland, Alaska. Department of the Interior. U.S. Geological Survey, Miscellaneous Geological Investigations Map I-590. 423 SNOWMELT RUNOFF 425 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 MODELLING SNOWMELT INFILTRATION AND RUNOFF IN A PRAIRIE ENVIRONMENT D.M. Gray, R.J. Granger and P.G. Landine* ~STRACT: The development of an infiltra- tion model for frozen soils and its ap- plication for predicting streamflow from snowmelt in a Prairie environment is dis- cussed. The model assumes frozen soils can be grouped to three broad classes ac- cording to their infiltration potential: Unlimited -cracked or highly porous soils capable of infiltrating all the snow water; Limited -the infiltration potential of a soil depends primarily on the snowcover water equivalent and the ice content of the soil layer, 0-300 mm at the time of melt; and Restricted - a soil containing an impermeable layer that in- hibits infiltration. Field measurements of soil water changes during snowcover ablation are used to support the concept and a prediction equation for soils of Limited potential is described. Esti- mates of infiltration quantities and run- off volumes determined by the model are compared with corresponding estimates which were either "observed" or "calcula- ted" by a mass balance analyses of data collected on areas ranging in size from 35 rn 2 to 11.4 km 2 • The agreement in values suggest the model is capable of giving estimates of areal infiltration of acceptable accuracy for operational pur- poses. It is demonstrated that the per- formance of the NWSRFS and SSARR opera..,. tiona! models in simulating streamflow from snowmelt on a Prairie watershed is markedly improved with the use of the infiltration algorithm. (KEY TERMS: infiltration model, frozen soils, snowmelt streamflow simulation) INTRODUCTION Water generated by the melt of a shal- low snowcover is an important component of the hydrological cycle of the Prairie region. It can be viewed as having either detrimental or beneficial effects. On the one hand, because of such factors as the poor relief and drainage develop- ment of the region, the presence of frozen soils at the time of melt, and the rapid ablation of the snowcover, direct runoff from snowmelt can lead to serious flooding, erosion and drainage problems. Conversely, snowmelt runoff provides a valuable source of water for domestic, livestock, and irrigation pur- poses as well as a wildlife habitat. It is estimated that 80-85% of the average annual volume of "local" surface runoff over much of the southern semi-arid re- gion of western Canada is derived from snow. A number of models have been developed for synthesizing streamflow from snowmelt. For example: the U.S. National Weather Service River Forecast System, NWSRFS (U.S. Department of Commerce, 1972; Anderson, 1973; Peck, 1976); the Stream- flow Simulation and Reservoir Regulation Model, SSARR (U.S. Army Corps of Engi- neers, 1972); the Hydrological Engineer- ing Center HEC-1 Flood Hydrograph Package (U.S. Army Corps of Engineers, 1973); the U.S. Department of Agriculture Hydrograph Laboratory Model USDAHL-74 (Holtan et al., 1975); the HBV of the Swedish Meteoro- logical and Hydrological Institute (Berg- str3m, 1979) and the UBC Watershed and Flow Model (Quick and Pipes, 1973). Of the existing systems no single model has *Division of Hydrology, University of Saskatchewan, Saskatoon, Saskatchewan, Canada, S7N OWO. 427 been adopted for universal use. Ongoing tests are being conducted by the World Meteorological Organization and other bodies on the performance of a number of the "better-known" models in different physiographic and climatic regions of the world. Each model differs from another, either as it calculates hydrological components or simulates the various processes involv- ed. A main factor contributing to the differences is that many models were de- veloped for a specific set of physiograph- ical conditions, e.g. climate, topography, vegetation and soil type. Consequently, a model developed in a mountainous area will not usually give reliable streamflow sim- ulations for a Prairie watershed; nor should it be expected. Direct runoff from snowmelt can be re- garded as the last step in a series of events beginning with snowcover accumula- tion and including the processes of inter- ception, ablation, evaporation, infiltra- tion, and the storage of meltwater by the snowcover, soil, surface depressions and channels. It is not surprising then that the simulation of snowmelt runoff has been most successful where there is abundant, uniformly-distributed snow and pronounced topographical relief, i.e., where there is the least opportunity for other proces- ses to be major components. The opposite situation exists on the Prairies. The snowcover is shallow and relatively large estimation errors of snowcover depth and water equivalent frequently occur, there is poor relief and the losses to evapora- tion, infiltration and depressional stor- age may be comparable in magnitude with the total water content of the snowcover. Thus, at least for the Prairie region, the performance of a model in simulating streamflow rates and volumes from natural catchments is often directly related to the accuracy with which infiltration is evaluated. In most operational systems infiltration is estimated by empirical equations such as those reported by Horton (1940) and Holtan (1961), soil moisture accounting routines, or from relationships that index antecedent groundwater storage conditions and the soil moisture storage potential to the base flow recession char- acteristics of the streamflow hydrograph. Two main problems arise in applying these procedures to watersheds in northern and 428 west-central Canada, namely; (1) no attempt is made to distinguish differences in the infiltration process to frozen soils and (2) many streams are ephemeral, i.e. flow occurs only following a rainfall or snowmelt event and therefore the reces- sion properties of the hydrograph do not properly index the soil moisture storage potential of a basin at the time of runoff. Anderson and Neuman (1984) used a frost index to calculate the reduction in perco- lation rates of frozen soils, and showed that this routine improved the performance of the NWSRFS Sacramento SMA model in sim- ulating runoff on a basin with frozen soil. They suggest, however, that further improvements in modelling runoff from areas with frozen ground will probably re- quire a more physically-based approach. The following material presents a model for describing infiltration to frozen soils based on such an approach and focus- ses on the effects of the model in improv- ing the simulation of streamflow rates and volumes from snowmelt in a Prairie environ- ment. Many of the results presented have been reported by Granger and Gray (1984) and Gray et al. (1984, 1985 b, c) and the reader is referred to these publications for more complete details. DEVELOPMENT OF A SIMPLE PHYSICALLY-BASED MODEL Granger et al. (1984) and Gray et al. (1985b) outlined the development and test- ing of a simple physically-based model describing snowmelt infiltration to frozen soils. They based their model on approxi- mately fifteen years of study by the Division of Hydrology, University of Sask- atchewan, of the snow hydrology of the Prairie region, including a comprehensive field investigation of infiltration to frozen soils in the Dark Brown and Brown soil zones of the region, and on the results of infiltration studies under simi- lar climatic regions of the USSR reported in the literature (Motovilov, 1978, 1979; Popov, 1973). The authors postulated that frozen soils may be grouped into three broad categories with regard to their in- filtration potential, namely: Restricted, Limited and Unlimited (Fig. 1). Restricted -infiltration is impeded by an impermeable layer, such as an ice lense at the soil surface or within the soil near the surface. For practical purposes the amount of meltwater that infiltrates can be assumed negligible and most of the snowcover water equi- valent goes to direct runoff and evap- oration. Limited -infiltration is governed prim- arily by the snowcover water equivalent and the frozen water content of the shallow layer of soil, 0-300 mm. Unlimited - a soil containing a high per- centage of large, air-filled, non-capil- lary pores or macropores at the time of melt and most or all of the snow water infiltrates. Examples of those soils are dry, heavily-cracked clays and coarse, dry sands. Granger et al. (1984) found for medium to fine-textured, uncracked frozen soils in which entry of meltwater is not impeded by ice layers (i.e. the Limited case) that: (a) the average depth water penetrated the soil during the melt period was 260 mm (standard deviation = 100 mm), (b) infil- tration was relatively independent of soil texture and land use and (c) the amount of snowmelt infiltration was inversely rela- ted to the average moisture content of the soil layer, Q-300 mm, at the time of melt (8p). With respect to the last point, it was observed that infiltration does not result in saturation of soil depth pene- trated by the infiltrating meltwater. The maximum degree of saturation of the layer can be approximated by the expression (. 6 + .48p), which reflects the trend for drier soils to reach moisture levels much less than saturation. The existence of an mverse relationship bewteen infiltration and the ice content of a soil has been demonstrated or postulated by several re- searchers (Will is et al. , 1961; Kuznik and Bezmenov, 1963; Gillies, 1968; Shipak, 1969; Romanov et al. , 197 4, Motovilov, 1979; Granger and Dyck, 1980; Kane, 1980 and Kane and Stein, 1983). Based on their findings, Granger et al. (1984) derived a set of equations defining the relation- ship between snowmelt infiltration (INF) snowcover water equivalent (SWE) and the premelt moisture content of the 0-300 mm soil layer (8p). Fo£ practical purposes and cases where SWE > INF, these results can be approximated by the equation: 429 INF = (1-8 )SWE 0 • 584 p (1) in which INF and SWE are in mm and ep is the degree of pore saturation cm 3 /cm 3 • The equation is based on measurements of soil water changes from 130 sites made over a 7-y period. It has a correlation coefficient of .85 and gives a standard error of the estimate between calculated and measured values of 5.5 mm. Equation 1 can be solved for INF when SWE and 8p are known. Snowcover data are generally available or obtainable. The premelt moisture content, however, is not routinely observed and thus must be estimated from other measured variables. Gray et al. (1985a) showed that changes may occur in the moisture regime of a Prairie soil over winter because of mois- ture losses from the soil surface layer, infiltration of water from mid-winter snow- melt or rain events, and the migration of water in response to soil freezing. They showed, however, that in the absence of mid-winter infiltration events the soil moisture of the 0-300 mm layer in the fall (8f) could be used to index the moisture content (ice) at the time of melt (8p)· The "best-fit" regression equations des- cribing the relationships between the two variables were, for fallow lands: eP = -5.05 + l.o5ef, and for stubble lands: eP = o.294 + o.957ef, (2a) (2b) in which ep and ef are expressed as a per- cent moisture by volume. The values of the correlation coefficient and the stan- dard deviation from regression of Eqs. 2a and 2b are 0.85 and 0.91, and 3.29 and 3.38% by volume respectively. Figure 2 shows field measurements of in- filtration plotted against snowcover water equivalent for the different categories of infiltration potential used by the model. In Figs. 2a, 2b and 2c the "x"-axis de- scribes the relationship for the restrict- ed case (INF = 0) and the 45° line, the unlimited case (INF = SWE). The infiltra- tion data given in Fig. 2a were obtained from measurements of soil moisture changes made at sites where there was visual evi- dence that an ice lense or impeding layer n= ~ Ca) RESTRICTED: inFiltration is low. runoFF pot~antial is high. tJ D;;ocr~aosing Sci 1 Moistur;;o Cont;;ont Cb) LIMITED: inFiltration gov;;orn~ad by ic;;o content oF sci 1 and snow wotar. (C) UNLIMITED• soil con inriltrot;;o oll Figure 1. Conceptual model of the infil- tration potential of frozen Prairie soils: (a) Restricted, (b) Limited and (c) Un- limited (after Gray et al., 1984). 430 80 " E E v 60 z 0 .. ~ .... t-40 l' < # 0:: t-<.; ...J .... 20 u. z .... 30 60 90 120 150 180 SNOW WATER EQUIVALENT <mm) Co) RESTRICTED 80 INF .. 5 < 1-Sp) SWE" 584 " E E v 60 z 0 .... t-40 < 0:: t- ...J .... 20 u. z .... 0 0 30 60 90 120 150 180 SNOW WATER EQUIVALENT <mm) (b) LIMITED • • •• 160 • " E E • v120 • z t • • • 0 •• • • .... t-80 < 0:: • t- ...J • ,.~ .... 40 u. _,t? z ~~ .... '\: 0 0 40 80 120 160 200 240 SNOW WATER EQUIVALENT Cmm) Cc) UNLIMITED Figure 2. Infiltration measurements for the three classes of frozen Prairie soils: (a) Restricted, (b) Limited and (c) Unlimited. (Note: 8p is the relative de- gree of saturation expressed in m3 /m 3 .) was present and the closeness of the points to the "x"-axis support the assump- tion of very little infiltration for this soil condition. Conversely, the data pre- sented in Fig. 2c, the Unlimited case, were obtained from measurements made in severely-cracked, heavy, lacustrine clays and represent moisture increases in the crack and adjacent soil. The fact that in- filtration is greater than the snowcover water equivalent at many sites is possible because of runoff from outside the area of moisture measurement entering the crack directly and lateral flow along a crack. It is believed the data confirm the as- sumption that INF = SWE for the soil con- dition. Figure 2b shows the family of curves described by Eq. 1, the Limited case. It is interesting to note in the figure that in a frozen soil at a given moisture content the incremental inc.rease in snowmelt infiltration per unit increase in snowcover water equivalent decreases with an increase in SWE; as well there is a decrease in the ratio of INF to SWE. TESTING OF MODEL The ability of the model to simulate areal snowmelt infiltration to landscape units of uniform landuse was tested by comparing the infiltration quantities from Eq. 1 with those calculated by a mass balance approach applied to data collected from eleven small runoff plots ranging in size from 35 to 4000 m2 • It was found that the infiltration quantities obtained by the two methods were associated with a oorrelation coefficient of .73 and a stan- dard error of the estimate of 7. 2 mm. Mso, it was noted that plot size appeared to have little effect on the difference between the estimates. The level of cor- relation obtained is encouraging consider- ing the small sample and the fact that the initial soil moisture was estimated from a single measurement within each plot. The initial tests of the infiltration model on a heterogeneous area involved ~ter balance calculations using data collected during 1974 and 1975 on the Creighton Tributary, a small watershed (11.4 km 2 ) located in the semi-arid region of western Saskatchewan. Certain features concerning the data available for the watershed and its physical characteristics 431 facilitated the approach: (a) in 1974 and 1975 comprehensive snow surveys had been undertaken to establish the snowcover water equivalent by procedures outlined by Steppuhn and Dyck (1974); (b) measurements of "fall" soil moisture content, made with a neutron gauge at 23 sites in fields lo- cated immediately adjacent to the water- sheds having the same soil type and crop- ping patterns, were available (Banga, 1981); and (c) streamflow from snowmelt was carefully monitored. An important property of the watershed is that it does not have large elements of depressional storage and therefore the gross area of the basin can be taken as a close approxi- mation of the contributing drainage area. The general topography of the watershed may be classed as rolling to gently undu- lating with approximately 85% of the area under the cultivation of cereal grains by dryland farming. The two winters, 1973-74 and 1974-75 had contrasting snowcover and premelt soil moisture conditions. The winter of 1973- 74 was a year of near-record snowfall, which produced an average depth of snow- cover on the watershed of 556 mm and a snowcover water equivalent of approxi- mately 143 mm. It was preceded by a warm, dry fall and the average moisture content of the surface layer of soil (0-300 mm) was very low (-15% by volume), especially in those land units cropped the preceding summer. Fallow and grass lands were classified in the "Limited" category while stubble lands were considered to have an "Unlimited" infiltration potential. In contrast, conditions during the winter of 1974-75 could be likened more closely to "normal". The average depth and water equivalent of the snowcover were 299 mm and 71 mm respectively, and the average fall soil moisture content was 27.4% by volume. All land use units within the watershed were classified in the "Limited" category. The tests compared the total volume of runoff? calculated as the difference be- tween the areally-weighted snowcover water equivalent and infiltration (Qcal = SWE- INF), with the measured volume determined from the streamflow hydrograph (Qmeas). The ratio of Qcal:Qmeas was 1.02 in 1974 and 1.17 in 1975. The residual (SWE-INF- measured runoff), expressed as an equiv- alent average depth of water over the area, was 1.0 rnm in 1974 and 5.8 mm in 1975. These values compare favorably with the expected error of the estimate found when the model was used to obtain infiltr- ation at a point and on small plots. INTERFACING WITH EXISTING OPERATIONAL SYSTEMS Computer programs were written to inter- face the infiltration model with the U.S. National Weather Service River Forecasting System, Sacramento Model (NWSRFS) and the U.S. Army Corps of Engineers Streamflow Simulation and Reservoir Regulation model (SSARR). Gray et al. (1984, 1985b, c) describe different interfacing procedures that have been considered and the reader is referred to these publications for greater detail. Only a brief description of the methodology employed in the current study is provided herein. Sequencing Infiltration Quantities In order to apply the model in opera- tional practice the variation in infiltra- tion rate with time during the melt period must be assumed. Granger et al. (1984) showed that the pattern of infiltration depends on many factors: the rates of snowmelt and snowcover runoff; the depth, temperature regime and water transmission characteristics of the snowcover; the amount and distribution of ice in the fro- zen soil; the soil temperature regime and others. Gray et al. (1985b) analysed the infiltration curves from melt of a Prairie snowcover and showed that two general patterns are dominant: an "advanced" pat- tern, in which most of the infiltration occurs early in the melt sequence, and a "delayed" pattern, in which the infiltra- tion rate progressively increases through the melt period. The advanced pattern typifies infiltration from the rapid melt of a shallow snowcover. The delayed pat- tern reflects an increase in infiltration with time primarily due to an increase in the melt rate because of increases in net radiation and the amount of energy advec- ted from snow-free areas and changes to the physical properties of the snow cover and frozen soil which occur over the period of ablation. 432 In view of the strong dependency of in- filtration on the melt process it is log~ cal to allow the output of meltwater from the snowcover generated by the ablation subroutine of a model to be a dominant factor in sequencing infiltration. Gray et al. (1985b), using the results from runoff simulations on the Creighton Tributary, suggest the use of an index approach and a snowmelt infiltration index to define the time of runoff. Runoff does not begin until the rate of release of meltwater from the snowcover is greater than the snowmelt infiltration index. They suggest an initial value for the "daily" snowmelt infiltration index equal to the snowmelt infiltration potential (from Eq. 1) di- vided by 6 days. When the daily meltwater flux exceeds the index, infiltration is calculated by multiplying the amount of meltwater produced by the ratio of the snowmelt infiltration potential to the snowcover water equivalent at that time. CONTRIBUTING AREA It is emphasized that the performance of any model in simulating streamflow from snowmelt is directly related to the snow- covered-area contributing runoff to the channel. Hence, it is important to have a reliable estimate of the volume of snow water, that is the product of contributing area * snowcover water equivalent, as in- put to the system. Runoff from a Prairie watershed is not generated uniformly from all parts en- closed by the topographical divide. Prai- rie lands are relatively flat and their natural drainage systems are usually poorly developed and unconnected. In many years there may be no contribution of snowcover runoff to streamflow from a large part of a watershed because of the shallow snowcover and large amounts of depressional storage. In low snow years the source of runoff is snow collected in the poorly-defined channels and depres- sions that feed the main drainageway. Conversely, meltwater produced by a deep snowcover will generally satisfy depres- sional storage and overflow into the channels. Thus, a Prairie basin has a variable "contributing" area whose magni- tude varies with such factors as the deptb of snowfall and antecedent soil moisture md surface storage conditions. In this regard hydrologists have made use of the concept of "Effective" and "Gross" drain- age areas. The "Effective" area is that portion of a basin which might be expected to entirely contribute runoff to the main stream during a flood with a return period of two years; the "Gross" area is the plane area enclosed by the drainage divide which might be expected to contribute run- off to the outlet under extremely wet con- ditions (Godwin and Martin, 1975). The "Effective" area includes the major drain- age channels and land immediately adjacent to defined drainageways. It is the snow- cover in these areas that needs to be surveyed and used as input to a model in low snow years; in snowier winters this component of the average basin snowcover becomes less important because of the larger area of a basin contributing to runoff. Gray et al. (1985c) studied the inter- action between runoff potential, runoff volume and "apparent" contributing area for a Prairie basin. Runoff potential was defined as the sum of the snowcover water equivalent (SWE) plus precipitation occur- ring during the melt period (PPT) less the volume of infiltration (INF) calculated by Eq. 1. The "apparent" contributing area was the area of the basin that produced the observed runoff volume with the unit runoff potential (i.e., "apparent" area = observed runoff / unit runoff potential). Their results suggest that a relationship does exist between these variables; how- ever, there was insufficient data to allow them to define the relationship. They indicate that further research is needed into the interaction of snowfall, snow- cover and topographical aspects, and con- tributing area for Prairie watershed. PERFORMANCE OF OPERATIONAL SYSTEMS WITH THE INFILTRATION MODEL NWSRFS An algorithm of the infiltration model was interfaced with the NWSRFS and the revised system was used to synthesize streamflow from snowmelt on the Wascana watershed at Sedley, Saskatchewan. The basin is located approximately 50 km southeast of Regina in the Dark Brown soil 433 zone. Approximately 85% of the watershed is under cultivation of cereal grains by dryland farming; the remaining area is pasture, woody vegetation, roads, farm- yards and townsites. The topography is flat to gently rolling and because of the poor relief and drainage development the portion of the "gross" area that contri- butes to annual streamflow from snowmelt can vary widely. The "effective" and "gross" drainage areas of the watershed are 125 km 2 and 350 km 2 , respectively. Figure 3 shows the observed streamflow hydrographs from a low flow year, 1972 (Fig. 3a) and a high flow year, 1982 (Fig. 3b) plotted with three hydrographs gene- rated with the NWSRFS: (1) Land -the system with its original "LAND" subroutine and a contributing area of 125 km 2 , the "effective" area; (2) REVISED, 125 -the NWSRFS with the LAND subroutine replaced by the infiltration algorithm and a drain- age area equal to the "effective" area and (3) REVISED, 144 (Fig. 3a) and REVISED, 296 (Fig. 3b) -the same as (2), but with "apparent" drainage areas of 144 km 2 and 296 km 2 respectively for the low and high flow years. In the simulations, all model parameters, except drainage area, were the same. The results in Fig. 3 show: (1) The NWSRFS with its LAND subroutine grossly underestimates the observed volume of runoff. For example, in 1972 the ob- served runoff, expressed as an equivalent depth of water on the effective drainage area of 125 km 2 , was 29 mm compared with a depth of 1.3 mm simulated by the LAND subroutine. (2) The infiltration model significantly improves the performance of the NWSRFS in simulating streamflow from snowmelt over that obtained with the LAND subroutine. A measure of the improvement is obtained by comparing the observed and simulated dis- charge rates using the non-dimensional paremeter R2 defined by the equation (Nash and Sutcliffe, 1970): R2 where: n (3) number of values at evenly-spaced time intervals, observed discharge rate, simulated discharge rate, and mean of the observed values. R2 , termed efficiency, can be likened to the coefficient of determination used in statistics. The closer the positive value is to unity, the closer the agreement be- tween observed and simulated hydrographs. With a drainage area of 125 km 2 , the R2 - values for hydrographs generated with the LAND routine in 1972 and 1982 were -0.32 and -0.15, respectively, compared with R2 - values of 0.70 and 0.48 for the correspond- ing hydrographs generated by the REVISED NWSRFS. The poor simulation obtained with LAND can be directly linked to manner in which the routine evaluates infiltration to frozen soils. As shown by Gray et al. (1984) the performance of the system can be markedly improved by adjusting the soil moisture accounting routine to reduce the Upper Zone soil moisture capacity and by setting the Lower Zone percolation rate to a very small value, if possible equal to zero. The agreement between the shapes and time elements of the hydrographs simulated using the infiltration algorithm and observed hydrographs vary in the two years, with the agreement being signifi- cantly better in the high flow year, 1982. Because the NWSRFS has a high degree of flexibility the "fit" could have been improved by changing different parameters in the model. However, such changes would only serve to mask the effects of the in- filtration model in improving model per- formance. Previous work with the NWSRFS (Division of Hydrology, 1977) has shown that a major factor contributing to poor agreement between simulated and observed hydrographs when the system is applied to prairie watersheds is an incorrect simula- tion of the ablation process. For example, the manner in which the system accounts for changes in the internal energy of a shallow snowcover; differences in the magnitude of the melt factor and water transmission properties of the snowcover; the use of a temperature index for calcu- lating melt and other factors. It is sus- pected that the larger differences between simulated and observed hydrographs in years of low snowcover reflects a problem 434 ~ g !5.0 u. >-.J ... 2.!5 < 0 eo. " ~ Ill E v AO. ~ 0 .J u. >-20. .J ... < 0 REVISED, 144 APR SO DATE (a) LOW FLOW YEAR. 1972 NAY 1~ DATE <b> HIGH FLOW YEAR. 1982 Figure 3. Observed and simulated hydro- graphs from snowmelt for Wascana Creek at Sedley, Saskatchewan: (a) low flow year, 1972 and (b) high flow year, 1982. LAND - the NWSRFS operated with its LAND subrou- tine and a contributing area of 125 km 2 ; REVISED 125, 144 and 296 -the NWSRFS with the infiltration algorithm and contribut- ing areas of 125 km 2 (effective area) and 144 or 296 km 2 ; areas that produced the observed volume of runoff with the NWSRFS. in simulating the meltwater release and water transmission properties of snow- filled channels. (3) In years of low flow the "effective" area of the watershed is a better estima- tor of the "apparent" area contributing to the peak discharge; in years of high flow a contributing area equal to or less than the "gross" area gives the better simula- tion. SSARR An algorithm of the infiltration model was also developed and interfaced with the U.S. Army Corps of Engineers Streamflow S~ulation and Reservoir Regulation (SSARR) model. The procedures followed in calculating and simulating infiltration quantities by the algorithm were identical to those used with the NWSRFS and discus- sed above. The revised SSARR was used to synthesize streamflow from snowmelt on the Wascana watershed. Figure 4 shows the observed streamflow hydrographs plotted with the simulated hydrographs generated by the SSARR and Revised SSARR models for a low flow year, 1972 (Fig. 4a) and a high flow year, 1982 (Fig. 4b); the same runoff events used in the tests with the NWSRFS (Fig. 3). The "apparent" drainage areas determined from the results of the tests with the NWSRFS were used, that is 144 km 2 in 1972 (Fig. 4a) and 296 km 2 in 1982 (Fig. 4b). For the low flow year (Fig. 4a) the agreement in shapes of observed and simu- lated hydrographs is poor and may be at- tributed to a host of factors: poor choice of melt factor, an incorrect base tempera- ture and, as with the NWSRFS, incorrect simulation of other factors affecting snowcover ablation in and runoff from in- channel accumulations. Despite the poor agreement in shapes of the curves it was found that at the end of active snowmelt the streamflow volume simulated with the ~VISED SSARR was only 4% less than the observed, whereas with the SSARR it was 48% less. In the high flow year (Fig. 4b) the agreement of the hydrographs in both shape and timing is markedly improved over the results obtained in 1972. Further, it is obvious that the simulation with REVISED SSARR is better than that obtained with SSARR. The R2 -value for the REVISED SSARR was 0. 79 compared with 0. 63 with SSARR and the volume of runoff simulated with the revised system was only 12% less than the observed whereas SSARR underestimated the flow by 34%. Simulation of streamflow with the two models was also undertaken for 1974, 1979 and 1980. Table 1 compares the "observed" and "simulated" volumes for the five years and it can be seen that in all years 435 DATE <a) LOW FLOW YEAR. 1972. eo " II ' Ill E v 40 :J: 0 _j l1. )-20 _j ..... < 0 DATE Cb) HIGH FLOW YEA~ 1882 Figure 4. Observed and simulated hydro- graphs from snowmelt for Wascana Creek at Sedley, Saskatchewan: (a) low flow year, 1972 and (b) high flow year, 1982. REVISED SSARR -SSARR with the infiltration algor- ithm and SSARR -the model operated in its original mode. The contributing areas were 144 km 2 in 1972 and 296 km 2 in 1982. except 1980 the relative error in predict- ing the runoff volumes was substantially smaller with REVISED SSARR than with SSARR. On average, the infiltration algor- ithm reduced the relative error in esti- mating volume by a factor of ~2, from 34 to 15%. Also the simulated volumes, except that calculated by REVISED SSARR in 1980, are less than the observed flow. This result was surprising inasmuch as the con- tributing areas used in the simulations were those values that gave the observed flow with the Revised NWSRFS and the same meteorological inputs, initial soil mois- ture contents and initial snow water Table 1. Comparison of observed flow vol- ume and volumes simulated with Revised SSARR and SSARR models for five years of snowmelt runoff on Wascana Creek near Sedley, Saskatchewan. Year 1972 1974 1979 1980 1982 Observed Flow (cmsd) 38.9 327.5 152.7 47.9 219. 7 Revised SSARR Flow Error (cmsd) % 37.4 4 254.2 22 127.5 17 58.1 21 193.3 12 Mean Std. Deviation 15.2 7.4 SSARR Flow Error (cmsd) % 20.2 48 231.7 29 85.7 44 40.4 16 145.9 34 34.2 12.7 equivalents. The underprediction of runoff volume by SSARR, compared with the NWSRFS, can be attributed to differences in the manner the two models treat snow accumula- tion and ablation. Because SSARR uses a low-level cut-off value of snowfall depth to eliminate very small events, in many years the maximum snow water equivalent (MAXSWE) used by the model is less than the corresponding value used by the NWSRFS. Table 2 compares values of MAXSWE given by the two "revised" systems. Only in 1980 was the MAXSWE with SSARR greater than for NWSRFS and this was due to the fact that SSARR predicted a later date of snowmelt and precipitation from a late-occurring event was added to the snowpack. The result, as shown in Table 1, was to cause the runoff volume simulated by the Revised SSARR to be greater than the observed. Table 2. Comparison of MAXSWE for the "revised" NWSRFS and SSARR. Year NWSRFS SSARR NWSRFS mm mm SSARR 1972 45 41 1.10 1974 137 129 1. 06 1979 83 77 1. 08 1980 51 57 .89 1982 99 94 1. 05 436 SUMMARY This paper presents a simple, physically based model of infiltration to frozen soils. The model is based on the concept that for practical purposes frozen soils may be grouped into three broad categories with respect to their infiltration poten- tial, namely: Unlimited -cracked or high- ly porous soils which are capable of in- filtrating most of the snow water; Limited -the infiltration potential of a soil depends primarily on the snowcover water equivalent and the ice content of the soil layer, 0-300 mm, at the time of melt and Restricted - a soil containing an imper- meable layer that inhibits infiltration. Field measurements of soil water changes made during snowcover ablation are pre- sented to support the delineation of "Restricted" and "Unlimited" classes and an empirical equation relating infiltra- tion to snowcover water equivalent and the moisture (ice) content of the soil layer 0-300 mm is given for the Limited class. Several validation tests of the model in estimating "point" and "areal" infil- tration amounts are given. As an example, it is shown that the potential volume of runoff, calculated as the difference : snowcover water equivalent less infiltra- tion for two years of widely-different snowcover and antecedent soil moisture conditions agreed within 17 percent of the observed streamflow from a heterogeneous 11.4 km 2 watershed. The results of the tests supported the proposition that the model would provide estimates of infil- tration suitable for practical management use. Procedures and methodologies of inter- facing the model with operational stream- flow forecasting systems are described. It is shown that the performance of the U.S. National Weather Service River Fore- casting System (NWSRFS) and the U.S. Army Corps of Engineers Streamflow Simulation and Reservoir Regulation model in simula- ting streamflow from snowmelt on a Prairie watershed is markedly improved with use of the model. The improvement being primarily due to a better estimate of runoff volume from that obtained with the original systems. For example, on average the infiltration algorithm in SSARR reduced the relative error by the system in esti- mating observed flow by a factor of ~2, from 34 to 15%. The need for correct evaluation of the snow-covered area contributing to stream- flow in order to obtain a good simulation of runoff from a Prairie watershed is em- phasized. LITERATURE CITED Anderson, E.A., 1973. National Weather Service river forecast system -snow accumulation and ablation model. Tech- nical Memorandum NWS HYDR0-17, Nat- ional Oceanic and Atmospheric Admini- stration (NOAA), Silver Springs, Mary- land. Anderson, E.A. and P.J. Neuman, 1984. Inclusion of frozen ground effects in a flood forecasting model. In: Proceed- ings of the 5th Northern Basin Sympos- ium and Workshop, Vierumaki, Finland. Banga, A.B., 1981. Spatial variability of soil moisture in a Prairie environment. M.Sc. Thesis, University of Saskatchewan, Saskatoon, Saskatchewan, 86 pp. BergstrHm, S., 1979. Spring flood fore- casting by conceptual models in Sweden. In: Proceedings of Modeling Snow Cover Runoff, 1978, S.C. Colbeck and M. Ray (Editors), U.S. Army Corps of Engineers Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire, pp. 397-405. Division of Hydrology, 1977. An examina- tion of the U.S. NWS River Forecast System snow accumulation and ablation model under Prairie snowmelt conditions. Unpublished Internal Report, Division of Hydrology, University of Saskatchewan, Saskatoon, Saskatchewan, 26 pp. Gillies, J.A., 1968. Infiltration in frozen soils. M.Sc. Thesis, University of Saskatchewan, Saskatoon, Saskatchewan, 136 pp. ~odwin, R.B. and F.R.J. Martin, 1975. Calculation of gross and effective drainage areas for the Prairie Prov- inces. In: Proceedings of the Canad- ian Hydrology Symposium:75. National Research Council of Canada, Ottawa, Ontario, pp.218-223. Granger, R.J. and G.E. Dyck, 1980. Report of the 1979/80 investigations of infil- tration and runoff from snowmelt. Int- ernal Report, Division of Hydrology, University of Saskatchewan, Saskatoon, 437 Saskatchewan, 67 pp. Granger, R.J., D.M. Gray and G.E. Dyck, 1984. Snowmelt infiltration to frozen Prairie soils. Canadian J. Earth Sciences 21(6):669-677. Gray, D.M., R.J. Granger and G.E. Dyck, 1985a. Overwinter soil moisture changes. Transactions Amer. Society Agricultural Engineers 28(2):442-447. Gray, D.M., P.G. Landine and R.J. Granger, 1984. An infiltration model for frozen Prairie soils. Paper No. 84-313 pre- sented at the 1984 Annual Meeting of the Canadian Society of Agricultural Engineers, Winnipeg, Manitoba. 15 pp. Gray, D.M., P.G. Landine and R.J. Granger, 1985b. Simulating infiltration into frozen Prairie soils in streamflow models. Canadian J. Earth Sciences 22:464-474. Gray, D.M., P.G. Landine and G.A. McKay, 1985c. Forecasting streamflow runoff from snowmelt in a Prairie environment. In: Proceedings of the 7th Canadian Hydrotechnical Conference, Canadian Society of Civil Engineers 1A:213-231. Holtan, H.N., 1961. A concept for infiltra- tion estimates in watershed engineering. U.S. Department of Agriculture, Agri- culture Research Service Report 41-51, 25 pp. Holtan, H.N., G.J. Stiltner, W.H. Henson and N.C. Lopez, 1975. USDAHL-74, Revised model of watershed hydrology. U.S. Dep- artment of Agriculture, Agricultural Research Service Technical Bulletin 1518. Horton, R.E., 1940. An approach to the physical interpretation of infiltration capacity. Soil Science Society of Am- erica Proceedings 5:399-417. Kane, D.L., 1980. Snowmelt infiltration into seasonally frozen soils. Cold Reg- ions Science and Technology 3:153-161. Kane, D.L. and J. Stein, 1983. Water move- ment into seasonally frozen soils. Water Resources Research 19(6):1544-1557. Kuznik, I.A. and A.I. Bezmenov, 1963. Infiltration of meltwater into frozen soil. Soviet Soil Science Number 6, pp. 665-671. Motovilov, Yu.G., 1978. Mathematical model of water infiltration into frozen soils. Soviet Hydrology 17(1):62-66. Motovilov, Yu.G., 1979. Simulation of meltwater losses through infiltration into soil. Soviet Hydrology 18(3):217-221. Nash, J.E. and J.V. Sutcliffe, 1970. River flow forecasting through concep- tual models. J. Hydrology 10:282. Peck, E.L., 1976. Catchment modelling and initial parameter estimation for the National Weather Service River Forecast System. NOAA Technical Memorandum NWS HYDR0-31, Office of Hydrology, Wash- ington, DC. pp. 1-24, plus Appendices. Popov, E.G., 1973. Snowmelt runoff fore- casts -theoretical problems. In: Role of Snow and Ice in Hydrology, 1972. Proceedings of the Banff Symposia, UNESCO-WMO-IAHS, 2:829-839. Quick, M.C. and A. Pipes, 1973. Daily and seasonal runoff forecasting with a water budget model. In: Role of Snow and Ice in Hydrology, 1972. Proceed- ings of the Banff Symposia, UNESCO-WMO- IAHS, 2:1017-1024. Romanov, V.V., K.K. Pavlova, and I.L. Kalyuzhnyy, 1974. Meltwater losses through infiltration into podzolic soils and chernozems. Soviet Hydrology Selected Papers Number 1, pp. 32-42. Shipak, I.S., 1969. Relationship between the runoff coefficient and the moisture content and depth of freezing soil. Soviet Soil Science, pp. 702-706. Steppuhn, H. and G.E. Dyck, 1974. Esti- mating true basin snowcover. In: Advanced Concepts and Techniques in the Study of Snow and Ice Resources. Interdisiciplinary Symposium, U.S. National Academy of Science, Washington DC., pp. 314-324. United States Army Corps of Engineers, 1972. Program description and users manual for SSARR model program 724-KJ- GOOlO. North Pacific Division, Port- land, Oregon. United States Army Corps of Engineers, 1973. HEC-1, Flood Hydrograph Package Users Manual. Hydrological Engineering Center, Davis, California. United States Department of Commerce, 1972. National Weather Service River Forecast Procedures. Technical Memo- randum NWS HYDR0-14, Silver Springs, Maryland, pp. 1-1 to 1-20. Willis, W.O., C.W. Carlson, J. Alessi and H.J. Haas, 1961. Depth of freezing and spring runoff as related to fall soil- moisture level. Canadian J. Soil Science 41:115-123. 438 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 USING REAL-TIME (SNOTEL) DATA IN THE NWSRFS MODEL Keith R. Cooley* ~STRACT: Physical process simulation models and real-time snowpack data show promise for enhanced water supply fore- casts yielding complete hydrographs rather than the traditional seasonal volume fore- casts based on regression techniques and monthly snow course readings. Hydrograph forecasts are especially important for efficient reservoir operation and flood peak estimates. The National Weather Service River Forecast System (NWSRFS) model which is capable of simulating snow accumulation and melt, and streamflow was tested using snowpack and climatological data from sites in Idaho and Montana. Several approaches for using real-time SNOTEL data in the NWSRFS model were also tested. Results indicated that using SNOTEL data did not improve forecasts in most cases. However, modification of standard calibration procedures to more nearly reflect the relationship between snow pillow data and streamflow could improve results. Results also emphasize the expedience of properly locating precipitation and snow water equivalent monitoring equipment. (KEY TERMS: simulation models; water supply forecasts; snowmelt; streamflow; snow pillow, precipitation.) INTRODUCTION Most of the annual streamflow 1n the western United States occurs during the spring and summer months, and is found to be a product of the water acc\lmulated during the winter in the snowpack and soil profile. Knowledge of the amount of water that will be available 1s essential for efficient use. Even a few months' notice allows managers to optimize reservoir operations for the multitude of competing users. Adequate water supply forecasts also let farmers decide how much acreage to plant and to which crops. Even more critical are accurate and timely forecasts during extreme years when droughts or floods occur. Water supply forecasts have tradi- tionally been made using multivariate regression techniques based on monthly snow course readings. These forecasts produced estimates of seasonal (i.e., April through July) runoff volume, and except 1n extreme years, were quite accurate. In recent years mathematical models designed to simulate physical processes have received more attention. These simulation models are of particular interest to streamflow forecasters because they produce a continuous record of streamflow volume with respect to time, which is essential to good flood prediction and reservoir operation. Another recent development that makes simulation models attractive is the avail- ability of continuous snowpack information provided by on-site sensors and trans- mitters like those used in the SCS SNOw TELemetry (SNOTEL) system (Crook, 1985). Snowpack snow water equivalent (SWE), accuumulated precipitation, and air temperature are now available for several hundred stations in the western United States at intervals of 15 minutes or greater. Simulation model forecasts could be updated or initiated at any time using *Hydrologist, Northwest Watershed Research Center, Agricultural Research Service, U.S. Department of Agriculture, 270 South Orchard, Boise, Idaho 83705. 439 SNOTEL data, as opposed to the once monthly forecasts normally provided by regression methods. The objectives of this study were to: 1) test a snow accumulation and melt model; and 2) evaluate methods of using real-time (SNOTEL) data in a simulation model for water supply forecasting. PROCEDURE AND METHODS Following a literature review, a snow model developed by Eric Anderson of the National Weather Service was selected for testing. The model, HYDR0-17 (Anderson, 1973), is a conceptual model of the significant physical processes affecting snow accumulation and snowmelt. It uses air temperature as an index to energy exchange across the snow-air interface and was selected because: 1) air temperature data are readily available; 2) the approach appeared to be technically sound; 3) the model has been tested in several climatic regions within the United States; and 4) an expected range of values for the calibration parameters was provided for a variety of conditions. The HYDR0-17 snowmelt model is part of the National Weather Service River Fore- cast System model (NWSRFS), which consists of modular units that can be used singly or in combination for simulating various parts of the hydrologic cycle. The NWSRFS model offers both a snow accumulation and melt algorithm (HYDR0-17), and a basis for evaluating the use of real-time data in water supply forecasting. Therefore it was used for both aspects of this study. Three evaluations of the HYDR0-17 model were made using snow water equivalent data from snow course or snow pillow sites in Idaho and Montana. In the evaluations, HYDR0-17 was calibrated by adjusting parameter values unti 1 simulated and observed SWE were esentially the same during one to five year periods. Runs were then made using the calibration parameter values for additional periods that represented above normal, normal, and below normal snow accumulation years. The entire model was calibrated by matching simulated and observed streamflow using 1973 to 1978 water year data from Lower Willow Creek, Montana. The modules of the NWSRFS model used were: 1) 440 HYDR0-17 -Snow accumulation and melt; 2) SAC-SMA -Soil moisture accounting; and 3) UNIT-HG -Unit hydrograph shape. Lower Willow Creek was divided into upper and lower zones (areas above and below 1860 m), and parameter values 1n HYDR0-17, SAC-SMA, and UNIT-HG were adjusted in both zones during the cali- bration runs as suggested by NWS model documentation (unpublished users manual). The SNOTEL data were used in the NWSRFS model 1n three different ways. These were: 1) After long-term calibration param- eter values were established, the snow correction factor (SCF) values were changed each year to produce simulated snow water equivalent (SWE) values equal to observed SWE values on April 1st. The mode 1 was then run for the April through July forecast period using observed temperature and precipitation data from Drummond, Montana. 2) The NWSRFS model was initiated on January 1st, February 1st, or March 1st each year using observed SWE on these starting dates and actual Drummond temper- ature and precipitation data for the forecast period from the starting date through July. 3) Actual SNOTEL SWE data were used (in place of HYDR0-17 simulated SWE) along with rainfall as input to the SAC-SMA soil moisture accounting subroutine. DESCRIPTION OF SITES AND DATA SETS Data for testing the HYDR0-17 snow model were obtained from an Idaho snow course and two Montana snow pillow sites. The snow course, Reynolds Mountain, is located at an elevation of 2073 meters on the upper end of the Reynolds Creek Exper- imental Watershed near Boise, Idaho. The surrounding area is predominantly sage- brush mountain meadows with a few scattered stands of willows, aspen, and Douglas fir. Mean annual precipitation is 1016 mm, most of which (60 -70%) occurs as snowfall, and is accompanied by southwesterly winds. In addition to the SWE data provided by the snow course readings, daily maximum and minimum air temperature and precipitation records are available for most of the period since 1966. The two snow pillows are located on the Lower Willow Creek basin near Hall, Montana (Fig. 1). The Combination site is located at an elevation of 1700 meters near a main tributary in a mixed conifer setting, and represents the lower end of a conifer forest. Average annual precipi- tation at the site is about 560 mm based on six years of record. Records of SWE are available for the 1973-1984 period, but precipitation data are available only for the 1979-1984 period. The Black Pine site is located at an elevation of 2164 meters on the southerly upper end of the basin. Average annual precipitation for a six year period (1979- 1984) was 716 mm. Records of SWE are available for the 1966-1984 period. The Black Pine site is located in a dense conifer setting and represents a contin- uous conifer forest. The Lower Willow Creek bas in was selected for evaluating the NWSRFS model simulations using SNOTEL data. This basin encompasses an area of 190 square kilo- meters above the reservoir. Elevation ranges from 1430 m above sea level at the reservoir to over 2400 m at the highest point. Average annual precipitation varies from 350 mm at the lower elevations to over 760 mm at the higher elevations. I 0 I 2 3 KILOMETERS Figure 1. Lower Willow Creek Watershed Near Hall, Montana. 441 Streamflow records are available for port ions of each year from 196 7 through 1984. Because only SWE and 1979-1984 precip- itation data were available from the Combination and Black Pine SNOTEL sites, Drummond temperature and precipitation data were used when needed as input to the model. Long-term c 1 imatological records from the Drummond weather station were the most complete and produced the best over- all results of three stations within the general area of Lower Willow Creek. Drummond is about 16 km northeast of the basin at an elevation of 1200 m. Drummond records were used as input to the HYDR0-17 model for both snow pillow sites and the upper and lower zones of Lower Willow Creek. TESTS OF HYDR0-17 SNOW MODEL Tests of HYDR0-17 using snow course data from the Reynolds Mountain site were described by Cooley, et al. (1983). They found that simulated SWE values could be made to essentially duplicate observed values for a given year. However, when the same parameter values were used for simulations in other years, results did not match as well . Even so, HYDR0-1 7 produced better results than four other models tested on the same data set (Huber, 1983). Figure 2 illustrates the results based on a five year calibration period at the Combination site (results for the Black Pine site were similar). These results suggest that when one set of parameter values is used for several years, simu- lated values will be high some years, low some years, and about the same as observed other years. The ambient temperature sometimes fails as a reliable index of the physical processes that cause the snow to accumulate and melt. The inclusion of additional variables such as solar radi- ation, wind run, and vapor pressure may be necessary to improve the mode 1. Since these data are seldom available at the sites requiring simulation, the temper- ature data must suffice and this 1 imi- tation of the model is recognized. 300 OBSERVED PREDICTED i: 250 ::E: 1- ~ 200 -' a: > ..... ~ 150 C3 UJ a: UJ ri; 1 DO 3: 3: c z en 50 0 1974 ~ I I I I ' I , .. 1 I I I I I' I I I I 1975 * II I I I I I I I 1976 1977 1978 Figure 2. Simulated Snow Water Equivalent at Combination Snow Pillow Using Drummond Weather Station Precipitation and Temperature Data, Compared to Observed Snow Water Equivalent at Combination Snow Pillow(-----Observed; Simulated). USING SNOTEL DATA IN THE NWSRFS MODEL In order to evaluate the effectiveness of using SNOTEL data to improve model simulated streamflow forecasts, it was first necessary to run the model using normally available data to produce streamflow forecasts which could be used as a basis for comparison. A twelve year record (1973-1984) of snow pillow and streamflow data at Lower Willow Creek, and corresponding temperature and precipi- tation data at Drummond, were available for the tests. In this case, a six year (1973-1978) calibration and a six year (1979-1984) test period were selected. Both six year periods contained high, low, and average years of streamflow. Relationships between precipitation at Drummond and the snow pillow sites were derived using the precipitation data collected at the pillow sites after the 1979 gage installations. Monthly and annual ratios of snow pillow precipitation to Drummond precipitation were used as first estimates for snow correction factor (SCF) and precipitation adjustment factor (PXADJ) values in the calibration runs. Precipitation-elevation relationships may also be used as a basis for estimating these parameters which alter volume but not time of occurrence. 442 Since the monthly ratios varied consid- erably during the year, seasonal ratios seemed to be more realistic than average annual values. However, it isn't possible to change parameter values during the year without stopping a run, re-initializing parameters, and starting again at the stopping date. Therefore, PXADJ was selected to represent the summer ratio between the snow pillow sites and Drummond precipitation, and SCF was used to adjust for the increased ratio noted during the winter when prec1p1tation was predomi- nately snowfall. The SCF parameter 1s thus raised above its normal range to account for not only snowfall loss from unshielded gages, but also to represent the greater seasonal ratio noted for the winter period. Other initial parameter values used in the model were based on actual data when possible, i.e., potential evapotrans- piration was based on data from Bozeman, Montana. The observed and model simulated runoff traces for the 1973-1978 cali- bration years are presented in Figure 3. In general, simulated values of runoff matched observed values better during the calibration period than during the test period. Simulated runoff volume was greater than observed runoff volume during most of the test years. The magnitude of 20 18 ---OBSERVED --PREDICTED 16 u 0 12 (fl :E: ~ 10 3: 8 c ....J u.. 6 :E: (! UJ II: • 1- (fl 2 0 1973 1974 1975 1976 1977 1978 Figure 3. April Through July Observed and Simulated Streamflow at Lower Willow Creek, Montana for the 1973-1978 Calibration Period. the overprediction during the test period tended to be greater than the differences noted for the calibration period. These differences could be due in part to a change in the relationship between precip- itation on the basin and that at Drummond during the two six year periods. Unfor- tunately, data are not available prior to 1979 to provide verification. Although differences between observed and simulated runoff volume were rather luge some years, the timing of major events (peaks) normally produced by spring snowmelt, was in quite close agreement. One problem noted several times during the s~mer or fall (late June through September) was the appearance of a simu- lated runoff event when there was no indi- cation of a change in observed runoff. These simulated events seem to be produced by thunderstorm rainfall recorded at Drummond and simulated by the model which, when adjusted by PXADJ, are significant enough to produce runoff from rainfall. Since there is no indication of an observed event, the storms either did not occur on the watershed or they were not of sufficient magnitude to produce overland flow and a change in streamflow. These descrepancies serve to emphasize the need for on-site precipitation data to improve model results. 443 Simulating Observed Snow Water Equivalent (SWE) on April 1st by Adjusting the Snow Correction Factor (SCF) Calibration values of the snow model parameters SCF and PXADJ were based on relationships between seasonal precipi- tation volumes at the snow pillow sites and Drummond. If these relationships changed from year to year because of variations in storm tracks, inversions, or other meteorological conditions, this could affect yearly simulation accuracy based on long-term averages. A possible method of accounting for this variability consists of adjusting the values of SCF and/or PXADJ each year while holding all other parameter values fixed. In this case, only SCF was adjusted since compar- isons were made on spring runoff which was mainly produced by snowmelt. The value of SCF was adjusted each year so that model simulated SWE on April 1st matched observed SWE at each snow pillow site. The simulated runoff hydrographs produced using the adjusted SCF values are presented in Figure 4, along with observed runoff for the April through July period for 1975, 1976, 1980, and 1983. The simulated runoff again exceeds observed runoff in most cases, while the timing of simulated and observed peaks matches 20 ---OBSERVED 18 --PREDICTED I 16 u c; 12 ttl I~ :::E: I ~ 10 I I I ~ Ia 1 I IJ ll I II 3:: 8 II I N I I EJ II I ....I II I u. II I 6 I I :::E: I I I I I a: I I \1,: I I UJ I I I I a: • I I ' I I 1-I 1 I I II : I '\I I ttl I J 1'\ I I I 2 I I \ I \ I '"' ,,J I v ' .. I 0 1975 1976 1980 1983 Figure 4. April Through July Observed and Simulated Streamflow at Lower Willow Creek, Montana for the 1975, 1976, 1980, and 1983 Water Years. SCF is Adjusted so that Simulated SWE Matches Observed SWE at the Snow Pillow Sites on April 1st Each Year. rather closely. The SCF values required to make simulated SWE match observed SWE on April 1st each year, are presented in Table 1 for the two zones used 1n the model. Also presented in Table 1 are statistical and runoff volume comparisons. The 1973 values are somewhat questionable, since October through December water year data were not available, and estimates of initial conditions on January 1st could affect these results. Omitting 1973, the range of SCF values required in the upper zone did not vary greatly from the value of 2.10 used in the calibration. However, the SCF values needed in the lower zone were all above the 1.25 value obtained during calibration, some by almost three times. These results suggest that April 1st SWE may not be a good index of spring runoff volume, especially from the lower zone where in most years some snowmelt has already occurred. Since snowpack conditions vary each year, matching maximum snow accumulation in each zone may provide a better index of later snowmelt runoff. However, it would be more difficult to use in forecast procedures since it would occur at different times in each zone and for each year. 444 Using SNOTEL SWE to Initialize the NWSRFS Model Another approach that could improve model simulated results would be to use SNOTEL snow water equivalent. data for updating the model whenever a forecast is desired. In this study, SWE as recorded at the pillow sites, on January 1st, February 1st, or March 1st was used to initialize the model. The model was then used to simulate April through July runoff, which was compared to observed runoff for the same period. The cali- bration values of model parameters were used, and the status of other conditions, such as soil water storage in the upper and lower soil zones, was estimated. Drummond temperature and precipitation data for the years tested were again used as input for the forecast periods. The four years which produced the greatest difference between simulated and observed runoff during the calibration and test periods (1975, 1976, 1980, and 1983) were selected for this test. The simu- lated runoff that was produced when the model was initiated on January 1st of the 1975 and 1980 years and the observed runoff for corresponding years is TABLE 1. Snow Correction Factors (SCF) Required to Match Observed SWE at the Combination and Black Pine Sites, and Simulated Runoff Compared to Observed Runoff at Lower Willow Creek, Montana. SCF April-July Runoff Daily Errors (nun) (nun) Lower Upper RMS 1 AVG-ABS 2 3 Zone Zone OBS SIM DIFF r 1973 5.90 3.20 25.7 44.4 18.7 0.16 0.12 0.935 1974 1. 75 1. 93 113.5 103.7 -9.8 0.35 0.26 0. 911 1975 l. 71 2.35 174.0 262.7 88.7 l. 06 0. 72 0.926 1976 1. 90 2.05 160.5 243.6 83.1 0.90 0.59 0.919 1977 2.84 2.46 16.4 37.5 21.1 0.18 0.13 0.913 1978 l. 75 2.42 83.6 125.7 42.1 0.34 0.25 0. 911 1979 l. 56 1.71 65.1 74.0 8.9 0.18 0.13 0.943 1980 1. 76 2.14 118.8 235.5 116.7 1.35 0.59 0.732 1981 2.93 2.02 86.54 156.54 70.04 0.58 0.45 0.863 1982 l. 30 l. 55 130.9 173.5 42.6 0.45 0.31 0.945 1983 3.42 2.35 66.9 1984 3.25 2.05 115.9 Total 24.17 23.03 Average 2.20 2.09 1 . Da1ly root mean square error. 2Daily average absolute error. 3 1 . ff. . f d . 1 fl Corre at1on coe 1c1ent or a1 y ows. 4April data missing. ~esented 1n Figure 5. Plots of the runoff traces produced when the model was initiated on February 1st and March 1st are not presented. These traces were essentially the same as the January 1st trace, in t 1m1ng, but were of greater magnitude in volume, especially peak values. Comparisons of runoff vol urne and statistical measures are presented in Table 2 for the various starting dates. hlso included for comparison are the results obtained during the calibration and prior test periods. In all cases tested, model simulated ·runoff exceeded observed runoff. It is 'interesting to note that snowpack conditions on the earliest starting date '(January 1st) produced the best results, i followed by the next earliest date. Also, this updating scheme improved results over 'iliose obtained in the calibration and 445 213.6 146.7 1. 22 0.61 0.752 174.1 58.2 0.90 0.46 0.661 20 ---OBSERVm -PREDICTED ~ 15 0 VI ::1: ~ 10 3: 10 _J 5 u.. ::1: a: UJ It: I- VI 0 1975 1980 Figure 5. April Through July Observed and Simulated Streamflow at Lower Willow Creek, Montana for 1975 and 1980. Model Initialized on January 1st Using Observed SWE at the Snow Pillow Sites. TABLE 2. Runoff volume and statistical comparisons for various model-run starting dates at Lower Willow Creek, Montana. April-July Runoff Daily Errors (mm) OBS SIM 1975 JAN 1st 174.0 184.8 FEB 1st 174.0 218.6 MAR 1st 174.0 215.8 CALIB. 174.0 184.4 1976 JAN 1st 160.5 170.9 FEB 1st 160.5 171.0 MAR 1st 160.5 180.7 CALIB. 160.5 187.5 1980 JAN 1st 118.8 208.4 FEB 1st 118.8 215.8 MAR 1st 118.8 227.5 TEST 118.8 198.6 1983 JAN 1st 66.9 77.1 FEB 1st 66.9 77.4 MAR 1st 66.9 93.7 APR 1st 66.9 132.3 TEST 66.9 77.2 1 See Table 1 for definitions. prior test runs in only the 1976 water year. However, results were essentially the same in the January 1st starting date runs for 1975 and 1983. Two factors that could have significant impact on the simulated results are the method of calibration used and the non- continuous data base available. If the model had been calibrated on the snow pillow data prior to calibration based on runoff, the SWE values may have been a better index of subsequent runoff. The problems associated with initiating a model run without a continuous record of input data such as runoff, soil moisture storage, etc., could also significantly affect the simulation results. In all cases where data are not available, the status of the various water storage and conveyance components must be estimated, thus impacting the results accordingly. (mm) DIFF RMS 1 AVG-ABS 1 1 r 10.8 0.56 0.34 0.924 44.6 0.70 0.48 0.930 41.8 0.69 0.46 0.928 10.4 0.54 0.33 0.929 10.4 0.35 0.26 0.973 10.5 0.34 0.25 0.974 20.2 0.40 0.30 0.965 27.0 0.38 0.29 o. 971 89.6 1. 26 0.62 0.703 97.0 1. 32 0.66 0.707 108.7 1.45 0. 72 0.699 79.8 1. 04 0.44 0.732 10.2 0.35 0.22 0.688 10.5 0.35 0.22 0.687 26.8 0.43 0.28 0.699 65.4 0.62 0.42 0.825 10.3 0.26 0.16 0.781 Using the NWSRFS Model with Daily SNOTEL SWE in Place of HYDR0-17 446 Simulated SWE One of the great advantages of the SNOTEL system over traditional snow course surveys is the ability to obtain real time data rather than biweekly or monthly SWE readings. Although the NWSRFS mode 1 was not originally designed to use SNOTEL information as input, it is possible to modify the normal procedures and make use of this more timely data. The procedure used in this study was to replace the simulated snowmelt plus rain output of the HYDR0-17 subroutine with actual daily snowmelt determined from the snow pillow readings and rainfall. In other words, only the soil moisture accounting and unit hydrograph submodels were used. The above procedure was tested on the 1975 water year. The model runs were initiated using April 1st watershed conditions as determined from SNOTEL data and calibration runs. Model parameters as determined for the calibration period were used in the first run and results are presented in Figure 6 and Table 3. Both timing and volume are noted to be quite different than observed. A second run was made in an attempt to correct these descrepancies. The precipitation adjust- ment factor PXADJ included 1n the soil moisture submodel was adjusted to compensate for the differences in runoff volume noted during the first run. ~ 0 (() I: 20 15 ~ 10 J: c ~ 5 I: CI ILl II: f- (1) 0 ---OBSERVED --PREDICTED J ----------'~ J II 1 I I \ 1 I 1 I 1 I 1 I I \ I \ I ' 1975 } II II II II I I I: I I I I I I I I \ ,_ Figure 6. April Through July Observed and Simulated Streamflow at Lower Willow Creek, Montana for 1975. Model Initial- ized April 1st Using Calibration Parameter Values, Observed SWE, and Actual Snowmelt at the Snow Pillow Sites Plus Drummond Rainfall as Input. Results of the second run are presented in Figure 7 and Table 3. As noted the observed and simulated runoff volumes are essentially identical, the daily errors are smaller, and the correlation coef- ficient ( r) has increased. However, the timing of runoff peaks is still not as good as that obtained during the cali- bration runs using HYDR0-17 results rather than observed snow pillow melt. Part of the difference observed could be due to the calibration being based on runoff uther than snow pillow data as previously mentioned. However, the large errors 447 noted at the end of May and first of June, may be greater than one could adjust for by changing soil moisture reservoir amounts and transmission times. If this was the case, it would again indicate that either the snow pillow sites do not represent the zones selected, or Drummond precipitation 1s not a good measure of precipitation on the watershed at this time. In either case, the importance of on-site climatological data and repre- sentative SNOTEL sites is imperative for increased modeling accuracy. Cl en ::E 20 15 ~ 10 ::!: D ~ 5 ::E a: w a: l-en o OBSERVED PREDICTED ----------'J 1975 I I It I I I I I I Figure 7. April Through July Observed and Simulated Streamflow at Lower Willow Creek, Montana for 1975. Model initial- ized April 1st Using "Best Fit" Parameter Values, Observed SWE, and Actual Snowmelt at the Snow Pillow Sites Plus Drummond Rainfall as Input. DISCUSSION The outcome of the studies was less than anticipated except for the cali- bration and test period results. However, the procedures were tested and proved to be operationally possible, and several main problems were identified that could be avoided in future studies. The first problem relates to the study site and the data set that was available. The lack of a continuous streamflow record made it difficult to develop calibration param- eters related to baseflow conditions. TABLE 3. Simulated and observed runoff comparisons for the April through July periods obtained from actual snowmelt measurements at the Black Pine and Combination snow pillow sites at Lower Willow Creek, Montana. PXADJ April-July Runoff Daily Errors Lower Upper Zone Zone OBS 1975 (1) 1.000 1.00 174 (2) 1.420 0.800 174 1 see Table 1 for definitions. Also, the discontinuous record made it necessary to assume initial watershed conditions at the beginning of each run and could have added to the magnitude of the errors. The second problem seemed to be a product of the method used to calibrate the model. In this case, the entire model was calibrated with respect to observed runoff. This may not have resulted in any problems were it not for the attempts to use actual snow water equivalent and snowmelt data as inputs for initializing and updating the model. A better approach might have been to calibrate the HYDR0-17 submodel with respect to data available at the snow pillow sites. Then using the calibration parameter values thus obtained for HYDR0-17, the parameter values in the soil moisture accounting and unit hydro- graph submodels could be adjusted (cali- brated) with respect to observed stream- flow. The tests described should then be more meaningful and produce better results. Acknowledgements. I wish to express my appreciation for the extra effort given by Dave Robertson in making computer runs, and providing suggestions while working with the NWSRFS model. I also wish to thank Bernie Shafer, Phil Farnes, Don Jensen, Shari Hennefer, Sue Jackson, Beth Pirrong, Apryl Wilson, and Leon Huber for their assistance. 448 (mm) (mm) SIM DIFF RMS 1 AVG-ABS 1 1 r 115.6 -58.4 1.26 0. 77 0.660 174.2 0.2 1.10 0.73 0. 755 LITERATURE CITED Anderson, E. A., 1973. National Weather Service River Forecast System--Snow Accumulation and Ablation Model. NOAA Technical Memorandum NWS HYDR0-17. U.S. Dept. of Commerce, Silver Spring, Maryland. 217 pp. Cooley, K. R., E. P. Springer, and A. L. Huber, 1983. SPUR Hydrology Component: Snowmelt, p. 45-61. In: SPUR-- Simulation of Production and Utili- zation of Rangelands: A rangeland model for management and research, J. R. Wight (Editor). USDA-ARS, Misc. Publ. No. 1431, Feb. Crook, A. G., 1985. Operational Experience in Meteor Burst Telemetry - Eight Years of SNOTEL Project Observations. In: Proceedings of International Workshop on Hydrologic Applications of Space Technology. Cocoa Beach, Florida. (In Press.) Huber, A. L., 1983. A Comparison of Several Snow Accumulation and Ablation Algorithms Used in Watershed Modeling, p. 76-88. In: Proceedings of Western Snow Conference, April 19-22, 1983. Vancouver, Washington. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 THEORETICAL BASIS AND PERFORMANCE EVALUATION OF CURRENT SNOWMELT-RUNOFF SIMULATION MODELS T.W. Tesche* ~STRACT: Since the early 1970's. comput- er simulation of snowpack accumulation. metamorphism, and snowmelt has emerged as a viable tool in water resources manage- ment. Snowpack modeling has also been employed in such diverse fields as the description of snow avalanche and simula- tion of the fate of acidic materials transported over large distances in the atmosphere prior to accumulation in seasonal snow cover (i.e •• acid deposi- tion). In this paper we review represen- tative examples of empirical and numeri- cal snowpack simulation models that have been developed since the pioneering work of Leaf and Brink (1973a). Present day snowpack simulation models are examined on the basis of their treatment of physical processes, operational features, data requirements, and results of past ~rification studies. Critical areas of. model development, data acquisition, calibration, and model evaluation are discussed. (KEY TERMS: snowmelt. modeling. perfor- mance evaluation, runoff) INTRODUCTION Hydrologic simulation models have developed over the last two decades into sophisticated, widely accepted tools used regularly in water resources management studies worldwide. Whether for flood prediction. reservoir operations. water supply forecasting or for studying water quality issues, these models have provid- ed engineers and scientists with a powerful means with which to study hydrologic processes. Watershed models, designed to predict or forecast watershed response to precipitation (solid or liquid), have been formulated using a variety of mathematical representations of the fundamental conservation relations for mass and energy. In the design and implementation of most watershed models. principal concern has been in relating basic precipitation to runoff. Simula- tion of the physical processes affecting snow accumulation, snow metamorphism. and the thermodynamic processes of snowmelt have been accorded secondary emphasis. In this paper. however, the focus is directed mainly toward models of the processes governing snowpack melt and the nydrologic processes leading to water delivery from a basin. The following section outlines the basic conceptual approaches that have been taken in developing snowmelt-runoff simulation models. Included is a survey of several representative snowmelt models exemplifying the different model types. While not exhaustive, the codes included in the review represent the range avail- able in sophistication and diversity of application. MODEL CONCEPTS Development of a simulation model of snowmelt-runoff requires (1) identifying the relevant physical processes and (2) determining the most appropriate mathe- matical representation for all or a subset of these phenomena. Radian Corporation, 10395 Old Placerville Road, Sacramento, California 95827 449 Basic Physical Processes Important physical processes in snowmelt modeling include snow accumula- tion on the ground, snow metamorphism, snowmelt, and runoff. Excellent summa- ries of the state of knowledge in each of these disciplines is available. Male (1980), for example provides a detailed overview of the processes of snow accumu- lation. Comprehensive treatments of theoretical and experimental aspects of snow metamorphism are presented by Colbeck (1982, 1983), Mellor (1964), Sommerfeld and LaChapelle (1970), and Yen (1979). Anderson (1968), Bengtsson (1982), Fohn (1973), and Colbeck (1973, 1978) present snow energy budgets and authoritative discussions on various aspects of snow melt processes. Finally, the collection of papers presented at the 1978 conference on Modeling of Snow Cover Runoff (Colbeck and Ray, 1978) amply summarize the physical processes govern- ing runoff (e.g., surface runoff, infil- tration, evapotranspiration and subsur- face flow). While the thermodynamics and microphysics of snow on the ground have been studied for decades and are now fairly well understood (Perla and Martinelli, 1976), the incorporation of this knowledge into operational models for estimating snowmelt discharge remains highly empirical. Types of Snowmelt Models Watershed simulation models are often classified according to one of three different categories transfer function, lumped parameter, or distribut- ed parameter -depending upon the degree of temporal and spatial averaging of the hydraulic processes governing runoff (Shafer and Skaggs, 1983). An alterna- tive scheme, adopted in this analysis, is based upon the degree of predictability assumed to exist in the system. Statis- tical models are predicated upon the assumption that the basic physical cause-effect relationships are somehow contained in the empirical data used to derive these models. These empirical models are based upon indices or correla- tions developed from observations. In 450 contrast, deterministic models are founded upon mathematical forms of the fundamental conservation relations governing mass, momentum, and energy. Through various spatial and temporal averaging procedures, the governing equations are reduced to a set that may be solved analytically, graphically, or numerically depending upon the particular model's formulation. The advantages and limitations of these alternative conceptual frameworks are well known. Based upon observational data, statistical models are attractive because they directly link routinely measurable quantities (e.g, the maximum or mean daily air temperature) with desired output information (the daily snowmelt). Among their limitations, the statistical models are poorly suited to the characterization of extreme hydrometeorological events. Deterministic models, while conceptually more appealing since they explicitly treat physically-based cause-effect relationships, are more complicated to use and require significantly more data for model operation. In the following section, we review a number of contemporary snowmelt-runoff models. Included are both statistical models (based upon indices and correla- tions) and deterministic models, formu- lated on the basis of either water or energy budgets (or both). SURVEY OF SNOWMELT-RUNOFF SIMULATION MODELS UC Davis Model (UCDM) Merritt (1978) developed a deterministic snow accumulation and ablation model based upon the heat balance equation approach first presented by the U.S. Army Corps of Engineers, (1956). The snowpack accumulation and ablation algorithm in the UCDM was taken from the Amorocho and Espildora (1966) model (a lumped parameter model applica- ble to a single hydrologic unit) • An energy budget was written for net longwave radiation, absorbed solar radiation, convective heat transfer. latent heat of vaporization, sensible heat change. rainwater heat content. ground heat conduction. and cold content of the pack. The model consists of three layers -an active upper layer. and upper and lower passive layers. Input require- ments of UCDM consist of initial snow depth and water equivalent, continuous air temperature. dewpoint temperature. wind speed and long-and short-wave radiation estimates. The model predicts daily basin runoff, snow depths and water equivalent in the hydrologic unit throughout the spring meltout period. Hydrologic Engineering Center (HEC-1) The HEC-1 is a deterministic, lumped parameter watershed model which includes a snowmelt-runoff calculation. The model is intended to simulate individual storms or events (i.e.. runoff from one sus- tained event). Snowmelt is computed from empirical equations governing periods with or without rain. For example. the daily snowmelt rate is written as a linear function of rainfall intensity. air temperature. surface layer wind speed, and an empirical constant to reflect the "exposure" of the snowcovered area to wind. The snowmelt calculation is thus based upon the degree-day or energy budget method. HEC-1 utilizes the unit hydrograph approach for flow genera- tion and any one of several routing equations are available. One of the limitations of the model is the inability to treat temporal and spatial variations in precipitation over the hydrologic basin. Ontario Ministry of the Environment (OME) Logan (1976) presented a lumped parameter deterministic snowmelt model for predicting melt in a single hydrolog- ic unit. The snowmelt computation is based upon an energy balance method. Energy and mass transfer mechanisms are simulated at the snow-ground and snow-atmosphere interfaces and across the average pack thickness. Although the model equations represent the energy balance and melting process at a point. numerical interpolation/ extrapolation procedures can be used to extend the spatial coverage of the model to a drain- age basin. Principal model out1,1ut 451 consists of ground surface free water production. Snowmelt Runoff Model (SRM) The Snowmelt-Runoff Model (SRM) is a deterministic. lumped parameter model formulated on an energy-balance concept. It utilizes the degree-day temperature index as the basic independent variable. In use for nearly a decade (see, for example Martinec, 1975; Martinec, et al •• 1983; Rango. 1980; Rango and Martinec, 1981), this model is intended for use under conditions where snowmelt is the major contributor to runoff in mountain basins of significant relief. A given day's runoff is estimated from the sum of the recession flow and a simple empirical relationship involving a runoff coeffi- cient, a degree-day factor. the number of degree-days. snow coverage and number of days. Although the model equation pertains. strictly speaking. to a specif- ic point. the calculation may be general- ized to cover a much broader range of geographic conditions through the use of detailed ground truth or remote sensing data. The model is capable of supplying forecasts from one day to several weeks depending upon the accuracy and range of the temperature forecasts. When the model is used in the forecast mode, it can be "nudged" or "updated" with measured discharge data at intervals throughout the simulation to improve the prediction. Tangborn Tangborn (1977. 1980) presented a lumped parameter snowmelt model based on the water-budget approach. The model is designed for short term predictions (1-15 days) although it may be used for the entire snowmelt season. The model, based upon an empirical treatment of snow storage and ablation, assumes that snowmelt-runoff may be determined from: • Calculated amount of water stored as snow in the basin • Estimated snow ablation just prior to the forecast period • Temperature and precipitation predictions for the forecast period. Model inputs include daily stream dis- charge, gaged precipitation, diurnal air temperature maxima and minima. From these synoptic inputs, the model. once calibrated using regression analysis, estimates the timing and volume of basin runoff. Possible refinements to the model include treatment of the storage and movement of liquid water in the pack, and separating the basin into distinct hydrologic subunits. Beven and Dunne The deterministic snowmelt-runoff model developed by Beven and Dunne (1982) is a distributed parameter model which focuses on the physics of meltwater flow through the pack. Mass transport in the slope parallel direction is treated in one dimension; in the slope normal direction, unsaturated flow to and into the soil mantle is simulated. Realistic hydrologic units are modeled by providing for a variable slope width. In princi- pal, the model can accommodate the three dimensional structure of a snow field over simplified topography. BURP2 The BURP2 model, developed by the U.S. Forest service (USFS, 1968) is a deterministic watershed/snowmelt model supplying continuous discharge forecasts on a daily basis for large scale forested watersheds. A distributed parameter model, BURP2 computes snowmelt through the use of a degree-day factor; the water balance is performed on a hydrologic basin comprised of several subunits of varying size. The model accounts for snowpack slope and aspect through a temperature adjustment factor. The model has several algorithms for calculating evapotranspiration and calibration is generally unnecessary. Routing of excess rainfall and infiltration processes are not treated in BURP2; the model predicts total water yield from a unit of land, but does not differentiate between surface runoff, streamflow, or ground water recharge. 452 MELTMOD Leaf and Brink (1973a,b) developed a deterministic, lumped parameter snowmelt model for applications to subalpine watersheds in Colorado. MELTMOD is based on an energy balance approach, simulating winter snow accumulation, heat energy balance, snow pack conditions in terms of energy content and free water state. and resultant snowmelt (temporally and spatially resolved), for a variety of terrain aspects, slopes, elevations, and vegetative characteristics. MELTMOD treats the snowpact as a dynamic heat reservoir. Heat is added algebraically during precipitation. Only when the pack becomes isothermal and saturated does melt occur. Unsteady heat transfer theory is used in the energy budget computation. The daily minimum tempera- ture is used as an index in segregating precipitation into snow, rain, or both. The model was designed primarily for mountain conditions where there are distinct accumulation and melt phases during the winter; the model has no prov~s~ons for a snowpack ripening and subsequently reverting back to a colder unsaturated pack. SNOWMELT SNOWMELT is basically a modified version of the MELTMOD code. Developed by Solomon et al, (1975, 1976) • SNOWMELT simulates winter snow accumulation. an energy balance, snowpack conditions, and resultant snowmelt. Its principal improvements over the MELTMOD model are that treats alternate freezing and thawing patterns during snow accumulation and melt phases. it does not tend to keep the snow pack excessively cold during the snow accumulation phase, and it has reduced input data requirements (e.g •• daily solar radiation). In addition, SNOWMELT may be applied in situations involving a discontinuous snowpack. The model is dependent upon four daily input variables: maximum and minimum tempera- tures, precipitation, and shortwave radiation. Only limited knowledge of local watershed and snowpack structure is needed for model initialization. Judson, Leaf and Brink As part of a long term research effort into avalanche prediction and hazard evaluation, Judson, Leaf. and Brink (1980) extended the capabilities vf the MELTMOD snowmelt model to include more details of snowpack physics. Specifically, the MELTMOD code was extended to simulate layer depth, age, densification, temperature gradient and equitemperature metamorphism, melt-freeze metamorphism, snow pack accumulation, and wind loading and redistribution. The output of this deterministic model includes, in addition to melt rates (if dictated by the conditions). snow pack stratigraphy, and weighted average loading rates from the previous one to three days. The model treats the vertical variation in snow pack water and energy exchanges in detail and can accommodate changing topography in the downslope direction. This model has the potential for routine application as an operational snowmelt-runoff model but the sophisticated treatment of the loading and metamorphic processes associated with snow slab instability (Judson and King, 1985), would require simplification in order to yield reasonable computing costs. Weissman Weissman (1977) developed a model of the melting of a snow pack due to warm air advection. Assuming a ripe, saturat- ed pack, this model calculates the heat and momentum fluxes in the thermal mternal boundary layer over a horizontal snowpack. Principal model output is the total average melt over the pack and the variation in melt as a function of downwind distance from the upstream edge of the snow field. Conventional mixing length theory is used to model the turbulent momentum and energy fluxes in the atmospheric boundary layer over the snow surface. Akan Recently, Akan (1984) formulated a time-dependent. distributed parameter 'mathematical model of snowmelt runoff based upon a numerical finite difference 453 solution of the diffusion equation. The model concept includes two physical layers resting on an inclined slope. In the upper unsaturated zone, the processes of heat and mass exchange, phase change, percolation, conduction, and vapor diffusion are simulated. In the saturat- ed zone, basal flow is modeled. The unsaturated zone is assumed incompress- ible and the melt layer is assumed not to settle. Detailed empirical treatment of microphysical processes are included in the model for the coefficients of hydrau- lic conductivity, thermal conductivity, and vapor diffusion in snow; variations in snow porosity and grain size due to phase changes; and constitutive relations for liquid water pressure, snow temperature, and water vapor content. The model is two dimensional (i.e., variation in the slope normal and parallel directions) and predicts melt and snowmelt runoff on 30 to 60 minute intervals. Jordan (1983) Jordan (1983) developed a distribut- ed parameter, deterministic model to study the physics of snow melt and vertical mass transport in deep, mature mountain snowpacks. The model is a one-dimensional, unsaturated, transient code which solves the diffusion equation in a homogeneous, rigid, porous medium. The principal objective of this model is the study of the propagation of meltwater flux waves down through the pack as a function of time. CALIBRATION AND EVALUATION In the literature the terms "model validity," "model evaluation," and "model verification, 11 are frequently inter- changed, at times obscuring the true nature of the comparison between predic- tion and observation. Here we provide the following definitions: 1. Model Calibration -the adjust- ment of empirical model constants or parameters in order to optimize the agreement between prediction and 2. 3. observation, with limited regard to the physical meaning or reasonableness of the value(s) ultimately chosen. Model Validity -the degree of agreement between model predic- tions and corresponding obser- vations, given perfect specifi- cation of model inputs. This concept refers to the scientif- ic soundness and implementation of the model's formulation. Model Evaluation -the process of examining and quantifying the performance of a snowmelt-runoff model. Use of the terms "validation" and "verification" to refer to this process should be avoided. 4. Model Verification the successful evaluation of the model; that is the imposed standards c·f performance have been met. In considering present snowmelt-runoff models, we are primarily interested in model calibration efforts and in subse- quent model evaluation results. Valida- tion of current generation models is unlikely due to the typical time and space averaging (lumping) that underlies most model's formulations. Verification of models is also unlikely at present since few if any standards of model performance have been established against which to judge quantitatively the accept- ability of a model for usage. A wide variety of snowmelt-runoff model evaluation studies are reported in the literature and most of them include some form of calibration. Solomon et al (1976) present evaluation results in three subjective categories good, average, and poor correlation between predicted and measured water equivalents. They note the difficulty in developing ;::tatistical ''goodness of fit" measures witi serially dependent data yet declare the model to perform "satisfactory." Price and Dunne (1976) compared predicted and measured daily runoff results from a .subarctic study plot us:ing a physically-based energy balance model. 454 The ratio of observed melt rates to predicted rates were examined as a function of time and the two rates were compared through linear regression. Based upon an r value of 0.85 and a sigma of 0. 71 em/ day. they conclude that the model's performance is "quite satisfacto- ry and within the ~urrer.tly accepted limits in snowmeLt hydrology." Baker and Carder (1977) compared the BURP, ECOWAT, SNOWMELT, and USDAHL-85 models. The models were first calibrated by varying threshold temperatures, temperature and precipitation "factors, 11 degree-day coefficients, and snowfall temperatures. Time series of model predictions of daily snowmelt were compared with data from the White Moun- tains snow course for 1966-1969. Only very limited data comparisons are pre- sented. Haverly, Wolford and Brooks (1978) present a similar, but slightly more rigorous evaluation study of the HEC-1, USAGE, and SNOWMELT models with data from northwestern Minnesota, Predicted and observed snowmelts were compared statis- tically for two seasons of data on open and forested plots. R squared values ranged between 0. 489 to 0. 924 prompting the researchers to conclude that the models were of comparable accuracy. In addition, the HEC-1 model was recommended for operational streamflow forecasting because it "requires little calibration time and fewer data." More recently, Gottlieb (1980) compared calculated and observed daily snowmelt discharges from the Peyto Glacier Basin using a detailed energy budget model. The r squared (explained variance) for the basin was computed for each of eight years based upon computed and observed daily runoff. The r squared values ranged between 0.56 and 0.83. Time series plots of daily calculated and observed discharge were also presented and were declared to be quite satisfacto- ry. Rango (1980) modeled snowmelt in the Wind River Range of Wyoming with SRM and matched seasonal runoff volumes with the model to within 5 percent. 82-86 percent of the daily variance in snowmelt was explained by model prediction, using "goodness-of-fit" statistics. An evaluation of the Bengtsson (1982) model of vertical snowmelt perco- lation through the pack was performed using data obtained in runoff experiments conducted over a 22 hour period. Hourly average runoff rates were compared with model predictions. The discrepancies in the hydrograph were attributed to either measurements errors or an incorrect ' energy balance. Rango (1985) also presents SRM model evaluation results for fourteen snow-covered bas ins in the U.S. and seven foreign countries. His statistical comparisons include absolute differences in calculated and observed seasonal discharge volumes, and coefficient of variation statistics based on daily flows. The model evaluation results are also stratified according to whether the snow-cover inputs are derived from visual or Landsat observations, snow course records. aircraft orthophotos. or NOAA satellite observations. The average variance across all basins simulated by s~. was 0.86 (0.73 to 0.96). Review of the snowmelt-runoff model evaluation literature reveals that qualitative comparisons of the computed and observed hydrographs is the basis for most summaries of model performance. ~ile statistical summaries of the hydrographs are important in assessing the overall accuracy of the model cali- bration and subsequent evaluation (on independent data sets). they by no means present the whole picture. Among the measures of performance that provide insight into the adequacy of a models performance are estimates of bias, error. accuracy of the peak discharge. total discharge volume, and temporal and spatial correlation in snowmelt-runoff rates. Furthermore. these statistical measures provide additional insight when calculated over daily. weekly. flow interval, and seasonal periods. Addi- tional insight into the performance characteristics of present snowmelt-runoff models can be developed through more rigorous statistical evalua- t~ons than have been reported to date. Current procedures for evaluating the predictive performance of air quality simulation models may be worth adopting (in suitably modified form) in the 455 development of more refined snowmelt model calibration and evaluation methods. SUMMARY In the dozen or so years since introduction of the distributed parame- ter. deterministic snowmelt-runoff models of Leaf and Brink (1973a,b) and Solomon et al (1975, 1976) few model developers have attempted to extend these simula- tion techniques to new models of fer ing greater temporal and spatial resolution in treatment of the governing physical processes. In other fields in the physical and earth sciences (i.e •• atmospheric chemistry and transport. ground water flow and contaminant disper- sion, surface water quality) the last decade has been one of intensive develop- ment of progressively more sophisticated mathematical modeling and simulation techniques. fostered largely by advances in computer technology. One reason, perhaps. for the less advanced state of development in snowmelt simulation models is the lack of a lead governmental agency in the U.S. to coordinate snow science research. Because studies of seasonal snow cover embrace so many geophysical disciplines the physical. earth, chemical, hydrological and atmospheric sciences -the problems associated with managing research in this field are apparently outside the scope of any one agency or department. Moreover, prospects for centralization of seasonal snowpack research within the government to not appear bright, especially in light of the growing trend of governmental agencies to deemphasize certain areas of snow hydrology research. The closing of the U.S. Forest Service Avalanche Re- search Station in Fort Collins is one recent example. Nonetheless. research and applica- tions in the study of seasonal snow cover will continue. The following areas continue to provide significant research opportunities: Snow Distribution and Accumulation Models Research opportunities exist in developing improved descriptions of the processes governing snow formation, transport, deposition, and redistribution by wind on the ground surface. Terrain and vegetative cover inhomogeneities play an important role in the local snowfall accumulation and redistribution patterns (Thomsen, 1980). Refinement of atmo- spheric transport models (for complex terrain). improved models of orographic snowfall. better descriptions of the effects of local topography and vegeta- tion on snow distribution, and improved measurement methods of snowpack proper- ties are all areas where refinements to snowmelt-runoff models and their inputs can be achieved. Snow Pack Energy Budgets Few previous studies have investi- gated in detail the several energy transfer processes simultaneously occur- ring within the snowpack at a given time and how they relate to mesoscale atmo- spheric circulations. Due to the large costs of data acquisition, it is impor- tant that field experiments be designed and study sites selected so as to maxi- mize the generality of point measurements to areas of larger scale. In particular, in-situ measurements of radiant, sensi- ble, latent, and geothermal heat fluxes and snowpack heat content properties need to be performed in such a way that they (1) support detailed analysis of evaporation, sublimation, condensation, sensible heat transfer, (2) permit the correlation between microscale energy transfers and mesoscale atmospheric circulations, and (3) aid in the investi- gation of concurrent snow metamorphoric changes. Snow Metamorphism The microphysics of snow metamorphism have been studied in detail in the laboratory and to a lesser extent in the field, the latter associated mainly with snowpack stability and avalanche release (Perla and Martinelli, 1976). There is a pressing need to extend the current knowledge of the metamorphic processes which occur at a point in the pack (and under idealized conditions) to the more general metamorphic changes which occur on 456 realistic, non-uniform packs of differing slopes, aspects, vegetative cover, etc. The processes of temperature gradient, equitemperature gradient, melt-freeze, and radiation-recrystallization metamorphism need to be formulated into mathematical models which can be operated with relatively simple climatic data inputs (either from routine measurements or supplied by numerical weather predic- tion or mesoscale meteorological models). Physical Properties of Snow The physical properties of snow are reasonable well understood, notwith- standing the extremely broad range of shapes and densities the substance may assume. However, improvements in present capabilities for in-situ and remote sensing (e.g., Rango et al 1979), of seasonal snowcover's physical properties remain one of the principal challenges in snow hydrology research. Needed are improved techniques for determining the aerial extent, depth, density, water content, and stratigraphy of the snowpack. These properties are directly related to improved knowledge of snowmelt processes. Related information on snowpack spectral reflectivity, infrared emissions, radiation scattering proper- ties, microwave emissions, and dielectric properties may also prove valuable. Physical Properties During Melt The thermodynamics of snowmelt have been studies intensively in the laborato- ry for early two decades. Advances in this area of snow microphysics need to be incorporated into the mathematical formulation of snowmelt, meltwater percolation, and runoff processes. Improved model formulation of the meltwater thermodynamics and percolation processes will require complete snow-profile descriptions including pack depth, density. freewater content, grain shape, grain size, and temperature in each layer of the pack. Complexity in Model Formulation The rapidly expanding capabilities of todays supercomputers has stimulated in some a tendency to promote signifi- cantly expanded complexity in model detail (e.g.. temporal and spatial resolution, number of physiochemical processes considered). This tendency should be avoided in those situations where added model complexity is obtained at the expense of physical insight into the system under study. Complex simula- . tion models play an important role in snow hydrology; an obvious example is in the estimation of and design for extreme snowmelt-runoff events, for which the more lumped-parameter empirically based techniques are poorly equipped. However, judicious selection is needed in obtain- ing the proper balance between mathemati- cal complexity of the simulated system and the conceptual understanding of the relevant physical processes of greatest importance in a given application. LITERATURE CITED Akan, A.O.. 1984, From Snow-Covered Resources Research, 707-713. "Simulation of Runoff Hillslopes," Water Vol. 20. No. 6 pp. Amorocho, J. and B. Espildora 1966, "Mathematical Simulation of the Snow Melting Process," Water Science and Engineering Paper No. 2001, University of California, Davis. Anderson, E.A. 1968, "Development and Testing of Snowpack Energy Balance Equa- tions," Water Resources Research, Vol. 4, PP• 19-37. Baker, M.B. and D.R. Carder, 1977. "Comparative Evaluations of Four Snow Models." proceedings 45th Annual Western Snow Conference. Beven, K. and T. Dunne, 1982, "Modeling the Effect of Runoff Processes on Snowmelt Hydrographs," International Symposium on Rainfall-Runoff Modeling. May 18-21, Mississippi State University, Mississippi State, MS. Bengtsson, L, 1982, "Percolation of Meltwater Through a Snowpack, 11 Cold Regions Science and Technology. Vol. 6. 'PP• 73-81. 457 Colbeck, S.C.. 1973, "Theory of Metamor- phism of Wet Snow," Res. Rep. 313, U.S. Army Cold Region Res. and Engr. Lab., Hanover, N.H. Colbeck. S.C.. 1978, "Physical Aspects of Water Flow Through Snow," Advances in Hydroscience II, Academic Press, New York. NY • Colbeck, S.C.. 1982, "An Overview of Seasonal Snow Metamorphism," Reviews of Geophysics and Space Physics, Vol. 20, No. 1, PP• 45-71. Colbeck, S.C. Metamorphism of Dry physical Research, 5475-82. 1983, "Theory of Snow," Journal of Geo- Vol. 88, No. C9, PP• Colbeck, S.C. and M. Ray 1978, "Modeling of Snow Cover Runoff." proceedings of a meeting on Modeling of Snow Cover Runoff. 26-28 September 1878, CRREL, Hanover, N.A., America-Geophysical Union and Americal Meteorological Society. Fohn, P.M.B •• 1973, "Short-Term Snow Melt and Ablation Derived From Heat-and Mass-Balance Measurements," Journal of Glaciology, Vol. 12, No. 65. Gottlieb, L.. 1980, "Development and Applications of a Runoff Model for Snow- covered and Glacierized Basins," Nordic Hydrology. Vol, 11, pp. 255-272. Haverly, B. A.. R.A. Wolford, and K.N. Brooks 197 8. "A Com par is on of Three Snowmelt Prediction Models," proceedings 46th annual Western Snow Conference. Jordan, P •• 1983, "Meltwater Movement in a Deep Snowpack 2. Simulation Model," Water Resources Research, Vol. 19. No. 4, pp. 979-985. Judson, A •• C. F. Leaf, and G.E. Brink, 1980. "A Process-Oriented Model for Simulating Avalanche Danger," Journal of Glaciology, Vol, 26, No. 94. pp. 53-63. Judson, A.. and R. King. 1985, "An Index of Regional Snow-Pack Stability Based on Natural Slab Avalanches," Journal of Glaciology. Vol. 31, No. 108. Leaf, C.F. and G.E. Brink, 1973a, "Computer Simulation of Snowmelt Within a Colorado Subalpine Watershed," USDA Forest Service, RM-99, Fort Collins, CO. Leaf, C.F. and G.E. Brink, 1973b, "Hydrologic Simulation Model of Colorado Subalpine Forest," USDA Forest Service, RM-103, Fort Collins, CO. Logan, L.A. 1976, "A Computer-Aided Snowmelt Model for Augmenting Winter Streamflow Simulation in a Southern Ontario Drainage Basin," Canadian Journal of Civil Engineering, Vol. 3, pp. 531-554. Male, D.H., 1980, "The Seasonal Snow- cover, 11 in .:::D:..~:y..!n::::am=..:i::..::c::.:s=---::::.o::.._f--=S:.::n:..:o:.::w:___;a=.:n=d=---=I-=c=e Masses, Academic Press. Martinec, J. 1975, "Snowmelt-Runoff Model for Stream Flow Forecasts," Nordic Hydrology, Vol. 6, No. 3, pp. 145-154. Martinec, J., A. Rango, and E. Major, 1983, "The Snowmelt-Runoff Model (SRM) User's Manual," NASA reference Publication No. 110. Goddard Space Flight Center. Greenbelt, MD. Mellor. M. 1964, "Properties of Snow," Cold Region Science and Engineering. Part III-A1, Hanover, NH. Merritt, R.G.. 1978, "Digital Simulation of Snowmelt Runoff," University of Nevada. Desert Research Institute. PB 293252. Reno, NV. Perla. R.I.. and M. Martinelli, 1976. "Avalanche Handbook." Rocky Mountain Forest and Range Experiment Station, Fort Collins, CO. Price, A.G. and T. Dunne. 1976, "Energy Balance Computations of Snowmelt in a Subartic Area. 11 Water Resources Research. Vol. 12. No. 4. Rango. A. 1980, "Remote Sensing of Snow Covered Area for Runoff Modeling. 11 pro- ceedings of the Oxford Symposium. IASH- AISH Publ. No. 129 •• PP• 291-297. Rango. A. and J. Martinec, 1981. "Accuracy of Snowmelt Runoff Simulation," Nordic Hydrology. Vol. 12, pp. 265-274. 458 Rango, A •• A.T.C. Chang, and J .L. Foster, 1979, "The Utilization of Space Microwave Radiometers for Monitoring Snowpack Properties," Nordic Hydrology, Vol. 10, pp. 25-40. Rango, A., 1985, "Assessment of Remote Sensing Input to Hydrologic Models," Water Resources Bulletin, Vol. 21. No. 3. Shafer. J .M. and R.L. Skaggs, 1983, "Identification and Characterization of Watershed Models for Evaluation of Impacts of Climate Change on Hydrology," Battelle Pacific Northwest Laboratories, PNL-SA- 11924. Solomon. R.M. et al, 1975, "Characteri- zation of Snowmelt Runoff Efficiencies," Watershed Management Symposium ASCE Irrigation Drainage Division Proc. 1975, pp. 306-326. Solomon, R.M. et al, 1976, "Computer Simulation of Snowmelt," USDA Forest Service. RM-174, Fort Collins, CO. Sommerfeld R. H., and E. R. LaChappele, 1970, "The Classification of Snow Metamorphism." Journal of Glaciology, Vol. 9, No. 55, pp. 3-17. Tangborn, W. V. 1977. "Application of a New Hydrometerological Streamflow Prediction Model," proceedings 45th Annual Western Snow Conference. Tangborn. W.V. 1980, "A Model to Forecast Short-Term Snowmelt Runoff Using Synoptic Observations of Streamflow. Temperature, and Precipitation," Water Resources Research, Vol. 16, No. 4, pp. 778-786. Thomsen. A. G.. 1980, "Spatial Simulation of Snow Processes," Nordic Hydrology, Vol. 11, pp. 273-284. U.S. Forest Service, 1968, "BURP: A Water Balance Program," Watershed Systems Development Unit, Berkeley, CA. U.S. Army Corps of Engineers. 1956, "Snow Hydrology." North Pacific Division, Portland. OR. Weisman. R.N. Two-Dimensional 1977. "Snowmelt: A Turbulent Diffusion Model." Water Resources Research. Vol. 13. No. 2. pp. 337-342. Yen. Y. C. 1979. "Recent Properties. 11 Advances Vol. 5, pp. 173-214. Studies in Snow in Hydroscience. 459 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 RECENT DEVELOPMENTS IN SNOWMELT-RUNOFF SIMULATION S B t .. 1 ten ergs rom ABSTRACT: Some tendencies in recent de- velopment of snowmelt simulation models are reviewed with special emphasis on the problem of model complexity versus data availability and model performance. The possibility to use new sources of areal snow water equivalent information is dis- cussed, and the need for reliable cost- benefit analyses is stressed. Finally a few important research needs for improved snowmelt simulation models are identified. (KEY TERMS: Snowmelt simulation; con- ceptual model; basin snow water equiva- lent.) INTRODUCTION Snowmelt-runoff simulation models are major components of hydrological forecast- ing systems and water management schemes in many parts of the world today. Their direct economic importance is maybe eas- iest to quantify in connection with stream- flow forecasting for reservoir operation, where damage can be avoided, and water can be put into productive use instead of leaving the reservoirs through the spill- ways. For design purposes, the models have a great value, as they are used to create longer runoff records and thereby more reliable runoff statistics in areas, where long climatological records are avail- able. Recently, interest in the use of runoff models for the interpretation of hydro- chemical records of flowing waters has grown, because these records show a wide natural variation, that strongly depends on the hydrological situation. For example, the analysis of pH or alkalinity can be very confusing, unless high and low flow situations are treated differently. Along with the development of handy desk-top computer systems and more cus- tomer-oriented soft-ware, conceptual run- off models are becoming more generally available for all categories of users. This has made it possible for local offices to run operational models independently, a fact, which is likely to have strong im- pact on the future use of conceptual models. CONCEPTUAL SNOWMELT MODELS A physically correct way to approach the snow modelling problem would be from the total heat budget of the snowpack, as discussed by, for example, U. S. Corps of Engineers (1956) and (1960), Kuz'min (1972) and Morris (1985). The general equation would be of the form: W +W1 +W +W1+W +W +Wt+W = 0 (1) sw w c g p m where: w g w p w m = absorbed short wave radiation, net long wave radiation, = convective heat flux, = latent heat flux (condensation and evaporation) , = heat flux from the ground, = contribution of heat from precipi- tation, = change in the energy content of the snowpack, = heat equivalent of the snowmelt. 1Swedish Meteorological and Hydrological Institute, S-601 76 Norrkoping, Sweden 461 Many modellers, who have attempted to model eq. 1 both at a point and on a basin scale, have experienced that its data requirements are difficult to fulfil and that some of the physical processes are very complex. This has led to the development of a number of simplified con- ceptual snowmelt models, which treat the meteorological variables more as indices of the physical processes that govern snow- melt rather than input to exact energy balance computations. The best known of these index methods is the commonly used and long-lived degree-day approach ac- cording to the following expression: M = C (T - T ) 0 0 (2) where: M = snowmelt (mm), c = degree-day melt factor (mm/°C), 0 T = surface air temperature (°C), T = threshold value of the air tern- 0 perature (°C). Equation 2 is normally used on daily data or data with higher resolution. When applied to runoff models, it is often combined with various options in attempts to make it more physically cor- rect. These options vary a lot from one model to another, but the following may serve as examples: -Liquid water holding capacity, which delays runoff from melting snow. -Seasonal variation of the degree-day factor, which accounts for variations in in solution, albedo and thermal pro- perties of snow. -Separation of rainy and clear days to account for differences in the energy exchange processes. -Introduction of wind speed and/ or va- pour pressure of the air into the equa- tion to obtain a more realistic expres- sion for energy exchange. -Use of daily air temperature range as an index of radiation. 462 -Differentiation into forested and open areas to account for differences in energy exchange. -Computation of the thermal conditions in the snowpack and/or in the ground. -Introduction of a statistical distribution function in order to account for redis- tribution of snow on the ground. -Use of elevation zones in combination with temperature and precipitation lapse rates. OPTIMUM COMPLEXITY OF SNOWMELT MODELS Many modellers have tried to find the optimum complexity of snowmelt models by comparative test runs on well controlled data bases. Anderson (1976) found that an energy balance equation was superior to an index model in well defined open areas and under extreme conditions. His results are to some degree supported by Braun ( 1985) , whose results indicate that some improvements may be obtained by energy balance models in small basins in Switzerland. Contradictory to these results, Kuu- sisto (1978) partly failed to show im- provements when moving from a simple index model to a more energy-balance- like approach on snowpillow data from Finland. Vehvililinen ( 1985) also in Fin- land, and Lundquist (1981) in Norway came to a similar conclusion when apply- ing a conceptual runoff model with snow- melt routines of increasing complexity to small basins in their respective coun- tries. Between 1976 and 1983 the World Me- teorological Organization, WMO, per- formed a well controlled intercomparison of models of snowmelt runoff in order to examine their performances under vari- ous conditions. Altogether 11 models, varying in complexity from simple de- gree-day approaches with a water hold- ing capacity option to more complex energy balance computations, were tested on data from 6 basins with snow- melt as a significant hydrological com- ponent (Tables 1 and 2). Table 1. Models participating in the WMO Intercomparison of Models of Snowmelt Runoff. 1. UBC Watershed Model (Canada) (Quick and Pipes, 1972) 2. CEQUEAU Model (Canada) 3. ERM Model (Czechoslovakia) (Charbonneau, Fortin and Morin, 1977) (Turcan, 1981) 4. NAM-11 (Denmark) 5. TANK Model (Japan) 6. HBV Model (Sweden) (Nielsen and Hansen, 1973) (Sugawara et al., 1974) (Bergstrom, 197 6) 7. SRM Model (Switzerland) 8. IHDM Model (United Kingdom) (Rango and Martinec, 1979) (Morris, 1980) 9. SSARR Model (USA) 10. PRMS Model (USA) 11. NWSRFS Model (USA) (U.S. Corps of Engineers, 1972) (Leavsley and Striffler, 1978) (Anderson, 197 3) Table 2. Test basins in the WMO Intercomparison of Models of Snowmelt Runoff. 1. Durance (France) 2 170 2. W3-Watershed (USA) 8.4 3. Dunajec River (Poland) 680 4. Dischma Basin (Switzerland) 43.3 5. Illecilleweat River (Canada) 1 100 6. Kultsjon (Sweden) 1 110 The models were calibrated to each basin over a six years period, and four remaining years were used for verifica- tion. This produced a wealth of informa- tion, varying from various statistical cri- teria of agreement between the models and the observations to linear scale plots of the simulations and flow duration curves. At a final meeting in Norrkoping, Swe- den, in September, 1983, the modellers had the opportunity to compare their results, and many valuable conclusions were drawn. On the question of optimal model complexity the following statement was made: "On the basis of available information, it was not possible to rank the tested models or classes of models in order of performance. The complexity of the struc- ture of the models could not be related to the quality of the simulation results." (WMO, 1986) AREAL VARIABILITIES One reason why often only moderate success is achieved when increasing snow- • melt modelling complexity on a basin scale . is lack of detailed knowledge of the phys- . ical processes involved. Another reason, • maybe more important, is the areal vari- ability of snow conditions and of the eli- 2 km 2 km 2 km 2 km 2 km 2 km 463 matological variables, which serve as input to the models. The latter problem can be illustrated by the following simple case: In most operational applications in mountainous areas, the climate stations are situated in the lower parts of the basins, as illustrated by an example from Sweden in Figure 1. Altitude (m as ll 1500 70 0 50 100% Figure 1. Example of poor altitude repre- sentativeness of climatological stations. Hypsometric curve and loc!tion of stations in the 1 109 km Kultsjon basin in Northern Sweden. The data therefore have to be extra- polated to higher elevations by the use of lapse rates. For air temP,eratures a con- stant lapse rate of -0.6 °C/100 m eleva- tion is often used, sometimes with a sea- sonal variation, as suggested, for exam- ple by Konovalov (1979). In fact, this temperature lapse rate shows strong day- to-day variation, depending on the meteorological conditions. The values 0 normally stay around -0.6 C/100 m as an average at temperatures around freezing and higher. At lower air tem- peratures, however, inversions fre- quently will turn the temperature lapse rate positive, as shown by an example in Figure 2. •• T lapse (0 C/100 m) ' +1.5 • • • • • +1.0 • • • • +0.5 Air temp. • at Alvdalen (OC) -10 -5 +5 +10 -0.5 • • .. • Figure 2. Example of the dependence of the temperature lapse rate on air temperature in Sweden. Data from the stations Krackelbacken at 670 m.a.s.l. and .Alvdalen at 250 m.a.s.l. The consequence of this is that even a very modest attempt to model the ther- mal properties of the snowpack or the ground will be physically obscure and is likely to fail at high elevations, unless a more realistic temperature lapse rate is adopted. Models without routines for thermal properties will suffer less from this problem and are therefore likely to be more physically correct in spite of their simplicity. 464 The problem of areal variability of the snowpack is particularly pronounced above the timber line, which shows up clearly, if snow water equivalent is plotted against elevation as in Figure 3. Water equivalent mm 700 600 I + 500 I I + + -tlr 400 I+ + + ~+ + + ++ 300 t+ 4t,. + 200 .,I c:l ++ i:l 100 ~I El 0 :;::I 600 700 800 900 1000 1100 masl Figure 3. Plotting of snow water equiva- lent against elevation in the Kultsjon basin in Northern Swe- den. The snow water equiva- lents are measured by gamma- ray technique (Bergstrom and Brandt, 1984). It is the redistribution of the snow on the ground that causes the great variatioml above timber line. Figure 3 illustrates that. a constant precipitation lapse rate can only be recommended in forested areas. This has1 for example, led Killingtveit and Aam (197!) to modify the HBV Model when applying it 1 to high elevation areas in No~way, by the~ troduction of a statistically distributed snow routine, as shown in Figure 4. The authors reported significant improvements by this modification, which resembles the areal depletion curve used by, for example, Anderson (1973). It is possible that we have to accept the above statistical approach in areas with very complex snow accumulation patterns, like in wind-blown rugged ter- rain. Fortunately, as shown by Rilling- tveit and Aam (1978), the distribution pattern is very consistent from one year to another, which simplifies the modelling approach. On the other hand, the great variability in these terrains makes it diffi- cult to justify melt equations that are too complex. ~ .... Qj +-" +-c I'OQI ~-, ~.::: O:J co- l/)Qj To ~ 0 As. so I ~ Snowmelt during time 6 T As = Snowcovered area ............... 100% Figure 4. Principles of the statistically distributed snow routine according to Killingtveit and Aam (1978). LOSSES When conceptual models are used for year-around simulation, the development of the snowpack is normally computed by an accumulation of all precipitation that falls at air temperatures below a given threshold value. Most models then use an empirically derived snowfall-correction factor to make the water balance match during the melt season. This correction factor, normally expressed as a percent- age of falling snow, accounts for all kinds of losses, lack of representative- ness of the precipitation gauge, poor estimates of the precipitation lapse rate and other systematic errors. The simple snowfall correction factor is thus the dominating model parameter for the fore- casts of total inflow to a reservoir dur- ing a snowmelt season. Considering the high value of cor- rect forecasts of the total inflow to a reservoir during snowmelt and the fact that these forecasts often show errors, which are difficult to explain, it is sur- prising that the simple correction factor is accepted by so many modellers, while greater efforts are spent on better de- tailed snowmelt computations. One reason for the lack of interest in losses may be a general feeling that there is a physical limit to possible eva- poration from a snowpack due to its low surface temperature (see, for example, Bengtsson, 1980). If we regard a basin 465 at mid-melt as a mix of snowfields, patches and bare ground, this physical constraint is no longer valid. On a sun- ny day, evaporation from warm patches of bare ground can very well reach high values, in particular as these areas are fed with water from adjacent melting snow. Better understanding of the losses that may occur from the snowpack or patches of snow and bare ground during extreme weather conditions, is an im- portant future field for research. If this research is successful, it will pay off in very short time by improved long term forecasts. ·NEW SOURCES OF BASIN SNOW DATA The lack of detailed information about the a~eal vari.ability of the snowpack and th~ cllmatologJ.Cal variables is probably the mam reason why rather few successful modifications of the established operational snowmelt-runoff models can be found in re- cent literature. There seems to be a con- census among modellers that we need bet- ter input data before the models can be improved further. As a consequence, there is growing in- terest in new techniques for obtaining in- formation on the snow condition in the field. In Western United States this has resulted in a very extensive network of automated snow observation stations, SNOTEL, as described by Barton and Crook (1980). The interest in use of sat- ellite imagery has also grown, and some success in modifying traditional snowmelt forecasting methods have been reported by, for example, Rango (1985). Attempts with satellite updating of conceptual snow- melt models have also been reported from Denmark (Danish Hydraulic Institute, 1985). Two main drawbacks of satellites for updating of conceptual models are their incapability under cloudy conditions and the fact that, for the time being, they can only give us information about the areal extent of snowcover. We will prob- ably have to wait for more advanced microwave sensors, until we get what we really want: direct information about the snow water equivalent over the basin. But when these data are available on a rou- tine basin, it will be a major break-through for remote sensing in snow simulation. Attempts to measure snow by gamma- ray techniques have been made in the USA, in Canada, Norway, Finland, the USSR and Sweden (see, for example, WMO, 1979), and an impulse radar system has been successfully applied by Ulriksen (1982) in Sweden and Norway. These techniques can be operated from airplanes at low elevation (gamma tech- nique) or from helicopters (impulsive radar) and are therefore very expensive. On the other hand, they provide snow water equivalents more or less continuously along the flight lines and are therefore produc- ing a wealth of data, as compared to con- ventional snowcourses. In the future we can expect that weather-radar systems will be more gen- erally available, which will provide hy- drologists with direct areal precipitation information. Problems remain, like the bright-band effect of melting snowflakes and the penetration of the radar-beam through low clouds, but a wise combina- tion of ground-information and the radar image may give us the opportunity to model areal snow accumulation in a better way than with the small point measure- ments of today. The new sources of basin-wide snow- data provide us with more detailed infor- mation about the basin snowpack than is possible from point climate observatons. If these data are to be used to update conceptual models, they have to be devel- oped in a more distributed way while re- 466 mammg simple enough to accept the rather incomplete data coverage that is generally available throughout the year. This bal- ance between complexity and areal distribu- tion on one hand and simplicity and limited data demand on the other will always be important and requires both field experience and a feeling for phys- ical realities. THE NEED FOR COST-BENEFIT ANALYSES The introduction of new sources of areal snow data is often connected with relatively high costs both for development of new routines and for operation. It is therefore of great importance that their cost-effectiveness is investigated along with their potential for improved forecasts. The cost-benefit analysis has to show favourable figures in comparison with existing, less expensive methods before we can expect that the new techniques will be fully accepted. This is a difficult task considering the very low costs (in Sweden in the order of $ 500/forecast) for a forecast based on a conventional conceptual model and an existing national climatological network. As a matter of fact, the Swedish programme for gamma-ray measurements in the Kultsjon basin (Bergstrom and Brandt, 1985) has now been terminated because the cost for this technique did not justify the rela- tively small and sometimes inconsistent improvements that were obtained when using the technique to update a conven- tional conceptual model used in the area. Gaining confidence in the use of new techniques finally means that they have to be verified in a forecasting mode where all possible conventional updating has been made, rather than in a simulation mode, where large volume errors in spring, for example, may remain due to poor simula- tion of the beginning of the accumulation phase in fall. CONCLUSIONS There is little doubt about the benefits of research in the field of snowmelt-runoff simulation in a society that is more and more\ trying to control its water resources. Im- portant areas are reservoir operation for water supply, power production, flood con- trol, navigation and recreation, but simu- lation models are also important tools for 'design studies. The development of conceptual snowmelt- runoff simulation models now seems to have reached a point, where further development is hindered by lack of knowledge about the areal variability of the climatological vari- ables and of the snowpack itself. New tech- niques are under way, and this will demand models, which have a more realistic areal distribution than present lumped or semi- lumped models. These models must repre- sent a balance between sophistication and simplicity, as they will have to be run on the conventional climatological network dur- ing periods when basin-wide snow data are not available. One basic research need is associated with the areal variability of climatological variables within a basin. It is, for exam- ple, necessary to avoid crude constant lapse rates of air temperature and preci- pitation, if a more advanced energy and mass balance is to be computed at vari- ous elevations. Due to the complexity of basin-wide snow conditions, it may be fruitful to introduce a statistical approach into the computations, in particular above · the timber line. One important task is then to find the optimum combination of snowmelt equations for a point and the proper statistical distribution of the snowpack. Another basic research need is better mderstanding of the processes associ- lted with losses from the snowpack and ~djacent ground during the snow season. The future use of satellite-based sen- ;ors, which are capable of informing about the basin snow water equivalent jirectly, might very well become a break- through for remote sensing in snow simu- lation. The research field is on the tech- nical side, but hydrologists will have to provide adequate ground truth for the verification of this new technique and will also be responsible for the develop- ment of proper models. The introduction of the new data- bases and models requires careful cost- benefit analyses based on verified fore- casting results instead of just simulations. 467 ACKNOWLEDGEMENTS The paper summarizes experience from both research and operational applications of conceptual snowmelt models at SMHI. Financial support from the Swedish Asso- ciation of River Regulation Enterprises is greatly acknowledged. LITERATURE CITED Anderson, E.A., 1973. National Weather Service River Forecast System - Snow Accumulation and Ablation Model. NOAA Technical Memorandum NWS HYDR0-17, Silver Spring, M.D. Anderson, E.A., 1976. A Point Energy and Mass Balance Model of a Snow Cover. NOAA Technical Report NWS 19, Silver Spring, M.D. Barton, M., and Crook, A. G., 1980. Operational Experiences with the SNOTEL System. ASCE Annual Conven- tion, Holywood-by-the-Sea, Florida. Bengtsson, L., 1980. Evaporation from a Snow Cover. Nordic Hydrology 11:221- 234. Bergstrom, S., 1976. Development and Application of a Conceptual Runoff Model for Scandinavian Catchments. Swedish Meteorological and Hydrological Institute, RHO 7, Norrkoping, Sweden. Bergstrom, S., and Brandt, M., 1984. Sno- matning med flygburen gammaspektro- meter i Kultsjons avrinningsomrade. Swedish Meteorological and Hydrological Institute, HO 21, Norrkoping, Sweden. Bergstrom, S. , and Brandt, M. , 1985. Meas- urement of Areal Water Equivalent of Snow by Natural Gamma Radiation -Ex- periences from Northern Sweden. Hydro- logical Sciences Journal, No.4, Decem- ber. Braun, L.N., 1985. Simulation of Snowmelt- Runoff in Lowland and Lower Alpine Re- gions of Switzerland. Ziiricher Geo- graphische Schriften, Heft 21, Zurich. Charbonneau, R. , Fortin, J.P. , Morin, G. , 1977. The CEQUEAU Model: Description and Examples of its Use in Problems Related to Water Resources Management. Hydrological Sciences Bulletin, IAHS, 22(1). Danish Hydraulic Institute, 1985. A Hydro- logical Modelling System for Operational Use of Satellite Snowcover Observations. Internal Report, Horsholm, Denmark. Killingtveit, A., and Aam, S., 1978. En fordelt modell for snoackumulering og -avsmeltning. EFI-Institutt for Vass- bygging, NTH, Trondheim, Norway. Konovalov, V.G., 1979. Calculation and Forecast of Glacier Melt in Middle Asia. (In Russian.) Gidrometeoidat, Lenin- grad, USSR. Kuusisto, E., 1978. Optimal Complexity of a Point Snowmelt Model. Proceedings: Modelling of Snow Cover Runoff, Hanover, New Hampshire. U. S. Army, CRRL. Kuz'min. P.P., 1972. Melting of Snow Cover. U.S. Department of Commerce, National Technical Information Service. Translated from Russian by the Israel Program for Scientific Translations, Jerusalem. Leavesley, G. H. , and Striffler, W. D. , 1978. A Mountain Watershed Simula- tion Model. Proceedings of Meeting I Workshop on Modelling of Snow Cover Runoff, Hanover, New Hampshire, U.S. Army, CRREL. Lundquist, D., 1981. Snomodellstudier i Dyrdalen. Norwegian Hydrological Committee, Internal Report No. 5, Oslo, Norway. Morris, E. M. , 1980. Forecasting Flood Flows in Grassy and Forested Basins Using a Deterministic Distributed Mathematical Model. Proceedings of the Oxford Symposium on Hydro- logical Forecasting, April, 1980. IAHS-AISH Publ. No. 129. Morris, E.M., 1985. Snow and Ice. In: Hydrological Forecasting, edited by Anderson, M.G., and Bart, T. P. John Wiley and Sons, Chichester. Nielsen, S.D., and Hansen, E., 1973. Numerical Simulation of the Rainfall- Runoff Process on a Daily Basin. Nordic Hydrology, Vol. 4, No. 3. Quick, M.C., and Pipes, A., 1972. Daily and Seasonal Runoff Forecasting with a Water Budget Model. UNESCO/WMO, International Symposium on the Role of Snow and Ice in Hydrology, Measure- ment and Forecasting. Banff. Rango, A., 1985. Assessment of Remote Sensing Input to Hydrologic Models. Water Resources Bulletin, Vol. 21, No. 3. 468 Ran go, A. , and Martinec, J. , 197 9. App- lication of a Snowmelt-Runoff Model Using Landsat Data. Nordic Hydrology, Vol. 10, No. 4. Turcan, J., 1981. Empirical-Regressive Forecasting Runoff Model. Proceedings of Conference VUVH, Bratislava, 1981. Sugawara, M., Ozaki, E., Watanabe, I., and Katsuyama, Y., 1974. TANK- Model and its Application to Bird Creek, Wollenski Brook, Bihin River, Kitsu River, Sanaga River and Nam Mune. Research Notes of the National Center for Disaster Prevention, No. 11, Tokyo. Ulriksen. C. P. F., 1982. Application of Impulse Radar to Civil Engineering. Doctoral Thesis, Lund University of Technology, Department of Engineer- ing Geology, Lund, Sweden. U.S. Corps of Engineers, 1956. Snow Hydrology. North Pacific Division, Portland. U.S. Corps of Engineers, 1960. Runoff froin Snowmelt. EM 1110-2-1406, Wash- ington, D. C. U.S. Corps of Engineers, 1972. Program Description and Users Manual for SSARR Model. U.S. Army Engineering Division, North Pacific, Portland. VehviHiinen, B., 1985. Snomodellering och prognoser -Finska erfarenheter. Vanni i Norden No. 4, 1985. WMO, 1979. Workshop on Remote Sensing of Snow and Soil Moisture by Nuclear Techniques. Technical papers presenteQ at the workshop convened by WMO and organized in cooperation with IAHS and Norwegian National Committee for Hydro logy, Voss, Norway. WMO, 1986. Intercomparison of Conceptual Models of Snowmelt Runoff. Operational Hydrology Report No. 23, Geneve. (In press.) GLACIER HYDROLOGY 469 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 THE ROLE OF GLACIERIZED BASINS IN ALASKAN HYDROLOGY c. Benson, w. Harrison, J. Goslnk, s. Bowling, L. Meyo and D. Trabent 1 ABSTRACT: This paper summerlzes the results end recommendetlons of the lnternetlonel Workshop on Alasken Hydrology: Problems Releted to Gieclerlzed Basins, held In Eegle River, Alaske, during Aprl I, ~Hydrology In Alaska Is strongly Influenced by problems essoclated with snow, Ice end permafrost whl ch will become lncreas I ng ly Important as econom1 c development proceeds. The Scendlnavlen countries, Swl tzer I and end Caneda heve preceded AI aska In developing hydrological resources that depend on glaciers for their sources, and Alaskans can learn from their experience. The lnternatlonel workshop sought to capitalize on this expertise and expe- rience from abroed, and to help bridge the com- munication gap between scientific and engineer- Ing/management groups by having both groups partici- pate. The subjects considered were: runoff from glaclerlzed basins, sediments In glacial streams, hazards associated with glaciers, lee problems on rivers and reservoirs, and selected aspects of permafrost hydrology In glaclerlzed basins. Two days of the workshop were devoted to presentations by the participants, followed by Intensive meetings of five subgroups. This paper contains the conclu- sions and recommendations of the slbgroups. (KEY TERMS: hydrology, glaciers, sediments, river Ice, glaciological hazards, permafrost.) INTRODUCTION This paper summarizes the results and recommen- dations of an International workshop on Alaskan Hydrology: Problems Releted to Glaclerlzed Basins. The workshop wes held In Eegle River, Alaske, from B-12 Aprl I 1985. It was sponsored by: The ·u.s. ~tlonel Science Foundation (Grant DPP84-17549), the Governor of Alaska, the Division of Geological and Geophyslca I Surveys <DGGS) of the Alaske Department of Natura I Resources, the Alaska Power Authority, the u.s. Army Cold Regions Research and Engineering ~boretory (CRREL), the Geophyslcel Institute of the University of Alaske, and the VIce Chencellor for Research and Advanced Study, University of Alaska, Fairbanks. It wes ettended by 49 people: 13 from Scandlnavle end Europe, 5 from Canada, 9 from the u.s. and 22 from Alaska. Hydrology In Alaska Is strongly Influenced by a broad range of prob I ems associated with snow, Ice end permefrost. Since relatively little economic development has occurred In Alaska, few of these problems have been faced by hydrologists, manegers and plenners. It Is timely to address these prob- lems now, because Alaska Is on the verge of develop- Ing hydrological resources that depend on glaciers for their sources. Norway, Swltzerlend, Iceland, Canada and Green land (Denmark) have preceded Alaska In these ventures and Alaskans can learn from their experiences. There Is also a need for glaciological and lee management Information, which Is prlmarl ly In the scientific and baste engineering I lterature, to be made more aval table to planners, managers and engineers. The workshop was motivated by such problems. Its Intent was to capitalize on expertise and experience from abroad, and to help bridge the communication gap between scientific and engineer- Ing/management groups by the participation of both groups. We hope that this will help us In Alaska to utilize our vast, glacier-derived water resources more effectively and to mitigate the assocleted hezards. The first two days were devoted to presenta- tions by the participants; summaries of these presentations are Included as an appendix to the report on the workshop (Benson et e 1., 1986). The remainder of the week was devoted to Intensive meetings of five subgroups, Identified In the next section, except for he If e day spent on a chertered flight over h~avlly glaclerlzed regions of south central Aleska. Plenary sessions were held each dey to communicate progress of the subgroups to all participants. The status reports of the subgroups comprIse fIve chepters of the f Ina I report on tho workshop. Each chapter contains a set of recom- mendations and an extensive bibliography. 1c. Benson, w. Harrison, J. Goslnk and s. Bowling, Geophysical Institute, University of Aleska-Felrbanks, Fairbanks, Alaska 99775-0800; L. Mayo end D. Trabant, u.s. Geological Survey, Cold Regions Hydrology Project Office, Felrbanks, Alaska 99701. 471 ALASKA'S SNOW, ICE AND PERMAFROST It Is useful to begin with some general Intro- ductory comments on the five selected facets of Alaska's snow, Ice and permafrost problems which were treated In the workshop. The Alaska-Yukon Glaciers For brevity, the glaciers of Alaska and adja- cent Canada (I.e., Yukon Territory and British Columbia) will be referred to as Alaska-Yukon g laclers. 2 These g laclers cover an area of about 102,000 km , three quarters of which Is In Alaska. For perspective, twelve states In the u.s. have sma Iter areas than thIs. About 80% of the fresh water on Earth Is In the form of glacier Ice. Most of this Ice~ by ~ar, Is contained In the ~ta~ctlc (1.3 x 10 km ) and Green land ( 1. 7 x 10 km > Ice sheets which are remote from large human populatJons:l The third and fourth largest Ice masses (10 km each) are the Queen Elizabeth Islands, of the Canadian high arctic, and the Alaska-Yukon g laclers. Either of the latter systems of glaciers covers more than twice the glaclerlzed area of the Himalaya and Karakorum ranges combined. The Alaska-Yukon glaciers are significantly different from the other large Ice masses for severa I reasons. They lie In a region of dynamic and ever-Increasing human activity. The mass flux of these glaciers Is among the highest In the world, and many of them are composed primarily of temperate Ice <0°C throughout). In contrast to this, the Queen Elizabeth Islands are nearly as remote as the large Ice sheets and their glaciers are polar, I.e., Ice temperatures are well below 0°C. The Alaska- Yukon glaciers are also located In an exceptionally dynamic physical setting. They are far more active than polar glaciers because many of them form a part of a major climatic barrier on the rim of the North Pacific and receive extraordinarily heavy precipita- tion from the Gulf of Alaska, one of the two most vigorous storm centers In the Northern Hemisphere <the other one being In the North Atlantic>. The climate of North America may be significantly affected by processes of storm and air mass modifi- cation which occur at this glaclerlzed border of the North Pacific. Also, the oceanographic effects of copious meltwater runoff along the southern Alaska coast are being Increasingly appreciated. In spite of the exceptlona I features of the Alaska-Yukon system of glaciers, they still must be considered largely unexplored: "There have been few attempts to dl scuss the glaciers of Alaska and adjoining parts of Canada on a regional basts. Most accounts tend to deal with single glaciers or groups of glaciers In rela- tively small areas; even general descrip- tions of the Ice cover of Individual mountain ranges are lacking." <Field, 1975, p. 3) 412 Given a random glacier In this system, the chances are that very II ttle Is known about Its most basic properties: mass balance (both Ice and water balances), the energy balance, recent changes In volume, mode of flow, facies and temperature distri- butions, sediment production, and so forth. Al- though we do not know a great dea I about most of these g laclers, we do know that they Inc I ude vastly dl fferent types. For example, In the coastal regions of Alaska the glaciers may receive 10m or more of snow and are temperate, while In the Brooks Range they receIve less than a meter of snow and the temperature of most of the Ice volume Is below 0°C. Therefore, within the system we have g laclers wIth comp Jete I y dl fferent hydro log! c, lce-f low, and sediment regimes. Glaciers and Hydroelectric Energy It has become abundantly clear In the last decade that Alaska will ultimately experience massive economic development. This means that resource management will become even more critical In the years ahead. Although most attention so far has been given to the petroleum Industry, It Is clear that there Is a vast hydroelectric potential. Most prom! nent Is the propos a I to dam the Sus I tna River at a cost In excess of 5 billion dollars. <11 the average, Alaska-Yukon g laclers and high- a Itt tude, snow-covered regions supply roughly a third of the total runoff, and exert a significant control on the flow and water qua I tty of all maJor Alaskan rivers except the Col ville. The value of knowledge about the hydrology of glaclerlzed basins Is Increasingly being appreciated by engineers and , planners Involved with hydroelectlc projects; yet In spite of the need for this basic knowledge, It does 1 not exist today. By misjudging the contributions fu long-term runoff from g laclers, one can make costly 1 errors (Bezlnge, 1979). Alaska's Seasonal Snow Snow forms a thIn veneer on the earth's surface over most of Alaska for 1/2 to 3/4 of the year. The Alaskan snow dl ffers from the hydrologlca lly Impor- tant mountainous snow of the western United States In that Its temperatures are lower, steeper tempera- ture gradients occur In It, and It lasts longer and enters more directly Into human activity as snow Itself, rather than serving primarily as a cold storage water reservoir. Although Alaska Is famous for Its glaciers, It Is especially well suited for the study of seasonal snow cover. Indeed, Alaska's glaciers, extensive as they are, cover only 5% of the tota I area-a II of which Is subjected to seasonal snow. Alaska has three distinct types of snow because It contains maritime, extreme continental, and Arctic climatic zones In proximity. Striking differences exist In the snow cover from one climatic zone to the next. Our knowledge of the amount of snowfall at high altitudes and on the Arctic Slope, and of the flux of wind-blown snow needs considerable refinement. Alaska's River and Reservoir Ice The site of virtually all water resource development projects In Alaska will be along the rivers. These projects will be subjected to severe Ice prob I ems for substantial portions of the year. Ice jams on the rivers can result In flooding, severe erosion, and destructive tee forces. The , construction of a reservoir can alter the natura I river regime adversely both thermodynamically and In ' terms of water qua II ty, and these prob I ems are exacerbated by the presence of tee for long periods of the year. It Is crttlca I that the I Imitations to our knowledge of river and reservoir tee processes, the state-of-the-art of current research, and the capab Ill ty for predt ctt ng the Ice prob I ems be delineated. For example, little Is known about winter stage-discharge relationships of Alaskan rivers. The patterns of recurring tee jams have never been compt led for any Alaskan rivers. No well-developed models exist for the prediction of erosion and scour In braided or tee-covered streams. Research models of aufets growth have been devised from laboratory studies, but field verification has not taken place. These are only a few of the examples of the need for continued study of river and reservoir Ice In Alaska. It Is clear that a knowledge of the available theory, practice and technology for the avoidance or mitigation of river Ice prob I ems Is essential for water resource planners, engineers and managers In Alaska In order to prevent costly, Inappropriate and ~sslbly dangerous development and construction. Alaska's Permafrost Both seasonally and perennially frozen ground (permafrost) give rise to scientific and engineering problems of great Importance to Alaska. About 80% of Alaska Is underlain by permafrost, of which 25% Is continuous; all of Alaska Is subjected to season- al freezing. Only selected aspects, pertinent to glaclerlzed basins In permafrost regions, were con- sidered In the workshop. These topics dealt mainly vlth the effects of permafrost on subsurface drain- age and runoff. Implicit In this concern Is the dearth of knowledge and the lack of a well-defined theory describing moisture migration In freezing and. thawing so I Is. Wh lie emp I rica I mode Is of the geotechnical aspects of permafrost (such as frost heaving) ext st, considerably less Is known regarding basic transport phenomena In a freezing soli. 473 THE EFFECTS OF GLACIERS ON RUNOFF AND RUNOFF FORECASTING A "glaciated" basin Is one which shows evidence of having had glaciers present In It In the past while a "glaclerlzed" basin has glaciers In It now. Glacterlzed basins tend to have less variabil- Ity In runoff than do non-glaclerlzed basins on a short-term, year-to-year bas Is. However, on longer time sea les of 30 to tOO years, changes In the mass balance of glaciers can significantly modify runoff from glaclerlzed basins as the amount of Ice In storage changes. The lag between precipitation and runoff Is measured In hours, days or posslb ly weeks where no snow cover Is present. With a seasona I snow cover, precipitation may accumulate over several months before It melts and appears as runoff. In a gla- clerlzed basin, precipitation falling as snow above the equt I lbrhan I tne on a glacier may not appear as runoff for years, decades or even centuries. Runoff from a glacier may be almost Independent of recent precipitation. Situations even exist where negative correlations exist between precipitation and run- off. For example, heavy winter snowfall may delay the exposure of relatively low-albedo glacier Ice, and thus reduce the total amount of runoff the following season. Precipitation during the ablation season Is normally associated with cloudy condi- tions, which again may reduce ablation. On the other hand, during hot sunny weather, when non- glacier sources of runoff tend to dry up, glacier melting accelerates. Thus, runoff from the glaciers themselves Is likely to be greatest when runoff from the rest of the basin Is least, and vice versa. If glaciers never changed their sizes, their existence would have no major effect on long-term changes In runoff. In the real world, glacier volumes do change, due to both climatic factors and Internal Instabilities. Records show that consider- able glacier variations have occurred In the past, and It Is only prudent to expect similar variations In the future. As a non-tidewater glacier shrinks, the Ice lost to the glacier appears as runoff over and above that due to recent precipitation. As It grows, some of the annual precipitation enters Into long-term storage as glacier Ice rather than appear- Ing as runoff that year. AI though u.s. hydropower development has taken place In areas with heavy seasonal snow packs, few projects have been In glaclerlzed basins. Also, the g lac fer I zed bas Ins In the conterminous u.s. lie In the Pacific Northwest where the climate has a winter precipitation peak and a relatively long ablation period. By contrast, the mainland of Alaska has a summer precipitation peak and a relatively brief ablation period. This means that a different approach to runoff model ling may be required for Alaska. Estimates of glacier depletion and Its effects on runoff have been given for Switzerland CKasser, 1973), Scandinavia U)strem, 1973> and Canada (Young and Ommanney, 1984>. The rate of wastage of glaciers varies from region to region (e.g., major losses of glacier volume In the Eastern Alps and only minor losses In the Western Alps) and during the course of time (e.g., glaciers can re-advance after many years of wastage). Over the past hundred years, glaciers In Alaska and around the world have suffered an overall loss of mass (Meter, 1984). This means that glaciers, as natural reservoirs, have been generally depleted, so less water will be available for the future. Meter also concluded that the loss of water from glaciers In Alaska during this century may have been greater than the world average. Water has been withdrawn from glacier storage and added to annual streamflow In varying amounts. Thus streamflow records obtain- ed during the last few decades Include this added component but do not specifically Identify tt. In planning projects which may have lifetimes of decades to a century or more, Information Is needed on the extent to which glacier variations or stabil- Ization may affect future water avallabl llty and quality. A summary of the recommendations follows: ONE MAINTAIN Nil EXPNI) UPON QJRRENT DATA COLLECTION EFFORTS Understanding of fundamental processes on both long and short time sea les Is very Important, and dependent on the ext stance of an adequate data base. Therefore, steps shou I d be taken to address this need by ensuring continuity of data collec- tion. In particular, the existing data collection efforts at Gulkana and lblverlne glaciers should be continued, and new stations should be Initiated In several other major climatic divisions of Alaska. Key areas to be added are southeastern Alaska, where Lemon Creek Glacier Is a possible choice, the Brooks Range, where some data exIst on McCa II Glacier, and at least one glaclerlzed basin near the lllest coast of Alaska. Companies or agencies operating power projects should be required to continue gathering hydrologlca I, meteorological and glaciological data within project areas after a project Is operational. TWO DEVELOP THE CAPABILITY OF FORECASTING RUNOFF f.bdels developed for other areas should be Investigated for their appl I cab I llty to Alaskan conditions. It Is anticipated that In most cases problems will arise from the lack of necessary Input data rather than from Inadequate or Inapplicable theory, and .that the results of this Investigation will therefore feed back Into the priorities of the data collection program recommended above. One area In which toea I research wl II be necessary Is the release of water from englaclal <or glacially- dammed) storage. 474 THREE DEVELOP NEW tEASUREJENT TEQIN IQUES WIERE EXISTING TEOINIQlES ARE INADEQUATE FOR TtE ATTAIIIENT OF CIJJECTIYE C 1) ABOVE At the present tl me, no good methods ext st for measuring winter precipitation with good time resolution In the coastal mountains of Alaska, or winter discharges whl le there are substantial amounts of Tee In streams. Improvements are needed In the continuous monitoring of snow and flrn temperatures and the telemetry of hourly or dally meteorological, hydrological and glaciological conditions. Reliable telemetered data are likely to be of great Importance to the application of fore- casting techniques for glacially-mediated flood events. FOUR ASCERTAIN REGIONAL MASS BALANCE PATTERNS FOR THE MAJOR GLACIER SYSTEMS OF N..ASXA This goal probably cannot be achieved even with the long-term data collection stations recommended In ONE. Short-term studies In other basins, lasting severa I years, are probably needed to provl de the basts for Interpolation between the long-term stations. FIVE AIW.YlE THE EXISTING CLIMATE/GLACIER YOlliE RECORD Comparison of old aerial photos of Alaskan glaciers with new ones permits volume change to be estimated. Comparl son with c llmatlc records from existing low-lying climate stations could greatly Improve our understanding of the variability of glacier response to climate over the state. NATURAL HAZARDS CAUSED BY GLACIERS As high-altitude storage reservoirs In the hydrological cycle, glaciers contain enormous potential energy. Relatively small amounts of this energy can be released In ways which produce major hazards. Once released, the potential energy Is restored by c II mate so the hazards are usua fly recurring events. Even though the existence of these hazards can often be foreseen, they generally cannot be predicted accurately. A "hazard" Is an event that Is potent I a II y dangerous. A "risk" arises when people place structures In a position where a hazard creates a threat to I I fe and property. The study of hazards Is a matter of science; the study of risks Involves science, economics, engineering, human aspirations, and po II tl cs. The primary hazards associated with glaciers are from floods, landslides, avalanches and Ice- bergs. Glaciers store water on top, within, beneath and beside them; the most obvious cases are glacier- dammed lakes. Post and Mayo ( 1971) ldentl fled about 750 Alaskan glacier-dammed lakes and their flood paths. When a glacier water reservoir discharges It produces an outburst flood for which the Icelandic term J~kulhlaup has become widely used. J~kulhlaups tnvo I ve large dl scharges for sever a I days and they are often highly charged with snow, Ice and rock. The retreat of a glacier may leave unstable slopes of debris which had been supported by glacier Ice; this magnifies the potential hazards from landslides which may be triggered by large earthquakes (Miller, 1960). Some attention has been paid to hazards from snow avalanches In Alaska but very little attention has been given to potential Ice avalanches from glaciers, yet these have caused loss of II fe and damage to property In SwItzer I and. Alaska has 90% of the explosive volcanoes In the United States. Large amounts of water, snow and tee are mobilized rapidly during an eruption. But It Is Important to recognize that the effects of volcanism on glaciers are not limited to spectacular eruptions. Variable volcanic heat flux produces basal meltwater at high altitudes which Interacts with the volcano In ways which are potentially hazardous but not well understood. Glaciers may take decades to recover from perturbations produced by eruptions, and the recovery process Itself may create secondary hazards (Sturm, Benson and f"acKelth, 1986). There Is much for Alaskans to learn from case histories In other parts of the lol)rld, especially from the Andes, from Iceland and from Kamchatka <USSR). In most areas of Earth, natural hazard research comes as a consequence of direct threat or damage to established economic Improvements already at risk. This provides only an emergency response where the damages are very much more costly than the re- search. In Alaska, the reverse situation exists In most cases. Studies can be used to direct develop- IOOnt away from known dangerous situations and thus avoid the hazards without having to endure the risks and costly damages. The economic and polltlca I justifications for the Alaska case require consider- able Insight, but the rewards for this are not sma 11. A summary of the recommendations follows: ONE f"tii...ISH ATLASES OF ALASKAN GLACIER HAZARDS Most types of glacier hazards are known to occur In Alaska. Statewide assessments showing the distribution of surging glaciers and outburst flooding have been published, but require periodic updating to report new knowledge and changing natural situations. No statewide atlases have been ptbllshed for other hazards. Special problems are associated with Alaska's many glacier-covered volcanoes, most of which are explosive. Our recom- ~ndatlon for attention to glacier-covered volcanoes 475 dovetails well with Resolution No. 7, on explosive volcanism, of the national workshlp on Seismic, Volcanic and Tsunami Hazards In Alaska <Davies, 1983). Projects to produce these atlases should be organized and funded. An advIsory group shou I d be assIgned to steer Individual research efforts along fruitful lines and ensure that the efforts of many lndlvl dua I projects over many years would produce a coherent set of published products that would be of direct benefit to the users. TWO COtDJCT tiJDEST RESEARat IN llfE NEAR FUllfiE. TO ENSURE THAT THE KNOWLEDGE TO ATTA<X EMERGENCY PROBLEMS WILL BE AVAIL.Ail...E llfEN NEEDED Research sites should be selected from the hazard atlases to conduct priority research on those glaciological processes which, once understood, could provide predictive methods. During this work, a series of selected research Individuals would evolve. To be available Immediately, and most effectively, they should be resident In Alaska. These scientists would be knowledgeable about the hazards and equipped to work on emergency situations should they occur. THREE ~TE EMERGENCY TRAVEL fUN)S TO BE AVAILABLE FOR SELECTED SCIENTISTS TO STUDY HAZARDOUS PROCESSES AS THEY ARE DEVELOPING The Alaska Council of Science and Techno logy, when It existed, considered carefully the problem of geologic hazards, Including glacier hazards. One outstanding suggestion was that mission-ready scientists be listed for contact when a short-term hazardous situation Is suspected. Any member of the team would be capable of charging transportation and susb I stance cost for a short time to an account estab II shed In the State of A I a ska government for the purpose of emergency scientific Investigations of short-term phenomena, especially when the evi- dence of what Is occurring would be otherwise lost. This list of Investigators and a modest supporting fund should be established. FOUR INTEGRATE GLACIER HAZARD IIFORMATION INTO LAN>-USE PLANNING An authority should be made responsible to bring general knowledge gained from geologic and hydrologic hazard and risk assessment research Into the public planning process. The u.s. Army Corps of Engineers and other agencies have limited author- Ity. The broader Authority will need a professional advl sory group estab II shed (poss lb ly the same group as recommended above) to work with them on reviewing required studies and setting priorities. The Authority would prepare overviews for making dec!- s Ions based on rl sk assessments, pi ann I ng defenses from natural events, and keeping legislators and the public Informed about the benefits to them of this research. FIVE DEVELOP mNTINUOUSLY TELDETERING RIVER STAGE ~S THAT OPERATE IN WINTER IN 1LASJ<A BELOW HAZARDOUS GI...ACI ER DAMMED LAKES AN> VOLCANOES Techniques for the operation of flood warning gages from glaciers during the long winters In Alaska do not exist. Such gages could a I so provide Important new Information about the norma I winter flow regime of rivers at standard river gage sites. GLACIERS AND SEDIMENT The sediment transported by Alaskan rivers poses problems for hydropower or other engineering developments, but Is also a valuable resource both as a construction material and as an Intrinsic part of the river environment, affecting hydrology and biological activity. Sediment In glacier and non- glacier rivers Is quite different, In quantity, variability, size distribution and mineralogy. This distinction Is obviously Important In Alaska where glaciers contribute to all the major rlve,·s except the Colville, glacier surges are common, the bedrock Is often friable and shattered, and there are Important glacier-volcano Interactions. The present understanding of the sediment regimes of Alaskan rivers Is unsatisfactory. The existing measurements (for example, U.S. Geological Survey, 1950-60a, 195Q-60b, 1961-84; Harrold and Burrows, 1983; Nelli et al., 1984; R & M, 1982; Lipscomb and Knott, 1985) are almost entirely well downstream from the glacier sources; their relation to the sources and to the transport processes In the Intervening stretches of river Is unknown. Recon- naissance measurements at a few glacier termini have been made (Gaddis, 1974) but the magnitude of the glacier sources remains virtually unknown. Measure- ments at the terminus of Variegated Glacier, Alaska <Humphrey, 1986; Humphrey et al., 1986) Indicate an Important connection between sediment supply and hydraulics and motion of the glacier, under both surge and non-surge condl tlons. The suspended sediment a lone during the 1982-3 surge of the glacier was equivalent to about 1/2 m of rock erosion under the glacier. It appears that a place as large and sparsely populated as Alaska stands to benefit from an understanding of sediment processes on a fairly baste level, because of the near Impossibility of evaluating every possible development In detail. As a start one needs to know, at least approximately, the magnitude of the glacier Input sources. This In turn would be facilitated by an understanding of the baste processes of sediment generation, storage, and transport within glaciers. Little Is known of these processes yet, but they are closely linked to 476 problems of glacier motion and Internal hydrology. If these are studied simultaneously with sediment production at the terminus, a deeper and probably more useful picture of the glacier sediment source will result. Even If the glacier and other source problems were completely understood, we would not understand the sediment regimes of the rivers without an under- standing of the downstream transport processes. Efforts by glaciologists to understand the source mechanIsms and those by others to understand sedi- ment dynamics In rivers therefore need to be coordi- nated. The observation of sediments produced by the surge of Variegated Glacier Is an Alaskan example of how glaciologists can contribute, but there Is need for similar measurements to be made at a glacier located on a major river, and for these to be related to sediment transport and channel changes downstream. Generally, very Intensive data collection Is required to define sediment-transport character- Istics adequately, especially where high variability exists. This Is certainly true of glaclerlzed basins, where there Is ample evidence that t~ glacier sources are extremely variable (fllstrem et al., 1967, for example). Whether carried out at glacier termini or downstream, sampling frequency should always be Intensive at the beginning of study, and adjusted only after some Idea of the variability Is obtained. Gurnell ( 1982) has review- ed the des I gn of g I ac I er sedIment samp II ng programs and Invest! gated the Imp II cations of spat! a I and temporal variations In suspended sediment and stream dIscharge In the desIgn of samp II ng programs. The sediment studies carried out at glacier termini by the Norwegian Water Resources and Electricity Board could possibly serve as models <Roland and Haakensen, 1985). Da II y, Inc I ud I ng sometimes hourly, sampling for an entire year would greatly Increase our knowledge of variability of suspended sediment In Alaskan rivers. As well as long-term studies to evaluate annua I variation, high frequency sampling of flow events provides better understand- Ing of the mechanics of transport, especially as pertains to bedload. Synoptic, accurate measure- ments of hydraulics and channe I geometry with suspended sediment and bedload sampling are essen- tl a I to further our understand! ng of the dynaml cs of sediment movement In natural rivers. Measurement of channe I geometry, for which aerl a I photography Is a usefu I method, and determin- ation of the composition of bed materia I over long periods of time provide a great dea I of Information on how rivers change. Simply resurveying a few cross-sections and eva luatlng the s lze distribution of bed materials In the channel only once a year for ten years may provide the necessary Information to tell whether a river reach Is aggrading, continually moving laterally In one direction, or that the bed material Is becoming coarser or finer over the long term. This, combined with discharge records alone helps predict future changes. Ideally, sediment data would also be collected. An excellent candidate for a fairly Intensive study of glacier sediment source would be Phelan Creek below Gulkana Glacier. Mass balance studies began In 1960 (Meter et aJ •• 1971; Tangborn et al •• 1977); stream flow records ex! st from 1966 to 1978, and from 1983 to present; some sediment data were coll.ected at the Phelan Creek gage In 1973 <Gaddis, 1974>; and this basin Is tributary to the Delta River which flows Into the Tanana River, where there Is fairly long-term suspended sediment data down- stream. Probably the best candidate for an Inten- sive study below a glacier with the Immediate potent! a I to surge wou I d be the short creek between Black Rapids Glacier and the Delta River. This study would also benefit from downstream sediment data on the Tanana River, and there Is considerable unpublished glacier data from Black Rapids Glacier from ongoing u.s. Geological Survey and Un Ivers tty of Alaska studies which Include dally motion. other promising sites would be iblverlne Creek below Wolverine Glacier, which has long-term glacier balance (Meter et aJ •• 1971; Tangborn et al •• 1977; Mayo et al.. 1985) and stream flow data but no downstream sediment Information, or one of the tributary basins of the Susltna, where stream flow, sediment and some glacier data are available (Knott and Lipscomb, 1983; Lipscomb and Knott, 1985; Harrison et al., 1983; Clarke et al., 1985; Clarke, 1986). It Is likely that much could be learned by comparing the sediment production of different glaciers with different sizes In different settings, and with different flow regimes (surge and non- surge. for examp I e > • A summary of the recommendations follows: ONE DEYEL<P AN Uti>ERSTAM>ING OF BASIC PROCESSES AS A LONG-TERM GOitl.. This long-term goal Is partl~ularly appropriate for Alaska because of the near lmposslbl I lty of study! ng every s I tuatl on In data II, and because of the I Imitations of statlstlca I methods. TWO INITIATE SEDIMENT STUDIES NEAR THE TERMINI OF REPRESENTATIVE GLACIERS. Nil RELATE THEM TO DOWNSTREAM SEDIMENT TRANSPORT STUDIES One of these glaciers should be likely to surge, and be located on a major river. The long- term purpose of the measurements should be to defIne the magnItude and character of glacier sediment sources In Alaska. 477 THREE RELATE SEDIIENT STUDIES AT GLACIER TERMINI TO STUDIES OF THE DYNAMICS. HYDROLOGY AND HYDRAULICS OF THE GLACIER ITSELF The measurements at glacier termini would therefore be related to lnterna I glacier processes and motion, and, as TWO Implies, to downstream transport processes. FOUR INITIATE STUDIES RELATED TO SPECIFIC PROPOSED DEYEU'MENTS WELL IN ADVANCE OF NEED This Is necessary because of the high variabil- Ity of sediment supply and transport. FIVE lfoFROYE JWT04AT IC SNFLI NG TEatN I QUES This Is particularly Important for Alaska because of difficulty and expense of field opera- tions. ICE PROBLEMS ASSOCIATED WITH RIVERS AND RESERVOIRS Water resource development In Alaska Involves special problems because of the presence of Ice tor long periods of the year on rivers. lakes, ponds and reservoirs. The Important physlca I phenomena associated with Ice on rivers and reservoirs In- clude: frazll Ice, aufels, anchor Ice, sheet Ice and the mechanical, thermal and chemical effects that these Ice forms have on water bodies and the sur- rounding basin. Frazl I Ice, which forms when water temperature supercools by a few hundredths of a degree below 0.0°C, are Initially observed as discs In the range from 0.1 to 5 mm In diameter. These discs agglomer- ate Into frazll floes and pans when river turbulence Is low. Frazll pans can exceed 1 or 2m In diameter In sma II rivers I Ike the Tanana River; In the Yukon River, the pans often exceed 10 m. Frazl I Ice Is known to deposit on the underside of an Ice cover where the river bottom slope de- creases substantially. Thus, frazl I Ice may accumu- late In a "hanging dam" at the entrance to slow reaches In the rIver or In poo Is downstream from rapIds or steeper reaches. These dams are a poten- tial hazard causing winter flooding, overflows and associated aufels formation. Aufels (meaning Ice on tee) formations develop when water flows over and freezes on surface Ice or frigid ground. Sub- permafrost groundwater and groundwater from thaw bulbs near streams are an additional source of water for aufels (Kane, 1981). Aufels formations ap- proaching 10m In thickness can be generated In this fashion (Sloan, Zenone and Mayo, 1976). Anchor Ice (Ice anchored or fixed In position In the flow, as on the river bed) Is generally believed to be Initiated by the attachment of frazll Ice crystals to the river bed and to objects In the flow (Osterkamp and Goslnk, 1983). In this manner, anchor Ice can cover extens lve sections of a river bed. The bonding between anchor Ice and the river bed may be released If sufficient heat from radia- tion or groundwater Is ava liable; then the buoyant anchor Ice will lift from the river bed, carrying with It sediment, gravel and on occasion, heavy rocks. This process causes an Ice run which can cause severe damage to downstream hydraulic struc- tures. Break-up flooding Is a major problem In Alaska which annually raises the prospect of loss of life and damage to property. Continued research Into baste river Ice dynamics, mitigation techniques and Ice Jam processes should eventually lower the risk of this flooding. For the present It Is essentla I that the early warning system of the National Weather Service for break-up flooding be maintained, and If possible expanded. Sheet Ice forms on slowly-moving or quiescent water as a consequence of heat loss to the atmos- phere. In this case the water turbulence Is Insuf- ficient to draw the Ice formed at the surface Into suspension, so that Ice growth propagates across the water surface rather than wl thIn the body of the water. Lakes and ponds may contain frazll Ice, formed during windy conditions, and sheet Ice, formed during calm periods, and snow Ice, formed by the flooding of a snow layer. Consequently the morphologlca I structure of the Ice may be complex, resulting In complex thermal, mechanical and optical properties. Most deep northern lakes can be classified as dlmlctlc, characterized by strong stratification In the summer and stratification under the Ice cover In winter. In spring, the melting Ice becomes porous and weakened with vertlca I channels and assumes an appearance known as "candled Ice." The stability of the water column Is altered by short-wave penetra- tion through the candled Ice cover, and, at least In one Instance, there Is evidence to Indicate that norma I sprl ng overturn was Impeded by the stable thermal gradient, resulting In the potential for hypollmnetlc oxygen depletion (La Perrlere, 1981>. In glaclerlzed basins, the turbidity of the water strongly affects short-wave radiative penetration as well as density structure. This also may have a significant effect on water temperature and quality. Water resource development proJects In cold climates and the subsequent effects on the Ice conditions and the hydrological regime were con- sidered In the workshop. A set of guidelines to define the relevant Ice problems for engineers and managers, to assess the state-of-the-art for predic- tion or modeling, and to consider mitigation tech- niques presently used In northern climates, Is presented In tab I u I ar form In the report on the workshop. 478 Recommendations Severa I Important documents have defined recommendations for research related to Ice problems In rivers, lakes and ponds, and reservoirs. These Include a National Academy of Science Report <National Research Council, 1983), and the results of a National Research Council Symposium In Canada <Natlona I Research Council Canada, 1979). These documents form a genera I framework for research needs related to Ice. In addition to the recommen· datlons of these documents, sever a I other points particularly relevant to water resource development In Alaska were proposed at the workshop. The Intent of these supplementary recommendations Is to obtain the most cost-effective use of limited resources and, simultaneously, to assure that critical con· cerns, related to safety, environment and economic well--being, are not overlooked. The recommendations particularly Important for Alaska water resource development fall Into 'hlo general categories: the first concerns the long· term state Interests, and the second Is related to site-specific Investigations during the planning and feasibility stage of water resource projects. A. Alaskan long-term Interests. Better use should be made of existing state and federal agencies and University of Alaska capablll· ties. Particular examples Include the following: ONE ~S OF DISCHARGE AS ..::u. AS STAGE SHOULD BE RECORDED TtftOUGHOUT THE YEAR. The u.s.G.S. presently records stage measure· ments at a number of sites throughout the year In Alaska, but determl nes stage-d! scharge re latlonshlps for the open water season only. For models of Ice Jam processes, break-up and winter flooding, It Is essential that the additional measurements be made. TWO SUPPORT SHOULD BE PROVIDED FOR TIE TRAINING OF GRADUATES STUDENTS IN ALL ASPECTS OF COLD REG ION HYDROLOGY. This Is an extremely cost effective use of state resources both over the short and long term. These students will In time become scientists and engineers resident In Alaska with respect for Alaska's best Interests, and well prepared for long· term water resource planning. THREE SOE OF 11£ DATA REQUIRED FOR 11£ CILIBRATION OF VARIOUS tllDELS IS PRESENlLY COLLECTED BY A VARIETY OF STATE AM> FEDERAL AGENCIES. ntESE DATA SHOULD BE COP I LED AM> ORGAN I ZED FOR BETTER ACCESSIBILITY. Two examples Include: the encoding of all u.s.G.S. data on floppy disc for easy and rapid access, and the determination of recurrent Ice bridging locations on Alaskan rivers through aerla I photographic surveys. FOUR TIE tEED FOR Lot«rTEAM CONTitiJOUS tiJNITORING OF SPECIFIC PHENCIENA SHOULD BE REmGNIZED AS AN ESSENTIAL PART OF PLANN lNG AM> DEYELCFMENT. These measurements are required both for the baste formulation of models and for the determina- tion of long term changes. Some specific examples Include Ice jam dynamics, erosion processes, perma- frost degradation and frazll Ice formation and growth mechanIsms. These efforts shou I d be dl rected toward the development of comprehensive river freeze-up and break-up models. B. Site-specific research. Planners, managers and eng I nears must be made aware that standards and models developed for temperate-zone water resource projects are often Inadequate In cold climates, and furthermore, that the data needed for ana lysis of projects In Alaska are often not available. In addition, It Is Impor- tant to note that remedl a I technIques to remove specific Ice problems may Inadvertently create new problems. ONE AS A FIRST STEP, AN INVESTIGATION TO DETERMINE Tl£ SCOPE OF TIE MAJOR ICE PROCESSES AM> PROILEMS SHOULD BE UfllERTAKEN. THIS STUDY SHOULD IElP TO DEFINE TIE lftJRQfR lATE LEVEL OF DETAI Lm fOlEL.ING REQUIRED AT THE SITE AM> SPECIFY THE DATA NEEDED FOR THESE fOlEL.S. For example, high turbidity levels may Indicate the need for a two-dimensional reservoir model rather than a more widely used one-dimensional oodel, or, the presence of heavy aufels deposits may suggest the need for extensive topographic and ground water studies of the flood plain. It may also be necessary to devise new models to assess local critical concerns, for example, erosion In braided streams, Ice formation In estuaries, frazll Ice discharge and Ice bridging processes. 479 TWO AI.. THOUGH DATA tEEDS DEPEtl> ON LOCAL QJN)I liONS AM> ON 11£ PARTIQJLM WATER RESOURCE PROJECT • tiJOELING CAN BE USED TO IElP ESTAII.ISH THE EXTENT AM> FREQUENCY OF THE REQUIRED DATA BY A PROCESS KNOWN AS •NlJERICAL EXPERIMENTATIQNII. With this process, a variety of field condi- tions and the associated Ice problems can be tested for sensitivity. If a particular situation Is found to produce critical conditions, then a clear mandate Is set for further field Investigations. For example, If a certain Ice discharge produces con- solidated and grounded Ice jams with flooding at a given site, then more Intense field work Is needed to define Ice bridging locations and frazll Ice production rates. Alternatively, If reservoir temperatures are particularly sensitive to light- extinction coefficients, then further Investigations of turbltldy levels and erosion processes are clearly warranted. It Is Important that these eventualities be considered to make the most cost effective use of the resources available, and to eliminate costly remedial procedures. THREE THE BASIC DATA tEEDS FOR EXISTING RIVER AM> RESERVOIR tiJDELS ARE: t. Rivers 11. Lakes Discharge: (dally> ~teorology: air temperature, wind ve Joe tty, precl p ttatlon, re latlve humidity, cloud cover, short-wave and long-wave radiation (dally) Basin geometry: adequate river cross-sections, river slope In stream: water temperature, water quality, water levels, sediment bed and suspended loads (dally> Ice conditions: bridging locations, anchor and edge Ice growth, Ice discharge (during freeze-up); leading edge advance, Ice cover thickness, aufels growth (during winter>; break-up locations and Ice jam locations (during break-up) Inflow and outflow: temperature, conductivity, suspended sediment concan tratl on Meteorological: a lr temperature, wind velocity, precipitation, relative humidity, cloud cover, short-wave and long-wave radiation (dally) Basin geometry: area, volume and depth relations, river slope and bed angle In-Jake: water temperature, conduc- tivity, suspended sediment, light extinction, water level (dally) Ice conditions: snow and Ice thick- ness (weekly> Ground conditions: extent of perma- frost, ground water table PERMAFROST The presence of permafrost Introduces a wide range of scientific and engineering problems. Since 1963, these have received attention In four lnter- natlona I permafrost conferences. The vast range of prob I ems extends we II beyond the scope of our workshop. However, some effects of permafrost on hydrology and water qua Jlty that are related to the primary theme of the workshop were considered. Although 5% of Alaska Is covered by glaciers, 80% of It Is underlain by continuous or discontinu- ous permafrost and all of It Is subjected to season- ally frozen ground. The one major drainage basin which Is underlain by continuous permafrost Is also the only one which Is not affected by glacier runoff; this Is the Colville River drainage on Alaska's Arctic Slope. The dl fferences In hydrographs between rivers lying In glaclerlzed and non-glaclerlzed basins within the zone of continuous permafrost may be more extreme than In non-permafrost regions. This Is expected because parma frost Impedes subsurface drainage, Increases runoff and virtually eliminates subsurface contributions. Hydrographic data for rivers on Alaska's Arctic Slope are rare and no data base exists for comparison of discharge from gla- clerlzed and non-glaclerlzed basins. So far the available hydrological measurements on the Arctic Slope have largely been responses to ad hoc, applied needs. When the lmmedl ate need Is removed, the observations cease. Such procedure would be more appropriate If It supplemented a designed, long-term regional monitoring program. Unfortunately, the long-term program with basic scientific goa Is does not exist. One special feature In the continuous perma- frost zone Is the formation of extensive aufels deposIts. Dean ( 1984) mapped these features throughout Alaska on Landsat Images and found the largest deposits and most frequent occurrences were on the north sIde of the B_£ooks Range. Some of these deposits exceed 400 km In area (Dean, 1984) and thicknesses of 5 to 6 m have been measured (U.S. Geological Survey, 1985). Upstream reaches of the Sagavanirktok River, where the Rlbdon, Echooka and Jvashak rivers Join It, are subjecte2d to a tota I areal aufels coverage of about 1000 km • This, plus high altitude snow packs and some glacier contribu- tion to runoff, are probab Jy responsible for the sustained base flow of the Sagavanirktok River at Prudhoe Bay which differs markedly from the sharply peaked hydrographs with low base flow of the Kuparuk and Putullgayak rivers (Car I son and Kane, 1975). Large aufels deposits act as glaciers In their 480 sustained contribution to river discharge; some of the deposits persist from year to year. The Colville River, which Is the major drainage of the Arctic Slope, has Jess effect from aufels and high altitude snow packs (and not any from glaciers) than smaller rivers to the east of It. There Is a need to determIne a I I bas I c hydro- logical parameters on the Arctic Slope. The amount of winter precipitation which comes as snow has apparently been underestimated by a factor of three; the wl nd transport of thIs snow and the amount lost by evaporatl on are a I so beIng reassessed (Benson, 1982). The contribution to runoff from long-term changes In the permafrost Ice mass Is virtually unknown. South of the Arctic Slope, the majority of the glaclerlzed basins lie within the zone of discon- tinuous permafrost, and In most drainages t~ glaclerlzed area covers only a small proportion of the total basin. Only a few straam gaging sites have been estab II shed In basins with more than 50% glacier cover and In the majority of the larger basins, the glaclerlzed area Is Jess than 10% (Lamke, 1979). Glaciers may have a larger I nf I uence on the hydrology of a basin than would be Indicated by proportion of glacier coverage. The glaciers may perform a moderating effect on streamflow by In- creasing runoff during hot, dry summers by glacier melt; during cool, wet St.lllmers glacier melt would be retarded <Krtmme I and Tangborn, 1974; Fountain and Tangborn, 1985). The permafrost areas of the basin have the opposite effect. The hydrologic response of the permafrost areas can be highly variable and Is dependent on the thaw season c II mate <01 ngman, 1975). Except for very dry seasons when runoff Is dramatically reduced, permafrost runoff responds very rapidly and has slow recessions when compared to non-permafrost areas <Dingman, 1971; Haugen et a 1., 1982). The hydrologic response to summer precipitation events may be attributable to thick organic soils associated with permafrost <Dingman, 1975; Slaughter and Kane, 1979). In basins with a large proportion of permafrost area relative to glaclerlzed area, much of the glacier Influence may be offset by the permafrost hydrology. To date, the study of permafrost hydrology has been limited to small research basins or study plots due primarily to the dl ff leu I ty In mapping permafrost In large basins (Dingman, 1971). A full understanding of the hydrology of Alaska's drainage basins cannot be achieved without considering the Influence of the permafrost areas. A sunmary of recommendations follows: ONE COFARE HYilROGR/iPHS OF RIVERS IN GLACIERIZED AND NON-GLACI ERIZED BAS INS IN THE CONTINUOUS PERMAFROST OF THE ARCTIC SLOPE The differences In hydrographs between rivers lying In glaclerlzed and non-glaclerlzed basins within the zone of continuous permafrost Is expected to be greater than In non-permafrost regions. The Jago and Meade rivers are excellent candidates for comparison of glaclerlzed and non-glaclerlzed ' basins, respectively. These rivers are situated In proximity to existing logistical facilities. TWO COFARE RUNOFF PROCESSES FR04 PERMAFROST AM) NOtH"ERMAFROST AREAS IN Gl.ACI ER I ZED BAS INS The hydrology of permafrost regions differs from that of non-permafrost regions and both con- trast sharply with the hydrology of glaciers. Often the extent of parma frost Is not known In a gIven draInage bas In; technIques for mappIng the extent of the various sub-units of a glaclerlzed basin must be refined as a first step In dealing with this recom- mendation. THREE DETEAMINE THE BASIC HYDROLOGIC PARNETERS ON N..ASV.1 S ARCTIC SLOPE The questions of winter precipitation need special attention because of uncertainties In Its quantity, the extent of wind transport and the evaporatl ve losses. FOUR DETERMINE THE LOt«rTEAM HYDROLOGIC CYCLE OF THE COLVILLE RIVER AS THE MAJOR DRAINAGE SYSTEM OF THE ARCTIC SLOPE In addition to the O:>lvllle, a river such as the tJeade River which drains the tundra, but does not Include mountain runoff should be selected for long-term monitoring. Hydrologic Information fr.om these rivers, lying completely within the zone of continuous permafrost, Is needed as a contrast to that from glacier-fed Alaskan rivers. REFERENCES ~nson, c., 1982. Reassessment of winter precipita- tion on Alaska's Arctic Slope and measurements of the flux of windblown snow. Geophysical Institute Report UAG-R-288, University of Alaska, Fairbanks, 26 pp. 481 Benson, c.. w. Harrison, J. Goslnk, s. Bowling, L. Mayo and D. Trabant, 1986. Workshop on Alaskan hydrology: problems related to glacier I zed basins. Geophyslca I Institute, University of Alaska, report In preparation. Bezlnge, A., 1979. Grand Dlxence et son hydrolo- gle. La collecte de donnees hydrologiques de base en Suisse, Association Suisse pour l'amenagement des eaux, Service hydrologlque natlona I, 19 PP• Carlson, R. F. and D. L. Kane, .1975. Hydrology of Alaska's Arctic, In: Climate of the Arctic: Proceedings of the MAS-NilS O:>nference, Fair- banks, Alaska, August 1973, Geophyslca I Insti- tute, University of Alaska, Fairbanks, P• 367- 373. Clarke, T. S., 1986. Glacier runoff, balance and dynamics In the upper Susltna River Basin, Alaska. M.S. Thesis, University of Alaska, Fairbanks, May 1986, 98 pp. Clarke, T. s.. D. Johnson and w. D. Harrison, 1985. Glacier runoff In the Upper Susltna and Susltna River basins, Alaska, In: Resolving Alaska's Water Resources O:>nfllcts, Institute of Water Resources Report IWR-108, University of Alaska, Fairbanks, AK 99775-1760, p. 99-111. Davies, J. N., 1983. Seismic, volcanic and tsunami mitigation In Alaska -an unmet need. Report of Investigations 83-11, Alaska Dept. of Natural Resources, Division of Geological and Geophysical Surveys, 13 PP• Dean, K., 1984. Stream-Icing zones In Alaska, Report of Investigations 84-16. Alaska Depart- ment of Natural Resources, Division of Geologi- cal and Geophysical Surveys, 20 PP• Dingman, s. L., 1971. Hydrology of the Glenn Creek watershed Tanana River Basin, centra I Alaska. USA CRREL Research Report 297. Dingman, S. L., 1975. Hydrological effects of frozen ground: literature review and synthe- sis, CRREL Special Report 218. Field, w. 0., 1975. Mountain glaciers of the northern hemisphere (2 volumes plus atlas), edited by w. 0. Field American Geographical Society, published by CRREL, Hanover, NH. Fountain, A. G. and w. v. Tangborn, 1985. The effect of glaciers on streamflow variations. Water Resources Research Vo 1. 21., No. 4, P• 579-586. Gaddl s, B. L.. 1974. Suspended-sediment transport relationships for four Alaskan glacier streams. M.S. Thesis, University of Alaska, Fairbanks, 102 PP• Gurnell, A. M., 1982. The dynamics of suspended sediment concentration In an Alpine pro-glacial stream network. lnternatlona I Association of Hydrologlca I Sciences, Pub II cation No. 138, P• 319-330. Harrold, P. E. and R. L. Burrows, 1983. Sediment transport In the Tanana River near Fairbanks, Alaska, 1982. U.S. Geological Survey Water Resources Investigations Report 83-4213, 53 PP• Harrison, w. D., B. T. Drage, S. Bredthauer, D. Johnson, C. Schock and A. B. Follett, 1983. ReconnaIssance of the g I ac I ers of the Susltna River Basin In connection with proposed hydroelectric development. Annals of Glaciol- ogy, Vol. 4, p. 99-104. Haugen, R. K., c. w. Slaughter, H. E. 1-bwe and S. L. Dingman, 1982. Hydrology and climatology of the Caribou-Poker Creeks Research watershed, Alaska. USA CRREL Report 82-26. Humphrey, N. F., 1986. Suspended sediment discharge from Variegated Glacier during Its pre-surge and surge passes of motion, In: Benson et al., Workshop on Alaskan Hydrology: Prob I ems fe lat- ed to Glaclerlzed Basins. Geophysical Insti- tute, University of Alaska, report In prepara- tion. Humphrey, N., c. Raymond and W. Harrison, 1986. Discharges of turbid water during mint-surges of Variegated Glacier, Alaska, submitted to Journal of Glaciology. Kane, D. L., 1981. Physical mechanics of aufels growth, Canadian Journal of Civil Engineering, Vol. 8, No. 2, P• 186-195. Kasser, P., 1973. Influences of changes In the g lactated area on summer runoff In the Porte due Sceux drainage basin of the Rhone, In: Symposium on the hydrology of glaciers, lASH Publ. No. 95, P• 221-225. Knott, J. M. and s. w. Lipscomb, 1983. Sediment discharge data for selected sites In the Susltna River basin, Alaska, 1981-1982, u.s. Geological Survey Open-File Report 83-870, 45 PP• Krlmme I, R. M. and w. V. Tangborn, 1974. South Cascade G I ac I er: the moderatl ng ef feet of glaciers on runoff, In: Proceedings of the Western Snow Q:>nference, 42nd Annual Meeting, Colorado State University, Fort Collins, 1974, P• 9-13. Lamke, R. D., 1979. Flood characteristics of AI askan Streams, USGS Water Resources I nvestl- gatlons 78-129. LaPerrlere, J. D., 1981. Vernal Overturn and Stratification of a Deep Lake In the High Subarctic under Ice, Proceedings of Int. Assoc. for Theoretlca I and Applied Limnology, Congress In Japan, 1980, p. 288-292. Lipscomb, s. w. and J. M. Knott, 1985. Sediment transport In the Susltna River Basin, 1982- 1983, In: fesolvlng Alaska's Water fesources Conflicts, Institute of Water Resources Report IWR-108, University of Alaska, Fairbanks, AK 99775-1760, p. 191-204. Mayo, L. R., R. s. M:lrch and D. C. Trabant, 1985. Growth of Wolverine Glacier, Alaska, determined from surface a Itt tude measurements, 1974 and 1985, In: Resolving Alaska's Water Resources Conflicts, Institute of Water Resources Report I WR-1 08, Nov. 1985, UnIversIty of AI aska, Fairbanks, AK 99775-1760, p. 113-121. Meier, M. F., 1984. Contribution of small glaciers to global sea level. Science 266, 1418-1421. 482 Meter, M. F., w. v. Tangborn, L. R. Mayo and A. Post, 1971. Combined Ice and water balances of Gulkana and Wolverine glaciers, Alaska, and South Cascade Glacier, Washington, 1965 and 1966 hydrologic years. Ice and Water Balances at Selected Glaciers In the United States, u.s. Geo I og I ca I Survey Profess I on a I Paper 715-A, 23 PP• with maps. Miller, D. J., 1960. Giant waves In Lltuya Bay, Alaska: u.s. Geological Survey Prof. Paper 354-C, P• 51-86. National Research Council, 1983. Snow and Ice Research, an Assessment. Committee on Glaclol· ogy Polar Research Board, Commission on Physi- cal Sciences, Mathematics and Resources, National Research Council, National Academy Press, Washington, D.C., p. 1-123. Natlona I Research Councl I Canada, 1979. Proceedings of Canadian Hydrology Symposium: 79 -Cold Climate Hydrology, Ottawa, Canada, KIA OR6. NeIll, C. R., T. S. Buska, E. R. Chacho, C. M. Collins and L. w. Gatto, 1984. Overview of Tanana River mon I tort ng and research studlel near FaIrbanks, AK. u.s. Army Co I d Regions Research and Engineering Laboratory Special Report 84-37. Osterkamp, T. E. and J. P. Goslnk, 1983. Frazll Ice Formation and Ice Cover Development In Interior Alaska Streams. Cold Regions Science and Technology (1983), P• 43-56. 0strem, G., 1973. Runoff forecasts for highly glaclerlzed basins, In: International Hydro- loglca I Decade --The role of snow and Ice In hydrology. Proceedings of the Banff Symposia, Sept. 1972. Paris, UNESCO; Geneva, WMO; Budapest, IAHS, Vol. 2, PP• 1111-1132. (Publi- cation No. 107 de I' Association lnternatlonale d'Hydrologle Sclentlflque.) 0strem, G., c. w. Bridge and w. F. Rennie, 1967. Glaclohydrology, discharge and sediment trans- port In the Decade Glacier area, Baffin Island, N.W. T., Geograflska Annaler, Vol. 49, Ser. A., P• 268-282. Post, A. and L. R. Mayo, 1971. Glacier dammed lakes and outburst floods In Alaska: U.S. Geological Survey Hydrologic Atlas HA-455, 10 pp., 3 pl. Roland, E. and N. Haakensen, 1985. Glaclologlske Unders.Ske I ser I Norge 1982. Norges Vassdrags- og Elektrlsltetsvesen Rapport nr. 1-85, with Eng II sh summary. R & M, 1982. Alaska Power Authority Susltna hydro- electric project; task 3 -hydrology; subtask 3.07 -closeout report -reservoir sedimenta- tion. Report for Acres American Inc., Buffalo, N.Y. Slaughter, c. w. and D. L. Kane, 1979. Hydrologic role of shallow organic soils In cold climates. Canadian Hydrology Symposium: 79 Cold Climate Hydrology, P• 38Q-389. Sloan, c. E., C. Zenone and L. R. Mayo, 1976. Icings along the Trans-Alaska Pipeline Route, u.s. Geological Survey Professional Paper 979. Sturm, M., C. Benson and P. MacKelth, of the 1966-68 Eruptions of Mt. flow of the Drift Glacier. Journal of Glaciology. 1986. Effects Redoubt on the Subm I tted to: Tangborn, w. v., L. R. Mayo, D. R. Scully and R. M. Kr lmme I, 1977. Comb I ned Ice and water balances of Maclure Glacier, California, South Cascade Glacier, Washington, and \lblverlne and Gulkana glaciers, Alaska, 1967 hydrologic year. Ice and Water Balances at Selected Glaciers In the United States. u.s. Geological Survey Professional Paper 715-B, 20 PP• with maps. u.s. Geological Survey, 195o-1960a. Compilation of records of quantity and quality of surface waters of Alaska. u.s. Geol. Survey Water Supply Papers 1372 and 1740. u.s. Geological Survey, 195o-1960b. Quantity and qua II ty of surface waters of Alaska. u.s. Geol. Survey Water Supply Papers 1466, 1486, 1500, 1570, 1640 and 1720. u.s. Geological Survey, 1961-1984. Water resources data, Alaska, water years 1961-1984, published annually. u.s. Geological Survey, Survey Open File Index: Stream flow to Sept. 30, 1983. 1985. u.s. Geological Report 85-332. Alaska and water qua II ty records Young, G. J. and C. s. L. Ommanney, 1984. Canadian glacier hydrology and mass balance studies; a hI story of accomp llshnents and recommendations for future work. Geograflska Annaler Vol. 66A, No. 3, P• 169-182. 483 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 GLACIER-CLIMATE RESEARCH FOR PLANNING HYDROPOWER IN GREENLAND Roger J. Braithwaite and Ole B. Olesen ABSTRACT: A program of glacier-climate research is being carried out in Greenland for the planning of possible hydropower stations. The program is based upon the ~rallel collection and correlation of glaciological and climatological data. The present results conf irrn the illi£X>rtance of air temperature for calculating ablation and have already been used for the simula- tion of runoff from glacierized areas. (~ TERMS: Greenland, hydropower, glaciers, climate, runoff.) INTRODUCriON Since the mid-1970s, Danish state agencies have been investigating the feasibility of hydropower stations in Greenland. Hydrological conditions in Greenland are, however, still poorly understood, e.g. because of sparci ty and shortness of records as well as the effects of the arctic climate upon the runoff cycle (Gottlieb & Braithwaite, 1985). Many of the possible hydropower sites are also influenced by glaciers. The program therefore includes glaciological investigations which are carried out by the Geological Survey of Greenland ( GGU) as described by Weidick ( 1984). An illi£X>r- tant objective of the glaciological work is to understand the effects of glaciers up:m runoff. For this purpose improved kno.vledge of glacier-climate relations is needed. TABLE 1. Locations of the three glacier-climate stations operated in West Greenland by the Geological Survey of Greenland. STATION Johan Dahl Land Qamanarssup serrnia Tasersiaq IDRTH 61°27' 64°29' 66° 7' WEST 45°22' 49°32' 50° 7' PERIOD 1977-83 1979- 1981- The basis of the glacier-climate pro- gram is the measurement of glacier accumulation and ablation at several field stations in parallel with climatological observations. The first station was esta- blished by O.B. Olesen in late-1977 in Johan Dahl Land, South Greenland, and was manned for six summers until closed in late 1983 (Clement, 1984). The other two stations at Qamanarssup sermia and Taser- siaq were opened by O.B. Olesen in late 1979 and late 1981 respectively and are still continuing (Braithwaite, 1985a; Olesen, 1985) . The locations of the three glacier-climate stations are given in Table 1 and fig. 1. Glaciological observa- tions are also made at several places by mobile teams, i.e. without detailed clima- tological measurements, but will not be discussed here. GI.ACIOra;y PR<X;RAM The glaciological measurements are made in a network of stakes drilled into The Geological Survey of Greenland, DK-1350 Copenhagen K, Denmark. 485 the ice which are visited at least twice yearly to measure transient balances, i.e. roughly at the beginning and end of the summer which essentially extends over the three months June to August. At Johan Dahl Land and at Qarnanarssup sermia, the main glaciological measurements are made upon large outlet glaciers from the Greenland ice sheet while the measurements at Taser- siaq are made on local ice caps but in all three cases the study areas cover hundreds of square kilaneters. The data therefore refer to balances measured in sparse stake networks. TABLE 2: Net ablation deviations for Qamanarssup sermia for various combina- tions of stakes. Units are meters of water equivalent. YEAR A1 A2 A3 A4 1979/80 0.3 -0.3 1980/81 0.4 0.5 0.5 0.4 1981/82 0.4 0.3 0.4 0.3 1982/83 -0.4 -0.5 -0.5 -0.6 1983/84 -0.4 -0.4 -0.4 -0. 1 1984/85 1.1 0.9 1.5 A1 = Based on 6 stakes for 6 years A2 = Based on 11 stakes for 5 years A3 = Based on 13 stakes for 4 years A4 = Based on 3 stakes for 6 years The stations in Johan Dahl Land and at Qamanarssup sermia are located beside the lower ablation zone and, because of the difficulties of travelling and the distances involved, the stake networks only extend to the lower parts of the accumulation zone. The accumulation zone is also poorly delineated at both places because of the inaccuracies of mapping the surface topography on the ice sheet. For these two reasons, the data cannot be expressed in the area-averaged form gener- ally used in glaciology (Anonymous, 1969). Braithwaite (1985b & 1986a) used a simpli- fied version of the method by Lliboutry (1974) to analyse such data. The results are shown in Table 2 for various combina- tions of stakes. Although there are sane substantial differences between deviations for the same year, it is clear that 1984/85 had exceptionally high ablation while 1982/83 had low ablation. 486 52' 48' 44' 40' Figure 1. Locations of the three glacier-climate stations in West Greenland The measurements are supplemented by daily readings of ice ablation at stakes located close to the field stations in Johan Dahl Land and at QamanarssO.p sermia. The readings on these stakes are used for detailed correlations with the climatolog- ical readings at the field stations. Similar measurements are also attempted at the Tasersiaq station. However, this sta- tion is located in an area with more durable snowcover so that the measurements refer mainly to snow ablation and are less detailed than the ice ablation measure- ments in Johan Dahl Land and at Qamanarssup sermia. CLIMATOLCGY PRcx:;RAM The climatological program is designed to be simple but comprehensive enough for glacier-climate research. Recordings of air temperature and humidi- ty, precipitation, wind, sunshine and global radiation are made with clockwork recorders at the base camps, supplemented by hand readings twice a day, i.e. in the morning and the evening. Such a program gives daily averages or totals for the various elements and leaves the field team free for other work during the day. The combined glaciological and climatological programs can therefore be carried out by only 2-3 persons while a full program of synoptic observations would be more demanding. Several remote thermohygrograph and raingauge stations are operated around the ffise camps, e.g. on glaciers at the same elevation as the base camps to measure the glacier "cooling effect". These stations are visited at intervals of 7-10 days to recover data. The core period for the climatologi- cal observations is the ablation season which falls roughly in the three months June to August. However, in most years the hand-collected climatological observations extend from mid-May to mid-September. A Danish-designed automatic climate station has also been tested at Qamanarss- up sermia in parallel to the hand observations ( Braithwaite, 1983) • Six silliDErs of comparisons of data from auto- ~tic and manned stations have shown excellent agreement for elements like air temperature and wind speed (note: there is little or no rime formation in this area). lliere are still problems with the measure- ~nt of precipitation however, especially in the form of snow. Despite this, auto- ~tic climate stations will be relied upon more in the future as their relatively law running costs, compared to manned sta- tions, offset their high capital costs. llie good performance of the automatic sta- tion at Qamanarssup sermia has allowed the ~asurement of temperature through most of the winter periods so that ablation can be c~red with temperature on an annual ffisis as, for example, in Table 3. 487 TABLE 3. Annual ice ablation A (rrm water equivalent) compared with annual melting degree-day total T ( deg d) at Qamanarssup sermia. The glaciological year is from 1 September to 31 August. YEAR A T A/T 1979/80 4090 635 6.4 1980/81 4690 617 7.6 1981/82 4660 599 7.8 1982/83 3740 426 8.8 1983/84 4260 594 7.2 1984/85 5880 728 8. 1 Mean 4550 600 7.6 c.v. +16% +16% +10% SCME RESULTS Although the field program is still continuing same results can be briefly surrmarized. There is a useful, although by no means perfect, relation between ice abla- tion and air temperatures expressed as melting degree-day totals. The degree-day factor relating ice ablation to tempera- ture is a fluctuating parameter rather than a constant (Braithwaite, 1986b). However, it shows no systematic variation between Johan Dahl Land and Qamanarssup sermia, e.g. no simple variation with latitude; nor does it shaw sinusoidal sea- sonal variation as suggested, for example, by Gottlieb (1980) or Lundquist (1982). The variations of annual ablation and degree-days at Qamanarssup sermia are shown in Table 3 where A/T is the degree-day factor in rrm/deg d. This annual degree-day factor has a year-to-year fluc- tation of about +10 % ·the reasons for which are still not fully understood although Braithwaite & Olesen (1986) sug- gest that they may be a result of covariations between the various sources of ablation energy. There is only a weak correlation between ice ablation and global radiation in West Greenland although global radia- tion is the main source of ablation energy. Braithwaite & Olesen (1986) sug- gest that the long-wave radiation balance acts in opposition to the short-wave radi- ation to suppress correlations between ablation and radiation. This means that there will only be a slight benefit to including radiation in future ablation models. The relation between snow ablation and climate is more complex than for ice ablation so that less progress has been made. For example, it is much more diffi- cult to measure mass changes in the lower accumulation area than on ice because meltwater can percolate to great depths before refreezing. It is likely that data from Tasersiaq will be more useful for snow ablation studies than those from the two older stations where most of the data refer to ice ablation. The ubiquity of refreezing means that there is little or no runoff from the accumulation area of the ice sheet, i.e. there is a "runoff line" located close to the equilibrium line. It also means that the effects of re-frozen meltwater must be included in ablation models. There is a glacier "cooling effect" whereby the air over a glacier surface is cooler than the air over land at the same altitude. There is also an "inland" effect whereby air is heated between the coast and the edge of the Greenland ice sheet during the sunmer and is cooled during the winter. Both of these effects must be taken into account when calculating abla- tion on the Greenland ice sheet using temperatures extrapolated from the coast. APPLICATIONS Although there are still problems to solve, sane results have been incoporated in runoff models using temperature and precipitation data from weather stations on the coast of Greenland. For example, Braithwaite & Thomsen (1984a & 1984b) and Thomsen & J0rgensen ( 1984) have calculated ablation for planning hydropower stations near Christianshab and Jakobshavn, central West Greenland, where runoff from the ice sheet is the chief water source. The present emphasis in Greenland hydrology is upon the planning of hydro- power stations. Runoff models, including 488 glacier effects, are being used to artifi- cially extend the short measurement series to provide a more reliable basis for dimensioning the planned hydropower sta- tions. A decision may be made soon to build hydropower stations in Greenland, e.g. at Jakobshavn, to came on-line in the early 1990s. The priority for Greenland hydrology will then shift to problems of forecasting runoff. The continuation of the present glacier-climate research will contribute to the development of suitable forecasting methods. ACKNaV.LEIXMENTS This paper is published by permission of The Director, The Geological Survey of Greenland. The field station in Johan Dahl Land 1977-1983 was partly funded by the European Economic Camn.mity (EOC) and by the Energy Ministry, Denmark while the other stations were wholly supported by the Geological Survey of Greenland. The work in Johan Dahl Land in the period 1979-83 was lead by Cand. Scient. Paul Clement. REFERENCES Anonymous 1969. Mass-balance terms. J. Glacial. 52(8), 3-7. Braithwaite, R.J. 1983. Comparisons between automatic and manual climate stations at Qamnan~rssup sermia. Gr0nlands geol. Unders. Gletscher- -hydrol. Meddr 83/5, 17 pp. Braithwaite, R.J. 1985a. Glacier-climate investigations in 1984 at Qamanarssup sermia, West Greenland. Report of Activities 1984. Rapp. Gr0nlands geol. Unders. 125, 108-112. Braithwaite, R. J. 1985b. Relations between annual runoff and climate, Johan Dahl Land, South Greenland. Gr0nlands geol. Unders. Gletscher-hydrol. Meddr 85/2, 25 pp. Braithwaite, R.J. 1986a. Assessment of mass-balance variations within a sparse stake network, Qamanarssup sermia, West , Greenland. J. Glaciology. (In press). Braithwaite, R. J. 1986b. Exceptionally high ablation in 1985 at Qamanarssup sermia, West Greenland. Report of Activities 1985. Rapp. Gr0nlands geol. Unders. 130 (In press). Braithwaite, R.J. & Thomsen, H.H. 1984a. Runoff conditions at Kuussuup Tasia, Christianshab, estimated by modelling. Gr0nlands geol. Unders. Gletscher- -hydrol. Meddr 84/2, 24 pp. Braithwaite, R.J. & Thomsen, H.H. 1984b. Runoff conditions at Paakitsup Akuli- arusersua, Jakobshavn, estimated by modelling. Gr0nlands geol. Unders. Gletscher-hydrol. Meddr 84/3, 22 pp. Braithwaite, R.J. & Olesen, O.B. 1986. Ice ablation in west Greenland in rela- tion to air temperature and global radiation. Zeitschrift fur Gletscher- kunde und Glazialgeologie 20(2} (In press). Clement, P. 1984. Glaciological acti vi- ties in the Johan Dahl Land area, South Greenland, as a basis for mapping hydropower potential. Report of Activ- ities 1983. Rapp. Gr0nlands geol. Unders. 120, 113-121 . Gottlieb, L. 1980. Developnent and applications of a runoff model for snawcovered and glacierized basins. Nordic Hydrology 11(1980), 255-272. Gottlieb, L. & Braithwaite, R.J. 1985. Greenland case study: water supply. In Techniques for Prediction of Runoff from Glacierized Areas. Int. Assoc. Hydrol. Sci. Publ. 149, 73-80. Lliboutry, L. 1974. Multivariate sta- tistical analysis of glacier annual balances. J. Glaciology 69 ( 13), 371-392. Lundquist, D. 1982. Modelling runoff from a glacierized basin. In Hydrolog- ical Aspects of Alpine and High Mountain Areas . Int. Assoc. Hydrol. Sci. Publ. 138, 131-136. Olesen, O.B. 1985. Glaciological inves- tigations in 1984 at Tasersiaq and Qapiarfiup sermia, west Greenland. Report of Activities 1984. Rapp. Gr0nlands geol. Unders. 125, 104-107. Thomsen, T. & J0rgensen, G.H. 1984. Hydrological data-model in Greenland. Nordic Hydrology 15(1984), 39-56. 489 Weidick, A. 1984. Studies of glacier beha- viour and glacier mass Greenland a review. Annaler 66A(3), 183-195. balance in Geografiska JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 A FORECAST PROCEDURE FOR JOKULHLAUPS ON SNOW RIVER IN SOUTHCENTRAL ALASKA David L. ABSTRACT: Sno-v.' River jokulhlaups are outburst floods caused by the rapid release of water from a glacier-dammed lake. These events occur at two-to four-year inter- vals, but it has not been possible to predict accurately when a jokul- hlaup will begin. The forecast procedure presented here will not predict when a jokulhlaup will start; however, once a jokulhlaup has begun and is recognized as such, the procedure will forecast the hydrograph of the remainder of the event. Usually, some five days advance notice of the severity of expected flooding can be provided to downstream residents. The fore- cast procedure is based upon the observation that the cross- sectional area of the glacier- dammed lake outlet is proportional to the amount of water that has passed through the outlet since the beginning of the event. A log-log plot of a function of the cross- sectional area versus accumulated outflow results in a linear rela- tionship that can be determined early in the event and used to forecast the discharge through the rest of the jokulhlaup. A tabula- tion of the hydrographs of all fourteen of the Snow River jokul- hlaups that have occurred since 1947 is also presented. (KEY TERMS: jokulhlaup forecast; glacier-dammed lake; outburst.) 1 Chapman INTRODUCTION Snow River 1s located at the head of the Kenai River on the eastern side of the Kenai Peninsula in southcentral Alaska. Figure 1 shows the locations of pertinent features in the Kenai River basin. Snow River jokulhlaups are outburst floods caused by the rapid release of water frore a lake which was formed by the main valley glacier damming a tributary valley. During sunruer, runoff from rainfall, snow- melt, and some glacier melt accumu- late in the lake. In winter, snowfall adds depth to the lake, as does continued drainage of glacier meltwater from earlier seasons. In the course of two to four years, the lake fills to a depth of 80 to 160 meters and creates a hydro- static head sufficient to initiate the self-dumping process. On at least one occasion, this was when the height of the lake surface was about 0.9 of the height of the ice dam, which suggests that lifting of the dam may take place. Once begun, the flowing water enlarges its escape route and the discharge accelerates until the head is insufficient to support accelera- tion. At that point, the lake is nearly empty and the discharge drops abruptly. The volume of water stored in the lake at the beginning of jokulhlaup evects has 1 Hydrologist, National Weather Service, 701 C Street, Box 23, Anchorage, Alaska Alaskan River Forecast Center, 9 9513. 491 ~ .A Streamflow oagin') station ----Sub-basin drainage divide ---Assumed location· of s~b- glacial tunnel Glacier []SEWARD 10 0 10 20 HHAHHI I I SCALE OF KILOMETERS KENAI RIVER MSIN LOCATION 3() ' Figure 1. Location Map of the Kenai River Basin ranged from 96 million to 240 mil- lion cubic meters. The elapsed time from the time the outflow be- comes recognizable as a jokulhlaup until it reaches its peak rate is five to ten days; however, many other glacier-dammed lakes dump much more rapidly. With five or more days advance notice, a valu- able forecast of river levels at downstream locations can be made, which provides time for property owners to move items to high ground if needed, or saves them the effort and expense if the river levels will not be so high. This r~quires an accurate forecast of the jokul- hlaup hydrograph, not just its peak discharge. GLACIER-DAMMED LAKE CAPACITY One of the tools needed for a 492 jokulhlaup forecast 1s a storage- elevation relationship. Unlike common lakes and reservoirs for which fairly accurate storage versus pool elevation relationships can be determined from surveys, such relationships can change in glacier-dammed lakes. The lake may change s1ze due to changes in the damming glacier. Floating ice is equivalent to a change in storage since large amounts of ice are usually left in the lake at the end of a dump. As the lake approaches its dumping level, the ice dam domes up and arcuate cracks appear 1n the glacier surface near the lake. A considerable volume of water may be stored in or under the glacier. Therefore, even though the lake might have been surveyed at some time when it was empty, no firm storage versus pool elevation relationship can be developed. Instead, a gross estimate was made. The lake geometry was simplified to a truncated triangular pyramid de- signed to have approximately the correct surface area, head, and volume vhen full. The storage versus head relationship was re- d~ced to the simple equations: S=5.06(h+228)3 -6.00xl0 7 (1) 7 1/3 b=[(S+6.00xl0 )/5.06] -228 (2) ~1her e: 3 S is storage 1n m ; and h is head in m. Accuracy to three significant figures is not implied. Very little accuracy is claimed but the equations are necessary tools in the procedures that follow. Fortu- nately, storage and head are not very sensitive parameters in those procedures. THE SUBGLACIAL OUTLET The Snow River glacier-dammed lake drains from the bottom of the lake and the water issues from under the glacier terminus about eight kilometers down the valley. Little else is known about the lake outlet. It is convenient to think of it as a tunnel, part of which may remain open between jokulhlaup events. In any case, some mecha- nism plugs the outlet and causes the lake to fill. Some other mech- anism triggers the jokulhlaup, be it lifting, fracturing, or any of several things, that allows the water to begin flowing out. Once begun, the tunnel, or more specif- ically the hy<lraulic control, is enlarged by melting with most of the heat being derived from the potential energy of the very high head. The water temperature in the lake is probably very near freezing because the lake is always in contact with the glacier and there is at least partial ice cover at all times. Some erosion must also 493 occur by physical scouring due to high velocity flow. Post and Mayo (1971) state that the cross- sectional area of the tunnel is proportional to the volume of water that has already passed through it, providing no collapse or closure takes place. A tunnel eight kilometers long could hardly be called an orifice. Nevertheless, considering the small outlet relative to the very high head, the orifice flow formula applies reasonably well and is certainly convenient. The formula states: Q=CA(2gh)l/2 where: Q 1s discharge in m3 /s; C is coefficient of discharge; A 1s cross-sectional area of the orifice in 2. m ' 2 g is 9.8 m/s ; and h is head in m, the depth of water that produces discharge. ( 3 ) The initial head can be deter- mined by flying over the lake and estimating the lake level based on visible markers. The coefficient, C, and the area, A, are both un- known, but it is not necessary to determine them separately. It is only necessary to determine their product, CA, and the procedure for doing that follows. COMPUTATION OF SNOW RIVER JOKULHLAUP HYDROGRAPHS The glacier-dammed lake drains under the glacier for a distance of eight kilometers to the glacier terminus. From that point the water flows down Snow River a dis- tance of 48 kilometers to Kenai Lake. During some of the jokul- hlaup events, the U.S. Geological Survey operated a stream gage at the railroad bridge near the mouth of Snow River. The floo~ wave at that location is necessarily very similar to the hydrograph at the glacier terminus because Snow River flow is supercritical through most of its length, slowing down only ror a few kiloueters through Para- dise Valley and arr1v1ng at the gage about three hours after leaving the glacier. For those events when the gage was operating, the jokulhlaup hydrographs were clearly superimposed upon the other Snow River flow. Separation of the flows into jokulhlaup flow and non- jokulhlaup flow vas simple. Two other gages were used to derive the other jokulhlaup hydro- graphs. Kenai River at Cooper Landing is located on the Sterling Highway bridge at the outlet of Kenai Lake. Trail River near Lawing is near the mouth of Trail River which empties into Kenai Lake. Both gages began operation 1n 1947 and although the U.S. Geo- logical Survey discontinued opera- tion of the Trail River gage, the ~ational Ueather Service has main- tained records during the open water season. For those events v1hen no Snow River record is avail- able, a reverse routing procedure vas applied to Kenai Lake outflows, 1.e., the flows recorded at the Kenai River at Cooper Landing gage, to obtain a Kenai Lake inflow hydrogrnph. The inflow was separ- ated into Snow River flow and non- Snow River flow by applying drainage area ratio to the Trail River gage records to determine the ungaged area flows. In most cases there was no significant rainfall during the events and the non- jokulhlaup flow was steady. llaving the Snow River flow determined, that part attributed to lake drain- age was separated. Table 1 lists the volutes and peaks of all the jokulhlaups since 1947 along with the peak flows at the Snow River gage and at the Cooper Landing gage. Table 2 lists the daily-flo~ hydrographs of those jokulhlaups along with accuEulated outflow, the water volume rehlaining 494 in the [:; lac i er-da m;; e d 1 a k e , the head as computed by equation (2), and CA COJr:puted as explained belo1,, Except for the events of 1982 and 1985, the jokulhlaup hydrogrdphs were published in Chapman (1981) along with the resulting reconsti- tutions of Cooper Landing hydro- [raphs as proof of the effectuality of the methods used. COflPUTATION OF OUTLET TUNNEL EXPANSION There is no practical ~..ray to or otherwise determine the measure outlet tunnel dimensions during a jokulhlaup. It is a simple proce- dure, hov.,ever, to compute the prod- uct CA, ~vbich is a function of the cross-sectional area, by transpos- ing equation (3) as: 1/2 CA=Q/(2gh) Table 2 shov.·s a strnightforward accounting of the volurae of ~..rater retaaining in the lake at the end of each day. Applying equation (2) provides the head, h. This is the head that produces the instanta- neous end-of-day discharge, Q. Entering these values into equation (L;.) yields the value of CA. All of the values mentioned above plus the 24-hour outflo"' volun:e and the accumulated outflo\-: at the end of each day are shoun in Table 2. A log-log plot of CA versus accumulated outflo\" results in a straight line up to t!1e point at which or1r1ce control is lost. Figure 2 (following Table 2) illus- trates that relationship for two events when the Snow River gage was operating, which eliminates r.>.ost of the possibility of author bias. The different slopes of the lines show that different events have different rates of 1ncrease tn outflow fror.: the glacier. The apparent straight line relationship verifies that the cross-sectional are a of the t ut: n e 1 is prop or t ion a 1 to the volume of water that hr.s passed tt-.rough the tunnel since the ~ = c.n TABLE 1 . Snow River Jokulhlaups Volume and Peak Outflow With Peak Flow at Snow River Gage and at Cooper Landing. ---------------------------------------------------------------------------------------------------Glacier Dammed Lal{e Snow R nr Seward Kenai R at Cooper Landing --------------------------------------------------------------------------------------Estimated Est. Peak Peak Peak Year Jokulhlaup Vo~u~e 0u3f1ow F~ow Date F~ow Date Incl. Dates ( 10 m ) (m /sec) (m /sec) (m /sec) --------------------------------------------------------------------------~----------------- 1949 18 Oct -29 Oct 143.6 422 464 (1) 27 Oct 328 (2) 28 Oct 1951 05 Nov -20 Nov 96.2 311 326 (1) 17 Nov 177 ( 5) 18 Nov 1953 OL~ Dec -19 Dec 98.8 198 207 ( 1 ) 15 Dec 125 (5) 17 Dec 1956 18 Oct -01 Nov 129.8 354 365 (1) 29 Oct 207 ( 5) 30 Oct 1958 06 Oct -20 Oct 128.9 394 402 ( 1 ) 17 Oct 236 ( 5) 17 Oct 1961 28 Sep -08 Oct 175.2 5L~4 566 (3) 07 Oct ( 3) 396 ( 2) 08 Oct 1964 15 Sep -27 Sep 154.9 450 507 (2) 23 Sep L~02 ( 2) 24 Sep 1967 26 Aug -02 Sep 150.5 760 810 (4) 31 Aug 609 (2) 01 Sep 1970 08 Sep -24 Sep 188.7 481 504 (2) 22 Sep 343 (5) 23 Sep 1974 09 Sep -22 Sep 240.5 708 748 (2) 20 Sep 654 ( 2) 21 Sep 1977 29 Aug -08 Sep 1 5 1 . 1 394 !~7 3 ( 2) 05 Sep 422 (5) 06 Sep 1979 18 Oct -24 Oct 127.0 419 445 (1) 24 Oct 362 (6) 24 Oct 1982 17 Sep -01 Oct 177.6 388 456 ( 1 ) 29 Sep 439 ( 2) 29 Sep 1985 21 Nov -05 Dec 159.2 334 340 (7) 02 Dec 233 (6) 03 Dec NOTES: (1) Inferred by downstream gages. (2) Published by U.S. Geological Survey. (3) Peal{ flow and date shown are inferred by ~ownstream gages. The 1964 Surface Water Records of Alaska show peak flow of 708 m /sec on 30 Sep 1961. That date cannot be supported by Cooper Landing flow records. (4) Peak flow and date shown are 3 inferred by downstream gages. The 1970 Water Resources Data for Alaska show 1,557 m /sec on 31 Aug 1967. (5) Daily flows, not necessarily peak flows. (6) From unpublished National Weather Service records. (7) Measured by U.S. Geological Survey. TABLE 2. Snow River Jokulhlaups -Glacier-dammed Lake Daily and Accumulated Outflow, Computed Head, and CA. I Lake Out(low : Acc:ua : Volume I Date tat End : 2q-hr :outflow :Remaln1ns: Head ; CA lolDay:Volur~~ef 63 63 :a 3 /:~ec I 10°mJ I 1011 11011 , 11 , m2 .................................. f ....................... ; ........................... f ""'""'""'""'""'""'""'""':'"''"''"'""'""'""'""'""''"'I""'""'""'""'""'""'""':""'""'""''"'""'""'""'""' : : I 1949 I I I 18 Oct 19 Oct 20 Oct 21 Oct 22 Oct 23 Oct 24 Oct 25 Oct 26 Oct 27 Oct 28 Oct 29 Oct I o.o I o.oo o.oo I I 9.9 I 0.66 0.66 : I 19.3 I 1.13 1.79 I I 38.2 I 2.67 4.46 I I 82.3 : 4.53 8.99 I I Jqq,q I 9.69 18.68 I I 214.0 I 15.85 34.53 I I 288.0 I 21. H 55.67 I I 364.0 I 28.63 84.30 I : 422.0 : 34.25 118.55 I : as.o I 22.70 1q1.25 I I o. o : 2. 32 14 3. 57 I I I I 1951 I I I 05 Nov : 0.0 : 0.00 0.00 : 06 Nov : 1.4 I 0.05 0.05 l 07 Nov , 3.4 I 0.24 0.29 I 08 Nov I 7. 1 I 0. 49 o. 78 l 09 Nov , 9.9 l 0.7) 1.51 I 10 Nov !, 19.8 l 0.98 2.49 l 11 Nov 40.9 , 2.43 4.92 I 12 Nov !, 78.2 I 4.62 9.54 13 Nov 100.8 I 8.89 18.4] 14 Nov l 117.6 8.51 26.94 15 Nov , 169.8 11.80 38.74 16 Nov I 253.0 17.54 56.28 17 Nov I 28].0 26.18 82.46 18 Nov l 56.6 12.48 94.94 19 Nov I 7.5 1.22 96.16 20 Nov I 0.0 0.06 96.22 I 195 3 I g; g:~ i ~:~ 06 Dec I 8.5 07 Dec I 20.5 08 Dec I 40.4 09 Dec I 63.7 10 Dec I 89.0 11 Dec ! 115.9 12 Dec , 140.9 13 Dec I 165.7 14 Dec I 184.1 15 Dec I 160.0 16 Dec , 103.4 17 Dec , 42.5 18 Dec l 9.9 19 Dec l 0.0 1956 18 Oct 19 Oct 20 Oct 21 Oct 22 Oct 23 Oct 2q Oct 25 Oct 26 Oct 27 Oct 28 Oct 29 Oct 30 Oct 31 Oct 01 Nov 1958 06 Oct 07 Oct 08 Oct 09 Oct 10 Oct 11 Oct 12 Oct 13 Oct H Oct 15 Oct 16 Oct 17 Oct 18 Oct 19 Oct 20 Oct I I I . : I I I I l I I I I . . i I . . 0.0 2.3 8.5 24. 1 46.6 75-3 11 q. 5 161.5 205.0 250.0 299.0 300.0 56.6 Jij. 3 o.o I 0.0 I o. 3 I 11. o I 29.7 t .qs.ll I 71.6 1 :~u I 220.0 : 280.0 ; 3~0. 0 : 164.2 : 39. 1 I 2. 4 i 0.0 : I o.oo 0.06 0.37 1.10 2.ij5 q .5] 6. qa 8.89 11. 13 1]. 21 15.41 15.65 11.99 5.87 1.47 0.20 0.00 0.02 0.44 1. 35 2.81 5.24 7.78 12.01 15.90 19.55 23.71 27.99 10.50 2. 10 o. 38 0.00 0.01 0. 69 1. 83 3. 30 5.06 7.32 11.62 16.69 21. 38 27.06 27. 11 6.43 0. 32 0.0') 0.00 0. 06 0.4] 1. 53 3. 98 8.51 14.99 23.88 35.01 48.22 63.63 79.28 91.27 97. 14 98.61 98.81 o.oo 0.02 0.46 1.81 4. 62 9.86 17.64 29.65 45.55 65.10 88.81 116.80 127.30 129. qo 129.78 o.oo 0.01 0.70 2.53 5.83 10.89 18.21 29.83 46.52 67.90 9ij.96 122.07 128.50 128.82 128.91 H3.57 142.91 1ij1.78 139.11 13ij. 58 12ij. 89 109.04 87.90 59.27 25.02 2.32 0.00 96.22 96.17 95.93 95.44 94.71 9].73 91.30 86.68 77.79 69.28 57.48 39.94 13.76 1. 28 0.06 0.00 98.81 98.75 98.38 97.28 9ij.83 90.30 83.82 7ij.93 63.80 50.59 35.18 19.53 7. 54 1.67 0.20 0.00 129.78 129.76 129.32 127.97 125.16 119.92 112. 1 q 100.13 84.23 64.68 40.97 12.98 2. 48 0.38 o.oo 128.91 128.90 128.21 126.38 123.08 118.02 110.70 99.08 82.39 61.01 33.95 6.84 0. 41 0.09 0.00 11 ~. 7 114. 3 113.6 112.1 109.5 103.8 94. 1 80.0 58.7 28. 1 2.9 0.0 85.7 85.7 85.5 85.2 84.7 8~.0 82.4 79.2 72.9 66.5 57.3 ij2.3 16.3 1.6 0.1 o.o 87.4 87.4 87.1 86. 4 84.8 81.7 77.2 70.8 62.3 51.6 37.9 22.5 9.2 2.1 0.3 0.0 106.7 106.7 106.5 105.7 104.0 100.8 • 96.0 88.3 77 .s 63.0 ij 3. 2 15.4 3. 1 0.5 0.0 106.2 106.2 105.8 10~. 7 102.7 99.7 95. 1 87.6 76.2 60. 1 ]6.8 a.~ 0.6 o. 1 0.0 0.00 0.21 0. 41 0. 81 1. 78 ].20 q. 98 7.27 10.73 17.98 o.oo 0.0] 0.08 0. 17 0.24 0.49 1.02 1. 98 2.67 3.26 5.07 8. 79 15.83 0.00 0.05 0.21 0.50 0.99 1. 59 2.29 ].11 4.03 5.21 6. 75 7.62 7.70 o.oo 0.05 0.19 0.53 0.89 1.69 2.64 3.88 5.26 7.11 10.28 17.27 o.oo 0.01 0.24 0. 66 1. 08 1. 62 2. 54 3. S5 5.69 8.16 12.66 12.80 496 take Outflow I Accullt I Volume : Date :at End 1 2'~-hr !Outflow lRe~alnlnsl Head loS Day : VoA•J,e I 6 3 I 106 ,.3 I :m /sec : 10 m I 10 ra : I m --... ------: -------1--------1--·-----1---------: ------- 1961 28 Sop 29 Sop 30 Sep 01 Oct 02 Oct 03 Oct 04 Oct OS Oct 06 Oct 07 Oct 08 Oct 196~ 15 Sep 16 Sep 17 Sep 18 Sep 19 Sep 20 Sep 21 Sop 22 Sep 23 Sep 24 Sop 25 Sop 26 Sep 27 Sep 1967 26 Aug 27 Aug 28 Aug 29 Aug 30 Aug 31 Aug 01 Sep 02 Sep 1970 08 Sep 09 Sep 10 Sep 11 Sep 12 Sep 13 Sep 14 Sep 15 Sep 16 Sep 17 Sop 18 Sep 19 Sep 20 Sep 21 Sep 22 Sep 23 Sep 24 Sep 1974 09 Sep 10 Sep 11 Sep 12 Sep 13 Sep 14 Sep 15 Sep 16 Sep 17 Sep 18 Sep 19 Sep 20 Sep 21 Sep 22 Sep 1977 29 Aug 30 Aug 31 Aug 01 Sep 02 Sep 03 Sep o• Sep 05 Sep 06 Sep 07 Sep 08 Sep : l I I I I I I I I I o.o I o.oo o.oo : 175.15 I 131.5 : 14.2 0.56 0.56 I 174.59 I 131.2 : 90.6 5.33 5.89 I 169.26 : 128.5 : 155.6 11.~0 17.29 : 157.86 I 122.5 • 185.9 15.49 32.78 • Jij2.37 : Jlij.O j 214.0 16.6~ 49.ij2 I 125.73 , 104.3 I 266.0 20.31 69.73 : 105.42 , 91.8 3ij9.0 25.59 95.32 ! 79.83 ! 7ij.3 499.0 ]ij.69 130.01 • 45.1ij • ij6.9 261.0 41.59 171.60 : ].55 I 4.4 o.o 3.55 175.15 I o.oo o.o 0.0 0. 7 • aU I ~~~:~ ~.: 350.0 397. o I 396.0 I 85.0 I 25.9 I II. 3 I o.o : 0.0 a. 5 175.6 391.0 617.0 719.0 22.7 o.o 0.0 0.8 3. q 7.4 13.7 26.2 45.9 76.4 106.2 157.2 229.0 31 q .0 ]86. 0 4 30.0 299.0 84. q o.o 0.0 7.5 19.5 -9.6 86.9 121.6 167. 1 2]1.0 ]2ij.Q 459.0 592.0 510.0 Joij.8 0.0 0.0 38.2 78 .• 1]2.2 205.0 287.0 ]46 .0 ]9ij. 0 21-.0 79.6 0. 0 i : : I l I I I : : I . . ' I . i I I I I I : : : . . I . . . . I I . I : . . 0.00 0.01 0. 12 q .65 9.17 16.76 27.87 32.66 35.92 23.29 3. 74 o. 73 0.02 0.00 0.24 8. 32 zq. 11 40.61 62.88 13.70 0.02 0.00 0.02 o. 15 0. 44 0. 83 1. 54 2. 98 4.9~ 7.39 10.96 16.20 2].29 30.97 35.72 ]8.51 13.09 1. 69 o.oo 0.24 1. 05 2.]2 6. 2~ 8. 7 8 12.2] 16.6ij 23.24 32.78 46.ij9 55.78 ]2. ]0 2.4] o.oo 1. 96 4.65 8.91 13.95 21.5 3 28. 14 ]2.05 26. 18 10.79 2. 94 0.00 0.01 0. 13 4.78 13.95 30.71 58.58 91. zq 127. 16 150.45 15~. 19 15q. 92 154. 9~ 0.00 0. 24 8.56 33.27 73.88 136.76 150.46 150.48 o.oo 0.02 0. 17 0.61 1. q~ 2.98 5.96 10.90 1li: 29 29.25 45. q5 68. 7ij 99.71 1]5. 43 17 3. 94 187.03 188.72 o.oo 0.24 1. 29 3. 61 9.85 18.63 ]0.86 47.50 70. 7~ 103.52 150.01 205.79 238.09 240.52 0.00 1. 96 6. 61 15.52 29. ij7 51.00 79. 14 111. 19 137.37 148. 16 151. 10 I I I . . I : : . . : I ! : : 15ij. 94 154.93 15ij. 81 150. 16 1~0.99 124.23 96.36 6].70 21.78 -.-9 0.75 0.02 0.00 150.48 15o. 2q 1Q 1. 92 117.21 76.60 13.72 0.02 0.00 188.72 188.70 188.55 188. 11 187.28 185.74 182.76 177.82 170.43 159.47 14].27 119.98 89.01 53.29 14.78 1. 69 0.00 zqo. 52 240.28 23?.23 2]6.91 2]0.67 221.89 209.66 19].02 169.78 137.00 90.51 34.7 3 2.43 O.Ov 151. 10 1~9. 1 ~ 1~ 4. ij9 1]5.58 121.6] 100. 10 71.96 39.91 13.73 2.94 o.oo 120.9 120.9 120.8 118. 3 113.2 103.4 85.8 62.2 30.9 5.6 1.0 0.1 0.0 118.5 118. ij 113.7 99.2 72.0 16.2 0.1 o.o 1]8.3 1]8.3 133. 1 138.0 1]7.6 1]6.8 135.4 132.9 129. 1 12].4 11 q. 5 100.9 80.8 53.8 17.4 2. 2 0.0 162.2 162. 1 161.6 160.6 157.8 153.9 148.] Jij0.4 128.8 110.9 81.8 37.5 3. 1 0.0 118.8 117.8 115.2 110. 1 101.9 88. 3 68.5 42.] 16. 3 3. 7 0.0 0.00 0.28 1. 81 3. 18 ]. 93 4. 73 6. ?.7 9. 15 16. 46 28. 11 0.00 0.01 0.17 1. 66 3. 19 5. 73 8. 53 11. 37 16.09 o.oo 0.18 3. 72 a. 87 16.42 40. 35 0.00 0.02 0.06 0. 14 0.26 0.51 o. ~9 1.40 2. i 1 ]. 20 -.8] 7.06 9. 70 1]. 24 16. 19 0.00 0. 13 0. 35 0.88 1. 56 2.21 3.10 q. 40 6. 45 9.85 Jij. 78 18.8ij o.oo o. 79 1. 65 2.85 4. 59 6.90 9. 50 1]. 6d Table 2 (continued) ---------:--t ;;; ; -a~~ r i ~:--i-; ~ ~~;--i -;~ i: ;; --i -------i -------- D•t• :at End 1 24·hr ;outrlow lRea~alnln&l Head I CA :or D•Y I VoAu~o I 6 3 I 6 3 I I 2 t•'/3ec : 10 ra : 10 11 I 10 111 I 11 I 11 .................. : -------: ................ : --------: --... ------1-------: ............... .. 1979 18 Oc~ 19 Oct 20 Oct 21 Oct 22 Oct 23 Oct 24 Oct 1982 17 Sep 18 Sep 19 Sep , 20 Sep ; 21 Sep ; 22 Sop , 23 Sep I 24 Sep ; 25 Sep ; 26 Sep : 27 Sep : 28 Sep : 29 Sep ; 30 Sop : 01 Oct : : 1985 ! 21 Nov 22 Nov 23 Nov 24 Nov 25 Nov 26 Nov 27 Nov 28 Nov 29 Nov 30 Nov 01 Dec 02 Dec 03 Dec oq Dec 05 Dec 100 N E c:( u 10 : I I I I I I I I I o.o , o.oo 1 o.oo : 127.0-I 105.1 I o.oo 56.6 I 3.18 1 3.18: 123.86 I 103.2 I 1.26 212.0 : 16.15 19.33 I 107.71 I 93.2 I -.96 zan.o : 20.55 39.88 1 87.16 I 79.5 1.09 351.0 : 27.89 67.77 I 59.27 I 58.7 10.35 408.0 : 32.78 100.55 I 26.49 I 29.6 16.94 0.0 : 26.49 127.04 : 0.00 : o.o I I I I I I 0.0 : 0.00 0.00 : 177.56 : 132.8 23.2 , 1.00 1.00 I 176.56 I 132.2 30.9 I 2.35 3.35 I 174.21 I 131.0 40.5 : 3.08 6.43 I 111.13 I 129.5 44.5 I 3.67 , 10.10 I 167.46 I 127.6 65.1 : -.75 14.85 I 162.71 I 125.1 93.4 I 6.85 21.70 I 155.86 I 121.- 133.1 I 9.79 31.49 I 146.07 I 116.0 184.1 I 13.70 45.19 I 132.37 I 108.2 255.0 I 18.96 64.15 I 113.41 96.8 326.0 : 25.08 89.23 I 88.33 80.3 382.0 I 30.58 119.81 I 57.75 57.5 340.0 I 32.40 152.21 I 25.35 28.5 169.9 I 22.72 174.93 I 2.63 3.3 o.o I 2.63 177.56 I o.oo o.o I I : I o.o I o.oo o.oo I 3.1 I 0.13 0.13 15.0 I 0.78 0.91 .28.6 ' 1.88 2.79 45.6 : 3.21 6.00 55.5 I -.37 10.37 98.3 I 6.64 11.01 HLO I 10.34 27.35 183.8 ' 14.03 ' 41.38 226.0 : 17.71 : 59.09 269.0 I 21.40 I 80.49 334.0 I 26.03 : 106.52 311.0 I 27.84 I 134.36 143.0 : 19.61 : 153.97 0.0 : 5.27 : 159.24 I I + 1964 0 1970 159.2- 159.11 158.33 156.45 153.24 148.81 142.23 131. 89 117.86 100.15 78.75 I 52.72 I 24.88 I 5.27 I o.oo I I 123.2 I 123.2 I 122.7 I 121.7 I 120.0 I 117.6 I 113.9 I 108.0 I 99.6 I 88.3 I 1].6 : 53.4 I 28.0 : 6.5 : o.o. I I o.oo 0. 46 0.61 0.80 0. 89 1. 31 1. 91 2. 79 4.00 5. 85 8.22 11.38 14.39 21.13 0.00 0.06 o. 31 0.59 0.94 1.16 2.08 3.06 q .16 5.43 7,08 10.32 13.28 0.1~------_.--------~------~ 0.1 10 Accumulated Outflow, 10 6m3 Fiqure 2. Glacier-dammed Lake Accumulated Outflow versus CA. 100 497 beginning of the jokulhlaup event. It also provides a basis upon which a forecast procedure can be devel- oped. JOKULHLAUP FORECAST PROCEDURE In 1969 and in subsequent years, several lake-level markers were installed which are visible from the air. National Weather Service hydrologists periodically fly over the lake to determine whether it is near the level at which it may dump. If it has reached a high level, a jokulhlaup may be immi- nent, but it 1s not possible to predict just when it will occur. The frequency of aerial observation is then increased, while noting the lake level, photographing whatever might be of value in iater studies, looking closely at the lake edges for evidence of falling stage, and observing the glacier terminus for signs of lake drainage. At the River Forecast Center, the opera- tional, i.e., current, Snow River hydrograph is watched for a change that might indicate the start of the jokulhlaup. When such a change appears, an aerial reconnaissance will be made, weather permitting, to verify that the jokulhlaup has started. The river observer will be asked to make two or more stage readings per day through the event. On the second or third day, while the lake outflow is still small, the CA versus accumulated outflow relationship will already have been established and the total remaining jokulhlaup hydrograph can be forecast. The forecast proce- dure is simple, but tedious if done manually. A computer program has been developed which requires as input the observed Snow River flows up to date, the forecast non-jokul- hlaup flows (same as routine daily runoff forecast), and the beginning lake elevation or storage. The program computes the equation of the CA versus accumulated outflow relationship based on observed values to date. The program fore- casts by selecting a discharge expected 24 hours after the last observation or determination. The average of that selected discharge and the previous disch~rge deter- cines the volume of outflov for the day, which in turn determines the remaining lake volume. Applying equation (2) yields the head which corresponds to the selected dis- charge. A value of CA is obtained from equation (4). Another value of CA 1s obtained from the CA versus accumulated outflow rela- tionship. If the two values of CA are alike (within 1%), the selected discharge is deeued correct. If they are not alike, another dis- charge is systematically selected anc the procedure repeated until agreement is reached. The program then proceeds to the next day and the next. Eventually, agreement cannot be reached because orifice control will have been lost due to low head. The program then pro- jects the hydrograph in a straight line to zero flow, the line being sloped to contain the remaining lake volume, which is small. The forecast is updated daily as better definition of the CA versus accumu- lated outflow relationship evolves. JOKULHLAUP FOR~CAST EXAMPLE The 1965 jokulhlaup was chosen to illustrate the results of applying the forecast procedure, primarily because it was free of unrelated precipitation events and most other perturbations in the hydrographs. As it was very late in the season, the Kenai River gage ct Cooper Landing was the only gage in the upper Kenai River system that was still operating; all other stations had shut down for the winter. On 25 November, a suall r1se on the Kenai River was unex- plained by weather, although the weather did prevent flying a recocnaissance of Snow River. Kenai Lake inflows were deter- mined that would produce the ob- served flows at Cooper Landing. 498 All streams 1n the vicinity \·lere kno\'m to be low and steady, so the Kenai Lake inflows were assumed to be composed of steady flows frou Trail River, Snow P,iver belo\v the glacier, and other intervening are as • A 11 of the increase in f 1 011' was attributed to dumping of the glacier-dammed lake. All flov values so determined were entered into the jokulhlaup forecast pre- gram along with an estiuate of the lake storage. From earlier flights over the lake, it was estimated that the head on 25 November was pro b a b 1 y 1 3 0 or 131 meters • H ov;- ever, it was also observed that the glacier had extended into the lake area, and the storage would be less than would be computed by equation (1). The storage "'as estimated to be 160 uillion cubic meters; and the resulting head computed by equation (2) and shown in Table 2 is not accurate. Post-event anal- ysis of the hydrograph at Cooper Landing showed that storage esti- uate to be nearly correct. Figures 3 and 4 show the hydro- graphs of the first forecast, whict was made on 25 November, compare~ with the actual hydrographs at the glacier terminus and at Cooper Landing. Figure 4 shows the minor rise of 23-25 6ovember that alerted the River Forecast Center of the beginning of a hydrologic event suspected of being a jokulhlaup. That r1se was sufficient to estab- lish the CA versus acccmulated outflow relationship upon which tte forecast was based. While the first forecast was not especially good, the peak flow from the glacier-dammed lake was predicted within 10 percent, and the crest stage at Cooper Landing was very close. Over the next two days, 26 and 27 November, high wind produced a considerable setup, or wind tide, and transported some of the Kenai Lake ice cover past the telemetered Cooper Landing stream gage, which produced an anomaly in its hydro- graph • 0 u 1 S !~ o v ember , c a 1m c o c d i- tions returned, and a new forecast 300 / First forecast I , \ \ Vl ' 1 comnut~d on / 25 ~lov 1985~1 \ \ \ E20() 1 I I I \ \ \ I I \ I I 100 I I I I I \ \ \ I- I!) > .... 0:: Inferred \ actua 1 hydro-\ grar>h computed \ after the event 1 o~~_.~~~~~._~~~~~ 24 26 23 30 2 4 6 22 NOV 1985 DEC Figure 3. First Foreca5t and Inferred Actual Hydrographs of Glacier-dammed Lake Outflow of 1985. 12 Stage comruted from first forecast on 25 Nov 1985---../', ~ I I I. ' I I I I I Observed stage Hind and ice induced anomaly Figure 4. First Forecast and Observed Stage Hydrographs of Kenai River at Cooper Landing. 499 vas made based upon a better defined CA versus accugulated out- flow relationship. That forecast is not shown as it was nearly perfect, mostly coinciding with the actual hydrographs. As shown in Figure 4, the river stage, or gage height, is the height of the water surface above an arbitrary datu~; it is not nec- essarily a measure of the depth of the water. The gage telemetry reports in units of feet and the records are kept in feet. The rise shown in Figure 4 is very close to 5.0 feet, or 1.5 meters. CONCLUSION It has been demonstrated that, during a jokulhlaup on Snow River, the cross-sectional area of the outlet of the glacier-dammed lake 1s proportional to the amount of water that has passed through the outlet. That proportion differs from event to event but it reveals itself early in each event, and that enables forecasting the re- mainder of the jokulhlaup several days in advance of the peak flow. REFERENCES Chapman, D. L., 1981. Jokulhlaups on Snow River 1n Southcentral Alaska. NOAA Technical Memoran- dum, NWS AR-31, National Weather Service, Anchorage, Alaska. Post, A. and L. R. Mayo, 1971. Glacier Dammed Lakes and Outburst Floods 1n Alaska. Hydrologic Investigation Atlas, HA-455, U.S. Geological Survey, Washington, D.C. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 SUSPENDED SEDIMENT BUDGET OF A GLACIER-FED LAKE -EKLUTNA LAKE, ALASKA Jeffrey H. Coffin and William S. Ashton* ABSTRACT: Suspended sediment data from Ek1utna Lake, a glacier-fed lake in southcentral Alaska, were analyzed to determine the annual sediment budget of the lake. The data may be more valuable than previously published methods for estimating suspended sediment yield from small to moderate glacier-covered basins ln the region or in similar areas. Streamflow and sediment measurements were made on two major inflow streams and the lake outflow. Instantaneous values of discharge and total suspended sediment concentration (TSS) were used to develop regression relationships for each of the inflow streams. Observed suspended sediment concen- trations ranged from 0.15 to 570 mg/1 in the inflow streams and from 0. 50 to 36 mg/1 in the outflow. The suspended sediment entering Eklutna Lake for the oo~year period from October 1983 through September 1984 was 42,000 metric tons (46,000 tons). The amount leaving the lake during the same period was 940 metric tons ( 1030 tons) , indicating a trap efficiency of 98%. Good agreement was found with Brune's trap efficiency rela- tionship, but poor agreement was observed with Guymon's sediment yield relationship. (KEY TERMS: suspended sediment; sediment budget; trap efficiency; glacial lake; Eklutna Lake; Alaska.) INTRODUCTION Suspended sediment and water dis- charge measurements were made on two major inflow streams and the outflow stream of Ek1utna Lake during a one-year period. Information on behavior of suspended sediment within a glacier-fed lake was collected to assess environmental impacts of suspended sediment within a proposed hydroelectric reservoir on the glacier-fed Susitna River in southcentral Alaska. A reservoir simulation model was tested on Eklutna Lake and then applied to the proposed Susitna reservoir (Wei and Hamblin, 1986). Values of reservoir trap efficiency and regional sediment yield were computed for Eklutna Lake and compared with published values (Brune, 1953; Guymon, 1974). STUDY AREA Eklutna Lake is a glacier-fed lake near Anchorage, Alaska (Figure 1). Twenty percent of the lake's 287 square km (111 sq mi) watershed is covered by glaciers. The glacier-covered area is drained by two primary tributaries which merge approxi- mately 1 km upstream from the lake's southeast end. One tributary originates from Eklutna Glacier and was designated Glacier Fork in this study; the other was called East Fork. Twenty percent of East Fork's 104 sq km (40 sq mi) basin is glacier-covered, and fifty-four percent of Glacier Fork's 70 sq km (27 sq mi) basin is glacier-covered. Although these two creeks drain about 50% of the lake's watershed, they provide about 90% of the lake inflow (based on flow data for the current study period). Elevations within the Eklutna Lake drainage basin range from 260 to 2440 m (850 to 8000 ft) above the mean sea level. Eklutna Lake's depth reaches a maxi- *Respectively, Senior Civl Engineer/Hydrologist and Hydrologist, R&M Consultants, Inc., 5024 Cordova Street, Anchorage, Alaska 99503 501 / 150° / / APPROXIMATE SCALE BOO KILOMETERS ~--~--~~~~~--~ 0 200 400 600 0 100 200 FIGURE 1. mum of 60 m (200 ft) in several areas around the center, at a full-pool elevation of 264 m (R&M Consultants, 1982). The lake is approximately 11 km (7 mi) long and 1 km (0. 7 mi) wide (Figure 2). The lake is a reservoir for a 33-megawatt hydroelectric powerplant. The powerplant is supplied by a power tunnel which has its intake at the northwest end of the lake. The only other lake outlet is an uncontrolled spillway which has discharged only very infrequently since its construction in the 1950's. The lake water surface elevation typically ranges from a high of 264 m (868 ft) above sea level during September or October to a low of 252 m (828 ft) during June. Peak outflow for power production occurs during the winter and spring months 300 400 !500 MILES LOCATION MAP 502 (January-June) o The average hydraulic residence time of the lake (lake volume divided by mean annual inflow) is 1.8 years (R&M Consultants, 1982), and the residence time was 2. 0 years for the 1984 study period. FIELD DATA COLLECTION METHODS A stilling well, float, and strip-chart water-level recorder were installed _in East Fork · and Glacier Fork and operated through the open-water season of 1984o Periodic discharge measuremenU were made to determine stage-discharge relationships o A one-liter, depth· integrated water sample was collected twice per week nt each gaging station .,,,.. ,, \ ' ·.,·j -~ '.• i\,. I ' ' . ... ' \ . 20 5 10 FIGURE 2. VICINITY MAP SHO\HNG EKLUTNA LAKE WATERSHED AND POWERPLANT TAILRACE from early June through freeze-up (about mid-November) and analyzed for total suspended solids (TSS) concentration. A hand-held DH-48 sampler was used for sampling, and TSS determinations were made ~ith 0.45-micron filters using standard procedures for detection of nonfilterable residue (APHA et. al., 1981). Measurements of daily lake outflows 503 were provided by the Eklutna Hydroelectric Project, Alaska Power Administration. Depth-integrated samples were taken twice per week from the lake outflow in the powerplant tailrace on the same dates the inflow streams were sampled, and TSS concentrations were determined in the laboratory in the same manner as for the inflow samples. ANALYTICAL METHODS The TSS concentrations determined for each of the inflow streams were matched with the instantaneous streamflows at the times of sampling. These data pairs were then related to derive a regression equation of TSS as a function of dis- charge. The regression equations were used to predict mean daily TSS concen- trations from mean daily discharges for each stream for the open-water period. The daily values for discharge and TSS were then used to compute each site's daily sediment inflow to the lake. Changes in flow and TSS were assumed to be negligible between the gaging sites and the lake. During the ice-covered season, different relationships were used because the TSS and streamflow were both very low. \.Jinter flows were estimated from extrap- olated recessions of the fall hydrographs and from occasional winter discharge measurements, made approximately once per month. TSS concentrations were estimated by interpolating between concentrations sampled at the times of discharge measure- ments. Monthly-averaged TSS concentrations for the outflow were computed from the sampled concentrations, which had been obtained twice per week during the open-water season and once per month during the winter season. The total release of sediment from the lake was computed from the monthly outflow volume and the average TSS concentration (Table 1) • RESULTS Good relationships were observed between TSS and discharge on both inflow streams, with correlation coefficients of 0.88 and 0.91 for East Fork and Glacier Fork, respectively (Figure 3). Measured suspended sediment concentrations ranged from 0.15 to 570 mg/1 in the inflow streams and from 0.50 to 36 mg/1 in outflow (R&M Consultants, 1985). The sampled inflow data for East Fork and Glacier Fork are plotted in Figure 4. The graphs show mean daily discharge and instantaneous values of TSS concentration for each site, indicating the period June 504 through August to be most prominent for contributing flow and sediment to the lake. Figure 5 shows TSS concentration of the outflow for the same period. Little variation is seen around the base level of 2-4 mg/1, except for a few "spikes" in the plot in July and August. The outflow'! TSS concentration varied fairly smoothly through the year, as the lake diluted the high concentrations observed in the inflo~ streams. Table 1 summarizes all the inflow anti outflow suspended sediment data for thE study period on a monthly basis. Ex- amination of the monthly totals of sedi- ment transport (thousands of kg) reveal! that over 90% of the inflow sediment occurs during July and August and over 99% occurs during June through September. The outflow of sediment, however, is relative- ly uniformly distributed throughout the year. DISCUSSION Eklutna lake's trap efficiency of 98% is in good agreement with generalized curves of trap efficiency developed by Brune (1953). Brune's curves, which relate trap efficiency to hydraulic residence time, give a median trap effi- ciency of 97. Si;, \lith envelope curves at 95 and 100% for Eklutna Lake's 2-year residence time. Suspended sediment data from the Eklutna Lake basin show differences from the regional suspended sediment yield relationship developed by Guymon (1974) by an order of magnitude. Eklutna Lake's 42,000 metric tons (46 ,000 tons) of sediment input from 174 sq km (67 sq mi) above the two gaging stations equates to 248 metric tons (tonnes) per sq km (707 tons per sq mi). When each tributary is considered individually, Guymon's rela- tionship indicates theoretical yields of about 182,000 tonnes (200,000 tons) per year from East Fork and about 364,000 tonnes (400 ,000 tons) per year from Glacier Fork (using distances to the glaciers of 10 miles and 3 miles for East Fork and Glacier Fork, respectively). The values of Guymon's (1974) A computed (SO and 5.5 miles, 31 and 3.4 b) are at and below the low limit of Guymon's data, indicating the relationship may not Ut Cl Ut Table 1. Eklutna Lake Sediment Inflow and Outflow for \ilater Year 1984 MONTH OCT 83 NOV 83 DEC 83 JAN 84 FEB 84 MAR 84 APR 84 MAY 84 JUN 84 JUL 84 AUG 84 SEP 84 TOTALS Avg. Water Fl~w Rate (rn /sec) 4.6 1.3 1.4 0.9 0.8 0.6 0.7 3.0 10.6 25.0 30.2 7.4 INFLOW Avg. TSS Concen. (rng/1) 21.0 4.0 4.0 3.0 3.0 3.0 3.0 4.4 78.9 215.4 295.7 61.7 Sediment Inflow (SED IN) (1000 kg) 259 13 15 7 6 5 5 35 2,168 14,423 23,918 1,183 42,037 Avg. \-later Flo~ Rate (m /sec) 3.3 4.0 6.3 10.9 12.3 13.1 13.0 9.4 10.1 5.8 4.2 4.3 *Tgs concentration estimated with data from Water Year 1985. [Trap Efficiency = (SEDIN-SEDOUT) (SED IN) 42,037-894 = 98%] 42,037 OUTFLOW Avg. TSS Concen. (mg/1) 3.8* 4.6* 3.0* 2.5* 2.5* 2.0* 2.5* 2.5 2.0 11.0 10.0 5.6 Sediment Outflow (SEDOUT) (1000 kg) 34 48 51 73 74 70 84 63 52 171 112 62 894 BALANCE Cumulative Net Sediment Sediment Input Input (SEDIN-SEDOUT)(SEDIN-SEDOUT) ( 1000 kg) ( 1000 kg) 225 225 -35 190 -36 154 -66 88 -68 20 -65 -45 -79 -124 -28 -152 2,116 1,964 14,252 16,216 23,806 40,022 1,121 41,143 41,143 1000 800 600 400 200 100 80 60 40 " 20 m E 1g Ul 6 Ul 4 1- 2 1 .8 .6 . 4 .2 . 1 (S) 1ooo F 800 600 ~ 400 200 ~ 100 80 " 60 m 40 E Ul 20 Ul 1-10 8 6 4 2 1 (S) EAST FORK (S) N + + + ++ + + + • + (S) (S) (S) (S) (S) (S)(S)(S) (T) <r L(1 !D "(l)(f)(S) • + + • • • • + (S) (S) N .. • -4 2.2. TSS: 1.23 X 10 Q R 2 = 0.88 n: 75 (S) (S) (T) (S) (S) (S) (S) (S)(S)(S) (S) (S) (S) (S) (S)(S)(S) <r L(1 !D "(l)(f)(S) INSTANTANEOUS Q Ccfs) (NOTE: 1.0 CU FT/SEC IS 0.0283 CUM/SEC) ++ GLACIER FORK + + + + + + + + TSS:1.61 a•·•• R2 a 0.91 n: 72 (S) (S) .sJ (S) (S) (S) (S)(S)(S) (S) (S) (S) (S) (S) (S) (S)(S)(S) N (T) <r L(1 !D "m mtSl (S) (S) (S) (S) (S) (S) (S)(S)(S) N (T) <r L(1 !D 1'-00altSl INSTANTANEOUS 0 Ccfs) Figure 3. Regression Relationships Between Total Suspended Sediment Concentration and InsLJ.ntc:meous 1-'low for East Fork and Gidcier Fork, Using 1984 Data. 506 -en ~ 1000j u >- 8oo -'w ;r(!) 800 cc: < ZJ: 400 <u wen 200 ~-c 0 400 ' EAST FORK E 200 (f) 0 (f) 1- MAY JUN JUL AUG SEP OCT NOV 1984 -en 1000 ~ u >- 800 ...IW -(!) 800 <a: c< 400 ZJ: <U wen 200 ~-c Q 0 800 ' GLACIER FORK Cl E 400 (f) 200 (f) 1- 0 MAY JUN JUL AUG SEP OCT NOV 1984 Figure 4. Mean Daily Discharge and Instantaneous Sampled Concentrations of Total Suspended Sediment for East Fork and Glacier Fork for 1984 Open-Water Season. TSS TAILRACE MAY JUNE JULY AUGUST SEPT OCTOBER NOV 1984 Figure 5. Instantaneous Sampled Concentration of Total Suspended Sediment for Eklutna Lake Outflow (Tailrace) for 1984 Open-Water Season. 507 be valid for drainage basins this small, with this high a percentage of glacier cover, or gaged thjs close to the glaciers (all factors which contribute to a lolo! A. and a high theoretical sediment yield and which were outside the values of Guymon's data). Another possibility is that 1984 was far below average in terms of sediment yield at Eklutna, but this would not likely explain the order of magnitude difference. Guymon's report was the only prior study found which had published values for regional sediment yield in Alaska. The data presented here may be more valuable than previously published methods for estimating yield of suspended sediment from glacier-covered basins of small to moderate size (e.g. less than 250 sq km or 100 sq ni) in southcentral Alaska or in an area with similar geological and climatological conditions. The presenta- tion of the data on quantities of suspend- ed sediment entering and leaving the lake rr:ay also be useful for making estimates of sedimentation in other glacier-fed reser- voirs. Comparison with previous studies indicates that trap efficiency of a glacier-fed lake may be more easily extended to other study areas than is annual sediment yield of a glacier-covered drainage basin. ACKNOWLEDGHENTS The field work and primary data reduction for this project were performed for the Alaska Power Authority under contract to Harza-Ebasco Susitna Joint Venture for the Susitna Hydroelectric Project. The Alaska Division of Parks and Chugach State Park rangers were very gracious for permitting installation and operation of the field instrumentation 't\ri thin the park. The Eklu tna Hyd roelec- tric Project, Alaska Power Administration, was very helpful in providing data on lake outflow through the powerplant. Analyses of suspended sediment concentration were performed by Chemical and Geological Laboratories of Alaska, Inc., in Anchorage. He thank Jim Munter and an anonymous reviewer for their review comments. 508 LITERATURE CITED American Public Health Association (APHA), American Water Works Association and Water Pollution Control Federation. 1981. Standard methods for the ex- arnination of water and wastewater. Fifteenth edition, 1980. APHA, Washington, D.C. Brune, C .M. 1953. Trap efficiency of reservoirs. Trans. Am. Geophys. Union, June, p. 884. Guymon, G.L. 1974. Regional sediment yield analysis of Alaska streams. Journal of the Hydraulics Division, American Society of Civil Engineers, Vol. 100, No. HY l, Proc. Paper 10255. January 1974. pp. 41-51. R&M Consultants, Inc. 1982. Glacial lake studies, interim report. Prepar~ for Alaska Power Authority under con- tract to Acres American, Inc. December. R&M Consultants, Inc. 1985. Glacial lab physical limnology studies, Eklutn~ Lake, Alaska (Draft). Prepared for Alaska Pm.;rer Authority under contract to Harza-Ebasco Susitna Joint Ven- ture. Wei, C.Y. June. and P.F. Hamblin. 1986. Reservoir water quality simulation in cold regions. In: Proceedings ~ the Cold Regions Hydrology Symposium, American Water Resources .Association. July. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 ANNUAL RUNOFF RATE FROM GLACIERS IN ALASKA; A MODEL USING THE ALTITUDE OF GLACIER MASS BALANCE EQUILIBRIUM * Lawrence R. Mayo ABSTRACT: Glaciers in Alaska occur in high precipitation areas where the runoff is difficult to measure, yet hydrologically important. The spatial variability of glacier runoff is understood poorly. The equilibrium line altitude (ELA) of glaciers is related inversely to the average precipitation rate. Therefore, information about the average runoff from individual glaciers is contained in ELA data. This newly evaluated information about runoff is available from topographic maps. An ELA runoff model proposed determines average annual runoff from basins in Alaska. As a test the runoff rate was calculated for the Knik River basin, Alaska, using the model is 2.0 m/yr which compares with the average rate of 2.03 m/yr measured from 1959 to 1985. Applied to an ungaged site in Alaska, the Bering Glacier drainage basin, the ELA model indicates that 34 km3 of water is produced annually from this basin which contains Alaska•s largest glacier. Furthermore, Bering Glacier is the source of 76 percent of the discharge from the drainage basin and the average discharge of the Bering Glacier drainage into the Gulf of Alaska is about 1080 m3/s. (KEY TERMS: glaciers; runoff; Alaska; snow and ice melt; estimation technique.) INTRODUCTION Glaciers in Alaska occur in areas of high precipitation, most of which arrives as wind-blown snow. Measuring that precipitation is difficult. Much of the precipitation remains in storage for time periods ranging from a few hours to centuries, but eventually runs off. Runoff from glaciers is almost as diffi- cult to measure as is precipitation because glacier-fed rivers are shifting- bed, braided streams that are rarely stable gaging sites. Even though areas with glaciers are widespread and therefore important, relatively little is known about their runoff. Estimates of runoff from basins with glaciers at ungaged locations are often needed. For example, most of the glacier runoff from areas such as the St. Elias Mountains, cannot be gaged because part of the water flows via innumerable short, braided rivers to the coast, and part of the water flows sub-glacially directly into the ocean. Knowledge of that runoff is important, however, to other scientific problems such as understanding the dynamics of the coastal North Pacific gyre circulation (Royer, 1982), and developing reliable weather forecasting models. Multivariate regression analyses used to generate equations for estimating annual river discharge in Alaska have been proposed (Parks and Madison, 1985, p. 20). Their ana)yses obtained the highest success (r2=0.98) for gaged basins using drainage area and precipitation as independent variables. Basin area, however, is not a truly independent variable because discharge is equal to runoff rate times drainage area. Furthermore, the--precipitation map they used (U.S. National Weather Service, 1972) is not independent from discharge data because it was drawn by the present author to conform with the runoff data. *u.s. Geological Survey, 101 12th Ave., Box 11, Fairbanks, Alaska 99701. 509 Input for a runoff model must be independent from existing runoff data and must be influenced strongly by precipitation. The model results would be most useful if they could be obtained from existing, readily available information for extensive remote areas where the cost of field investigations is high. Runoff data from three research glacier basins, Wolverine, Gulkana, and McCall (Figure 1) are analysed in this paper to provide a simple model whereby glacier runoff can be estimated reliably in Alaska. A L A 5 K A 16cf Figure 1. Locations of Glaciers used in this study. GLACIER MASS BALANCE EQUILIBRIUM ALTITUDE 65" Information about spatial variations of precipitation and runoff rates is poten- tially contained in analyses of variations of glacier equilibrium line altitude. The equilibrium line altitude, or ELA, is the altitude at which snow accumulation is equal to snow melt. Meier and Post (1962) recognized that ELA is a function of precipitation. Increased snowfall tends to lower the ELA. Glaciers in a high precipitation climate receive considerably more snow than do glaciers in a drier, more continental climate. For a limited range of latitude, the average ELA in high precipitation areas is at relatively low altitude, and higher where drier. 510 Pewe and Reger (1972} and 0strem et al. (1981) used the relation of glacier ELA in Alaska to analyse precipitation patterns. However, ELA information has not been applied previously to studies of runoff. DETERMINATION OF ELA FROM TOPOGRAPHIC MAPS Time-averaged equilibrium line altitudes (ELA) of glaciers can be determined from topographic maps with varying degrees of precision depending on quality of the maps and the presence of mappable geomorphic indicators of glacier mass balance equilibrium. Using U.S. Geological Survey topographic maps of Alaska, the highest quality information is available at the largest scale, 1:63,360. These maps a~ readily available, accurate, and cover almost all glacierized areas of Alaska. The ELA is the altitude of the boundary between a glacier's accumulation zone and its ablation zone. These two zones can be identified on high-quality topographic maps. Snow accumulation together with glacier flow in the accumulation zone cause dust and rock debris on the glacier from surrounding mountains to submerge into the glacier. The surface is therefore almost always relatively clean, smooth snow. Friction at the glacier edge retards this submergence, leaving a thin up-turned glacier edge against the adjoining mountain slope. Ice from the accumulation zone flows into the ablation zone replacing (only approximately) the melt losses from the glacier. In this zone, the previously- deposited snow mass and rock debris emerges to the glacier surface. In 1 ate summer when the ablation zone is bare of snow and the old ice, mantled with accumulated rock debris, is exposed. This accumulation of rock is called supra- glacial till and forms medial moraines. On the air photos the topographic maps are compiled from, this rock debris usually appears black, in contrast to the white color of the ice, and is mapped with a stipple pattern. Friction at the glacier ablation zone edge retards emergence flow, leaving a narrow valley between the glacier and the adjoining mountain slope. The rock debris and edge valleys of the ablation zone, and the upturned glacier edges in the accumulation zone are not small features; they show prominently on high-quality topographic maps. The central part of Gannett Glacier (Figure 2) near Knik Glacier illustrates these features. 'E=:==:!i::_·S ==::E:==:====:===ii=:====::i3 KILOMETERS Figure 2. Central area of Gannett Glacier showing topographic features used to determine the glacier equilibrium line. Contour interval, 100 feet (30.48 meters). The glacier zones determined from topographic maps represent the average glacier mass balance conditions over a period of many years, probably decades, because significant glacier flow responses to changing climatic conditions requires time. It is, therefore, useful for analysing only average runoff conditions. RELATION OF ELA TO ANNUAL RUNOFF Research by the U.S. Geological Survey at Wolverine Glacier, Kenai Peninsula, and Gulkana Glacier, Alaska Range; and by the University of Alaska at McCall Glacier, Brooks Range provide measurements of the glacier component of runoff in Alaska. Runoff from the glaciers originates from snow and ice ablation, rainfall, and condensation. Annual ablation, the dominant source of runoff, was determined from glacier mass balance measurements. Runoff from alpine (non-glacier) areas of the basins is small because part of the 511 precipitation which falls on these areas as snow is redistributed onto the glacier by wind and avalanche action. Continuous streamflow data are available for the three glacier research basins. In this paper, runoff rate is expressed as a vertical velocity, m/yr (meters per year) for the average over the surface considered. This is similar to the practice of reporting precipitation and glacier mass balance rates as a vertical velocity. Runoff flux rate, or discharge, is runoff rate times area, expressed here as km3/yr (cubic kilometers of water per year) for the area specified. All runoff rates are considered to be averages in both space and time. Runoff measurements are available for several glaciers in Alaska. The U.S. Geological Survey has monitored Gulkana and Wolverine Glaciers from 1966 to the present. Continuous streamflow informa- tion is available during the period 1967 to 1978, so this analysis is limited to that period for those glaciers. The average measured runoff rate from all sources at Wolverine Glacier basin in southern Alaska, from 1967 to 1978, was 3.14 m/yr (meters per year) (U.S. Geological Survey, 1979). During this same period, the measured average ablation rate from snow, old firn, and ice on the glacier was 2.72 m/yr, or about 62 percent of the runoff. Rain and condensation on the glacier produce 25 percent of the basin runoff, and the alpine area contributes only 13 percent to the basin runoff. The average ELA, from yearly measurements during the period, was 1150 meters. At Gulkana Glacier, in the Alaska Range of interior Alaska, the measured average runoff rate from 1967 to 1978 was 1.97 m/yr (U.S. Geological Survey, 1969) indicating that the climate at Gulkana is drier than at Wolverine, but still quite maritime compared with most of interior Alaska which typically has only 0.25 m/yr average runoff. The average ELA at Gulkana during the same period was 1180 meters. The total glacier area in the basin is 22.2 km2 of which Gulkana Glacier is 19.3 km~. Measured average ablation of Gulkana Glacier was 1.88 m/yr for the same period. This ablation with rainfall added resulted in the glacier runoff rate of 2.38 m/yr, 85 percent of the measured runoff. Rain on the glaciers was the source of 18 percent of the runoff. The alpine area produced only 15 percent of the runoff. Glacier ablation and runoff rates at McCall Glacier in the Brooks Range were measured in 1969 and 1970 by Wendler et al. (1973). Although the data cover only about half the ablation season for each of only two years, the information is quite useful because it indicates the magnitude of glacier runoff in a climatic setting that is different from that at either Gulkana or Wolverine Glaciers. The glacier ablation volume rate reported here (Table 1) is calculated from data presented by Wendler et al. (1973, p. 421- 422). They also measured the runoff rate, which they reported was 9.1 liters per second per square kilometer, which is equivalent to 0.28 m/yr. An error in determining glacier area may have been made (Trabant, oral communication, 1986), so the glacier and basin areas reported here were determined by this author. The total area of glaciers in the basin is 10.1 km2, of which McCall Glacier is 7.2 km2. McCall Glacier produces only 0.62 m/yr runoff and the non-glacier area produces only 0.11 m/yr. In this rela- tively dry climate, the ELA of 2050 m is much higher than at either Gulkana or Wolverine Glaciers. Table 1. Equilibrium line altitude and runoff measurements for three glaciers in Alaska. The alpine area is the non- glacier area. Measurements at Gulakana and Wolverine Glaciers by the author; at McCall Glacier by Wendler et al. (1973. ============================================================================= Wolverine Gul kana McCall 1967-1978 1967-1978 1969-1970 Equilibrium Line Altitude 1150 m 1780 m 2050 m Glacier m/yr km3/yr % m/yr km3/yr % m/yr km3/yr % Measured Ab 1 at ion 2. 72 .0482 62 1.88 .0417 67 0.57 .0058 68 Estimated Rainfall 1.10 .0195 25 0.50 .0111 18 0.05 .0005 6 Ca 1 cul a ted Runoff 3.82 .0676 87 2.38 .0528 85 0.62 .0063 74 Basin Est. Alpine Runoff 1.40 .0097 13 1.00 .0094 15 0.11 .0022 26 Measured Runoff 3.14 .0773 100 1. 97 .0622 100 0.28 .0085 100 km2 % km2 % km2 .. Glacier Area 17.7 72 22.2 70 10.1 33 A 1 pine Area 6.9 28 9.4 23 20.1 67 Basin Area 24.6 100 31.6 100 30.2 100 512 The glacier and alpine components of the measured runoff were estimated (Table 1) from measurements of R, the basin runoff rate based on the equation: (1) where: N is non-glacier alpine runoff rate; Sn is the non-glacier alpine area; G is glacier runoff rate; and Sg is the glacier area. If glacier mass balance data are available, the glacier runoff,G, can ~ estimated as follows: G = A-K+C+Pr ( 2) where: A is measured annual ablation; K is freezing of water in the glacier; C is condensation on the glacier; and Pr is estimated annual rainfall. Freezing of water in a glacier removes liquid from that available for runoff, and condensation adds to runoff. Both quantities are relatively small. In this analysis, condensation and internal freezing are not considered further, because they tend to cancel each other for runoff, although they can be important to glacier mass and heat balances. The estimated quantities, N and Pr, are not uniquely determined, but the total of the two estimated quantities is known, so the estimates are constrained by the data. The estimated values of N and Pr were adjusted (Table 1) so that the calculated runoff is equal to the measured runoff. An assumption made in the process is that rain runoff from glacier and alpine areas are approximately equal because rain is not redistributed by wind to the extent that snow is. Furthermore, alpine runoff must be larger than glacier rain runoff because there is, in addition, snow melt runoff from alpine areas. The resulting uncertainty in the calculated glacier runoff is small because the snow and ice ablation from glaciers dominates the runoff from these basins. The interpreted runoff and ELA data (Table 1) from the research glaciers indi- ~tes a well-defined inverse relationship of runoff rate to ELA exists in Alaska, as shown in Figure 3. "' <: w > "' 4 ::' "' "' ~ 3 w l! z "- 0 z :J "' 0 GLACIER' ~WOLVERINE ALPINE -----r ... .,._ ---... ,, I McCALL ' 1000 2000 EQUILIBRIUM LINE ALTITUDE, IN METERS 3000 Figure 3. Relationship of average glacier equilibrium line altitude to the average runoff rate. Dashed line is the runoff from the alpine part of the basin adjacent to the glacier. PROPOSED ELA RUNOFF ESTIMATION MODEL It is proposed here that the relation- ship of runoff rate to ELA shown by Figure 3 is a general relationship over much of A 1 as ka and it is, therefore, use- ful for calculating glacier runoff. The ELA Runoff Mode 1 proposed is the app 1 i ca- tion of the measured relationship between [A and runoff to other basins in Alaska. The ELA Runoff Model is developed as follows: The basin average runoff rate, R, is calculated by first determining the glacier and alpine runoff rate from the graph (Figure 1) for each hydrologically different area in the basin; determining the area of each unit from the map; multi- plying runoff rate by area to obtain runoff volume rate; and then dividing by the total basin area. If a basin has forested lowlands, the runoff of that part of the basin must be estimated from other studies. Thus, equation applies, for (1) that: a basin stated with glaciers, previously here ( 1) 513 For a single glacier or a set of glaciers with a representative ELA, the annual glacier and alpine runoff rates, G and A, are determined from Figure 3. ELA variations within a drainage basin are so large in some cases that no single ELA is representative. In this situation, the basin can be sub-divided into zones where representative ELA values can be obtained. The basin average runoff rate is calculated by first determining the runoff volume rate for each subdivision; then the total runoff volume rate is accumulated before dividing by basin area. If a basin has forested lowland areas, the runoff from those areas must be estimated from other studies. The ELA runoff model provides time-and area-average runoff rates, which is new hydrologic information for vast ungaged areas of Alaska. It should be recognized, however, that the is quite limited. In this model, for example, annual variations of glacier runoff have not been demonstrated to be related to annual variations in ELA. Furthermore, runoff calculated using the ELA model cannot be used directly to construct contour maps of runoff rate because glacier runoff varies inversely with altitude on individual glaciers. Snow and ice melt is the primary source of glacier runoff and the melt rate generally decreases with altitude. Additional techniques must be developed before glacier runoff contours over large areas of Alaska can be drawn using ELA as part of the argument. The estimated accuracy of the model is not known, but cannot be more accurate than about0.2 m/y ELA RUNOFF MODEL TEST AT A GAGED BASIN As a test of the ELA runoff model, runoff was calculated for the Knik River basin near Palmer, Alaska. Water flowing from the Knik River Basin has been gaged continuously since 1959 (U.S. Geological Survey, 1959-1985). In the Knik basin, glacier equilibrium altitudes range widely from 670 m in the southern part of the basin to 2100 m in the northern part of the basin (Figure 4). This large variation of ELA indicates that a strong 1340 1500- EXPLANATION Equilibrium line altitude, in meters Equilibrium line altitude contours; interval 100 meters Knik River Basin boundary (above gaging station) --------c=c-=:,.:> ...... - _ Glacier boundary // k::...::::J.!f=.=:=i=""""'=:i<10======15 KILOMETERS +61°00' 148' Figure 4. Distribution of the equilibrium line altitude in the Knik River Basin. Altitudes in meters. precipitation shadow occurs in the Marcus Baker Glacier area and that precipitation and runoff are probably quite high in the southern part of the basin. The Knik basin can be divided into three parts (Figure 4), the North sub-basin with the highest ELAs, the Knik Glacier sub- basin dominated by Knik Glacier, and the south sub-basin which has low ELAs. Each sub-basin contains a major glacier, other small glaciers, rock with alpine tundra vegetation, and forested lowland. To apply the ELA model, the areas of glacier, alpine, and forested terrain in each sub-basin were measured from the map and shown in Table 2. The average ELA's 514 of glaciers in each sub-basin were determined from the ELA graph (Figure 4); then used to determine runoff rates (Figure 5). ELA contours represent an imaginary surface above which glaciers form only where the land or glacier surface is higher. Runoff from the forested lowlands was estimated to ~ approximately half that from alpine areas because the annual precipitation measured nearby at Palmer is only 0.35 m/yr. The accuracy of this estimate is relatively unimportant, however, because the runoff from the forested lowlands is only 6 percent of the basin total. GLACIER ..... LAKE GEORGE .... COLONY "'i~LVERINE "'•KNIK ~GULKANA ~ARCUS BAKER ALPINE -----I--._ FORESTED --.... ! LOWLAND•••• .. ••• ' •••"•"•••••••M••• ':·~l(McCALL ... I .. 0 ~o~~~~1~000~~~~=2000~~~~~~00 EQUILIBRIUM LINE ALTITUDE, IN METERS Figure 5. ELA model interpretation of runoff rate for Knik River Basin using glacier equilibrium line altitude informa- tion (see Figure 4). During the calculation process (Table 2) high accuracy is shown only to avoid committing rounding errors during the calculation. Reportable results should be rounded. The ELA model calculation (Table 2) indicates that the basin average runoff rate is 2.0 m/yr. The measured 24-year average runoff rate for the Knik River from 1960 to present is 2.03 m/yr. This remarkable agreement, at least for this one case, indicates that the technique potentially can produce valid results. Other interesting results, such as the percentage of runoff from each identified source, also come from the ELA model at Knik River basin. Glaciers occupy 47 percent of the basin area yet contribute 77 percent of the runoff. Glaciers occupy high local precipitation areas and also serve as effective precipitation traps in the windy climate. Furthermore, the south sub-basin, which is closer to the coast, contributes more water than the larger north sub-basin. Knik Glacier (14 percent of the basin) is the source of 23 percent of the river flow. ESTIMATION OF GLACIER RUNOFF FROM AN UNGAGED BASIN The ELA runoff model can be applied to other hydrologically important areas in Alaska, such as the St. Elias Mountains, where runoff measurements are sparse yet considerable runoff occurs. The basin chosen for an application example is the 515 Table 2.--Average annual runoff from Knik River Basin, Alaska, calculated by the ELA runoff model. ELA and area data from Figure 4 (map); estimated runoff (one decimal) from Figure 5 (graph); calculated runoff (two decimals). All calculations carried out to full accuracy. Apparent errors in totals are not real, but are due to rounding. ============================================================ Data from map Model predicted A~EA ELA RUNOFF RATE km % m m/yr km3/yr % NORTH SUB-BASIN Marcus Baker Glacier 270 9 1830 2.1 0.57 9 Other Glaciers 160 5 1500 3.2 0.51 8 Alpine 760 25 0.9 0.68 11 Forested lowland 500 17 0.45 0.23 4 Subtotal 1690 56 1.18 1.99 33 KNIK GLACIER SUB-BASIN Knik Glacier 430 14 1470 3.2 1.38 23 Other Glaciers 20 1 1450 3.3 0.07 1 Alpine 60 2 1.2 0.07 1 Subtotal 510 17 2.97 1.51 25 SOUTH SUB-BASIN Colony Glacier 160 5 950 4.1 0.66 11 Lake George Glacier 200 7 830 4.3 0.86 14 Other Glaciers 180 6 1170 3.7 0.67 11 Alpine 140 5 1.6 0.22 4 Forested lowland 140 5 0.8 0.11 2 Subtotal 820 27 3.07 2.52 41 KNIK RIVER BASIN SUMMARY Glaciers 1420 47 3. 31 4.70 78 Alpine 960 32 1.02 0.98 16 Forested lowland 640 21 0.53 0.34 6 TOTAL 3020 100 1.99 6.02 100 Measured runoff 2.03 m/yr Bering Glacier basin. Runoff from the basin would be particularly difficult if not impossible to gage because the glacier runoff forms 12 rivers that flow into the Gulf of Alaska along 100 km of shoreline before they can gather into a single large river. Bering Glacier, 6540 km2 area, is a complex system of ice streams (see Figure 6) that includes the Bagley Icefield and the Stellar Glacier as a tributary. Even without considering the Stellar tributary, 830 km2, the Bering is still the largest glacier in North America. The ELA rises dramatically with increasing distance from the ocean (Figure 6), indicating that precipitation decreases significantly over a relatively short distance. A small, low altitude basin, Dick Creek, at the west edge of the basin, gaged from 1970-1981, had an average of 5.9 m/yr 1340 Equilibrium line altitude, in meters -··--Bering Glacier Basin boundary ~?::.-_:-Glacier boundary --800 Equilibrium line altitude contours; interval 200 meters --~ Figure 6. Equilibrium line altitude in the Bering Glacier basin. Contour interval, 200m. runoff (U.S. Geological Survey, 1982). Precipitation at nearby Yakutat (2.1 m/yr) and Yakataga (3.1 m/yr) average 2.6 m/yr. The sparse precipitation and runoff measurements indicate that the average runoff rate from the forested lowlands in this region is probably very high, between 3 and 4 m/yr. To apply the ELA Runoff Model to the Bering Glacier basin, the area was sub- divided into 3 glacier and 2 non-glacier areas. The Bering Glacier and Stellar Glacier tributary were separated because the ELA of the Bering is 1100 m, whereas it is only 640 for the Stellar. Other small glaciers (210 km2), alpine, and forested lowland areas were also consider- ed separately. ELA model calculations (Table 3) indicate that the Bering Glacier basin produces a considerable amount of water, 34 km3/yr (1080 m3;s, or 38,000 ft3/s), 79 percent of which comes from the glaciers. The average runoff rate indicated by the ELA model for the Bering basin is 3.8 m/yr. Royer (1982, p. 2018) estimated the average runoff rate to be 2.4 m/yr for the same region based on sparse precipitation and oceanographic data, but concluded that this was probably an underestimate of the actual amount. 516 Table 3. Average annual runoff calculated for the Bering Glacier drainage area. The equivalent average river discharge rates are also given in units of cubic meters per second, m3/s, and cubic feet per second, ft3/s, for comparison. GLACIER SYSTEM Bering tributary Stellar tributary Bering/Stellar Other glaciers TOTAL NON-GLACIER AREAS Alpine Forested lowland TOTAL BASIN TOTAL AREA ELA RUNOFF RATE km2 % m m/yr km3/yr % m3/s ft3/s 5710 63 1100 3.9 22.3 830 9 640 4.5 3.7 6540 72 4.0 26.0 210 2 910 4.1 0.9 6750 74 4.0 26.9 550 6 900 1.6 0.9 1770 20 3.5 6.2 2320 26 3.1 7.1 65 707 . 25000 11 117 4100 76 824 29100 3 29 1000 79 852 30100 3 29 1000 18 196 6900 21 225 7900 9070 100 3.8 34.0 100 1078 38000 DISCUSSION The relationship of runoff to ELA, Figure 3, suggests that the highest expected ELA in Alaska is about 2100 m. No glaciers in Alaska have ELA•s higher than 2200 m. In a study of the refreezing of water in glaciers, Trabant and Mayo (1985) showed that almost all glacier meltwater above 2100 m altitude refreezes in cold, permeable glacier firn where it remains as glacier accumulation, which explains the altitude limit. Rock areas above about 2100 m altitude shed most precipitation by avalanching rather than ~ liquid runoff. Thus, little liquid runoff occurs from glaciers or alpine areas above that altitude. Most precipi- tation above 2100 m flows by avalanching and glacier ice motion. Precipitation in mountainous areas tends ~ increase with altitude. This general- ization does not, however, apply well to Bering Glacier. The ELA runoff model of the Bering area indicates that precipi- tation and runoff from this large glacier are less than the runoff measured at Dick Creek, the small, lower altitude sub- basin. Thus, the zone of maximum precipi- tation adjacent the Pacific Ocean in this part of Alaska may be as low as only 500 m altitude. The ELA runoff model for Alaska can be expected to give good estimates of runoff from ungaged basins that contain glaciers. The model as calibrated in this paper is restricted to the latitude band from 600 to 700 North in northwestern North America, and can be expected to over- estimate runoff north of this band and underestimate runoff to the south. Where glacier ablation and runoff data are available for other areas, the ELA runoff oodel could be recalibrated. Among the five basins studied in Alaska, snow and ice melt produced substantially oore runoff than did rainfall, and glaciers produced from 74 to 87 percent of the tota 1 runoff. The ELA model provides a means of esti- ~ting the runoff from all glaciers in Alaska, which can be compared to the tota 1 runoff of the nation. Glaciers occupy 73,000 km2, or about 5 percent of Alaska. The average ELA of these is not known precisely, but is between that of Gulkana and Wolverine Glaciers. Thus, the average runoff rate from Alaskan glaciers is approximately 3 m/yr, or 220 km3/yr. Average runoff from the conterminous states has been estimated to be 1,230 billion gallons per day (U.S. Geological Survey, 1984), or 1550 km3;yr. R.D. Lamke (oral comm., 1986) estimated the average runoff from Alaska to be 620 km3/yr. Bering Glacier produces 4.2 percent of Alaska • s runoff and 1. 2 percent of the 517 national runoff including Alaska. Glaciers in Alaska produce approximately 35 percent of Alaska's runoff; an amount equal to 14 percent of that from the conterminous states; and 10 percent of the total runoff from the nation. ACKNOWLEDGMENTS The author thanks T.C. Royer, Chester Zenone, and R.M. Krimmel for offering many helpful suggestions in this study. Krimmel, using Landsat images, verfied the Knik basin ELA map (Figure 4). LITERATURE CITED Meier, M.F., and A.S. Post, 1962. Recent variations in mass net budgets of glaciers in western North America. Int. Assoc. Scientific Hydrology Pub. 58/A:251-260. 0strem, G., N. Haakensen, and T. Eriksson, 1981. The Glaciation Level in Southern Alaska. Geograf. Ann. 63/A:251-260. Parks, B.P., and R.J. Madison, 1985. Esti- mation of Selected Flow and Water- Quality Characteristics of Alaskan Streams. U.S. Geological Survey Water- Resources Investigations Report 84-4247. P~w~, T.L., and R.D. Reger, 1972. Modern and Wisconsinan Snowlines in Alaska. 24th International Geological Congress, Montreal, Proceedings 12:187-197. Royer, T.C., 1982. Coastal Fresh Water Discharge in The Northeast Pacific. Jour. Geophysical Research 87:2017-2021. Trabant, D.C., and L.R. Mayo, 1985. Esti- mation and Effects of Internal Accumula- tion on Five Glaciers in Alaska. Annals of Glaciology 6:113-117. U.S. Geological Survey, 1959-1985. Water Resources Data, Alaska. U.S. Geological Survey Hydrologic Data Series, Published Annually. U.S. Geological Survey, 1984, National Water Summary 1983--Hydrologic events and issues: U.S. Geological Survey Water Supply Paper 2250, 243 p. U.S. National Weather Service, 1972. Mean Annual Precipitation-Inches; Alaska. Map 73-2446. 1 sheet. Wendler, G., D. Trabant, and C. Benson, 1973. Hydrology of a Partly Glacier- Covered Arctic Watershed. Int. Assoc. REMOTE SENSING 519 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 SEASONAL AND INTERANNUAL OBSERVATIONS AND MODELING OF THE SNOWPACK ON THE ARCTIC COASTAL PLAIN OF ALASKA USING SATELLITE DATA Dorothy K. Hall, Alfred T. C. Chang, James L. Fosterl ABSTRACT: Snow is a dynamic com- ponent of the global hydrologic cycle. Measurement of approximate snow extent and depth on a global scale is possible using passive microwave data from the Nimbus-7 Scanning Multichannel Microwave Radiometer (SMMR). Improvement of snow depth measurement is dependent partly upon our ability to recognize the influence of snow structure on microwave signatures. Multiple linear correlations of microwave brightness temperature (TB) from the SMMR, snow depth and air temper- ature have been performed for the Arctic Coastal Plain of Alaska. The portion of the TB variability that cannot be explained by snow depth and air temperature is most likely due to physical differences in the snowpacks among the 4 years studied. Results from a snow energy balance model show that the temperature profiles of the snow- packs on the Arctic Coastal Plain can be quite different between years. The depth hoar layer which is comprised of large crystals is influenced by the snowpack tempera- ture gradient. The presence and variability of a depth hoar layer may be detected by passive microwave satellite sensors and can be modeled using a radiative transfer model and a snow energy balance model. (KEY TERMS: depth hoar, passive microwave, SMMR, snow depth.) BACKGROUND The world-wide snow cover is one of the most dynamic and poorly understood features on the Earth's surface. It is quite variable in terms of depth, density, duration and stratigraphic characteristics. Through the use of remote sensing techniques we are gaining an improved understanding of snow extent and depth on local, regional and especially global scales and the role of snow in the Global Hydrologic Cycle. Physically-based models have been developed which allow us to estimate snow depth, and to simu- late changing snow structure and snowmelt processes. In this paper, some of the major sources of error in the determination of snow depth on the Arctic Coastal Plain of Alaska using microwave radiometry are assessed. Model- ing is shown to be an effective tool in accomplishing this. The Arctic Coastal Plain of Alaska is a good region in which to study annual and interannual snow characteristics because it is a large area that is flat and treeless. The snow cover on the Arctic Coastal Plain of Alaska is quite shallow and generally begins to form by early October and lasts through May. Depth hoar which is common in the snowpack on the Arctic Coastal Plain is comprised of large lLaboratory for Terrestrial Physics, Code 624, NASA/Goddard Space Flight Center, Greenbelt, MD 20771 521 (1 to 15 mm in size) crystals that form as a result of a temperature gradient in the snowpack (Benson et al., 1975). The physical tempera- ture at the base of the snowpack is warmer than at the surface of the snowpack. Vapor diffusion (a result of the temperature gradient) occurs from the lower to upper portions of the snowpack where sublimation from crystals in the lower layers gives way to redepo- sition in the upper layers. The thickness of the depth hoar layer and the size of the depth hoar crystals are dependent upon the meteorological conditions during snowpack build-up and the degree of snow surface cooling throughout the winter. Date and amount of first snowfall, and ground surface temperature at the time of the first snowfall will help to determine the amount of heat that will be retained by the ground and will thus influence the temperature profile within the snowpack. A steeper temperature gradient will lead to a better developed depth hoar layer in a given year. Satellite passive microwave observations of the snow on the Arctic Coastal Plain provide the opportunity to study inter- annual changes in snow cover and depth.on a regional scale, and the use of a radiative transfer model and a snowmelt energy balance model allow inferences to be made about snowpack structure. PREVIOUS WORK Passive microwave data from the Nimbus-7 Scanning Multichannel Microwave Radiometer (SMMR) are useful for estimation of global snow extent (Chang, 1986) and snow depth (Kunzi et al., 1981). ,::;-cuuie::~ nave snown -cna-c tnere is an inverse relationship between snow depth and passive microwave brightness temperature (TB) as measured by satellite sensors under dry snow conditions (Foster 522 et al., 1984). Snow crystals scatter the upwelling emission and thus deeper snow which contains more snow crystals will permit more scattering and will result in a lower TB. Recent studies have shown that the microwave response of certain types of snow structure can be modeled successfully using a radi~ tive transfer model (Hall et al., in press) leading to an improved ability to estimate snow depth in selected regions using microwave radiometry. Radiative transfer modeling has shown that snow depth) temperature and grain size are important parameters which govern the TB, where the TB is a function of emissivity and temperature of t~ snow. In addition, it has been shown that large snow grain sizes reduce the microwave emission frorna snowpack and increase scattering which results in a lower TB (Chang et al., 1982; Hall et al., in press: The large snow crystals in the deptl hoar layer contribute to a low microwave TB• Thus, two snowpacks of equal depth may have different TBs due to different grain size distributions. Use of a two-layer radiative transfer model (Chang et al., 1976) permits the effect of grain size of the upper and lower layers in a snowpack to be analyzed. Larger grain sizes can be assigned to the lower layer to simulate a depth hoar layer. The thickness of the depth hoar layer can be increased through time when modeling a time series of data. Hall (in press) found that the coefficients of correlation between observed and modeled TB correlated well when the thickness of the depth hoar layer was allowed to increase by 0.5 em per week during a 3 month study perioa in each of 4 years studied. ~nergy balance modeling of snm has been accomplished by several authors (Choudhury et al., 1980; Anderson, 1976; Dozier, 1984). The energy balance model used in this paper is a one-dimensional, time-dependent numerical integration model based on the surface energy balance, heat transport within the snowpack and snow density variation equations. It incorporates the effects of snow melt, liquid water retention, liquid water percolation and refreezing. However it does not incorporate different snow structure effects except as can be inferred by temperature profiles which are produced by the model. Numerical integrations are carried out for specified periods of time by dividing the snowpack into finite layers. Input parameters which influence the thermal structure of snowpacks are: air pressure, air temperature, humidity, wind speed and solar heat- ing. Air temperature, wind speed and solar heating are the major controlling factors (Choudhury et al., 1980). The temperature and density profiles are simulated. As mentioned previously, temperature profiles are important because they influence depth hoar development. METHODOLOGY In this paper, the snow cover on the Arctic Coastal Plain is studied during each of 4 winters: 1979-80, 1980-81, 1981-82 and 1982-83 using horizontally-polarized passive micro- wave SMMR data at the 0.81 em wavelength (37 GHz frequency) obtained from 5 day TB averages from late September through early June in each year. These data are plotted and shown in Figure 1. The SMMR data represent the average of 15 TB grid cells each of which represents an area that is l/2° latitude x l/2° longitude in area. The TB data were located approximately between 70 - 70.5 °N and 152 -159 °W. The TBs are used in single and multiple linear correlations with air temperature (which is used to infer snow temperature) and snow depth. Air temperatures and snow depth meas- urements are obtained from the Barrow, Alaska meteorological station (NOAA, 1979a-l983a; NOAA, l979b-l982b). 523 The energy balance model which was described previously, employs meteorological data from Barrow, Alaska. The model is employed in order to determine if the negative temperature gradient in the snowpack could be maintained given initial snowpack conditions that were assumed for early in the snow season for the first 5 days of November in each year of the study. A negative temperature gradient allows the depth hoar layer in a snowpack to be maintained. RESULTS Note in Figure 1 that just prior to snow accumulation (mid- September), the TBs in each year are 210 ± 5 °K. Similarly, just after snowmelt in early June, the TBs are 220 ± 5 °K. As soon as the snow begins to accumulate each fall, significant changes in TB appear due to the presence of snow. For example, in mid-February there is a 35K difference in TB between the 1981-82 snowpack (166K) and the 1979-80 snowpack (201K). Analysis of a time series of Nimbus-7 SMMR TB data of snow on the Arctic Coastal Plain shows that seasonal variability in TB is correlated with snow depth and air temperature (Tables 1 and 2). In some years, these correlations can be very good. For example, the correlation between snow depth and microwave TB was R = -0.84 in 1981- 82 (Table 1). In other years, the correlation is poorer, e.g. R = -0.52 in 1980-81. Linear and multiple linear regressions were performed using the TB data and meteorological data obtained from Barrow, Alaska. Part of the interannual variability in TB can be explained on the basis of snow depth and temperature dif- ferences between years. Table 2 shows the results of the multiple linear correlation using air temp- erature, snow depth and TB. Note the addition of air temperature ~ 2300 w ~ 2200 !;( a: w 0.. ~ 20()0 l- en ID 1900 z ~ 180° Q a: al 170° 37GHz(H) Ts ARCTIC COASTAL PLAIN OF ALASKA o---o 1979-80 1980-81 1981-82 1982-83 \ \ \ \ \ ' \' '• ~ 2300 2200 210° 20()0 19()0 180° 170° L----~--~~--~~--~--~~--~~--~----~----~--~=---~----~---..~--~~--~~100° 100'240 200 280 300 320 340 360 15 35 55 75 95 115 135 155 175 SEPT OCT NOV DEC JAN FEB MAR APR MAY JUNE Figure 1. Plots of horiztonally-polarized 0.81 ern (37 GHz) SMMR brightness temperature versus time, 1979-1983. (Table 2) improves the correlations in 3 out of the 4 years studied. Especially in some years, there is a substantial portion of the TB variability that cannot be explained by snow depth and/or air temperature (Table 2). Major factors which contribute to this variability are: atmospheric effects, variability in snow depth within a SMMR resolu- tion element, and snow structure variability. Atmospheric effects are considered to be quite minor during the winter in northern Alaska because the liquid water in the atmosphere which would affect the microwave emission is negligible. Ice crystals in clouds do not have a measurable effect on the microwave TB at the 0.81 em wavelength (Wilheit, 1972). Other factors also may affect the microwave emissivity of snow but are not considered here. TABLE 1. Correlation of Tn and Snow Depth Standard Winter R R2 Error 1979-80 -0.76 0.58 8.041 1980-81 -0.52 0.27 12.313 1981-82 -0.84 0.71 9.471 1982-83 -0.64 0.40 4.708 TABLE 2. Multiple Linear Correlation of TB, Snow Depth and Air Temperature Standard Winter R R2 Error 1979-80 0.77 0.59 7-990 1980-81 0.69 0.47 5.872 1981-82 0.84 0.71 4.600 1982-83 0.70 0.49 2.524 The variability in snow depth in a resolution element of the SMMR data (in this case, 1/2° latitude x 1/2° longitude) is a source of error. The SMMR will record an "average" TB for each resolution element. Benson 525 (1982) reports that some snow driftE along river banks can be > 10 m dee} Snow on the flat tundra is generallJ less than 40 em deep. This large areal variability in snow depth will contribute to the unexplained errorE in the linear correlations between TB and snow depth because of the need to assign one snow depth value to a large (1/2° X 1/20) area. Thus the average snow depth is under- estimated by using the reported point source values. This is especially true if the meteorologica station lies in a precipitation shadow. Finally, there is an error which is due to snow structure variability. The snow structure will change if the thickness of the depth hoar layer and the grain size distribution within the snowpack change. The variability in the thickness of the depth hoar layer and the size of the depth hoar crystals contributes to TB changes since greate~ depth hoar development causes more microwave scattering and thus lower TBs (Hallet al., in press). In order to study the inter- annual differences in snow structure, energy balance modeling was performed in addition to the modeling of the microwave emission using a radiative transfer model. Temperature profiles simulated from the energy balance model are shown in Figures 2, 3, 4 and 5. The initial (given) temperature profile and the temperature profile for the end of each 6-day period (total = 4 years) is shown. Note that the steep, negative temperature gradient is maintained throughout the 6 day period in each year according to results from the energy balance model. A steep, negative tempera- ture gradient, if maintained, permits increased development of depth hoar layer thickness and size of the depth hoar crystals as the winter progresses. Initial snowpack conditions were estimated and based on a snowpack with a steep negative temperature profile which is typical 30 NOVEMBER 1979 E () - I 20 1-a.. -INITIAL w Q --DAY 6 s 10 0 z CJ) 0 -30 -20 -10 0 TEMPERATURE (°C) Figure 2. Temperature profiles of the snowpack on the Arctic Coastal Plain of Alaska. The initial snowpack profile (based ?n mea~ured snow depth) was assumed for November 1, 1979 and 1nput 1nto the energy balance model. The profile on day 6 November 6, 1979, was simulated by the model. ' 30 NOVEMBER 1980 -E () ,.. -. ....-: I 20 I .... 1-.... a.. -INITIAL I w --DAY 6 /. Q • I s 10 /. 0 • z /. CJ) • • / 0 -30 -20 -10 0 TEMPERATURE (°C) Figure 3. Temperature profiles of the snowpack on the Arctic Coastal Plain of Alaska. The initial snowpack profile (based on measured snow depth) was assumed for November 1, 1980 and input into the energy balance model. The profile on Day 6, November 6, 1980, was simulated by the model. 526 30 NOVEMBER 1981 -E (.) - :I: 20 ~ -INITIAL w --DAY 6 Cl s 10 0 z en 0 -30 -20 -10 0 TEMPERATURE ( ° C) Figure 4. Temperature profiles of the snowpack on the Arctic Coastal Plain of Alaska. The initial snowpack profile (based on measured snow depth) was assumed for November 1, 1981 and input into the energy balance model. The profile on Day 6, November 6, 1981, was simulated by the model. 30 NOVEMBER 1982 -E (.) - :I: 20 1- 0.. -INITIAL w --DAY 6 Cl s 10 0 z en 0 -30 -20 -10 0 TEMPERATURE (°C) Figure 5. Temperature profiles of the snowpack on the Arctic Coastal Plain of Alaska. The initial snowpack profile (based on measured snow depth) was assumed for November 1, 1982 and input into the energy balance model. The profile on day 6, November 6, 1982, was simulated by the model. 527 for the Arctic Coastal Plain as discussed in Benson et al. (1975). Comparison of the simulated temperature profiles among the 4 years reveals considerable dif- ferences in the steepness of the gradient. This can be attributed to varying snow depths and variable meteorological conditions among years. Thus, one can see that this would lead to significant dif- ferences in snowpack structure (in terms of depth hoar development) from year to year. This snow structure variability, which can be observed in the field or modeled using energy balance modeling, helps to explain the annual variability in TB which is evident in the SMMR data. Large variations in TB between years are apparent (Figure l) particularly beginning in mid-January and continuing until snowmelt in the spring. DISCUSSION AND CONCLUSION Analysis of the errors that are present in the simple and multiple linear correlations between microwave TB and snow depth and air temperature has revealed that the major sources of error are primarily due to vari- ability in snow depth and the physical conditions within the snowpack on the Arctic Coastal Plain of Alaska. Atmospheric effects are probably not signifi- cant during this study period. Unrepresentative snow depth measurement and snow depth variability due to drifting are problems that should be solvable with better in-situ snow measure- ment techniques. Areal variability in snow depth would be less of a problem with higher resolution sensors which should be considered for future satellites. The snow structure variability that affects the microwave signal on the Arctic Coastal Plain is largely a result of changes in the thickness of the depth hoar layer 528 and is believed to represent a large part of the unexplained error' from coefficient of determination studies. As described, this depth hoar variability results from meteorological conditions which differ from year to year. Use of a snowmelt energy balance model has allowed us to simulate snowpack temperature profiles given a set of initial conditions. Results of the model show considerable variability in snowpack temperature profiles between years resulting from inter- annual variations in snow depth a~ meteorological conditions, thus resulting in interannual variabilty in depth hoar development. Snowpack structure and depth vary from region to region due to differing meteorological and topo- graphic conditions. In snowpacks of equal depth, the microwave emissivity can vary due to snowpack structure. In the case of depth hoar, a common component of snowpac structure, the effect on the micro- wave emission can be understood through modeling. Snowpack structure effects such as ice lense and layers are not as easily understood in terms of their microwave response. Additional analysis and modeling may help to explain these and other snow structure effects in the future. Though errors remain in the measurement of snow depth from microwave satellite sensors, analysis of these errors through radiative transfer modeling and snow energy balance modeling is providing valuable information on regional snow depth and structure characteristics. Extensive depth hoar development is known to be present in seasonal snowpacks throughout the world. As we gain a better understanding of the inter- actions between microwaves and snow structure, the ability to determiru regional snow depth using passive microwave satellite data will improve dramatically. REFERENCES Anderson, E. A., 1976. A Point Energy and Mass Balance Model of a Snow Cover. NOAA Technical Report NWS 19, 150 pp. Benson, c., B. Holmgren, R. Timmer, G. Weller and s. Parrish, 1975. Observations on the Seasonal Snow Cover and Radiation Climate at Prudhoe Bay, Alaska during 1972. In Ecological Investigations of the Tundra Biome in the Prudhoe Bay Region, Alaska (J. Brown, ed.), Biological Papers of the University of Alaska Special Report Number 2, October, 1975, pp. 13-50. Benson, C. S., 1982. Reassessment of Winter Precipitation on Alaska's Arctic Slope and Measurements of the Flux of Wind Blown Snow. Geophysical Institute University of Alaska UAG R-288, 26 pp. Chang, A. T. c., P. Gloersen, T. Schmugge, T. T. Wilheit and H. J. Zwally, 1976. Microwave Emission from Snow and Glacier Ice. Journal of Glaciology, V. 16, pp. 23-39. Chang, A. T. C., J. L. Foster, D. K. Hall, A. Rango and B. K. Hartline, 1982. Snow Water Equivalent Esti- mation by Microwave Radiometry. Cold Regions Science and Tech- nology, V. 5, pp. 259-267. Chang, A. T. C., 1986. Nimbus-7 SMMR Snow Cover Data. Proceedings of the Snow Watch 1985 Workshop on C02 Snow Interaction, 28-30 October 1985, College Park, MD. Choudhury, B. J., C. S. Cheng and A. T. c. Chang, 1980. Numerical Simulation of the Thermal Struc- ture of Snowpacks using an Energy Balance Model: Model Description and Parameter Test. NASA Techni- cal Memorandum 82044, 30 pp. 529 Dozier, J., 1984. Snow Reflectance from Landsat-4 Thematic Mapper. IEEE Transactions on Geoscience and Remote Sensing, V. GE-22, pp. 323-328. Foster, J. L., D. K. Hall, A. T. C. Chang and A. Rango, 1984. An Overview of Passive Microwave Snow Research and Results. Reviews of Geophysics and Space Physics, V. 22, pp. 195-208. Hall, D. K., A. T. C. Chang and J. L. Foster (in press). Detection of Depth Hoar Layer in the Snow- pack on the Arctic Coastal Plain of Alaska Usini Satellite Data. Journal of Glaciology. Hall, D. K. (in press). Influence of Changing Depth Hoar Structure on Microwave Emission from Snow in Northern Alaska 1980-1983. Submitted for publication. Kunzi, K. F., S. Patil and H. Rott, 1982. Snow-cover Parameters Retrieval from Nimbus-7 Scanning Multichannel Microwave Radiometer (SMMR) Data. IEEE Transactions on Geoscience and Remote Sensing, GE-20 (4), pp. 452-467. NOAA, l979a-l983a. Climatological Data -Alaska, NOAA, Asheville, N.C. NOAA, l979b-l982b. Local Climato- logical Data, Barrow, Alaska, NOAA, Asheville, N.C. Wilheit, T. T., 1972. The Elec- trically Scanned Microwave Radiometer (ESMR) Experiment. The Nimbus-5 User's Guide, pp. 59-105. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 OPERATIONAL DEMONSTRATION OF MONITORING SNOWPACK CONDITIONS UTILIZING DIGITAL GEOSTATIONARY SATELLITE DATA ON AN INTERACTIVE COMPUTER SYSTEM Milan w. AI len and Frederick R. Mosher 1 ABSTRACT: The Nat I on a I Severe Storms Forecast Center ( NSSFC) collects and maintains real-time data bases of atmospheric and surface observations In addition to full resolution GOES data with the Centra I I zed Storm I n form at t on System CCSIS). These data sets and analyses derived from these data sets are then combined with topographic, hydro I og lc and/or ~eo p o I I t I c a I b o u n dar I e s to ass I s t nydrologlc users as well as the pr I rna ry severe storm forecastIng users. DurIng the 1985 snow me 1 t season, the Sate II tte F tel d Se r v Ices Stat I on ( SF S S ) began utilizing CSIS capabilities to derive operational digital snowcover maps from 1 km. visible WES data. The snowcover mapping technIque Is based on the fact that snowcover I ncr eases the brIghtness of a I and area. Snow free/c I oud free data sets are adJusted to ~count for dally and seasonal so I a r I I I u m I n at I on a n g I e differences between the reference time and the snowmapplng time. In 1986, snowmapp I ng servIces w t II be ex p a n d e d f rom 3 5 0 , 0 0 0 s q • k m to 1 • 2 1111 I on sq. km to meet IncreasIng user requIrements. An opt I on a I 1986 serv tee w I II be snowcover mappIng ~y elevation zones while deve I opment work Is conducted to determine the feaslbll lty of deriving solar Insolation, skin temperature over snowcovered areas and free atmospheric conditions (temperature, relative humidity, wind speed and direction) at specified elevations. CKEY TERMS: Interactive processing; digital data; satel I tte hydrology, snowcover mapping.) INTRODUCTION Traditionally, water resources forecasters have rei led upon point measurements of snow water equivalent, temperature, and prectpttatton collected from manual snow courses, cooperative observ- ers, cl lmatologtcal stations, and remotely sensed platforms for snowpack monitoring. These point measurements are extrapolated to derive mean areal values for Individual hydrologic units (river basins). Regression analyses, lapse rates and empirical Jy derived curves are also appl led to these data sets to account for the effects of large topographic and physical features such as mountains and I akes. Spatial data sets became available with the launch of the 1Respectively, Satell tte Hydrologist SFSS and Director, Techniques Development Unit, National Severe Storms Forecast Center, Room 1728, Federal Building, 601 E. 12th Street, KAnsas City, Missouri 64106. 531 first meteorological sate! I ltes. Snow cover maps derived from sate! I tte Imagery provided forecasters with Information on snowpack location and extent. During the last five years there have been two Important development phases In the transition from manual photolnterpretatlon techniques to digital techniques. The original digitally derived maps were alphanumeric computer printouts utilizing relatively coarse 4 km GOES visible data sets. Processing required a batch processing system to access the National Environmental Sate! I tte Data and Information Services CNESDJS) data bases on the IBM 360/195 computers In Washington, D.C. Currently, snowmapplng operations uti I lze real-time ful I resolution 1 km GOES data on the Interactive Central tzed Storm Information System CCSJS) at the National Severe Storms Forecast Center CNSSFC) In Kansas City. CSIS processing capabll ttles have enabled the NSSFC sate! I lte hydrologist to begin developing more diversified products and services to monitor snowpack conditions In addition to expanding areal coverage with the more detailed snowcover maps. Visual Snowcover Mapping Techniques Prior to 1978, NESDIS employed photolnterpretatlon techniques to produce operational snowcover maps. The technique uti I lzed a Zoom Transfer Scope (ZTS) to register a rectified and enlarged visible Image from NOAA polar orbiting satel I ltes to a basin outline by aligning landmarks (lakes, rivers, etc.). The snow- 1 tne was traced onto the basin map and percent snowcover was calcu- lated using a planimeter or e 1 ectron lc densIty s I leer. These visually derived analyses were labor Intensive and accuracy was directly related to the analyst's familiarity with each basin's topography, canopy, and skills 532 Interpreting the effects of varying II lumlnatlon angles on features (forests, etc.) within each basin. Snowcover Mapping with Digital Data From 1978 to 1981, NESDIS began conducting snowmapplng tests utilizing digital GOES data on Interactive and batch processing systems to Increase the timeliness, effectiveness, accuracy and expand the areal coverage to meet user requests. Results of tests conducted on Interactive systems using GOES Imagery and basin boundary overlays displayed on video terminals were somewhat mixed. This visual technique enhanced the analyst's abi I tty to eliminate clouds and measure snowcovered area by pixel counting. However, complex snow- 1 tnes and discontinuous snowcover required superior eye-hand coordi- nation by the analyst. Limitations with existing hardware and software resulted In emphasis being placed on the digital snowmapplng tech- nique being concurrently developed on the NOAA main computers. This batch processing alI digital procedure was tested opera- tionally during the spring of 1981. This procedure developed snowfree/ cloudfree data sets (masks) for each basin. The darkest (lowest count value) pixel within each basin was chosen as the "threshold value" used to reference every pixel In each basin. During the snowmapping season a new threshold value was selected to brighten or darken the mask to account for varying solar Illumination angles. On cloudfree days the program was submitted through a Remote Job Entry (RJE) terminal to the NOAA main computers in Suitland, MD. Repetitive sub- missions are required to exactly al tgn the mask and the visible data and then produce a series of snow- cover maps. The Initial threshold value was added to each pixel In the mask and compared to each pixel retrieved from the GOES data set. If the mask plus threshold value Is less than the GOES value, the pixel Is assumed to be snowcovered. The number of snowcovered pixels was divided by the total number of pixels in the basin to derive per- cent s nowcover. Between December, 1981 and July, 1984, the SFSS generated and distributed 691 snowcover analyses covering 16 basins In the western UnIted States. Because the proce- dure required visible data from a GOES satel I tte at 135°W, the pro- gram was significantly affected by the failure of three GOES satel- lites beginning In November, 1982. The failure of GOES-5 and the decision to move GOES-6 to a more centra I I ocat I on threatened to leave the SFSS without the abtl tty to generate digital snowcover ana I yses. After months of revIew- Ing the digital snowcover mapping software It was decided the man- hours required to modify existing software and write new software for the reduced resolution data would not be worth the effort. CURRENT SNOWMAPPING OPERATION Central tzed Storm Information System CCSIS) CSIS is a developmental system which exists in an opera- tional environment and supports the operational mission at the National Severe Storms Forecast Center CNSSFC>. CSIS is a spin-off of the University of Wisconsin's Space Science and Engineering Center's CSSEC) Man computer Interactive Data Access System (Me I DAS). Wh tIe the ultimate goal of CSIS Is to Improve NSSFC's abl I tty to forecast severe weather, It a I so represents another phase In the NWS effort to develop a system to Ingest, analyze, Integrate and display real 533 time data from at I available sources. The CSIS system consists of a GOES receiving antenna system, four Harris slash-6 computers, five Interactive video terminals, conven· ttonal ~urface and upper air data, dial-up radar inputs, 1600 bpi magnetic tape drives used for data archiving and four 300 mb hard disks, an interface to other NSSFC computers and a phone It nk wIth the SSEC MciDAS for VIsible Infrared Spin Scan Radiometer CVISSR) Atmospheric Sounder (VAS) data access. System flextbil tty gives the analyst a diverse menu of options and capabilities Including: real- time stretched data from the GOES VISSR; color enhancement of satel- lite imagery; the ab II tty to super- Impose data and analyses on satel- 1 lte imagery; locating of points on the imagery using earth, satel I lte, and screen coordinates; a macro language facility for defining a sequence of commands; color graphics; the capabll tty to overlay a variety of maps over satel I lte data, to change the gray scale of imagery and remap data to different map projections; a statistical analysts of data; data archival on magnetic tape; and black and white hardcopy of graphics and/or imagery. CSIS Snowcover Mapping The CSIS snowcover mapping technique Is based on the fact that snowcover Increases the brightness of a land area. The following are the procedures used in the CSIS snowcover mapping technique: o Identify areas to be mapped and select data sectors. o Develop an automated system to col teet ful I resolution data. o Set up seven day rotating files to archive the data. o Input United States Geological Survey (USGS) catalog units for basin del lneatton over areas to be mapped. o Prepare snowfree/cloudfree reference data bases (masks). o Remap GOES data Into mask projection. o Exactly al tgn the mask and GOES data. o Adjust the brightness of the mask to reflect the difference In solar Illumination angle of the GOES data. o Composite two Images to el lmlnate clouds. o Compare each pixel In the mask and Image to produce a snow/no snow Image. o Sub-divide each sector Into mapping scenes. o Eliminate cloud contamination manual Jy with the cursor. o Overlay basin boundaries and digital Jy compute percent snowcover. o Overlay percent snowcover tables for each basin. o Hardcopy the snow/no snow Image with basin boundaries and tables. o Archive GOES data used for mapping on magnetic tape. o Disseminate maps via first class mall and raplfax. Real time ful I resolution GOES data are received directly from satel I tte and stored on the CSJS Data Base Manager CDBM). A clock control Jed scheduler then "sectortzes" these data and sends them to another CPU where they are stored for real time or future display on an Interactive terminal. Snowcover mapping data are stored In two seven-day rotating files, 534 one for each time period. During the winter and early spring, data are archived at 1700Z and 1800Z (GMT). From early Spring through July, data are archived earlier at 1600Z and 1700Z to reduce the possibl Jlty of small cloud con- tamination. Software was written to allow the analyst to manipulate file parameters and display the ft Jes with one I lne key ins. The USGS basin boundary tape has 27,000 sectors representing regions, sub-regions, accounting units, and catalog units for the contiguous 48 states. The highly detat Jed boundaries are defined by latitude/longitude pairs of points that are line segments between Individual basins. Software developed by the NWS Office of Hydrology COH) was modified to read the data Into CSJS. Additional software was written to convert the data to CSIS geographical format, and display various configurations of regions, sub-regions, accounting units, and catalog units In addition to the snowmapplng basins. Ful I resolution GOES-5 visible data for September 24 and 25, 1983 were al lgned and compos- ited to create snowfree/cloudfree Images. These Images were then remapped Into a projection of the satell tte at Its winter longitude C108°W). Navigation coordinate data were then exactly al lgned with the Imagery using landmark coordInates derIved from 1 : 1000000 Lambert Conformal Operational Navigation Charts. The sate I I I te vIs I b I e I mages need to be corrected for brightness changes due to the dally and seasonal change of the sun angle. It has been assumed that the land and snow reflect sunl tght In an I sotrop lc manner. ThIs Is a good first order approximation to these surfaces as long as extreme sun angles are avoided. The visible brightness counts on the GOES satell tte have been digitized with a square root dlgltlzatlon tech- nlque.The solar zenith angles e were calculated for the center of the basin of Interest for the reference Image and the current Image and the entire satel I Jte Image was then multiplied by the correction factor ~.Q~1 cos e 2 • To reduce the amount of time required to manually align the GOES-Imagery a reference mask software was developed to al lgn the Imagery with the mask resulting In the al Jgnment of the digital data b~ses. The entire operation re- quires one command and two single letter keylns. GOES data and the reference mask are loaded on "opposite frames" of a terminal. This allows the analyst to flicker between the two Images using a one letter keyln as the toggle switch to check their al Jgnment and look for landmarks visible on both Images. A cursor Is pI aced directly over the landmark In the GOES Imagery by moving a pen over a data tablet connected to the terminal. The one Jetter keytn Is entered and the reference mask Is displayed. The operation Is repeated on the reference mask. The GOES data are repositioned and all pixels are now given the navigation, satel I Jte, and earth coordinates of the reference mask. The repositioned Image Is loaded and al Jgnment Is checked by fl Jckerlng between the real Jgned Image and the mask. Composttlng two Images to el Jmtnate clouds Is a combination of the two previous steps. Bright- ness levels of one Image are scaled to match the II Jumtnatton angle of the other. Pixels on each Image contaminated by cloud shadows are set to zero using a cursor data function. The two Images are 535 al Jgned and every pixel In each Image Is compared. The darkest pixel value Is then selected to produce the composlted Image. Snowcover sectors are produced by comparing every pixel In the sector w Jth every p lxel In the scaled mask. It the sector pixel Is brighter than the mask pixel, the sector pixel Is classi- fied as being snowcovered and Its count value Is used to build the snowcover sector. If the sector pixel Is Jess than or equal to the mask va I ue, the pI xe I In the snow- cover sector Is given a count value equal to zero. Each snowcover map In a sector Is then displayed and checked again for cloud contami- nation. If contamination exists, It Is removed Interactively using a cursor data function which sets the contaminated values to zero or the threshold stripping program can be rerun with the background Increased by a smal I delta to determine the snow/no snow threshold. Each catalog unit boundary Is overlayed over the snowmap and a statistical discrimination function Is used to derive percent snowcover by divid- Ing the number of non-zero pixels by the total number of pixels Inside each boundary. Percentages are then entered Into a macro, written to overlay a table of percent snowcover values on each map. An Interactive enhancement function Is used to convert alI non-zero pixels to white and change catalog boundaries to gray. Hard copies are made using a Honeywel I VGR 4000 Video Graphics Recorder. Snowcover maps are disseminated via first class mall and raptfax. GOES Image sectors used to derive snow- cover maps are archived on magnetic tape. During 1985, the 16 basins mapped using reduced resolution NESDIS data were subdivided Into 68 USGS hydrologic catalog units and service was expanded to Include 17 additional catalog units In Idaho. The catalog units range In size from 800 to 12,000 sq. km. and cover 350,000 sq. km. From late February through June 1985, the SFSS derived and disseminated 773 snowcover analyses for 85 catalog units In the western U.S. CSJS Snowcover Mapping In Perspective Many factors affect the accuracy of satel I lte areal snowcover measurements, Inc I ud I ng data resolution, methodology, exactness of basin boundaries, precision of registration of boundaries to Imagery, measuring techniques, ~nd the ski I I of the analyst applying the tools and procedures. There are four critical time components that affect the operational uti I tzatlon of these data: frequency of observation, the time between data collection and availability for analysts, time required to do the analysts, and the time required to del tver the product to the users. The principal costs of the product are data acquisition, data process- Ing and product delivery. In terms of accuracy, the GOES Interactive digital snowcover mapping technique has many distinct advantages over manual photolnter- pretatlon procedures except In the area of data resolution. LANDSAT's 30 m resolution Is clearly superior to the NOAA polar orbiting data (1000 m at Nadir) and GOES (1000 m at the Equator). The digital technique has the abil lty to overcome the complexity of re- flected I tght on various types of terrain from constantly changing solar Illumination angles. Navigation and geographic data bases In CSIS enable the analyst to register Imagery and basin bound- aries more accurately (±1 km) than a ZTS, and pixel counting Is far more rei table than a planimeter or density slicer. The digital tech- 536 nlque Is not affected by complex, broken, or spotty snow fields. The digital technique also has conslde~ ably shorter processing times per basin than planimeter techniques. AI I other factors can be considered eq u a I • "Finding the clear spots" between cloud cover Is one of the greatest advantages of using GOES data, because visible Imagery Is co I I ec ted every 1 I 2 hour d u r I n g daylight hours and Is available for analysis within five minutes time. The wide range of possible col tec- tlon times at lows more frequent and reliable data collection. Analyses for 29 basins can be completed with· in 40 minutes of the data col lec- tlon time. Maps can be dissemi- nated within one hour of data col Jectton time using a rapt fax and within 24 hours using overnight express mall service. Whi Je the CSJS technique has many advantages over other proce- dures, Its' accuracy In heavily forested areas must be verified and documented through a cooperative program with UH and the users. SERVICE REQUIREMENTS Background During the 1970's, the National Aeronautics and Space Administration (NASA) sponsored the Appl Jcatlons Systems Verification and Transfer (ASVT) snowmapplng program. This five year user cooperative effort used four test sites In Arizona, California, Colorado and the northwestern United States to perform opera- tional evaluations of the effects of technological capabll ttles (In existence at that time) on water resources forecasting. Study results show that slxtyelght percent (68%) of the surface water In the 11 western states Is derived from snowmelt runoff. The tot a I va I ue o ·. snow- melt runoff water Is subd 1 rJded Into the following catego1 es: Irrigation & hydrodelec energy Municipal & Industrial Flood Damage Other Uses -87% 9% 4% -<1% Test results show that a minimum 6% Improvement In forecast accuracy was achieved using satel- lite derived snowcover maps. This six percent Improvement In fore- casting snowmelt runoff would re- sult In a total annual benefit of $36.5 million (1981) dollars for Irrigated agriculture and hydro- electric energy for the 11 western states. Est lmatzd annua I cost for the 2,195,250 km area rmpacted by snowm~lt forecasting was $0. 23/km • These costs were derived from the Colorado ASTV and are based upon acquiring eight sets of Landsat Imagery ($400), 16 man-days to Interpret the Imagery ($800), eight man-days to Implement the data ($600) and the cost of a zoom transfer scope CZTS) ($10,000) with 25% uti I lzatlon and amortized over 10 years. CSIS snowmappln~ costs during 1985 were $0.14/km of area mapped. User Survey The user community for snowcover maps Is diverse In mission and, therefore, product utilization varies. Field and research offices of the National Weather Service CNWS), Corps of Engineers CCOE), Bureau of Reclamation CBOR), Soli Conservation Service CSCS), Agricultural Research Service (ARS), and the Forest Service can uti I tze these data for real-time, medium range and seasonal stream- flow forecasting, reservoir regulation and a variety of other water resources and agricultural services at regional and local levels. Additionally, many 537 agencies wtl I uti I lze these data to develop, expand and ca I i brate hydrologic models. During the summer of 1985, the SFSS surveyed existing and potential satel I lte hydrology users to prioritize expansion and develop- ment work on CSIS. Products and services I tsted for operational testing and development were per- cent snowcover per basin, percent snowcover per elevation zone per basin, solar Insolation and precipitation, skin temperature over snowcovered area, atmospheric data (temperature, wind speed and direction), and historic data sets of alI these products. Additional Information was also sol tclted for user require- ments for product format, resolu- tion, frequency of observations, seasonal coverage, and timely dissemination. Survey Results -Non-Weather Service Users Surveys were distributed to existing snowcover users and new agencies that had requested coverage during the 1986 snowmelt season. Therefore, the results cannot be Interpreted as being representative of the entire hydrologic community. In fact, the results stated here wll I only cover non-NWS users. In general, percent snowcover was requested by operational users plus historic data sets, while nonoperatlonal users engaged In model development and cal tbratlon requested alI or parts of the products surveyed. New areal snowcover mapping service was requested for 561,000 sq. km In the Western United States, 516,000 sq. km In the Great Lakes Basin and 176,000 sq. km In the Appalachian mountain range from Pennsylvania to Alabama. Elevation zone snow- mapping services were generally confined to basins above 6,000 ft. Solar Insolation was the most requested product In terms of total areal coverage and regional diversity. Data was requested for 2,823,000 sq. km, Including large areas such as the Great Lakes Basin, Great Plains, and Appalachians, In addition to requests on the sub-regional and basin levels from West Virginia to Ca I I for n I a. Skin temperature data was requested for 979,000 sq. km In the Great Lakes, Missouri headwaters, Appalachians, Rio Grande headwater, Salmon River, Death Val ley, and Salt River Project. Half of the Interest In the data Is to see If It will help Identify those por- tions of the snowpack where melt has begun. The other half Is for year-round monitoring of skin temperatures for vegetative and soil moisture stress on a real-time basis. Upper air data was requested for 753,000 sq. km In the Great Lakes, Missouri headwaters, Sierras and Death Val ley areas. These data will be used to develop, calibrate and drive operational m~dels year- round. These data are needed on a real-time basis. One of the most significant results of the survey has been requests by at least one office from the ARS, BOR, COE, USFS and USGS for alI or most of the pro- ducts surveyed. These data sets will be utilized to develop, calibrate and test a variety of hydrologic models. Output from these models wll I enable us to assess the value of these data sets and the val ldlty of the scientific principles and techniques used to derive these data. These studies will play an Integral role In prioritizing ~nd Implementing the data operationally. 538 CURRENT SERVICE OBJECTIVES Areal Snowcover Mapping During the 1986 snowmelt season, snowcover mapping services wll I be expanded to Include an additional 561,000 sq. km In the Western United States and 128,000 sq. km for Lake Super lor draInage. These new areas are almost double the 350,000 sq. km mapped during 1985. This expansion of service area Is made possible by new software that wll I enable the hydrologist to display, combine and, analyze Imagery and geographical data bases more efficiently. Selected basins wll I be analyzed under snowfree and snowcovered conditions to assess the combined e f f ec t s of v a r I o us so I a r I I I u m !- nation angles, slope orientations, and terrain on visible brightness levels. Snowcover Mapping by Elevation Zones The Rio Grande River above Del Norte, Colorado, Salmon River above Whltebird, Idaho and possibly the Boise River above Lucky Peak Dam wll I be mapped by elevation zones. Digital elevation data sets will be remapped Into the satellite projection of the snowfree/cloud- free mask. Each pixel In the mask which has a corresponding pixel In the e I evat I on data set w II I then be level discriminated with the elevation pixel to produce a gray scaled elevation Image. The de- rived snowcover Image, elevation range and the hydrologic boundary data set wll I be combined to produce Images of snowcover area and the percent of snowcover by elevation zone per hydrologic unit. These maps wll I only require one additional step and one minute of time to the existing procedure. Solar Insolation The metamorphosis of a snowpack can be modeled by monitoring the energy balance of Incoming solar radiation, escaping longwave radiation, sensible heat advection, and precipitation. Techniques developed by Gauthier~± u. (1980), and Gauthier (1982) have shown the feasibt I tty of measuring solar insolation from ,geostationary satel I tte data and detecting mesoscale vartabi I tty In energy ba I ance. One of the prImary Influences on the radiation balance Is cloud cover, and one of the 'techniques used to determine snow- ,cover was developed to Identify c I oud cover. E I evat I on data sets 'WI I I also be uti I tzed to account for terrain orientation (snow lasts longer on the north slope of a mountain than on the south slope) and to adjust the data sets to more accurately estimate these parame- ters. We plan to Investigate the possibility of monitoring the conditions on CSIS. CONCLUSIONS Interactive digital snowcover mapping with 1 km visible GOES data Is an accurate, timely, and cost effective means of acquir- Ing basin snowcover area to assess, allocate and forecast water resources from snowmelt runoff because: o It can overcome the complex- Ities of reflected I lght on various types of terrain from constantly changing solar Illumination angles. o Imagery and basin boundaries can be accurately registered <± km). o Pixel counting Is a fast, accurate and rei table measu- r tng techn tque. o Data col lectton is more frequent and rei table. o Cloud contamination can be minimized more effectively. o Up to 30 snowcover analyses can be completed within 40 minutes of the data col lee- 539 tton time. o Digital data can be archived easily and inexpensively. o Analyses only cost $140 per 1000 sq. km. per year during 1985. Today's long and short range snowmelt runoff forecasting systems are dependent upon a variety of point hydrometeorologlcal observa- tions which are then statistically weighted and/or extrapolated to derive areal ly distributed data sets. Escalating costs for data col lectton have made snowpack monitoring with an interactive geo- stationary satel I tte system the most cost-effective means of acquiring temporally and spatially complete data sets to enhance the value of alI other data. REFERENCES Allen, M.W. and F.R. Mosher, 1985. Interactive Snowcover Mapping with Geostationary Satel I lte Data Over the Western United States. 1 n: .s.wn.m.a.r:l~4-.Nl.nn~~.n.ilLl.n±~.c=­ n~±l~~l_.s~~~l~m_on_~~±~ .S~nli.n.g_.Q.f_f.ll.Y~~.n.i. p. Gautier, C., G. Diak, and S. Masse, 1980. A Simple Physical Model to Estimate Incident Solar Radiation at the Surface From GOES Sater I Jte Data. 1L-A~~lL-M~~LL~11: 1005-1012. ______ , 1982. Mesoscale Insolation Varlabil tty Derived From Satel I lte Data. M~±~LL~Zl: 51-58. Tarpley, J.D., Schneider, S.R., and Danaher, E. J., 1979. An All Digital Approach to Snow Mapping Using Geostationary Sate I I I te Data. In Proceed- Ings of the Final Workshop on the Operational Appl !ca- tions of Satel I lte Snowcover Observations (Sparks, Nevada), NASA Conference Publication 2116, National Aeronautics and Space Administration, Washington, D.C. Castrucclo, P., Loats, H., lloyd, D., and Newman, P., 1981. Appl !cations Systems Verification and Transfer Prqject Volume VI 1: Cost/ Benefit Analysts for the ASVT on Operational Appl !ca- tions of Satel I tte Snowcover Observations, NASA Technical Paper 1828. 540 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 APPLYING A SNOWMELT-RUNOFF MODEL WHICH UTILIZES LANDSAT DATA IN UTAH'S WASATCH MOUNTAINS Woodruff Millerl ABSTRACT: The Martinec-Rango Snowmelt- Runoff Model (SRM) was applied to the 130 km2 American Fork, Utah basin in order to simulate the daily streamflow. The year 1983 was chosen to be investigated because of the record-breaking snowpack and peak discharges which uniquely tested the model. Landsat images were used to determine the snow-cover areas within the basin zones. While all model parameters were generally within the suggested numerical range and temporal sequence, some variations were necessary in order to generate the best correspondence between simulated and measured discharges. Because the recession coefficient is a function of historical streamflow data, an extensive analysis was required in order to approximate the all-time maximum flowrate. The simulation accuracy was very good as evidenced by the high hydrograph visual correspondence, the large daily discharge R2 value of 0. 85, and the small runoff volume Dv value of 3.7%. These statistics are in line with SRM average values from other basins. (KEY TERMS: snowmelt-runoff model; snow cover; Landsat; simulation; recession coefficient.) INTRODUCTION Over the past two decades many stream- flow models have been developed in various countries which utilize snowmelt as the major runoff source. These usually compute day-to-day flows taking advantage also of the daily cycles of temperature and solar radiation. The models are generally operational in a simulation and/or forecast mode. Simulations are used to produce discharge in ungaged basins or to compute snowmelt runoff for hypothetical climatic conditions. The forecast mode uses real time data to predict the runoff into the future. Applicability of these models has developed and improved as they have been tested on river basins with differing physical characteristics and climatological conditions. The snowmelt-runoff model (SRM) developed by J. Martinec and A. Rango is one of these models. SRM is designed to simulate and forecast daily streamflow in mountain basins where snowmelt is a major runoff component. A unique feature of SRM is that Landsat imagery can be utilized to identify snow extent (Rango and Martinec, 1979). Description and information on application of the model are given in a published user manual (Martinec et al. , 1983). The model has been applied successfully to basins ranging in size from 3 to 4000 km2 but may not be as applicable to non- mountain basins. SRM has been used for climate conditions from humid to semi-arid but tends to be less accurate when there are significant amounts of rainfall during the snowmelt period (Martinec et al., 1983). In a recent comparison among eleven snowmelt-runoff models conducted by the World Meteorological Organization lAssociate Professor, Department of Civil Engineering, Brigham Young University, 368 K Clyde Building, Provo, Utah 84602. 541 (WMO), SRM performed very well. The results were consistent with earlier tests, the statistical evaluation parameters were good, and the potential for forecasting and year-round simulation were found to be promising (Rango, 1985). BACKGROUND The Snowmelt-Runoff Model was applied to the American Fork, Utah basin for the 1983 runoff period in order to evaluate the model under somewhat extreme conditions. The American Fork River basin is located near the center of the Wasatch Mountain range which runs north-south almost the entire length of Utah (Figure 1). This semi-arid basin has an area of 130 km 2 and ranges in elevation from 3580 m at the summit of Mount Timpanogos to 1820 m at the gaging station called "American Fork above Upper Power-Plant". American Fork River flows southwest into Utah Lake which drains to the north into Great Salt Lake through the Jordan River. Figure 1. Location of American Fork Basin in Central Utah's Wasatch Mountains. 542 Snow depths in the American Fork basin have been observed to be well over 300 ~ near the summits. High-elevation sno1 course maximum depths average approximately 180 em and low-elevatior. snow course maximum depths average approximately 80 em. The snowpack generally lasts well into June or July and on occasion remains throughout the swnmer at the higher elevations. Snowmelt is the major runoff component as evidenced by the fact that the American Fork River becomes almost dry in the late summer and fall. Snow depths in the basin during the winter of 1982-83 were among the greatest on record and both the American Fork River volume of runoff and peak flowrate were the all-time maxima. The flooding was aggravated by the rapid temperature rise and above normal rainfall in late May along the Wasatch Mountains. Discharges from American Fork River and other tributaries caused Utah Lake to rise 2 D above its normal elevation and flood thousands of acres of prime agricultural land. Parameter values used in the SRM which are based upon historical data were uncertain because of the record-breaking snowpack and runoff. This resulted in a unique use of the model and will contribute to the growing set of applications. The results of this test should be helpful in applying the SRM on bas ins with deep snowpacks which remain well into the summer months. MODEL PARAMETERS SRM is a deterministic runoff model with all parameters to be predetermined. Four elevation zones of approximately 440 m were established and the areas of each zone were variables ratio of area (S), to runoff determined. The other measured are: number of degree-days (T), snow-covered area to the total and precipitation contributing (P). For the American Fork study, T and P data were obtained from the weather station within the basin at Timpanogos Cave National Monument. Landsat scenes for May 30 and July 17, 1983 were used to develop S data (Figure 2) . The percent of snow-covered area in each zone on these dates was plotted and curves were drawn to estimate the daily percent of snow-covered area for the simulation period (Figure 3). More Landsat data would have been useful in establishing these curves but no other cloud-free scenes were available. Figure 2. May 30 and July 17, 1983 Landsat Scenes with the Snow-Cover on Americ a n Fork Basin. 543 The model parameters to be calculated are: snow and rain runoff coefficients (Cs & Cr), degree-day factor (a) indicating the snowmelt depth from one degree-day, temperature lapse rate (~). critical rain/snow threshold temperature (Tc), recession coefficient (k) indicating the decline of discharge, and the lag time (L) . Typically Cs values are suggested to be near 1.0 at the start of the snowmelt season, decrease to 0. 5 or 0. 4 in midsummer and may increase somewhat at the end of the runoff period. The Cr values are usually slightly lower than Cs values and decrease similarly but without a late season increase. (Martinec & Rango, 1986) In the American Fork case, during April and May, the soil profile was relatively dry and there was significant recharge along with snowpack retention which called for lower than usual Cs values. There are also two small reservoirs in the basin which were holding a portion of the early runoff. During June, the Cs values were slightly higher than suggested because of the saturated conditions. Low values appropriate for late runoff were used for July and August. Cr values paralleled the Cs pattern at slightly lower magnitudes . eo •o 20 20 30 10 20 30 10 20 30 10 20 30 Figure 3. Landsat-Derived Snow-Cover Depletion Curves for American Fork Basin. In SRM computations, a is determined by the empirical relation; a = l.l psjpw, where ps and pw are snow and water densities. Factors u s ually range from 0.35 to 0.6 cm/°C-day with the lower values earlier in the season (new snow) and at higher elevations (Martinec & Rango, 1986). Values for the American Fork study varied from 0. 3 to 0 . 6 through the snowmelt season, but were constant for all elevations. The density ratios as calculated from snow course data in the lower two zones were also used for the higher two zones. Varying the a's with elevation decreased the runoff simulation accuracy. Without temperature stations at different elevations, the lapse rate must be evaluated with regard to climatic conditions. SRM simulations in the Rocky Mountains (of which the Wasatch are a part) have successfully used 0.75 to 0.95 oc per 100 m which increase throughout the season (Martinec & Rango, 1986). In the American Fork study, -y was held constant at 0. 95 except for a ten day period in late May. During this time, there was an unseasonal warm period which apparently altered the temperature profile requiring a much lower lapse rate value. Critical threshold temperature determines for each day of the snowmelt period whether measured precipitation was rain or snow. SRM retains rain in the snowpack in the beginning of the snowmelt season so that only the rain falling on snow-free areas contributes to runoff. Later, SRM allows the rain to penetrate the snow cover and become part of the runoff. Tc is suggested to decrease from +3oc to +0.5°C from April to August. Values of +2°C until mid-June and +1 oc thereafter for the American Fork simulation resulted in the best correspondence. The recession coefficient is a particularly sensitive parameter in the SRM, but it can be determined quite accurately when streamflow data are available. The discharge of one day (Qn) is plotted against the discharge of the following day (Qn+l) when the hydrograph is falling. An envelope curve is established which includes most of the points but not the extreme values. Then an average line between the lower envelope and the 1:1 line is drawn. By determining two k values from the usual equation, k = Qn+l/Qn, at two different discharges on the plot and substituting into the equations; values of x (Martinec et along with log x + y log Ql and log x + y log Q2, and y can be calculated. al. 1983). SRM uses x andy the current discharge to 544 determine a variable recession coefficient from the equation, Figure 4 shows the recession flow plot for American Fork from 10 years of recession data. The envelope and average lines are shown and the values of x and y are 0. 916 & -0.026 respectively. Because of the record-breaking streamflows, the determination of x and y was difficult. The final values are in the appropriate range and are the result of an extensive calibration process involving changes in the envelope and average recession lines. This x and y set produced the most accurate runoff simulation. 10 1 1:1L\NE LOWEA ENVELOPE LINE AVERAGE LINE FOR RECESSION COEFFICIENT kn kn= X Qny X= 0 916 Y= -0.026 ~~--~------~5----~10~--~2~0-------5~0~--~100 Oi rm 3isl Figure 4. Recession Flow Plot for American Fork Basin with the Lines Used to Derive Recession Coefficients. The time lag can be determined from characteristic daily fluctuations of snowmelt runoff. For example, if the discharge begins rising at noon, it lags behind the temperature rise by about six hours. It was found that the appropriate lag for the American Fork basin is twelve hours. RESULTS The SRM was run many times with different parameter values as part of the calibration process. Each iteration evaluation was based on three criteria. First, the graphical plot of simulated versus observed daily discharge was examined for overall correspondence and to identify major simulation errors. Second, the coefficient of determination (R2) value was examined for relative accuracy of daily discharge values. Finally, the percentage difference between the total measured and simulated runoff (Dv) was examined. Figure 5 shows the final plot of the simulated and observed snowmelt season hydro graphs. The temporal correspondence of streamflow peaks is good, although the magnitudes are not always exactly comparable. The R2 of 0.90 and Dv of 1.7% are also very good compared to other SRM results. In the WMO evaluation, the average calibration R2 and Dv for the four basin data sets analyzed were 0.85 and 5% respectively. Rango (1985) states that these statistics are exactly in line with SRM average values from all the other basins previously tested. "' ;, E ~ 0 ...J ... :I c( w a: 1- (/) 30 10 3 1 .3 APRIL MAY CONCLUSIONS Based upon the American Fork study, the conclusion is made that the SRM can fairly accurately simulate streamflows and runoff volumes from a basin of deep and long- lasting snowpacks. A total runoff volume which had never been accumulated before was closely computed and a peak discharge which had never been gaged before was nearly reached by the model. This was accomplished with all model parameter values within the physically realistic ranges and without the benefit of more complete Landsat snow-cover area information. Continuation of this investigation is ongoing. Some alternatives which are being pursued include choosing a year with more Landsat images, choosing a larger basin, choosing a year with minimum snowpack, and running the model for several years and on a year-round basis if data are available. Within Utah's Wasatch Mountains are ideal river basin locations for these future studies. JUNE __ MEASURED STREAM FLOW ____ SIMULATED STREAM FLOW J U LV AUGUST Figure 5. Daily Simulated and Observed Hydrographs during the Snowmelt-Runoff Season for American Fork Basin. 545 REFERENCES Martinec, J. , A. Rango, and E. Major, 1983. The Snowmelt-Runoff Model (SRM) User's Manual. NASA Reference Publication 1100. 118 pp. Martinec, J. Parameter Modeling. and A. Rango, 1986. Values for Snowmelt Runoff Journal of Hydrology. Rango, A. and J. Martinec, 1979. Application of a Snowmelt-Runoff Model Using Landsat Data. Nordic Hydrology 10:4:225-238. Rango, A, 1985. Results of the Snowmelt- Runoff Model in an International Test. Proceedings of the Western Snow Conference, Boulder, Colorado. 546 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 INITIATION OF SPRING SNOWMELT OVER ARcriC LANDS David A. Robinson1 ABSTRAcr: The springtime initiation of snowmelt in the Tanana River Basin and over the North Slope of Alaska was found to correlate with daily regionally averaged values of parameterized absorbed shortwa!2 radiation at the ~~ound (Q) of from 6-8 MJm and 2.5-4.5 MJm , respectively. These values are a function of solar insolation, surface albedo and atmospheric screening factors. Results are based on analyses of satellite imagery and ground station data from 1978-1985. During these springs, the date melt started ranged from 3/27 to 4/23 (ave. 4/10) in the Basin and from 4/13 to 5/28 (ave. 5/10) on the Slope. An early or late melt in the Basin was not paralleled on the Slope. Regional albedo at the start of melt was approximately 0.35 in the Basin and 0.75 on the Slope. As the two regions differ by only several degrees of latitude, this difference in albedo appears to be the primary explanation for the considerably later date at which me 1 t coiTI!lenced on the Slope. In both regions, an early start of melt appeared to be caused by abnormally warm air advec ted in to a region. On the Slope, regardless of the starting date, the most rapid interval of decreasing albedo ~~urr~~ once a secondary threshold of 6-8 MJm day was reached. Correlations between Q and loca 1 and regional temperatures appeared strongest on the Slope. (KEY TERMS: snowmelt, absorbed shortwave radiation, albedo, temperature, runoff) INTRODUcriON The impact of spring snowme 1 t on hydrologic regimes in the arctic and subarctic depends on the water content of the snowpack and on the timing and duration of the melt period. An early, rapid, melt might result in spring flooqing, followed by drought in summer. While considerable effort has gone into improving estimates of the potential runoff from a snowpack (eg. Rango et al., 1975; Ostrem et al., 1979; Shafer et al., 1979), less attention has been paid to the prediction of the start and duration of snowmelt, with some notable exceptions {eg. Carlson et al., 1974; Rasmussen and Ffolliott,-197"9). Here, using shortwave satellite and ground data, we report on variations of spring snowmelt in the Tanana River Basin and on the North Slope of Alaska from 1978 to 1985. We found that during the study period the initiation of melt in these regions was related to the amount of shortwave energy available for surface heating, which in spring was primarily controlled by insolation and surface albedo. The latter, in turn, is a function of relief, the height and density of the vegetation canopy and of the extent, depth and physical characteristics of the snowpack {Kung et al., 1964; Robinson and Kukla, 1984). The latter properties depend on the age of the upper layers of the pack. The timing and duration of snowmelt was monitored by observing regional changes in surface brightness using satellite data. 1Lamont-Doherty Geological Observatory of Columbia University, Palisades, New York 10964~ 547 DATA Imagery from NOAA TIROS and Defense Meteorological Satellite Program (DMSP) near-polar orbiting satellites was examined. The Advanced Very High Resolution Radiometer onboard the NOAA satellites provided imagery with a resolution of 1 km in both visible (0.55-0.68um) and infrared (10.5-11.5um) channels. Daily data covering Alaska were available throughout the study period. Direct readout DMSP imagery was used for sea ttered intervals during the period. Spectral ranges are 0.4-1.1um and 10.5-12.5um. Resolution is 0.6 km. Daily data of surface air temperature and snow depth were taken from Climatological Data for Alaska (NOAA, 1978-85). METHOD Daily sets of surface albedo, temperature, snow depth, cloudiness and absorbed shortwave radiation at the surface were assembled for the Tanana Basin and North Slope regions (Figure 1) from March 15 to May 15 and April 15 to June 15 of each study Figure 1. State of Alaska, showing the locations of the North Slope (NS) and Tanana Basin (TB) study regions. The locations of climate stations used in the study are marked with dots. 548 year, respectively, as follows: 1. Surface albedo: Clear-sky shortwave AVHRR scenes close to the satellite subtract and showing different stages of progressing snowmelt were selected from the eight year data base to create a "master melt progression set" for each region. The average surface brightness of each scene of the sequential set was measured on an image processor. Specular reflectance is minimal over the surfaces measured at spring solar zenith angles in Alaska and at the satellite viewing angle (Taylor and Stowe, 1984). Surface albedo was computed from brightness data by linear interpolation between homogeneous bright and dark snow-covered or snow-free targets in each scene, whose surface albedos were estimated from data gathered in Alaska and elsewhere (eg. Larsson and Orvig, 1962; McFadden and Ragotzkie, 1967; Weller et al., 1972; Maykut and Church, 1973; Robinsonand Kukla, 1984). This method effectively eliminates differences in brightness between images resulting from variations in image production. Shine and Henderson-Sellers (1983), using a radiative transfer scheme, showed that this approach works well with satellite imagery, regardless of channel width. Albedos so calculated on an independent image may be up to 0.10 too low or 0.05 too high (Robinson and Kukla, 1985) • However, monitoring the time progression of brightness histogram statistics throughout the melt period, such as standard deviation, skewness and kurtosis, narrowed the error range. The albedo of a study region on a specific day during the eight year period was obtained by visual comparison of a clear-sky image for that day with the measured "master melt progression set". Where clouds prevented analysis, albedo was estimated by linear interpolation from the closest clear-sky dates. 2. Temperature and snow depth: Daily regional averages of high and low temperature and snow depth were calculated by averaging reports from stations evenly distributed within the study regions (Figure 1). The Tanana Basin stations included Big Delta, Fairbanks, McKinley Park and Tanana. Barrow, Barter Island and Umiat comprised the North Slope network. 3. Absorbed radiation: Daily parameter- ized shortwave radiation available for surface heating or evaporation was computed from a simple model, whose inputs were regional surface albedo (a), insolation at the top of the abnosphere (I) and a parameterized regional atmospheric screening factor (S) (equation 1). Q = (1-a) I S (1) Since an analysis of cloud cover from the polar orbiter imagery showed no seasonal trends in either region, the average observed cloud coverage during the eight melt seasons was used to derive a mean atmospheric screening factor of 0.47 in the Basin and 0.44 on the Slope. The preceding values were obtained from once-a-day visual estimates of areal cloud cover in each study region, following a method previously employed over arctic sea ice (Robinson et al., 1985). This involved a visual inspec tionand charting of clouds in shortwave and infrared imagery, determining a fraction of cloud-covered area in the study region and estimating relative cloud thickness from the degree to which the clouds obscured the underlying surface contrast. Clouds were classified as thin, moderately thick or thick. Features under thin clouds were clearly recognizable but with reduced contrast. Features were marginally recognizable through moderately thick clouds and fully obscured under thick clouds. Cloud-class screening factors used in calculating regional screening factors were 0.6 (thin clouds), 0.4 (moderate) and 0. 2 (thick) and clear skies 0. 7. These highly approximate values were based on measurements in other regions at solar zenith 0 0 ( angles of between 40 and 70 Robinson, unpublished), the range of spring midday angles in the study regions. As only one morning or midday image per day was available for analysis, the cloudiness statistics do not take into account any potential diurnal cycle of cloud cover. RESULTS North Slope In the early spring each year, regional surface brightness was homogeneous, with surface albedo close to 0.80. Subsequently, rivers began to appear as dark ribbons through the snow-covered region, as snowmelt runoff from the foothills of the Brooks Range reached the area (cf. Holmgren et al. (1975) 549 for a detailed description of breakup on the North Slope). As a result, regional albedo decreased to about 0. 75. Once the paths of streams originating in the tundra began to appear, albedo fell below 0.75. At this time we considered regional melt to be underway. When albedo reached approximately 0.25, only a few patches of snow and numerous frozen ponds were recognized on the imagery as being brighter than the surrounding snow-free tundra. At this point the major period of melt was considered over. The timing and duration of the North Slope melt varied by up to 45 and 38 days, respectively, during the 1978-1985 period (Table 1). When melt began in 1a te April to mid-May (Figure 2) it tended to last longer than when it conmenced later in May (Figure 3). TABLE 1. Dates on which the major period of snowmelt started (albedo fell below 0. 75) and ended (albedo reached 0.25) on the North Slope. The duration (Dura.) of melt and the daily parameterized absorbed ~2or~ive radiation at the ground (Q) (MJm day ) in the region on the date melt conmenced and at the start of the most rapid 10 day period of albedo decrease are also given. Year 1978 1979 1980 1981 1982 1983 1984 1985 Start Date Q 5/24 4.4 4/19 2.8 5/21 4.2 5/6 3.6 5/29 4.5 4/13 2.5 5/28 4.5 5/16 4.2 End Dura. (days) 6/13 20 6/4 46 6/15 25 6/13 38 6/18 20 6/7 55 6/14 17 6/3 18 Rapid start: Date Q 6/5 7.2 5/28 10.9 6/6 7.3 5/28 8.1 5/31 6.1 5/29 7.8 6/6 6.9 5/24 6.7 Early melt was associated with abnormally warm temperatures. On the average, highs were near 0°C during the period of initial slo.J albedo decreases of from about 0. 75 to 0.55. Following this, an interval of stable albedo and colder temperatures was conmon. In the last week of May, the albedo again dropped at an increased rate. The major melt .II) .10 .II) ; .w :;;;;! ,«l .3D .20 ·ID 1111 liD l2ll 130 1<111 lSD Ull 110 JLUAN DAY 16 ~ ID 5 ~ 0 ~ -5 ~ -ID •15 I -m -25 -3D .. 1111 liD 1Zil 130 1<10 lSD Ull 110 JULIAN DAY Figure 2. Daily regional surface albedo (top) and regionally averaged high and low temperatures (bottom) from April 15 (Julian day 105) to June 15 (Julian day 166), 1979 on the North Slope. ended between June 3 and 13. High 0 temperatures were close to +5 C during the last active phase of melt and frequently rose above 5°C following the melt. Lgw temperatures generally remained below 0 C until melt was complete. In years with a later start of melt, temperatures prior to, during and following the melt were similar to those for the early start years. However, in late years melt was most active in early June and ended between June 13 and 18. There was no relationship between snow depth measured at the s tart of melt at the three Slope stations and the satellite- derived timing and duration of the regional melt. It is uncertain, however, to what degree these stations represent regional conditions. Daily parameterized absorbed shortwave radiation at the ground at the conmencement 550 ·Ill .10 .II) ~ ·50 ,.f) ·liD ·10 ·ID 1111 liD 1Zil 130 1<10 lSD Ull 110 Jl.UAN DAY 16 ~ ID 5 ...... /'-~ ~ 0 ,-../ ,-~ -5 ,.1\f"., ... , I -ID M ,;J ,. ' -15 ' \. , r"\ ! \1 -fD I ' ~~ " -26 \ I \J -Sl JV .. 1111 liD 1Zil I3D 1«1 lSD Ull 110 JLUAN DAY Figure 3. Same as Figure 2, but for 1984 (Julian days 106 to 167). -2 of regional melt ranged from 2.5-4.2 MJm (1J=2.239cal) in early melt years and 4.2-4.5 M.Jm-in late melt years (Table 1). At the beginning of the most rapid 10 day decline in surface albedo, albedo values ranged from 0.53 to 0.65 and_~ily Q values were between 6.1 and 8.1_2llm • The only exception was the 10.9 MJm value in 1979, when albedo was 0.37. There was a large range between the daily values of Q over the North Slope in the earliest and latest melt cases (Figure 4). On average, over twice as much radiation was absorbed on a given date in late May or early June in 1979 than in 1984. Correspondingl6, in May 1979, Slope temperatures were 4 C higher than in 1984. Within the region, hiijh temperatures in May 1979 differed by 1g C between Umiat, where the high averaged +5 0 C, and Barrow, where the high averaged -5 C. Albedo during this month was considerably lower in the vicinity of Umiat than at Barrow. In 1984 the Slope remained snow 18 I J ______ JO_, ___ .. __ ,.-----J' ---~-___! o~~~~~~~~~~~~~~ m m ~ ~ ~ oo w m JULIAN DAY Figure 4. Daily parameterized absorbed shortwave radiation at the ground in the North Slope test reglon from April 15 to June 15 in 1979 (top curve) and during the same period in 1984 (bottom curve). Results based on a model which incorporates realistic surface albedo, top of a t:mosphere insolation and parameterized mean eight year surface radiation income. Bottom axis shows Julian days in 1979. Add one day to get Julian days in 1984. covered throughout Mad. and the mean Umiat high (-1°C) was only 4 C higher than Barrow's (-5°C). Tanana Bas in In late March and early April of the eight studied years, surface albedo in the Tanana Basin ranged between 0.32 and 0.38. The higher value appeared to come close to the Basin's potential maximum, while the lower value was the result of meager snow cover in upstream valley regions. Once the surface brightness began to continuously decrease, the melt was considered to be underway. The rate at which albedo decreased remained relatively constant until the regional surface albedo was close to 0.20. This was considered the active period of melt. When albedo reached 0.20, snow cover was only visible on the higher ground and occassionally in downstream portions of the 'Basin and the major melt period was considered over. The timing and duration of Basin melt varied by up to 27 and 19 days, respectively, between 1978-1985 (Table 2). Melt commenced between April 6 and 14 in six of the eight 551 TABLE 2. Dates on which the major period of snowmelt started (albedo fell and stayed below the spring maximum) and ended (albedo reached 0.20) in the Tanana Basin. The duration (Dura.) of melt and the daily parameterized absor~zd s~~rtwave radiation at the ground (Q) (MJm day ) in the region on the date melt commenced are also given. Year Start End Dura. Date Q (days) 1978 4/9 7.2 4/28 19 1979 4/14 7.9 5/2 18 1980 3/27 5.9 4/29 33 1981 4/14 8.1 4/28 14 1982 4/6 6.6 5/13 37 1983 4/9 7.2 4/30 21 1984 4/10 7.5 5/8 28 1985 4/23 8.5 5/20 27 years, with 1980 (March 27) (Figure 5) and 1985 (April 23) (Figure 6) being the only exceptions. There was some indication that the melt took longer when it began earlier, but snow depth, which on a regional scale was poorly known, may also have affected the duration. In 1981, when the depth of the snowpack at the four Basin stations averaged the lowest ( 17 em) , the melt duration was the shortest (14 days). Conversely, in 1985, when the pack was deepest ( 64cm) , the duration was relatively long (24 days), despite having the latest starting date. High temperatures during the early stages of melt were generally within several degrees of freezing. Only in 1980, when there was a persistent period of abnormal highs of about +5°C in late March and early April, were temperatures noticeably higher at the beginning of melt. In all years, highs were 0 0 usually between +5 C and 10 C during the middle and later stages of melt and regularly 0 above 10 following melt. Low temperatures 0 generally remained below 0 C until melt was complete. Daily parameterize~2 values of Q were between 6.6 and 8.1 MJm at the commencement of melt in six of the ~2ght years (Table 2). Onl~2 in 1980 (5.9MJm ) and in 1985 (8.5 MJm ) did they fall outside of this range. These two years exhibited the largest ·411 •IU .41! ! ·liD ·«l ,31J ... .II lU 110 Ill llll liD I2D 1:11 l-«l JUL!AN DAY 211 ~ l5 Ill li f\. 1\r, ~"' ~ 0 J .../'1 II r, ,.., , " {'' "i J\ .s I /'; '"\ '0; -ID i\J " \ -15 I f -211 I -25 I I ·:Ill v ~ lU 110 Ill IIJI m 1211 1:11 140 JWAN DAY Figure 5. Daily regional surface albedo (top) and regionally averaged high and low temperabJres (bottom) from March 15 (Julian day 75) to May 15 (Julian day 136), 1980 in the Tanana Basin. differences in daily Q amongst the years s bJdied (Figure 7). In the latter half of Apri~, daily values differed bf close to 2.5 MJm-, with 1985 values lagging about 2 weeks behind those in 1980 (Figure 7). In April 1980, the mean temperabJre in the region was +1°C, while in April 1985 it was -8°C. DISCUSSION The approximately six degree difference in latitude of the Tanana Basin (64°N) and North Slope (70°N) s rudy regions results in the Slope receiving 16% less daily insolation at the top of the atmosphere than over the Basin on April 1 and 6% less on May 1. Thus, there would be a difference of only several days in melt initiation be tween the two 552 . ., -10 .eo ~ .m .«J .311 ... .l!J 711 811 Ill :lfJI liD I2D 1:11 l-«l JULIAN DAY 211 l5 ~ ID & eli 1\ r/\ ~ ,-, f\ '"' \I ~ .s V-' '"' . ,...., / -iO V, I \ I I~ t. ''{" ~ -i5 I • {I f I \Jt \ I I -211 t' "v\t v'J -25 -SJ lU 110 Ill llll llll 1211 1:11 1..0 Jtl.lAN OAY Figure 6. Same as Figure 5, but for 1985 (Julian days 74 to 135). 1.8 l 14 ~ 12 ] Ill ~ ~ i Figure 7. Daily parameterized absorbed shortwave radiation at the ground in the Tanana Basin sbJdy area from March 15 to May 15 in 1980 (top curve) and in 1985 (bottom curve). See Figure 4 for model description. Bottom axis shows Julian days in 1985. Add one day to get Julian days in 1980. regions in April if regional snow-covered albedos were the same. By late May, the Slope receives more insolation at the top of the atmosphere than the Basin. Therefore, the difference in the regional albedos of the two regions when snow covered appears to be the primary reason why the most active period of melt in the open country of the Slope began an average of 51 days later than the start of the active period of melt in the forested Basin. In each region, there was a thresholg2 range of daily Q of approxilila tely 6-8 MJm reached before the most active period of decreasing albedo began. This threshold range was reached in mid-April in the Bas in but in the last week of May or first week of June on the Slope. An early or late start of melt in the Basin did not necessarily correlate with the relative timing of initial melt on the Slope. The starting date of initial melt in the two regions differed by as few as 4 days and as many as 55 days. This range was generally a reflection of the widely ranging starting dates on the North Slope. There was also no relationship between the two regions regarding the time at which melt ended. This differed from 14-57 days. An early commencement of melt on the North Slope in April or early May did not trigger the rapid demise of snow cover over the entire Slope, apparently because of the 1~ values of absorbed radiation which exist at that time of spring. Insolation is considerably below its late spring values and the albedo, despite a partial decrease from its maximum, remains relatively high. Correlations between absorbed radiation and the most active period of melt and between absorbed radiation and local and regional temperab.Jres appear strongest on the North Slope. This is likely due to the greater simplicity of the boundary layer dynamics on the Slope compared to the Basin. In the latter, the water content of the snowpack is more variable in an absolute sense in a given year as well as between years. The forests in the Basin add to the complexity of the radiation budget, as does its more variable topography and frequent inversions. The radiation budget of the Slope is more singularly and directly affected by the snowpack. Several of the s rudy results stand out as being of importance to the hydrologic conm.mi ty. 1) There is not a parallel 553 relationship between an early or late melt in the Basin and on the Slope. 2) The major period of snowmelt runoff can not be expected to occur in either region until the initial shortwave radiation threshold is reached. 3) Whe!!z tEf secondary threshold of 6-8 MJm day is reached on the Slope, ·the subsequent runoff should be reduced in years when the melt starts early. 4) The duration of melt in the Basin appears to be more sensitive to the initial water content of the snowpack than on the Slope. Aa<NOWLEIX;MENTS Appreciation is expressed to Mark Serreze and George Kukla for their contributions to this s rudy. This work was supported by Air Force grant AFOSR 86-0053 and National Science Foundation grant ATM 85-05558. This is LOGO contribution 3972. LITERATURE CITED Carlson, R. F. , W. Norton and J. McDougall, 1974. Modeling Snowmelt Runoff in an Arctic Coastal Plain. University of Alaska Instib.Jte of Water Resources Report IWR-43, 72pp. Holmgren, B., C. Benson and G. Weller, 1975. A S rudy of the Breakup on the Arc tic Slope of Alaska by Ground, Air and satellite Observations. In: Climate of the Arctic, G. Weller and S.A. Bowling (Editors). Twenty-Fourth Alaska Science Conference, 15-17 Aug. 1973, 358-366. Kung, E. C., R.A. Bryson and D. H. Lenschow, 1964. Study of a Continental Surface Albedo on the Basis of Flight Measurements and Strucb.Jre of the Earth's Surface Cover Over North America. Mon. Wea. Rev. 22:543-564. Larsson, P. and S. Orvig, 1962. Albedo of Arctic Surfaces. Arctic Meteorology Research Group, McGill University, Publication in Meteorology No. 54, 33pp. Maykut, G.A. and P.E. Church, 1973. Radiation Climate of Barrow, Alaska, 1962-66. J. Appl. Met. 12:620-628. f1cFadden, J.D. and R.A. Ragotzkie, 1967. Climatological Significance of Albedo in Central Canada. J. Geophys. Res. 72: 1135-1143. NOAA, 1978-1985. Climatological Data, Alaska, v.64-71. Ostrem, G., T. Andersen and H. Odegaard, 1979. Operational Use of Satellite Data for Snow Inventory and Runoff Forecast. In: Satellite Hydrology, M. Deutsch, D.R. Wiesnet and A. Rango (Editors). American Water Resources Association, 23Q-234. Rango, A., V.V. Salomonson and J.L. Foster, 1975. Employment of Satellite Snowcover Observations for Improving Seasonal Runoff Estimates. In: Operational Applications of Satellite Snowcover Observations, A. Rango (Editor). NASA, 157-174. Rasmussen, W.O. and P.F. Ffolliott, 1979. Prediction of Water Yield Using Satellite Imagery and a Snowmelt Simulation Model. In: Satellite Hydrology, M. Deutsch, D.~. Wiesnet and A. Rango (Editors). American Water Resources Association, 193-196. Robinson, D.A. and G. Kukla, 1984. Albedo of a Dissipating Snow Cover. J. CU. and Appl. Met. 23:1626-1634. Robinson, D.A. and G. Kukla, 1985. Maximum Surface Albedo of Seasonally Snow-Covered Lands in the Northern Hemisphere. J. Cli. and Appl. Met. 24:402-411. Robinson, D.A., G.J. Kukla and M.C. Serreze, 1985. Arctic Cloud Cover During the Summers of 1977-1979. Lamont-Doherty Geological Observatory of Columbia University Technical Report LDG0-85-5, 175pp. Shafer, G.A., C.F. Leaf and J.K. Marron, 1979. Landsat Derived Snow Cover as an Input Variable for Snow Melt Runoff Forecasting in South Central Colorado. In: Satellite Hydrology, M. Deutsch, D.R. Wiesnet and A. Rango (Editors). American Water Resources Association, 218-224. Shine, K.P. and A. Henderson-Sellers, 1983. Cryosphere-Cloud Interactions Near the Snow/Ice Limit: Sensitivity Testing of Model Parameterizations. Final Report NSF Grant ATM 8Q-18898, University of Liverpool, 158pp. Taylor, V.R. and L.L. Stowe, 1984. ~eflec tance Characteristics of Uniform Earth and Cloud Surfaces Derived From Nimbus-7 ERB. J. Geophys. Res. 89:4987-4996. Weller, G., s. Cubley, S. Parker, D. Trabant and C. Benson, 1972. The Tundra Microclimate During Snow-melt at Barrow, Alaska. Arctic 25:291-300. 554 RIVER ICE HYDRAULICS 555 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 FORECASTING mE EFFECTS 00 RIVER ICE IXJE TO THE PROF03ED SUSITNA HYIH>EUX:I'RIC PinJEX:l' Ned w. Paschke and H.W. Coleman* ABSTRACT: River ice processes affect the physical and hydraulic properties of many of the world's rivers. Although winter flaws are characteristically low, the additional friction and ice displacement within an ice-covered river can greatly increase the water surface elevation. The Susitna River, located in south- central Alaska, is generally subject to river ice processes for 6 or 7 months of each year. Environmental studies in connection with the proposed Susitna Hydroelectric Project (Alaska Power Authority, 1985) included documentation of natural (pre-project) river ice condi- tions and forecasting the effects of the project on river ice. In this regard, a numerical river ice model was calibrated and applied to an 85~ile reach of the Susitna River downstream of the proposed project. This paper presents a summary of the river ice modeling process, ob- served trends in natural ice conditions and the expected effects of the proposed project. (KEY TE~S: cold regions: river ice: winter hydro operation, river ice model- ling.) IN'I'R>IXJCI'IOO Proposed Susitna Hydroelectric Project The proposed Susitna Hydr~lectric Project includes the constt~ction of two large dams on the Susitna River (Figure 1). Watana Dam, an earthfill structure with an ultimate planned height of 885 feet, would be located 184 river miles upstream from the river mouth at Cook Inlet (i.e., "RM 184"). I:evil Canyon Dam, a 645-foot high concrete arch structure, \<Tould be located at RM 152, i.e., 32 miles downstream of Watana Dam. The project is planned for construc- tion in 3 stages as follows: Stage I -Watana Dam would be constructed to an initial height of 700 feet. Stage II -Devil Canyon Dam (full height) would be constructed. Stage III -Watana Dam would be raised to its ul- timate height of 885 feet. * Respectively, Hydraulic Engineer, Harza Engineering Company, 150 South Wacker Drive, Chicago, Illinois 60606 (Presently, Assistant Director of Engineering and Planning, Madison Metropolitan Sewerage District, 1610 Moorland Road, Madison, ~tisconsin 53713); and Head, Hydraulic Analysis and Design Section, Harza Engineering Company, 150 South Wacker Drive, Chicago, Illinois 60606. 557 Stage I is planned to begin operation in the year 1999. Stages II and III would be added in accordance with energy demand. Figure 1. Susitna River Location Map Environmental Background Environmental concerns regarding ice processes on the Susitna River include potential effects on the salmon popu- lation. A number of slough and side channel areas along the river provide habitat for spawning and juvenile over- wintering. These areas are generally isolated from the rnainstem by a natural berm at the upstream entrance to the slough or side channel. During the winter, these areas are often warmer than the mainstem due to upwelling of rela- tively warm (e.g. 3°C) groundwater. Ice- induced stage increases periodically overtop some of the berms under natural conditions, flooding the slough with 0°C rnainstem water and possibly harming the developing salmon. With the proposed 558 project in operation, changes from the natural flows and stream temperatures will affect the river ice conditions and the frequency and severity of the slough overtopping events. The river ice rrodel therefore focuses on the timing and magnitude of ice-induced river stage variations at the slough and side channel locations. METHOOOI.DGY Study Reach River ice modeling was limited to the "middle reach" of the Sus i tna River, i.e., the 85-mile reach from the Watana damsite to the "three-river confluence" forned by the Susitna, Chulitna and Talkeetna Rivers (Figure 1). Downstream of this confluence, the substantial incoming tributaries are expected to lessen the relative effects of the future project. Typical natural river flew rates in the vicinity of the damsites range from 30,000 cfs in June to less than 2,000 cfs in March. With the pro- ject operating reservoir releases would generally be 8,000 to 13,000 cfs year- round. Susitna River Ice Observations Ice observations on the Susitna River have been documented for the past five winters (R&M Consultants 1981, 1982, 1983, 1984, 1985}. Natural ice processes on the Susitna typically begin in early October with the generation of frazil ice, i.e., small ice crystals probably formed in supercooled surface water exposed to subfreezing air temperatures (Ashton, 1978}. Frazil ice is first observed in the middle and upper reaches of the Susitna which are subject to the coldest air temperatures. Reaches of low solar radiation and high water turbulence appear to be highest in frazil ice pro- duction. As the frazil is carried downstream, it coalesces into pans or rafts of "slush" ice which are often 2 to 5 feet in diameter and which may cover as much as 80% of the river surface. Border ice is observed to form as some of the slush ice pans come to rest and freeze together in law velocity zones along the river banks. Typically late in October, an accumu- lation of slush ice becomes jammed and freezes together near the river mouth, forming a stationary ice bridge across the river. Formation of the ice bridge appears to be triggered by a high con- centration of slush ice pans, law air temperatures and a high tide event in Cook Inlet which substantially reduces river velocities for several ndles up- stream. Following formation of the ice bridge, slush ice pans accumulate against its upstream edge and thereby advance the ice cover in an upstream direction. Same slush is observed to be swept beneath the ice front and is apparently deposited downstream on the underside of the cover, thereby thickening the ice cover. Pe- riodic mechanical oampression or "shoving" of the advancing ice cover (Pariset et. al., 1966) is observed, whereby as much as 2000 feet of the slush ice cover consolidates and thickens. The advancing ice cover often reaches the three-river confluence (RM 98) in November and the vicinity of Gold Creek (RM 137) in late I:ecember or January, rut varies with weather and flow conditions. Observed ice front progression rates in the middle reach typically range from 0 to 2 miles per day. Ice front progres- sion generally becomes undefined upstream of Gold Creek, where intermittent bridg- ing of border ice precedes the arrival of the ice front. River stage increases of 2 to 6 feet in the middle reach are ccmnon during progression of the ice front, and over- topping of some slough and side channel areas has been observed. Slush ice cover thicknesses are observed to vary substan- tially along the river and particularly across the channel width. Often little or no ice is observed in a central, high ~locity core area whereas accumulations as great as 12 feet thick reaching the channel bottom have been observed closer to the river banks. Following progres- sion, the upper surface of the slush ice cover begins to freeze into solid ice. The solid ice portion of the ice cover is observed to reach typical thicknesses of 559 2 to 4 feet by February or March. Spring breakup of the ice cover typi- cally occurs in early May and is largely due to the natural flow increases which lift and fracture the cover. Sporadic ice jams caused by blocks of the frac- tured ice cover are observed to cause greater stage increases and more frequent slough overtoppings than those of the initial ice cover progression. River Ice Model The numerical river ice model was based on the work of Calkins (1984) on the ottauquechee River and was modified and calibrated by Harza-Ebasco ( 1984) for application to the Susitna River. The model provides. a daily summary of hydrau- lic and ice conditions throughout the study reach during the period from November 1 to April 30. A detailed description of the model and its govern- ing equations has been presented by Calkins (1984). The general features of the model are briefly summarized as follows: 1. Hydraulic profiles are computed daily based on the Bernoulli and Manning equations (standard step method). Computations include the effects of the ice cover and border ice where appropriate. 2. Frazil ice production within reaches of 0°C open water is com- puted by a heat transfer coeffi- cient approach (Ashton, 1978). Cumulative frazil flaw rates are tabulated as the ice travels downstream. 3. Border ice growth is oamputed as a function of air temperature and water velocity and is calibrated to Susitna observations. 4. Hydraulic conditions at the ice front determine if the slush ice pans are swept beneath the ice cover or accumulated at its upstream edge, thereby advancing the ice front (Figure 2). repo- sition of slush ice beneath the cover is conputed based on the ice supply and water velocity under the ice cover. Thicknesses of the advancing ice cover are computed in accordance with OAM Pariset, et. al. ( 1966). Ice front progression rates are based on river geometry and the supply of ice reaching the front. 5. Solid ice growth within the slush ice cover is conputed (Figure 2). 6. Melting of the ice cover and re- treat of the ice front are c~ puted when warm water (i.e. above 0°C) reaches the ice front. Water temperature decay beneath the ice cover is also computed. Mechanical breakup of the ice cover is not simulated by the m:xiel. WATER r;MP > I'J°C I WATER TEMP= 0°~ OPEN WATER FLOW PLAN PROFILE Figure 2. River Ice Schematic SOLID ICE GROWTH Surveyed river cross-sections at 102 locations between Watana damsite (RM 184) and the "three-river confluence" (RM 98) were used in the m:xiel. Manning's "n" values ranging from .022 to .065 were selected to calibrate the open-channel portion of the model to stage-discharge measurements. Daily air temperatures and wind speeds recorded at 3 locations along the study reach were used for the various ice processes in the model. For simula- tions of natural (pre-project) condi- tions, daily flow rates and frazil ice discharges were input at the upstream boundary based on observations at Gold 560 Creek (RM 137). For with-project condi- tions, flow rates and water temperatures upstream of the ice front were provided by corresponding reservoir and stream temperature simulations. Starting dates for the simulated with-project ice front progression at the three-river confluence were based on tabulation of the total ice volumes supplied to the lower Susitna River and the time required to advance the lower Susitna ice front from Cook Inlet to the three-river confluence. The river ice model is primarily in- tended to simulate the timing and mag- nitude of river stage variations as- sociated with ice. Simulated natural ice conditions shaw reasonably good agreement with field observations (Figure 3). Limitations of the model relate primarily to its one-dimensional nature. Veloc- ities and ice cover thicknesses computed by the model are mean or characteristic values intended to represent an entire cross-section. Actual velocities and ice thicknesses are likely to be quite non- uniform within the cross-section. z 0 150 140 ~-130 8o!! ~~ 120 Z.!: oa: ~-110 w !,! w 14 "' ~ 12 ~ 10 ~-; 8 -I ~~ 6 i 4 x 2 ~ 0 NOV DEC JAN FEB REFERENCE LINE J 3000 cfs OPEN WATER SURFACE PROFILE I I I .I I 1\ .....J..--; 100 _... I 110 . 120 RIVER MILE "' MAR APR J I I 130 140 1r~-~---w ----++----+----, MF---+--------i--f4l--+----¥N------If--+----+--+-t+-ll,lllll ~ ~1oo~----~110~~u_~1ro~~~~~130~~~~140~~ RIVER MILE LEGEND -SIMULATED • OBSERVED Figure 3. Sample River Ice Calibration Winter 1982-1983 RESULTS Sample results of river ice simula- tions for natural conditions and the three stages of the project are shown in Figures 4 and 5. These simulations are based on weather conditions of 1981-82 (an average winter in terms of mean air temperatures) and shaw typical trends. with Stage I in q;>eration, ice front progression at the "three-river confluence" during an average winter is expected to be delayed until rrdd- December, about 3 weeks later than that of natural conditions (Figure 4). With the operation of the project Stages II and III, respectively, the ice cover progression is expected to be further delayed until late December or early January (Figure 5). Spring meltout in the middle reach of the Susitna River with Stage I operating is expected to be ccrrpleted by late April (Figure 4), about 2 weeks earlier than the natural breakup. With q;>eration of -130 ~ ~ i 120 ~ > ! 110 100 14 12 10 100 REFERENCE LINE· _r:__)ooo ef1 OPEN WAi-ER SURFACE PROFILE I ....., .... I . ~· \. I !/ . ./ -""'., ·-· " , I 110 120 l\1.~~ \ II' -, --I 130 140 Stages II and III, the meltout would be further advanced (Figure 5) , occurring in late to early March, respectively. The delayed ice front progression and the earlier-than-natural ice meltout with the project operating is due primarily to warmer-than-natural water temperatures released from the project reservoirs during the winter months. The maxLmum upstream extent of the ice cover during an average winter is expected to be in the vicinity of RM 139 with Stage I operating. This ice cover extent would be reduced to near Rt1 133 with Stage II q;>erating and further reduced to the vicinity of RM 114 with Stage III operating (Figure 5). Little or no ice is expected upstream of these locations with the project operating, whereas under natural conditions these reaches become covered by border ice growth. The total thickness of the river ice cover with Stage I operating is expected to be generally sirrdlar to that of natu- ral conditions (Figure 4). Ice cover 150 140 ~ ;:: 130 "-go! .J:E !Z.t 120 Oa: ff-110 100 2 0 REFERENCE LINE / 3000cft OPEN WATER SURFACE PROFILE I I I --lo;. ~ ... ------- t 100 110 ~~ -~ ...... 120 RIVER MILE ~-.--...\ r\ "' /'-~' --~ ~. 130 140 ·~I~---·---lM:l)U.Jfi,JIIIII li1~-ijtt11}lfl-l.llll IOO 1)0 I20 I30 ,10 RIVER MILE fUVER MILE LEGEND' -NATURAL CONDITIONS ·-·-STAGE I OPERATING • NATURAL SLOUGH BERM ELEVATION Figure 4. Simulated River Ice Conditions Stage I vs. Natural 1981-1982 Weather Conditions 561 LEGEND ---COLD WINTER 1971-72 --AVERAGE WINTER 1981-82 ·-•••• WARM WINTER 1976-77 • NATURAL SLOUGH BERM ELEVATION NOTE' ALL SIMULATIONS ASSUME WATANA DAM IS ON-LINE Figure 5. Simulated River Ice Conditions Stages I, II and Ill 1981-1982 Weather Conditions thicknesses are expected to be progres- sively reduced with the addition of Stages II and III (Figure 5). The reduced extent and thicknesses of the ice cover with the project operation again primarily reflect the warmer-than-natural reservoir release temperatures. Maximum river stages within the ice- covered reaches during operation of Stages I, II and III are expected to be generally several feet higher than those of natural conditions. This is due primarily to the greater-than-natural flow rates with the project in operation. The frequency and magnitude of the slough overtopping events upstream of the ice front with the project operating are therefore expected to be less than or equal to those of natural conditions. The simulation results discussed above are based on the average winter weather conditions of 1981-82. Simula- tions were also made for a cold winter (1971-72) and a warm winter (1976-77). Although these sinulations (Figure 6) were based on a different construction and operational schedule than Figures 4 and 5, they serve to indicate the sensi- tivity of the simulated river ice processes to weather conditions. With the project in operation during a cold winter, for example, the ice front would be expected to begin several weeks earlier and would extend several rrdles further upstream than for an average winter. Maximum ice cover thicknesses and river stages during a cold winter would also be about 2 feet greater than those during an average winter. During a warm winter, conversely, ice cover thicknesses and river stages are likely to be about 2 feet less than for the average winter. StHtARY A numerical river ice model was applied to the Susitna River to forecast the effects of the proposed Susitna Hydroelectric Project. The model simula- tions predicted delayed ice cover formation, reduced ice cover extent and earlier and rrdlder spring meltout rela- tive to natural conditions. Greater than natural river stages were predicted for 562 ,. 12 10 100 ~ ,.... REFERENCE LINEo 3000cfs OPEN WATER SURFACE PROFILE 1 __.. ~------'1- I 110 _, .. .. .......... / ., . 120 RIVER MILE . ' ' ""~ ... "\ lo"i . ·~ , ~\! ' 130 140 1'1 I b-kiHflfH,J 1111 0 I I 100 110 120 130 1CO RIVER MILE LEGEND: STAGE I OPERATING STAGE II OPERATING STAGE Ill OPERATING • NATURAL SLOUGH BERM ELEVATION Figure 6. Simulated River Ice Results For Various Weather Conditions sane reaches, and mitigation rreasures will be proposed therein. Weather condi- tions and project stage were shown to substantially affect the expected river ice conditions. REFERENCES Alaska Power Authority, 1985, Amendment to the Application for Major License, The Susitna Hydroelectric Project (Draft), prepared by Harza-Ebas~ Susitna Joint Venture. Ashton, George D., 1978, "River Ice", Annual Reviews in Fluid Mechanics, Vol. 10, pp 369-392. Calkins, D.J., 1984, "Nurrerical Simulation of Freezeup on the ottauquechee River", Workshop on Hydraulics of River Ice, June 20-21, 1984, Frederickton, New Brunswick, w 247-277. Gerard, R., 1984, Notes fran short course "Ice Engineering for Rivers and Lakes", University of Wisconsin. Harza-Ebasco Susitna Joint Venture, 1984, "Instream Ice Sinulation Study" for Alaska Power Authority. Harza-Ebasco Susitna Joint Venture, 1985, "Instream Ice Sirrulations; Supplemen- tary Studies for Middle Susitna River", for Alaska Power Authority. Pari set, E. , Rene Hausser and Andre Gagnon, 1966, "Fonnation of Ice Covers and Ice Jams in Rivers", ASCE Journal of Hydraulics Division, Vol. 92, HY6. R&M Consultants, Inc., 1981, "Ice Obser- vations, 1980-81", for Acres American for Alaska Power Authority. R&M Consultants, Inc., 1982, "Winter 1981-82, Ice Observations Report", for Acres American for Alaska Power Authority. R&M Consultants, Inc., 1983, "Susitna River Ice Study, 1982-83", for Harza- Ebasco for Alaska Power Authority. R&M Consultants, Inc., 1985, "Susitna River Ice Study, 1983-84", for Harza- Ebasco for Alaska Power Authority. R&M Consultants, Inc., 1985, "Susitna River Ice Study, 1984 Freeze-Up", Draft Report for Harza-Ebasco for Alaska Power Authority. 563 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 A STRUCTURE TO COHTKOL ICE FOIQiATIOlJ AlW ICE JAN FLvOiiiNG ON CAZENOVIA CllliEK, NEW YORK Steven R. Predmore 1 ABSTRACT: Cazenovia Creek, in \/estern r~ew York State, frequently ice jam floods the coronmnities of West Seneca and Buffalo. This report describes t:he ice character- istics of Cazenovia (.;reek and an ice control structure (IC3) proposed by the U.S. Arr;;y Corps of Engineers to ( 1) tnocJ.ify the creek's ice formation, (2) restrict and control Cazenovia Creek ice runs, and (J) eliLdnate ice jam flooriing in downstream residential and business areas. 'llH:! lCS, tO ·::,e locateu jUSt Upstream or the urt,an oevelopllient in \~est Seneca, consistu of a 6 ft (l.b m) high, 250 ft (76 m) long 111eir (ci.arr.). Extenuing 6 ft ( i. 0 ;:;l) a.tove the weir are nine concrete piers, spaced 25 ft (7 o 6 n.) apart, ~1hich \vill vrevent ice runs frorll moving cor..nstreau1. The weir ana its concrete piers corrprise the principal spillway for the ICS. Acijacent to this spillway are a ,:2.5 ft (6~ m) long elevat:ed weir and a as f t ( G~ t:,) long elt:va ted floociway for pa!:!8ir;g bigh flC\iS. \LLY 1LhL~: ice control structure; ice retention structure; ice jam flooding.) IhTLlluuGTivi~ The v1estern l'-iew rork conr:unities of i·1est ~enec.a imri I.:uff£~lv, located. along the lower reache~ oi Lazenc;.via ~reek, <.ire 1 annually ,,r: H;ctc.e: r.::y the creeK's lc.te 1.1nter cui <:!UI:i.y SJ'ring ice ja.L flooos. ' l'i•e cn.>(::t; 'L• peaK uiscr,4tr}·es, vnich arc often less than channel capacity uncer free-flow conditions, cause significant residential and business flooding as a result of ice jams which raise stream stages above bank level. In response to the ice jam flooding, the U.S. Army Corps of Engineers has designed and physically nodelled an ice control structure (ICS) which will reduce ice formation in i~est Seneca and Euffalo and elininate ice jam flooding in these communities. 'l'his report describes the ice formation anci ice-run charact:eristics of Cazenvc1.a ~reek ana the design anci operation of the l~J which will elioinate ice jar;;. floco uan,age. (.; ... -i.ZLhOVIA C~li:.LK ilatersheci Description The Cazenovia Creek watcrsr,e:.c. is about 30 1ni ( 4b km) long anci nas a rdlXii:ur.l \>.~"idth of 10 L1i (16 km) as shewn in Figt:.re 1. The creek is composed ot nio na.ir1 Lranches anri a r.ain otcr.:; .. >•,il.ich drain lJo rd 2 (357 kr}) into tr.e tuff&lo hiver mu C:.mro.strcaw L·rancnes, watersbec., \:hile tLe Lake ~:.rie. The t.&st ana ~.€:St in the t~eaaw"t~rs cr the are st:eer.ly slo]:eu ar.a rurG".l, lmio·er reacnes along tt,e nain stem in \lest Seneca a.no downstream .ourfc.lc are flat anci urban ci.evelopec. '.1l.e strea~ varies in \dt:..t:L trou, 100-200 ft (:i0-60 n;). Its averc;6e c.ischarge ever 44 years of reccra is :.i.Ju tt3/s (6.·~ u.J/s) zr.e: it::; maxi~:..ur_ reccrcec ii.ycrolo!Jist, u.S • .iirl"Y ~orps of ~ngineers, i.iuf:.:clo L,istri.ct, 1776 tiia~ara ;;)treet:, ' t.uf t nlo, rir llii.u 7. 565 ' 1.6 0 !'2:o.Oi. 0 1.6 3.2 BUFFALO l ' 4.8 64 3 4 WATERSHED 8.0 9.6 ·=-=a BOUNDARY-·· KILOMETERS MILES ONTARIO (/JToCA y 0 R FIGURE 1. CAZENOVIA CREEK WATERSHED. 566 z " .. discharge is 13,500 ft 3 /s (382 m3 /s). A reach of very still water extends from the mouth of the creek to a point about 1 mi (1.6 km) upstream. This still- water reach is created by a deepened channel and the back-water effect of Lake Erie. Stream velocities in this reach are very low. Upstream from the still-water area, Cazenovia Creek is only 2 ft (0.6 m) deep at low discharge and exhibits stream velocities of 1-4 ft/s (0.3-1.2 m/s). The different velocity characteristics of the still-water area near the mouth and the upstream fast water area result in different ice formation characteristics in the two areas. Ice Formation The ice cover on Cazenovia Creek first forms as sheet ice in the 1 mi (1. 6 km) long still-water reach near the mouth. Frazil ice generated in the faster waters upstream, is transported to the upstream end of the sheet ice where it is carried beneath the ice to form a hanging dam. Eventually, the frazil ice bridges the creek at the upstream end of the sheet ice and with its downstream movement halted, forms a complete ice cover which extends upstream for 7-30 mi (11-48 km), depending on the severity of the winter. Cold tem- peratures then solidify the surface of the frazil ice, resulting in 1.5-9 in (4-23 em) of solid ice underlain by up to 2 ft (0.6 m) of frazil. lee Break-up Because the headwaters of Cazenovia Creek are composed of two branches, the Cazenovia Creek ice break-up is often characterized by two separate ice runs. The East Branch of Cazenovia Creek experiences break-up first. The East Branch ice run subsequently enters the main stem, frequently jamming and flooding as it moves downstream. The jams cause no damage in the undeveloped upper reaches but cause significant residential and business damage when, at 6 mi (10 km) above the mouth, they reach West Seneca and then Buffalo. Following the first run, a second ice run occurs as ice va- cates the West Branch and moves relatively unrestricted to the mouth. 567 The average annual damage due to Cazenovia Creek flooding is $427,000, of which over $300,000 is caused by ice jam flooding (USACE, 1985). The remainder is due to free-flow flooding. The ice control structure (ICS) proposed by the the U.S. Army Corps of Engineers will alleviate the ice jam flooding in the urban-residential areas adjacent to the creek. ICE CONTROL STRUCTURE Design The ice control structure was designed and modeled in 1984 by David S. Deck of the U.S Army Corps of Engineers, Cold Regions Research and Engineering Laboratory (CRREL), in Hanover, NH, with design assistance from the U.S. Army Corps of Engineers, Buffalo District, Buffalo, NY. The model, which simulated about 4,400 ft (1,341 m) of the creek at the proposed res location upstream of West Seneca development, was constructed using the conventional method of sandfill, templates, and mortar skim, and sealed with epoxy paint (Figure 2) (Deck, 1985). Constructed in CRREL 's refrigerated laboratory, the model was designed using Froude criterion scaling, and used real ice doped with urea to properly scale the flexural strength of the ice. The pump used to supply water to the model provi1ed a model discharge of up to 6, 000 ft /s (170 m3/s) which was fully adequate for designing the structure. A variety of ICS designs were tested, with the final ICS design shown in Figure 3. The ICS is a three stage weir. The lowest weir (dam) is 6 ft (1.8 m) high, 250 ft (76 m) long and has a crest eleva- tion of 641 ft (195.4 m) National Geodetic Vertical Datum (NGVD). Associated with this low weir are nine concrete piers, spaced 25 ft (7.6 m) apart, which are 12 ft (3.7 m) high, 3 ft (0.9 m) thick, and extend 6 ft (1. 8 m) above the top of the low weir. The piers will restrict ice movement downstream. Together, the low weir and piers comprise the principal spillway of the structure. The second stage is a 225 ft (69 m) eo' 5 (24) ~~-----------------------------------160' (49)--------------------------------------~·~1 .---Water Supply way c::C> Ice Supply Flume Excavated NOTE: NOT TO SCALE NUMBERS WITHIN () ARE METERS FIGURE 2. PLAN VIEW OF CAZENOVIA CREEK PHYSICAL MODEL (DECK, 1985). ~ ~ " !?; f1} ~ ~ ~ en ~ a) CD t,., \;'] 655 ii45 635 LOW FLOW GATE NOTES: NUMBERS WITHIN () ARE METERS NOT TO SCALE FIGIJRE 3. CAZENOVIA CREEK ICE CONTROL STRIJCTORe (DECK, 1985 MODIFIED). wide side-channel floodway at elevation 643 f t (196. 0 m) NGVD, and will be uti- lized when the streamflow exceeds 3,300 ft 3 fs (93 m3 js). The third and highest stage of the res is a 225 ft (69 m) long weir which bridges the principle spillway (low weir and piers) and the side-channel floodway. The elevation of the third stage weir is 649 ft (196.6 m) NGVD. Additional featur~s include (USAGE, 1985): (1) A pool area, just upstream of the ICS, excavated to elevation 635 ft (193.5 m) NGVD. It is 400 ft (122 m) wide at the dam and tapers linearly to the natural stream width at a point 600 ft (183 m) upstream of the dam. (2) A gated low-flow opening, located at one end of the low weir. The gate will be closed early in the winter to create the 6 ft (1.8 m) deep pool behind the dam. The gate will be opened to drain the pool each Spring when the threat of ice jamming has passed. The sill elevation of the opening will be 1 ft (0. 3 m) lower than the pool bottom elevation. A low-flow channel will be excavated to direct sum- mertime flows through the excavated pool to the opening. (3) Training levees which confine excess discharges to the floodway channel. (4) A 20 ft ( 6 m) long concrete splash apron immediately downstream of the low weir. The apron inhibits erosion caused by water falling from the top of the weir. (5) An after bay excavated to 635 ft (193. 5 m) NGVD extending 600 f t (183 m) downstream of the dam. ( 6) An access road used for construction and subsequent maintenance of res. Operation The ICS will (1) reduce downstream ice formation, (2) initiate ice jamming in the undeveloped reach upstream of the structure, and (3) restrict ice runs from progressing to downstream developed areas. 570 During the winter, the 6 ft (1.8 m) deep, 30 ac (12 ha) pool will trap frazil ice which is generated in upstream reaches. By preventing the frazil ice from moving downstream, the res will effectively reduce the ice cover in the lower reaches of Cazenovia Creek. In addition, the stable ice cover on the pool will restrict the downstream movement of the ice run and cause the jam to occur upstream of the structure and deposit ice in the undeve- loped flood plain. Ice movement past the ICS will be restricted by the piers on top of the low weir. Further, when inflow to the struc- ture exceeds 3,300 ft 3 /s (93 m3/5), and water and ice begin flowing to the flood- way adjacent to the main spillway, natural vegetation (trees and brush), located just upstream of the floodway, will restrict ice movement through the floodway. Buffalo District hydraulic studies show that the res piers will be overtopped when discharge reaches 10,500 ft3/s (298 m3 ;s). However, no ice jam flooding downstream from the res is expected to occur at discharges greater than 10,500 ft3/s (298 m3js) for two reasons: model- ling of the structure revealed that most of the retained ice will melt before pier overtopping occurs, and any ice passing downstream of the structure will exper- ience relatively unrestricted flow to the mouth of the creek due to the reduced ice formation downstream of the structure and the evacuation of the ice cover from the channel prior to 10,500 ft3/s (298 m3/s). Cost The cost of constructing the ICS is $1.8 million. The benefit-to-cost ratio for the project is 1.6 to 1.0. Construction may begin in 1988 and be completed in 1989. CONCLUSION Based on physical model tests, the proposed ICS for Cazenovia Creek will be effective in eliminating ice jam flooding in the urban developed areas of West Seneca and Buffalo, NY. The ICS design may alSO prOVe USeful for ICS IS in other similar cold region streams. REFERENCES Deck, D.S., 1985. Cazenovia Creek Physical Ice Model Study, U.S. Army Corps of Engineers , Cold Regions Research and Engineering Laboratory, Hanover, NH. U.S. Army Corps of Engineers (USACE), Buffalo District. 1985. Draft Detailed Project Report and Environmental Impact Statement for Cazenovia Creek, West Seneca, NY, Section 205. 571 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 FREEZEUP PROCESSES ALONG THE SUSIT:NA RIVER, ALASKA Stephen R. Bredthauer and G. Carl Schoch * ABSTRACT: Operation of the proposed Susitna Hydroelectric Project in south- central Alaska would significantly alter the flow, thermal, and ice regimes of the river dow"'llstream of the project!':, poten- tially causing significant environmental impacts. Consequently, the ice regime of the Susitna River has been monitored since 1980 to document natural ice processes and their environmental effects, and to obtain calibration data for ice modelling of certain segments of the river. This paper describes the different freeze-up charac- teristics along the river's length which result from the significant variations in climate, morphology, and gradient along the river. (KEY TERMS: river ice; Alaska; Susitna River.) INTRODUCTION The Susitna River basin is located in southcentral Alaska, originating from glaciers on the southern flank of the Alaska Range (Figure 1) • The drainage basin covers 19,600 sq. mi. (50, 7 60 sq. km.), and is the sixth largest in Alaska. The upper basin upstrea~ of the damsites is in the Continental climate zone, with climate characteristics of cold, dry winters and warm, moderately wet summers. The lower basin is in the Transition climate zone (between the Continental and Maritime zones), where temperature is less variable and pre- dpitation is greater than in the Conti- nental zone. The Susitna River travels a distance of about 318 mi (512 km) from its glacial headwaters to Cook Inlet. Just downstream from the headwater glaciers, the river is highly braided. About 18 mi (29 km) downstream from the glaciers, the river develops a split channel configuration which continues for 53 mi (85 km). This initial reach, known as the upper Susitna River, has colder air temperatures than the downstream reaches due to its higher elevation and latitude. However, it also receives a substantial amount of solar radiation during freezeup because of its north-south orientation. The river then flows through a series of steep-walled canyons for about 96 ~i (154 km) to the mouth of Devil Canyon. This reach, known as the impoundment zone, contains the Watana and Devil Canyon damsites at river mile (RH) 184.4 and RM 151.6, respective- ly. The river then emerges from the canyon into the middle Susitna River, which flol>TS through a broad glacial U-shaped valley to the confluence with the Chulitna River (RM 98), about 50 mi (80 km) downstream. The Talket:~tna River enters about one mile downstream (RM 97). Steep canyon walls along the impoundment zone and the middle Susitna River tend to shade the turbulent water surface for much of the winter. Average winter flow (November through April) in this reach is 1,600 cfs (45 ems). The river again becomes highly braided at the confluence uith the Chulitna River. Average winter flow downstream of the Chulitna and Talkeetna rivers is 4,500 cfs (127 ems). The braided pattern continues for the 98 mi (157 km) downstream to the mouth of the river, with a few intermittent reaches of well-defined single or multiple channels. The Yentna River, the largest tributary to *Respectively, Senior Civil Engineer and Hydrologist, R&M Consultants, 5024 Cordova Street, Anchorage, Alaska 99503. 573 0-t. co EXPLANATION 15292000 .. STREAM GAGING STATION (FLOW AND SEDIMENT DATA) 15292100 V SEDIMENT SAMPLING STATION DRAINAGE BASIN BOUNDARY 10 20 30 40 50 MILES 10 20 30 40 50 Kl LOMETERS FIGURE 1. SUSUTNA RIVER BASIN the Susitna River, enters at RJ1 28. ICE COVER FOID·1ATION PROCESSES Progression of an ice cover on the Susitna River begins in late October near the mouth at Cook Inlet. Frazil ice pans from the Yentna River, the middle and lower Susitna Rivers, and the Talkeetna and Chulitna Rivers jam to form a bridge near the mouth. This occurs during a high tide period when the air temperatures are significantly below freezing throughout the basin and frazil ice discharge is high. After the ice bridge forms, the incoming frazil ice accumulates at the upstream or leading edge of the ice cover, or at natural lodgement points such as shallows or islands, and causes the ice cover to progress upstream. The ice cover advances upstream by different mechanisms, depending on the air temperature, volume of incoming ice, and river hydrodynamics. The mechanisms of upstream progression have been described by Calkins (1983). Additional descrip- tions of these mechanisms may be found in Pariset et al (1966) and Ashton (1978). The mechanisms are described below, along with observations of processes on the Susitna River. (1) Progression by juxtaposition of arriving floes with no subsequent thicken- ing, leading to a rapid ice cover develop- ment. This occurs at water velocities less than some critical value required to submerge incoming ice floes below the ice cover. On the Susitna River, this has been observed to be approximately 2 ft/sec (0. 6 m/ sec) . Ice cover thickness equals slush floe thickness. On the Susitna River, this is the predominant process of progression upstream to about RM 25. Slush ice floes drifting through this reach have been on the water surface and exposed to cold air temperature long en0ugh to form a solid surface layer, significantly strengthening the floes so that they resist crushing or breaking apart. (2) Hydraulic thickening, in which slush floes arriving at the leading edge thicken to a greater value than the original ice floe thickness. The ice thickness is sufficient to transmit hydraulic forces to the banks. The ratio 575 of ice cover thickness to flow depth is usually less than 0.33. A related process also described by Calkins (1983) and often observed on the Susitna River is mechan- ical thickening or shoving of an ice cover already in place. This apparently occurs due to an instability within the ice cover relative to water velocity and increased upstream water levels which increase the pressure on the ice cover. A portion of the ice cover that has progressed by juxtaposition or hydraulic thickening may suddenly fail and move downstream, thickening as the surface area decreases. The ice cover thickness created by hydrau- lic processes is not sufficient to with- stand the forces acting on it during its progression. The ice cover thus breaks and moves downstream, being mechanically thickened until it can withstand the forces imposed on it. The momentum of the moving ice mass may cause the ice thicknesses to be greater than the hydrau- lic stability requirement. The ratio of ice cover thickness to flow depth usually exceeds 0. 33. Shoving usually causes a downstream progression, sometimes moving the leading edge downstream as far as 1 mile (1.61 km). This process usually occurs where water velod.ties exceed 4 ft/sec. (1.2 m/sec.). Hydraulic thicken- ing and shoving are the primary processes of ice cover advance from RM 25 upstream to near RM 130. Compressions may occur repeatedly, creating higher upstream water levels and lower velocities, until pro- gression can resume. (3) Arriving ~lush floes are com- pressed and added to the cover, but some also submerge and break apart, eventually being deposited underneath the ice cover further downstream if lower velocities occur. {4) Arriving slush floes do not accumulate at the ice front, but are subducted beneath the cover. They may be deposited some distance do"t-mstream. The process of undercover deposition is difficult to document, but most likely occurs on the Susitna River. Juxta- position and hydraulic thickening seem to be the dominant progression processes on the Susitna River, with undercover deposi- tion and shoving the primary thickening processes. Two other processes are also common, but do not significantly affect ice cover progression in the reach between the river mouth and Gold Creek (RM 137). These are anchor ice and border ice formation. Anchor ice formation is common in shallows throughout and downstream of a turbulent reach. Anchor ice is particu- larly prevalent upstream from RM 120, where the river may not develop an ice cover until late December. Anchor ice dams up to 2 feet thick have been docu- mented between RM 130 and RM 149. Border ice forms along the banks of the river as a result of (a) freezing of water in shallow areas, (b) accumulation of frazil pans in eddies and on obstruc- tions such as bars or tree limbs, or (c) shearing of moving frazil pans on the river banks or on the border ice shelves. Border ice does not generally close the river downstream of m1 137, but may result in raising of water levels and obstruc- tions to the downstream passage of frazil ice pans. This may lead to intermittent bridging of the river, resulting in the ice cover progressing upstream of the bridge prior to the downstream :f.ce cover completely forming. Border and anchor more dominant in the Gold Creek (RM 137), velocities and to the ice processes are reach upstream of due to the high fact that the ice cover normally does not progress upstream to this reach. SEQUENCE OF ICE COVER PROGRESSION LOWER SUSITNA RIVER Frazil ice usually first appears by October in the upper Susitna River. This ice drifts downriver, often accumulating into loosely bonded slush ice floes, until it either melts or exits from the lower Susitna River into Cook Inlet. The initiation of ice cover progression usually occurs in late October. An ice bridge forms near the mouth of the Susitna River during a period of high tide and high slush ice discharge. Initial ice bridges have been observed at RM 1. 9, RM S, and RM 9. During the freeze-up period, the Yentna River (RM 26) often contributes from SO to 60 percent of the total es- timated ice volume below the Yentna-Susit- na confluence (R&M Consultants, 198Sa,b). Upstream of the Yentna River, about 80 576 percent of the ice is contributed by t~ middle Susitna River, with the Chulitna and Talkeetna Rivers contributing only about 20 percent (R&M Consultants, 198Sa,b). The ice cover progression the lower Susitna River occurs primarily by juxta- position to about RM 25 and by hydraulic thickening upstream to about RM 130. Intermittent bridging may occur at natural lodgements points, such as shallows and islands. When this happens, the ice cover may progress upstream before the river downstream is fully ice covered. Depend- ing on weather conditions, the ice cover will reach Talkeetna between early Novem- ber and early December. As the ice cover progresses upstream, the water level increases (stages) due to the increased resistance of flow and the displacement of the ice. Water levels generally increase about 2 to 4 feet (0.6 to 1.2 m) in the lower Susitna River due to ice, although increases of up to 8 feet (2.4 m) have been observed at the mouth of Montana Creek (R&M Consultants, 1985ab). The increased water levels due to ice are illustrated for a number of sites on the lower Susitna River in Figure 2. The increased water levels often result in the overtopping of previously dewatered or isolated side channel en- trances. The increased water flow caused by overtopping may wash out the snow cover and fracture existing ice. Slush ice from the mainstem will generally not flow into the side channel unless the overtopping depth at the overtopped upstream berm exceeds about one foot (0.3 m). If slush ice flows into the side channel, an ice cover forms rapidly in a manner similar to that described for the mainstem. Other- wise, the ice cover forms by border ice growth, which may take several weeks. Many of the side channels dewater prior to freezeup. Others have separate water sources from tributaries or ground- water seeps. However, the groundwater seepage is greatly reduced from summer levels due to the lower flow and water level in the mainstem. During ice cover progression, the increase in main stem water levels raises the groundwater levels in the river alluvium. Consequently, even if the entrance to a side channel is not overtopped, the increased groundwater levels may result in seepage flow in the RELATIVE STAGE LEVELS AT SELECTED SITES DURING 1983 5 SUSITNA RIVER FREEZEUP .... 4 w CHULITNA ~ 3 CO NFL (RM9B.5) 2 10 -----------------~ 9 B 7 MONTANA .... 6 CREEK ::: 5 (RM76.9) II. 4 3 OPEN LEAD 2 5 8 ------________ ..,__ ~ 4 ~ .... Iii KASHWITNA w3 RIVER w z II. 2 H (RM 60) p;:j 5 {.9 ____ _} OPEN LEAD ES DELTA ISLANDS 4 Ul SIDE CHANNEL 3 ~ ENTRANCE :;;2 ( RM 4B) WI H II. E,. 5 :s DESHKA RIVER/ 4 § KROTO SLOUGH 3 (RM 40.1) .... w 2 w II. I 5 -+-----} ---------J ALEXANDER 4 SLOUGH :;; 3 (RM 14) w 2 II. OCTOBER NOVEMBER D~CEMBER MONTH LEGEND ----INTERPRETATION BASED ON OBSERVATION --INTERPRETED STAGE BETWEEN DATA POINTS SURVEYED DATA POINTS FIGURE 2. RELATIVE STAGE LEVELS AT SELECTED SITES DURING 1983 SUSITNA RIVER FREEZEUP 577 LOWER RIVER TYPICAL CROSS SECTION JJA % _..--FROZEN SNOW-- ~~''•· . /SLUSH ICE COVER ICE COVER FROZEN SLUSH BORDER ICE FIGURE 3. TYPICAL ICE COVER DEVELOPHENT, LOWER SUSITNA RIVER. channel. Major tributaries of the Susitna River (such as the Yentna, Deshka, Tal- keetna, and Chulitna rivers) form an ice cover by surface accumulations of frazil slush ice after their mouths are blocked by the ice cover on the Susi tna River. Smaller tributaries generally develop an ice cover by border ice and anchor tee accumulations. These minor tributaries are generally too shallow and turbulent to form a stable ice cover. Following freezeup on the mainstem, the ice cover sags due to a gradual decrease in discharge, ice cover erosion, and bank storage. Open leads may persist in the mainstem and side channels due either to high velocity or to the thermal effects of warm groundwater. Typical ice cover development on the lower Susitna River is illustrated in Figure 3. The days numbered on the left indicate the approximate passage of time since the leading edge of the progressing ice cover advanced upstream past each cross-section. These cross-sections are only schematics and do not represent the actual river. The lower river is much 578 wider than shown here, with widths exceed- ing 6,000 feet (1 ,829 m), so that the depth-to-width ratio is not representa- tive. The schematic illustrates many of the processes of ice cover development documented on the Susitna River. Day 1 shows slush ice rafts drifting downstream in the mainstem. The discharge has dropped low enough to dewater secon- dary channels and side sloughs. The drifting slush ice rafts have accumulated in low-velocity flow margins or eddies and subsequently frozen to form border ice. Little additional border ice growth occurs until water velocities decrease further. Open water exists in side sloughs, since this water is generally warm, flowing from seeps or springs. The ice front progress- es to the area on Day 2, resulting in a rapid increase in water level, flooding of the surrounding gravel bars, and overtop- ping of the side slough. The secondary channel is inundated and now conveys water that bypasses the ice-choked mainstem. Snow on the floodplain is saturated, eventually freezing into snow ice. The ice accumulation and compressions in the mainstem have fractured the existing border ice, which was either shoved laterally or incorporated into the cover. By Day 10, the slush ice cover has probably frozen solid, black ice has grown under the new ice, and the side channel is beginning to freeze over by border ice growth. Within about one month after the ice cover has formed over the mainstem, few additional changes will occur for the remainder of the winter. Open channel leads will typically erode through the ice cover. Depressions over the secondary channels are typical. The side channels are essentially ice-covered, but may retain an open lead. SEQUENCE OF ICE COVER PROGRESSION MIDDLE SUSITNA RIVER When an ice bridge forms at the Chulitna confluence (RM 98.6) , ice cover progression continues upstream to the vicinity of RM 137. Depending on flow rate, ice concentrations, climatic con- ditions, and channel morphology, this bridge may form either when ice cover progression on the lower Susitna River reaches the confluence, or else indepen- dently of the lower river progression at a point just upstream of the Susitna-Chu- litna confluence. Flow in the middle Susitna River during this period is typically about 2,000 -3,000 cfs (610 - 914 m). In very cold years, one or more secondary bridges may form upstream of this bridge, resulting in secondary pro- gressions of the ice cover. Ice cover shoving, sagging, open lead development and secondary ice cover progression predominate through the reach from the Chulitna confluence to about RM 137. The ice cover progresses by juxta- position and hydraulic thickening until encountering a critical velocity, which causes leading edge instability and failure of the ice cover. The subsequent consolidation results in ice cover stabi- lization due to a shortening of the ice cover, substantial thickening as the ice is compressed, a stage increase, and lateral expansion (telescoping). As the stage increases, the entire ice cover lifts, and pressures are then relieved by lateral expansion of the ice across the river channel. This process of lateral telescoping can continue until the ice 579 cover has expanded bank to bank or else has encountered some other obstruction (such as gravel islands) on which the ice becomes stranded. Ice cover sag, collapse, and open lead development usually occur within days after a slush ice cover stabilizes. A steady decrease in streamflow gradually lowers the ice surface along the entire river. Prior to breakup, much of the ice rests on the channel bottom. The typical ice cover development on the middle Susitna River is shown on Figure 4. The sequence is essentially the same as on the lower Susitna River, with the primary difference being the higher degree of staging and compression of the ice cover. The slush ice cover is shoved laterally, often to the top of the bank and vegetation line. Some ice may be eroded in high velocity areas, and rede- posited where velod.ties are lower. As the ice is redistributed into a more hydraulically efficient cross-section, the water level recedes, causing the cover to sag, often conforming to the configuration of the channel bottom. Open channel leads are typical through this reach, but often freeze over by early March. The progres- sion rate decreases as the ice front moves upriver, due to the increasing river gradient and the decreasing amounts of ice flowing downstream as the upper river freezes over. The reach from RM 137 to Devil Canyon (RM 150) gradually freezes over, with complete coverage occurring much later than further downstream. The reach has a steep gradient, high velocities, and a single channel in winter. The most significant freezeup characteristics include extensive anchor ice, wide border ice layers, ice dams and snow ice. Anchor ice dams have been observed at several locations which are constricted by border ice. The dams and constrictions create a backwater area by restricting the streamflow, subsequently causing extensive overflow onto border ice. The overflow bypasses the ice dam and re-enters the channel further downstream. Within the backwater area, slush ice accumulates in a thin layer from bank to bank and eventual- ly freezes. The processes of ice cover pro- gression described for the reach down- stream of RM 137 generally do net occur in MIDDLE RIVER TYPICAL CROSS SECTION FLOOD PLAIN I BANK FIGURE 4. TYPICAL ICE COVER DEVELOPMENT MIDDLE SUSITNA RIVER. this reach. There are only minimal water level increases due to anchor ice growth on the channel bottom. Sloughs and side channels are generally not breached. Open leads exist in the main channel, primarily in high-velocity areas between ice bridges. Ice processes in Devil Canyon (RM 150 toRN 151.5) create the thickest ice along the Susitna River, with observed thick- ~esses of up to 23 feet (7 m) (R&M Consul- tants, 1981). Large volumes of slush ice enter the canyon, generated by upstream rapids or by heavy snowfall. Additional frazil ice forms in the extreme turbulence within the canyon. The slush ice repeat- edly jams in a plunge pool near ~~ 150 and an ice cover progresses upstream, even- tually staging nore than 25 feet (7. 6 m) above the open water level. However, slush ice has litt]e strength, and the center of the ice cover rapidly collapses after the downstream jam disappears and the water drains from beneath the ice. Some slush ice freezes to the canyon walls, increasing in thickness with each staging repetition. The ice cover forms and erodes several times during the winter 580 (R&M Consultants, 1984). Upstream of Devil Canyon, the Susitna River generally has a steep single channel with banks rising gradually from the water surface to the vegetation trim line. Low discharges through the winter result in generally shallow water. Numerous boul- ders exist along the channel margin, providing anchors for slush ice that drifts along the banks. Shore ice devel- ops rapidly into the channel unti.l water velocities exceed about 1-2 ft/sec (0.3 - 0.6 m/sec). As streamflow decreases, there is a gradual filling of the narrow open channel into a continuous ice cover. Anchor ice thicknesses exceed 2 feet (0. 6 m) in some areas, raising the water level accordingly. The rising water either fractures the border ice or over- flows on top. When overflow occurs, snow on the shore ice is flooded and eventually freezes, significantly thickening the border ice. CONCLUDING REMARKS The paper discusses the various types of ice formation processes documented along the length of the Susitna River, a major Alaskan river being considered for hydroelectric development. Operation of the proposed Susitna Hydroelectric Project would significantly alter the flow, thermal, ar.d ice regimes of the river downstream of the project. The studies have been conducted to document the natural physical processes on the Susitna River, both to determine their environ- mental effects and to provide calibration for ice modelling of with-project con- ditions. Since ice processes play a major role in the natural regime of northern rivers, knowledge of the effects of water resource development on the ice regime is necessary before any assessment of the environmental impacts of the project can be made. ACKNOWLEDGMENTS Ice studies of the Susitna River were funded by the Alaska Power Authority for licensing and design of the Susitna Hydroelectric Project. Field work was performed by R&M Consultants, Inc. under contract to Acres American, Inc. (1980-82) and to the Harza-Ebasco Susitna Joint Venture (1983-1986). Ashton, G. Review 369-92. REFERENCES 1978. River ice. Annual of Fluid Mechanics. 10: Calkins, D.J. 1983. Hydraulics, mechanics and heat transfer for winter freezeup river conditions. Class notes for: Ice Engineering for Rivers and Lakes, University of Wisconsin, Madison, Wisconsin. Pariset, E., Hauser, R. and A. Gagnon. 1966. Formation of ice covers and ice jams in rivers. Journal of the Hydraulics Division ASCE. 92: 1-24. R&M Consultants, Inc. 1981. Ice obser- vations 1980-1981. Alaska Power Authority. Susitna Hydroelectric Project. Anchorage, Alaska. R&M Consultants, Inc. 1982. Ice obser- vations 1981-1982. Alaska Power Authority. Susitna Hydroelectric Project. Anchorage, Alaska. R&M Consultants, Inc. 1984. Susitna River ice study, 1982-1983. Alaska Power Authority. Susitna Hydroelec- tric Project. Anchorage, Alaska. R&M Consultants, Inc. 1985a. Susitna River ice study 1983-1984. Alaska Power Authority. Susitna Hydroelec- tric Project. Anchorage, Alaska. R&M Consultants, Inc. 198Sb. Susitna River ice study, 1984-1985. Alaska Power Authority, Susitna Hydroelec- tric Project. Anchorage, Alaska. 581 JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 GROWTH AND DECAY OF RIVER ICE COVERS Hung Tao Shen and A.M. Wasantha Lal 1 ABSTRACT: In this paper a mathe- matical model for calculating the thermal growth and decay of river ice covers is developed. The model takes into consideration the insulation effect of snow and frazil ice, the formation of snow ice, and is able to provide a continuous description of the ice cover thick- ness throughout the winter. Applications of the simulation model to ice covers at two different locations are also presented. (KEY TEPJiS: ice cover thickness; snow ice, capillary rise; black ice; heat exchange.) INTRODUCTION Since the classical work of Stefan (1889), studies of the thermal effect on river ice cover have generally been limited to approximate analytical solutions and empirical degree-day methods (Hichel, 1971; Pivovarov, 1973). In recent years, simple finite- difference models have been developed for calculating river ice cover thickness (Ashton, 1982). These models, which correctly treated the boundary condition at the air-snow or air-ice interface and the insulation effects of snow and frazil ice, are relatively easy to use. However, these models are still restricted in that~ a) all of these models consider only the growth of the ice cover, and are not able to describe the variation of the cover thickness during the melting period; b) the formation of snow ice due to the freezing of snow slush are not considered. Studies by Lepparanta (1983) and Bengtsson (1984) considered the effect of snow ice formation. However, further refinements to these studies are needed. The existence of the capillary fringe in the snow slush, which can affect the snow ice thickness, were neglected in both of these studies. In Lepparanta's model snow slush was directly transformed into snow ice whether frozen or not. The snow surface temperature was ass~ to be the same as the air tempera- ture. Bengtsson correctly included the effect of the surface thermal resistance in calculating the black ice growth. The growth of single layer snow ice was calculated by considering the heat transfer process in the cover neglecting the surface thermal resistance. Snow slush between layers of white ice was directly transformed to snow ice. In this paper, a generalized model for the thermal growth and decay of a river ice cover is developed and compared with field observations. The model takes into consideration the insulation effects due to the snow cover and the frazil ice as well as the possible fonration of snow ice. The model also provides a continuous description 1Respectively, Professor and Research Assistant, Department of Civil and Environmental Engineering, Clarkson University, Potsdam, N.Y. 13676. 583 of the ice cover thickness variation and its composition throughout the entire winter. ANALYTICAL TREATMENT The thermal growth and decay of an ice cover is governed by the heat exchange at the air-snow or the ice-air interface, the heat transfer through the snow and ice layers and the composition of the cover (Shen, 1985). For the ice cover shown in Fig. la the sub- mergence depth hw can be calculated by: hw = [(1-e )p.h + p.h. + s l. s l. l. (l-ef)pihf + efpwhf]/pw (l) in which, hw = submergence depth defined as the difference between the water level and the elevation of the bottom surface of the frazil accumulation; hs, hi, hf = thick- ness of snow, ice, and frazil layers, respectively; Pi' Pw = density of ice and water, respectively; es, ef = porosity of snow and frazil respectively. In this equation, the density of the solid fraction of snow, and frazil are assumed to be the same as the density of ice. If the subrrergence depth is greater than the combined thickness of the ice cover and the frazil ice, then part of the snow will be submerged to become snow slush. This critical condition, hw > (hi+hf)' is equivalent to > ~p(hi+(l-ef)hf) hs ( 1-e ) p. s l. (2) in which ~P = Pw-Pi· For es = 0.67, and hf = 0, Eq. 2 gives hs > 0.27 hi, which indicates that a relatively small snow cover thick- ness is required in order for an ice cover to be submerged. Once submerged, a snow slush layer will form. The thickness of the slush layer can be calculated 584 from the hydrostatic balance by including the effect of the capillary fringe above the phreatic surface. The slush layer will freeze from the upper surface of the capillary fringe downward to form white ice. The thickness of the slush layer is hsw = {CRpw + (l-es)pihsd -~p[hib-(1-ef)hf]}/ (pw-psw) (3) in which Psw = density of slush = (1-es)Pi+s esPwi CR =capillary rise; s = water saturation of the slush; hsd = thickness of the dry snow. Due to the small difference between Pw and Psw 1 the slush thickness hsw will be much larger than CR. With additional snow fall, the white ice may be flooded with another slush layer forming on top of it. If the submergence of the white ice occurs before the slush layer underneath it completely turns into ice, a new layer of snow ice will be formed from the top of the new slush layer above the flooded white ice layer. This process may continue during the growth period leading to the formation of multiple layers of white ice sandwiched between slush layers. If the submergence of the white ice occurs after the slush underneath the white ice completely turned into ice, this layered white ice-slush structure will not form. Fig. lb shows the structure of a river ice cover consists of layers of dry snow, snow slush, white ice, black ice, and frazil ice. Surface Heat Flux To calculate the rate of ~ge of the cover thickness, the heat exchange with the atmosphere at the top surface of the cover must first be determined. Using the linearizro form of Dingman and Assur (1967), the surface heat exchange rate can Air Dry Snow Black ice Flow Bed (a) Dry Snow Slush White ice Black ice Frazil qa t ·.·· 'fq · .... ·TS .... s ... ·T ·::::_ .. m.._u·R. 0./1 7/Ut.OIWIN II . T . . . . . . . h Wl/li/tlil/711!/i.t wl h :ff~=: j t qwi Flow Bed (b) - Figure 1. Thermal growth and decay of river ice cover: a) black ice growth without snow slush; b) white ice growth with snow slush and layered white ice-slush structure. be expressed as: (4) in which qa = heat flux to the atmosphere from the cover surface; ¢R = solar radiation; Ts = surface temperature; Ta = air temperature; and a,S =coefficients which can be derived from the complete surface heat exchange processes including radiation, evaporation and sensible heat transfer. To further simplify the analysis, the solar radiation effect is sometimes lumped into the coefficients a' and S', thus (5) Since ¢R varies with latitude, Eq. 5 can only apply to sites located in the same latitude for a given set of a' and S' values. Ice Thickness Variations The thickening and melting of the cover can be analyzed by assuming a quasi-steady linear temperature distributions in each layer of the cover. The quasi- steady assumption has shown to be acceptable for river ice covers (Greene, 1979; Ashton, 1982). In the following discussion, the formulation for the ice cover thickness variation will be presented separately for cases with or with- out snow slush. The surface heat flux qa is expressed in the form of Eq. 4. The resulting formulas can be applied when qa is approximated by Eq. 5, by setting ¢R to be zero and a and S to be a' and S'. Ice Thickness Variation without Slush Layer -In this case, the surface temperature, T 5 , can be determined by the conservation of heat flux across interfaces between various layers in the cover. 586 h. h ¢ -a T [ (__!. + ~) (T + R + __!!!J I h k a -s-s s s h. h l) (__!. + ~ + ( 6) k. k s l. s in which, Tm = the melting point, 0°C; and ki, kg = thermal conductivities of 1 snow and black ice 0.3 wm-1 oc-and 2.24 wm-1 oc-i, respectively. When the ice layer consists of both the white ice and the black ice, then hi and ki becomes hi = hwi + hbi and ki = hi/(hwilkwi + hbi/kbi), in which subscripts wi and bi represent white ice and black ice, respectively. In the absence of frazil accumulation on the underside of the ice cover, the rate of growth of black ice, when T 5 <Tm is given by dh. l. piLi dt = [(Tm-Ts)/ h h. s l. (k + k.)] -qwi (7 ) s l. in which Li = the latent heat of fusion of water, 3.4 x 10S J kg-1, and, qwi = heat flux from the water to the 1.ce. For the present analysis, this heat flux is taken as (Ashton, 1983) q . = h . (T -T ) w1. -wl. w m ( 8) in which, Tw = the water tempera- ture; and hwi = a heat transfer coefficient which can be evaluated by the formula (Ashton, 1982) h . = c . uo.8 o-0.2 Wl. Wl. W W (9) in which Cwi = 1662 ws 0 •8 m-2 •6 oc-1; Dw =flow dept~m; and Uw = the average flow velocity, m/s. The coefficient Cwi may increase by up to 50% when relief features form on the underside of the cover. Eqs. 6 and 7 are also valid when there is no snow cover by setting hs=O. When the surface heat flux qa becomes negative, the surface temperature, Ts, calculated from Eq. 2 will be greater than T • Under this condition, Ts sho~ld be set equal to Tm, and melting of the snow cover and the ice cover can be calculated by Eqs. 10 and 11. dh (1-e )p.L. dts = s 1 1 -<P + a R + S (T -T ) s a - h . (T -T ) w1 w m (10) ( 11) When the snow cover is melted completely, melting will occur on the top and the bottom surfaces of the ice cover, the melting rate when there is no frazil accumulation becomes -<PR + a + S(T -T ) s a -hwi<Tw-Tm) (12) In the above analysis the water resulting from the melting of the snow or ice cover is assumed to be drained through the cover. Wake and Rumer (1979) have examined the magnitude of the error introduced by the assumption of well-drained ice surface compated to an un- drained ice surface. Their analysis indicates that the effect of accumulated water over melting ice on the melting rate may be insignificant. In the presence of a frazilice layer on the underside of the ice cover, the growth of ice cover thickness will be accelerated (Calkins, 1979). This is because the frazil ice layer insulates the ice cover from the warm water below, as well as the fact that only the porewater in the frazil layer needs to be solidified for the downward growth of the ice cover. During the growth period, when Ts<Tm, the rate of growth of 587 the ice cover is dh. 1 dt = h h. (Tm-Ts)/[efpiLi(ks+k~)] s 1 (13) The rate of reduction of the frazil layer at its bottom is dh:f dt = - h . (T -T ) I w1 w m [PiLi (1-ef)] ( 14) Hence, the total rate of reduction of the frazil layer is dh ' dh f i dt -dt (15) During the melting period, when Ts > Tm, the rate of reduction of the snow cover and the frazil layers can be calculated using Eqs. 10, 11 and 14, respectively. The melting at the bottom of the ice cover can occur only in the absence of the frazil layer and can be determined by Eq. 11. The rate of melting at the upper surface of the ice cover in the absence of a snow cover is dh. 1 dt = Ice Thickness Variation with Slush Layer -The major consequence of the presence of the slush layer formed by the submergence of the snow cover is the formation of white ice. In this case the • I port1on of the cover below the top ~urface of the slush layer is 1sothermal, and maintains a temperature at the freezing point. For the case shown in Fig. lb white ice will grow downward from the top surface of the slush layer. The cover surface tempera- ture, Ts, is calculated with Eq. 17 by considering the balance of the upward heat flow in the dry, snow cover and at the snow-air interface h <I> -a T = [ sd (T + _R_) s k a s s T h .!) + --.!!!] I ( sd + (17) S ks s If Ts < Tm, then the growth of white ice can be calculated by the equation In addition to the case shown in Fig. lb, other cases with slush layer are possible. When the phreatic surface is close to the snow surface all of the snow will turn into slush due to the capillary rise. In this case white ice will grow downward from the top surface of the snow cover, at a rate dh . W1 crt= ( a-<1> T -T + __ R)I m a S 1 h . [ P L e S ( + W1)] . . -s 1 1 s ~i (19) When the water level is located below the top surface of the upper- most white ice layer, the growth rate of the white ice from the bottom of this layer is a-<I> (Tm-Ta + __ R) I s 1 hs h 1 [p.L.e s(-8 +-+ ~)l 1 1 s ks k . W1 (2 0) in which, hwl = the thickness of the uppermost white ice layer. It should be noted that the black ice growth can occur only after the entire snow slush layer turned into white ice. The growth rate of black ice can then be calculated by Eq. 13. The melting of snow cover, 588 frazil layer, and the ice cover at both the upper and the lower surfaces can be calculated as discussed in the case with no slush layer. MODEL APPLICATIONS The thermal growth and decay of river ice covers can be computed through stepwise numerical inter- grations. The model is applied to the ice covers in the upper St. Lawrence River near Massena, N.Y. (Shen and Chiang, 1984, Shen and Yapa, 1985), and the ice cover on the Finnish coast of Bothnian Bay at Virpinieme (Lepparanta, 1983). Fig. 2 shows the comparison between the observed and simulated ice thickness variations for four stations along the St. Lawrence River for the winter of 1977-78. In the simulation the surface heat exchange at the snow-air interface is calculated from Eq. 5 with a' = !4:12 wm~2 and S' = 12.62 wm-2 C • F1g. 3 shows the comparison between the observed and simulated cover thickness at Virpinieme for the winter of 1976-66. The surface heat exchange is calculated based on Eq. 4. The a and B values on the snow-air interface can be calculated from a= 196.77 + 6.62 v a -1.12 RH-0.28 cc 2 8 = 0.185 T + 4.61 V a a + 0.11 RH On the ice-air interface a = 154.90 + 2.88 V a -1.11 RH-0.32 cc 2 B = 0.098 T + 2.47 V a a + 0.08 RH in which Ta = air temperature, (21) (22) (23) (24) e: en en Q) J:: ~ 0 ·r-1 ..c: E-1 !-! Q) :> 0 C) Q) 0 H 1.5~------------------------~~------------·-------------, Station E1 Station E2 1.0 0.5 0 Station F1 Station F2 1.0 0.5 0 0 50 75 100 125 0 25 50 75 100 125 150 t, No. of days from Dec. 1 Figure 2. Comparison between observed (e) and simulated (--) ice cover thickness, St. Lawrence River, Winter of 197~-78. 589 1.00 ~------------------------------------------------------~ 0.75 0.50 0.25 0 0 Observed · • Tota 1 ice thickness 0 White ice thickness • Snow thickness 25 Simulated Total ice thickness Black ice thickness Snow thickness White ice thickness 50 • / .,, ,- 1 , ...... -- 75 • • • • -· ____ _,.. ,/ .-/ -·-·...J ,... • • ·*'~ ... ~ •• c;> ... $. .. •·•···•••• . ..·· • .. • • ® •• e .. e c:> s • 100 125 150 175 No. of days from 1 Oct. 1976 Figure 3. Comparison between observed and simulated cover thickness, Virpinieme, Finland, Winter of 1976-77. 590 200 v = wind velocity, m/sec; RH = rglative humidity, %; and CC = cloud cover in tenths. Other constants used in the c~mputations are: kwi = 2.1 wm-1 oc-I Pw = 1.0 gcm-3, Pi= 0.916 gcm-3, es = 0.8, s = 0.94, CR = 1.75 em, and_~wi = Pi[esS + (1-es)l = 0.872gcm • These results show that the model can provide good simulations for the growth and decay of an ice cover. Refinements by including variable snow properties due to the packing effect, (Lepparanta 1983) and the effect of the refreezing of undrained meltwater may be considered for further studies. ACKNOWLEDGEMENT This study is partially supported by the U.S. Army Cold Regions Research and Engineering Laboratory under Contract No. DACA 89-84-K-0008. The writers would like to thank S.C. Colbeck, C.A. Knight, and M. Lepparanta for valuable discussions. M. Lepparanta provided data at the Virpinieme site. REFERENCES Ashton, G.D., 1982. Theory of Thermal Control and Prevention of Ice in Rivers and Lakes. Advances in Hydroscience. Vol. 13:131-185. Bengtsson, L., 1984. Forecasting Snow and Black Ice Grwoth from Temperature and Precipitation. IAHR Ice Symposium, Hamburg. 173-185. Calkins, D.J., 1979. Accelerated Ice Growth in Rivers. CRREL Report 79-14, Cold Regions Research and Engineering Laboratory, Hanover, NH, 4 pp. 591 Greene, G.M., 1981. Simulation of Ice Cover Growth and Decay in One-Dimensional on the Upper St. Lawrence River. NOAA TM ERL GLERL:36, Great Lakes Environmental Research Lab., Ann Arbor, MI, 82 pp. Lepparanta, M., 1983. A Growth Model for Black Ice, Snow Ice and Snow Thickness in Subarctic Basins. Nordic Hydrology, 14: 5 9-70. Michel, B., 1971. Winter Regime of Rivers and Lakes. Cold Regions Science and Engineering Monograph, III-Bla, U.S. Army Cold Regions Laboratory, Hanover, NH, 131 pp. Pivovarov, A.A., 1973. Thermal Conditions in Freezing Lakes and Rivers. John Wiley and Sons, New York, NY, 136 pp. Shen, H.T. and Chiang, L.A., 1984. Simulation of Growth and Decay of River Ice Cover. Journal of Hydraulic Engineering, ASCE, 110: 958-971. Shen, H.T., and Yapa, P.D., 1985. A Unified Degree Day Method for River Ice Cover Thickness Simulation. Canadian Journal of Civil Engineering, Vol. 12:54-62. Shen, H.T., 1985. Hydraulics of River Ice. Report No. 85-1, Department of Civil and Environ. Engineering, Clarkson University, Potsdam, NY, 80 pp. Stefan, J., 1889. Uber die Theorien des Eisbildung insbesondere uber die Eisbilding in Polarmure. Wien Sitzunber. Akad. Wiss., Ser. A, Pt. 2: 965-983. Wake, A., and Rumer, R.R., Jr., 1979. Effect of Surface Melt Water Accumulation on the Dissipation of Lake Ice. Water Resources Research, 15(2): 430-434. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 ICE JAMS IN REGULATED RIVERS IN NORWAY, EXPERIENCES AND PREDICTIONS Randi Pytte Asvalll ABSTRACT: Developing water-courses for power production may cause severe ice problems in the areas involved. The general theory of ice formation in rivers is described. Possible influences of power develop- ment on ice formation are discussed. Of special importance is the amount and temperature of the discharged water flow, and sudden increases in discharge flows during the ice period. Experiences from selected regulated rivers in Norway are then described. In the upper Glomma river and in Otta river there has been an increase of winter discharge in order to transport reservoir water from the upper reservoir to intake dams further down the river. This caused initially various problems with ice runs, ice jamming and flooding. By modifications of the scheme of water discharge the main problems have been overcome in most cases. In the river Nea there has been an increase in discharge downstream from a power station in the upper part of the water-course. The problems here, which to a certain degree were expected, could only be dealt with by further developing the river downstream. In the river Alta, which is presently being developed, and Stj !llrdal, which is being planned, recommended schemes of water discharge to avoid ice problems downstream from the power stations are described. These schemes will be re-- viewed following a 5 year trial period. (KEY TERMS: ice conditions, frazil, ice run, ice jam, flooding.) ICE FORMATION IN RIVERS In slow flowing rivers with a vertical temperature gradient the formation of an ice cover is similar to that in lakes. When the flow is turbulent, however, the temperature differences within the water body are very small. In cold weather the whole water body will be cooled down to the freezing point. The water surface will be supercooled and this supercooled water film will traverse through the 0°C water body. When it comes in con- tact with any nuclei, frazil will be formed. When the supercooled film meets the bottom, frazil will form there and fasten to the bottom as bottom ice (Figure 1). If there are large rocks or other ob- stacles on the river bed these may then be natural foundations for ice dams, often referred to as anchor ice dams. Such ice dams may develop at more or less regular intervals and divide a rapid stretch into a staircase-like system. The dams will form small reservoirs which will soon be covered by surface ice. The ice cover then stops the supercooling. Following this, a slow thaw, due to the Figure 1. Formation and accumulation of frazil. !Norwegian Water Resources and Energy Administration. 593 energy change, gradually opens a narrow passage through the ice dams. After this stage the ice condition is said to be stabilized. However, before the process has reached the stage where most of such a stretch has become ice covered, large amounts of ice have been produced. The ice masses are carried further along and frequently cause problems downstream. Ice formation in rivers with moderate current takes place mainly by the ice growing out from the shore edges or up- wards from already established ice fronts. The critical values of the water velocity for this type of ice formation are about 0,4 m/s and 0,6 .m/s for shore ice and front ice, respectively. At greater velocities drifting ice or frazil, often produced in turbulent flowing water upstream, does not adhere to or remain with former ice fronts. Instead it dips down under the ice and is transported further downstream. When reaching areas where the water current is reduced, the ice will accumulate under the ice cover and reduce the available cross section for water transportation. These accumulations are frequently called hanging ice dams. The whole river might gradually be clogged by ice and thus cause flooding upstream. With changes in meteorological con- ditions, particularly increased tempera- tures or with increased water discharges, an ice dam may break down, causing step bursts, and initiating an ice run in the river. The floating ice may then accumu- late further down, often where the river narrows or in a river bend, or where the slope gets smaller. This may cause an ice jam and subsequent flooding upstream. Such occurrenoes are known to take place in several natural rivers in Norway, and more frequently in some rivers than others. CHANGES IN ICE FORMATION DUE TO POWER DEVELOPMENT A power development will cause in- creased winter flow in rivers downstream from the power station. Normally the water temperature of the discharged water is also increased. Between the reservoir and the power station the river winter 594 flow may often be reduced. Generally there is a danger of more severe ice conditions when the winter discharge is increased and the water temperature is close to 0°c. The tempera- ture of the discharged water depends on the design of the power development. The size and relative location of the reser- voirs are of special importance. Dis- charged water temperature in the range of 3-1oc in the beginning of the winter, gradually decreasing to 0, soc or colder is common to many Norwegian power develop- ments. Substantial variations in dis- charge temperature between plants and also from year to year are frequent. The evaluation of future discharge water temperature following a water power development is therefore of large import- ance in judging the consequences on the ice formation. This is often done by modelling the temperature in the future reservoirs, based on observations of actual temperatures in existing lakes. Temperature above ooc in the river down- stream of the discharge will result in ice free stretches downstream. The extent of these stretches is often very limited, and as soon as the water masses are cooled down to 0°c ice production com- mences. It may be very important to evaluate the area of such an open water stretch following a power development. In recent years several theoretical models describ- ing the cooling of water masses under various meteorological conditions have been developed. For practical purposes, however, we have found the following simple formula presented by 0. Devik in 1931 to be adequate: F·S = Q·t where F = cooling surface, S = specific heat loss, Q = water discharge and t = discharge water temperature. Based on calculations and measurements in several Norwegian rivers Devik also established values for specific heat loss from a water surface at ooc for various meteorological conditions. For meteoro- logical conditions referred to as medium cold, e. g. air temperature -10°C, no clouds and no wind, or air temperature -2ooc, cloudy and no wind, an average specific heat loss in the order of 200 W/m2 is used. For meteorological con- ditions referred to as strong cold, e.g. air temperature -20°C, clear sky and no wind, or air temperature -3ooc, cloudy and no wind, an average specific heat loss in the order of 400 W/m2 is used. The area of open water (F) needed to cool a stream from 1oc down to the freezing point by medium cold, strong cold and given flow (Q) will then be as follows: Q m3/s 1 10 50 I I I I I,O'N I I I I 1_.- I 1"5 I I I I I I 60 N- I Medium cold F(1o3m2) 20 200 1000 -- 20°E I __ M·1=5000000 Strong cold F(103m2) 10 100 500 I I I ~ ---1 c\.:. ...... c'!--I ",c,., ... ~-~-c__----· I I I J \ I I -;1 I I I I J Figure 2. Location of the rivers dealt with. 595 As mentioned an increased winter flow of water with a temperature close to freezing will generally increase frazil formation in a turbulent flowing river. The increased flow will also cause longer stretches of the river to get turbulent flow conditions and thus expose larger areas for frazil formation. The building up of ice dams sufficiently large to stabilize the river will also take a longer time. The risk for ice dams to break down before the river is stabilized will consequently increase. With a relatively small winter dis- charge the breakdown of an ice dam may often be a local event. With an increased winter flow, however, a series of ice dams may simultaneously break down. Due to the larger flow and larger ice masses stored in the river bed the damages may then be more severe. Upstream from the discharge of the power plant, the winter flow may often be reduced. This reduction depends on the design of the power plant, but is fairly common to larger and newer developments. The river stretches thus getting a re- duced flow frequently have experienced turbulent flow and subsequently large ice production before the regulation. In such areas the ice production will diminish after the regulation. NORWEGIAN EXPERIENCES ON RIVER ICE PROBLEMS following some examples of experiences regarding ice in regulated rivers are de- Expected ice conditions for currently being built or planned commented on (Figure 2). In the Norwegian problems scribed. projects are also Glomma Glomma was the first river in Norway where changed ice conditions due to power development was experienced to be a problem (Figure 3). Glomma is the longest river in Norway. The valley runs north- south in the eastern part of the country. The upper part, which is dealt with here has a typical continental climate and often very cold winters. Aursunden, a lake in the very upper part of the water-course, was regulated for power production on January 1. 1924. Average water discharge from Aursunden before regulation was normally about 12 m3/s on November 1. and gradually decreasing to about 5 m3/s by the first of January. The natural water flow from Aursunden just before the reservoir was taken into use, was 4 m3/s. Upon start- ing on January 1. the release of water was gradually increased to 20 m3/s by January 25. An ice run was initiated, this caused the upper 8 km to be ice free and flooding of the ice further down- wards. Due to this, winter transport on the ice, previously common, was prevented down to Os. The next winter the discharge was kept steady at 13 m3/s from November 15. un- til April 15. No ice problems were reported this winter, which was extremely mild. The following winter the release from Aursunden was started November 15. at 13 m3/s, which was kept unchanged until December 17. From then on the release was increased gradually to 21 m3/s by January 2. , and from then on kept con- stant until April 10. The ice situation was, however, not stabilized from Hummel- . -8~0 f-l ----"-.,,~ ~ . c 60 ~ i . . voll to Eidsfoss, where the river is very turbulent. Several ice runs took place in connection with the increase of dis- charge. The next two winters the release from Aursunden was started in October/November at 18-19 m3/s. This flow was kept until the end of January, and from then on gradually reduced. Several severe ice runs were experienced in December these two winters, both in this upper part of the river (Tolga area) and in another turbulent river stretch further down the valley. After these years with severe damages due to ice problems it was claimed by the inhabitants in the adjacent area that the increased winter flow had increased the possibilities for ice runs in the river. A state commission was then established to study the problem. Each breakdown of an ice dam and following ice run was analysed and com- pared with the meteorological situation. In all cases there had been meteoro- logical changes that theoretically could have promoted breakdown of ice dams. Com- parisons were also made with the situ- ation in the neighbouring water-course . 40 r\ ~ 111 : ~ 0 g 20 ~--r:_~--r....:~::_"""T"".,.:: ] ! ~ ~ l. ~~ i ~ ~ 6~~ +--+---+---+--+--+---'"""'-.,..--N ~ 60 :-..... ! 40 ~ • :::1 1'\. .,~ ~ 20~-+---r--+--+-_,--r--+--+--~ - 500 '\--1 80~-+--~-~-~--4-~~-~-+--~-~-~~ 560 554 548 542 536 530 524 518 512 506 500 494 km Figure 3. The upper part of Glomma river. Below the elevation and distance from the ocean. 596 Trysil, which was not regulated. A great similarity could be seen regarding break- down of ice dams within the two water- courses these two years. The formation of an ice dam was not found to be much influenced by the size of the water flow. During formation ice dams are being built just strong enough to keep the water masses accumulated by each dam, and the strength thus adjusted to the water flow. An increase in flow after the dams are established, however, will often break them down. The com- mission therefore arrived at the con- clusion that the breakdown of ice dams and following initiation of ice runs was not due to the larger winter flow, but to changes in the meteorological conditions. The consequences of the ice runs, how- ever, were realized to be much more severe with the increased winter flow. Therefore the commission proposed to build lenses or thresholds across the river on critical points. These should promote an early formation of a stable ice cover, reduce the slope of the river, and thus prevent possible ice runs further down the river. Another problem in connection with the regulation was increased flooding and 1c1ng, especially in two areas, Os and Koppang (Koppang is outside the map area). The slope at both places is very gentle. The commission found that the increased winter flow had created more open river stretches, thereby increasing the formation of bottom ice and frazil, and causing jamming of the river. This increased the water stage and flooding of the river beds. Flooding of cellars was also frequent in these areas. To avoid these problems the commission proposed to reduce the discharge from the lake Aursunden to previous unregulated winter conditions. This has later been changed to a release of 10 m3/s during the period from the start of ice for- mation until the river is stabilized. This normally takes place mid December. From then on it is allowed to gradually increase the release. This has been the general scheme for regulation of lake Aursunden since then, and major ice problems have generally been avoided. 597 Nea. The river Nea originates in Sweden and runs north-west-wards through Selbusj~&en and further towards Trondheim draining into Trondheimsfjorden. The upper part of the watershed has generally a continental climate, but westerly winds may cause mild coastal weather, even with rains high up in the valley (Figure 4). Even before regulation the river Nea experienced fairly large ice production. Problems associated with this were accepted by the population. The large and sudden changes in the meteorologic situ- ation which the valley often experiences during winter further promote ice problems. During the period 1941-1950 various lakes in the upper part of the water- course were regulated for power pro- duction purposes in existing power plants downstream from Selbusj~&en, in the very low parts of the water-course. The corre- sponding increase in winter flow was fairly small and did not give noticeable changes in the ice conditions in the valley. mas1 / ./'" j_NE /--GRESS I FOSS 360 320 280 --c. v ,.--HEGSE FOSS 240 200 __..+.----- I I I I 160 DISTANCE 10 20 30 40 50 60 km Figure 4. The Nea river with the locations of existing power stations and levelling of the main river. It was the introduction of Nea power station, which was the first project in the upper part of the water-course that gave tremendous changes in the ice con- ditions in the river. Large reservoirs were established as a part of this pro- ject, and the winter discharge from the power plant was increased to about SOrn3js. The natural winter discharge was in the order of 10 m3/s. The water was diverted through tunnels from the reservoir to the outlet of Nea power station, a river stretch that previously had experienced a considerable frazil production. Even so, it was pre- dicted that there would be severe ice problems in the river downstream, and which could only be solved by developing a larger stretch of the river. Planning of Hegsetfoss power station was therefore started even before Nea power station was carm:i.ssioned. The first winter of operation of the Nea power station (1960) the ice pro- duction increased. There were severe ice runs, ice jamming and subsequent flooding downstream from the power station. Farms were overrun by ice, roads blocked and large fields flooded. One main reason for building the Hegsetfoss power station was to solve the ice problems in the valley. The winter discharge was by this development taken away from a part of the river with very large frazil formation, thus reducing the already severe ice production in the river. To cope with the ice situation in the valley it was found necessary to keep an even winter discharge in the river. Diurnal variations in the production, which of course would give corresponding variations in discharge, is desirable to satisfy the power demand. When Hegsetfoss was taken into use (1963) it became possible to allow variations in pro- duction in Nea and keep an even discharge downstream from Hegsetfoss power station. Variations in discharge from Nea was then evened out in the intake reservoir to Hegsetfoss. This gave a satisfactory ice situation downstream. Further development of the water- course, especially the building of Gresslifoss power station, have improved the ice situation even more. 598 Otta Otta is part of the Glomma water- course. It drains the mountainous area eastwards of the main watershed in Norway (Figure 5). Before regulation the river Otta was usually covered with stable ice all winter. The contributary rivers from Aur- sj II) en ( Skj ak I) and Tesse are now developed, and also a local fall, Eide- foss, downstream from the lake Vagavatn in Otta river. The lakes Rauddalsvatn and Breidalsvatn are, after regulation, controlled by dams for power generation downstream in Eidefoss, and further down in Lagen. The winter discharge from these lakes has increased the winter flow in the upper part of Otta and caused serious ice problems. These problems have mainly been flooding and icing in the area down- stream from Skjak, particularly the first years after the reservoirs were taken into use. It has been realized that only a very gentle increase in winter dis- charge after the ice situation is stabilized is acceptable to avoid ice problems. In most years severe ice problems have been avoided by this scheme of operation. Downstream from Eidefoss to Otta the winter discharge has been increased from OTTAVASSDRAGET c 800 ~ c: 0 700 :~~ -~ c:~ ~ 600 t---t--1----'\.l:l!••O!JII'~\. :t ~ ~~ ~ ~ ~ ~ ~ soo o ~~mo ~ 400 " -g E 0 oo E o~ ,_ > • i 300 +--+-t---t--+-t--+---+-+--+-t-+--+---"=='=1-...... lOO t--t--t-+--+-t--t--t-t--t--f--t---t--f----J 1 00 1+-:40-'-----:-:12'-:-0 --'------,1~00----'--::l:---'--60-1----__j__----l--..L--2-l-O ----+-, m_JO Figure 5. The otta water-course with existing power plants. Below levelling and distance from the main river Lagen. about 10 to about 40 m3/s by management of the existing reservoirs. The water temperature at Eidefoss is very close to ooc. Every winter there is frazil production, large bottom ice formation and building up of ice dams with follow- ing increased water stage along the river. The river is not stabilized every year. Under special meteorological con- ditions ice runs may occur, occasionally causing serious threats to the village Otta. This demonstrates that the present winter discharge is very close to the maximum acceptable level for this river. Only minor variations in discharge or meteorological conditions may conse- quently affect the ice conditions. The power plant management is well aware of this, but even though special care always is taken damages due to ice jams do occur. Several alternatives for further exploitation of the upper part of the water-course have been investigated. Possible influences on the ice situation have been a part of these investigations. The present alternatives will eliminate the ice problems in the Skj ak area, but introduce new problems in the Otta area unless the river from Eidefoss and down- wards is also utilized. Being aware of mast 360 340 320 300 280 260 2~0 220 200 180 160 1~0 120 100 80 6 ~ 2 0 0 0 l! .. s .. N ::r ~\T'6~·~~ \ \ \ \ \ 1<-ou\o\r.e\noe\~o r--t- ~\\oe\~o ~ ] 0 ~ 0 \ .;a!: '(;<l -~a. ' -.. 8> ~ 0 1\U ' ' -1'1--. I ' I : I I I I I I i I I I I I : I ' I I ' I I r--h--- I I I I this, all plans for further power pro- duction of the upper part of the water- course include simultaneous utilization also of the lower otta river. Alta In the river Alta the Alta power plant is now under construction. During the planning stage of this power development ice problems were also of major con- sideration. Although the area has a fairly gentle climate considering the latitude, the winter temperature is generally low, and being north of the arctic circle there is no daylight for part of the winter. Occasionally there are intrusions of milder weather in the lower part of the valley, particularly during the ice- forming period. The water-course drains part of the Finnmark plateau. It runs · northwards through the lake Vir'dnejav'ri (250 m.a.s.l.) descending a steep narrow canyon to Sav'co (80 m.a.s.l.) in the Alta valley. Alta power station utilizes the fall between Vir'dnejav'ri and Sav'co. The lake Vir'dnejav'ri is the only reservoir, and there is about 46 km J 0 "' > .. -;; :>: ~ ... N ::r 'i'lf ne n..,n IL Avstand fro Altatlva FM 2 _ . . 0 ; ·;;. 0 ~r-! t-! r c!l ~ ko--"' .. II i h;-' ;; II h-r-> II II I I 1--0 I I :z: II I II I 0 62 60 58 5 6 5~ 52 50 ~8 ~~ ~2 ~'ol 38 136 ~Iii I I II I 3~ 32130128 2'6 I 2~1122°20 18 16 1t 12 10 16 I ~ 2 0 ! : : : : I :: ! : : Distance frcJn t+ +ean in km : Slop~ m/km I I I I I ; 1,1 ! 0,2 !,: 2,4 ! 1,7 9,1 Figure 6. Levelling of the Alta river. Average slope of separate sections are shown below. 599 from the power station discharge to the river's outlet in Altafjord, located at approximately 70°N. The average slope is 1,7 m/km, varying between stretches with white water and more gently flowing river (Figure 6). The question raised concerning the ice conditions was how much water could be utilized during the winter period. Every winter there is large ice pro- due tion in the river, and the rise in water stage due to ice is considerable (Figure 7). Ice jams of some kind form every winter. Normally the river will develop a stabilized ice condition, and generally be ice covered during November or December. Occasionally, however, the river has experienced severe ice runs with subsequent flooding and icing, which have delayed the stabilizing of the river. Experiences from other water-courses in Norway have, as already mentioned, indicated that power plant generation with an increased winter flow may affect the ice conditions, often causing serious damage. The Alta river was considered to be very sensitive in this respect, mainly due to the extremely cold weather that often occurs in the winter. The known ice runs in the river were analysed with regard to the discharge and ::E;Q "" L{) N q N ~ C> L{) c:i C> c:i JUN Figure 7. Characteristic values (min., 25 pc, med., 75 pc and max.) of rise in water stage due to ice in the river (1914-1968). 600 the meteorological situation. Only 6-8 larger winter ice runs were reported in the lower part of the river. It was known, however, that breakdown of ice dams that did not give larger ice runs were often not reported. There are certain areas where ice jams formed in this manner often occur. Water temperature measurements were made in the Vir'dnejav'ri reservoir, and the discharge temperature was judged to be 1-2°C in the beginning of the winter season. This will give positive water temperatures a distance of some 6-10 km downstream from the discharge, depending somewhat on the meteorological conditions and the release of water. There is how- ever in this area a fairly long river stretch with an average slope of 2,4 m/km. A stronger current here following increased flow will further retard a stabilization of the ice conditions. In the more gentle flowing areas of the river considerable amounts of ice often accumulates at present. By in- creasing the winter flow, and thus creating larger ice free areas, it was feared that the ice production might be so large that it would cause severe problems, especially in the lower and most populated part of the valley. By analysing the ice situation experienced so far no connection was found between the discharge and the registered ice runs in the river, as in the Glomma case. Even a smaller increase in discharge during the ice-forming period may, how- ever, easily break down ice dams that are being established. It is therefore con- sidered of major importance to avoid sudden increases in the discharge, even if these are small, at any time after ice production has started. It was also con- cluded that it was preferable to maintain fairly high discharge within the natural range of variation during the ice-forming period. The main discharge pattern should be similar to that during natural con- ditions. The reservoir should be used so that no artificial increases in discharge occur. Natural increases during this period should be eliminated or reduced as much as possible. On these grounds it was recommended to permit a discharge during the ice-forming period in the order of the 75 percentile m3/s 140 120 100 80 60 40 20 ---Recommended water discharge SON OJ FMAM MIN Figure 8. Recommended regulated dis- charge compared to natural discharge. of natural discharge until the river is stabilized. This is expected to happen by the middle of December at a discharge of about 30 m3/s (Figure 8). This dis- charge may then be kept throughout the winter. If the discharge for some reason is reduced during the winter it can not be increased again without special con- siderations of the ice conditions. This larger winter discharge is ex- pected to give an earlier ice break-up in the spring and thus give the possibility for an earlier increase in discharge than the normal spring flood. In this way possible unused water in the reservoir may be used at this time. The Alta river is also a very important salmon river, and this winter scheme of discharge was acceptable also for these interests. Usage of the reservoir according to these recommendations were concessioned for a trial period of 5 years. Upon this the control scheme will be reviewed. During the trial period the ice situation of course will be studied closely. The power plant is planned to be commisioned in 1988. The jurisdictional survey evaluating damages and incon- veniences for the population has already started. Due to the general hydrological situation and the reservoir capacity it is agreed to maintain a winter discharge somewhat lower than the official per- 601 mission stated above. This scheme will impose fairly large annual variations in discharge. This may be of particular interest for the coming studies. Stj~rdal The river Stj-rdalselva drains the mountainous areas close to the Swedish border, and runs westwards towards Trondheimsfjorden (Figure 9). Similar to the neighbouring Nea river described earlier, this area has generally a cold winter climate. Mild weather with rain up to fairly high altitude may occur during the winter period. This is of course very promoting for ice runs, and severe dam- ages due to ice problems do occur. The upper part of the water-course was regulated in 1890-1915. The existing regulation consists of several smaller power plants along the Koppera river. The river stretches and intake reservoirs within the various plants are generally not ice covered. The water temperature is gradually decreased so that the discharge water from the lowest plant, Nustadfoss, normally has a temperature of o,soc or colder during winter. The discharge is fairly even and amounts to 10-12 m3/s during the winter period. This is approximately a doubling of the natural discharge. The upper river stretch, from Nustadfoss and somewhat downstream from Funna, has a very gentle slope. Then mas.! ·~ L1 4kMZft7~:~. 0 DISTANCE 0 10 20 30 40 50 lcm Figure 9. Stj -rdalsel va with planned power development. Below elevation and distance from the ocean. there is a relatively steep river stretch down to the area of Flornes. The somewhat newer development of the river Funna has a winter discharge of 3 m3/s with a temperature normally in the order of 2°C the first part of the winter. This gives an open area down- stream of the inlet to the main river. Before any regulation was introduced in the water-course the main river was ice covered and used for winter transport all the way up to Nustadfoss. The larger ice problems that did occur in the river was mainly due to ice runs originating in the tributaries in the lower part of the water-course. When the ice flowed into the main river it caused jamming and subsequent flooding upstream in the main river. The introduction of Nustadfoss power plant, and thus increased winter flow downstream, reduced the ice cover on the river. The relatively steep river stretch down to Flornes became an area of fairly large frazil and bottom ice production. Series of ice dams are now being formed, and the river stretch is normally stabi- lized very late in the winter. Only in winters with very cold weather and a small natural discharge this river stretch will now be completely ice covered. In recent years several ice runs have started in this river stretch which have caused jamming with subsequent flooding and other damages, especially in the area around Flornes. Further development of the power potential in the upper part of the water- shed is now being planned. Larger reser- voir capacity is being introduced. The planned new power plant is designed to take the reservoir water mainly through tunnels, thereby reducing the heat loss of the water. The discharge from the lowest power station, Meraker, will be approximately at the same location as the present discharge of Nustadfoss power station. The question to be answered in this development, as was the case in Alta, is the consequences on the ice situation of a larger winter discharge downstream from the power station. How large a winter discharge that can be accepted without causing severe ice problems is of par- ticular interest. As this is a salmon 602 river proper care has also to be taken of the fishing interests of the river. This new power development is expected to give a discharge water temperature in the order of 2oc for the first part of the winter and then gradually decrease. Depending somewhat on the climatological conditions and the discharge this will prevent the ice cover on a river stretch downstream the station. It is however expected that the water temperature will be reduced to the freezing point by reaching the area of steeper slope. Here there will be frazil production and for- mation of ice dams. One main question is whether it is possible to run the power plant without breaking down ice dams and initiating an ice run in this area. Because of this it is extremely important to avoid any further increases in the discharge during the ice period. Meraker power station will be controlling the discharge from a far larger part of the watershed than Nustadfoss power station is today. By a careful management of the Meraker power station it is expected possible to avoid or at least reduce even natural increases in discharge during the ice-forming period. By this the possi- bilities for initiating ice runs on the river stretch above Flornes should be reduced. On these grounds it will be re- commended for a scheme of regulation similar to that for Alta. The discharge is gradually to be decreased according to the pattern of natural discharge during the ice-forming period. When the river is stabilized it is expected that a constant winter discharge can be kept through the winter. A winter discharge in the order of 25-30 m3 Is wi 11 probably be proposed for a trial period of 5 years also in this case. The ice situation will then be observed closely. A final scheme of regulation based on actual experiences may then be established. JULY COLD REGIONS HYDROLOGY SYMPOSIUM AMERICAN WATER RESOURCES ASSOCIATION 1986 HYDROLOGIC ASPECTS OF ICE JAMS Darryl Calkins* ABSTRACT: The hydrologic aspects of ice jams have received very little attention. During the last 30 years, the major empha- sis has been placed on understanding the hydraulics and mechanics of ice jams and determining their "flood" levels. How- ever, a parameter that should be known with reasonable accuracy is the flow dis- charge at the ice jam location. This paper examines hydrologic information that is important for analyzing ice jam flood- ing problems, such as flow measurements under the ice cover and winter stage rat- ing curves, frequency analysis of winter flow records, watershed cooling and nat- ural river thermal regimes, ice discharge and snowmelt runoff prediction. The significance of each of these areas is ad- dressed and suggested research opportuni- ties are examined. (KEY TERMS: ice jams, river ice, snowmelt, thermal analysis, winter hydrology.) INTRODUCTION The first question one might ask is: What is really meant by the term "hydro- logic aspects" of ice jams? The Inter- national Association for Hydraulic Re- search, after years of heated debate, pub- lished with still much dispute the defini- tion of an ice jam as "a stationary accu- mulation of fragmented ice or frazil which obstructs a waterway." The hydrologic aspects within this definition can prob- ably be as broad and include the "quantity and quality of water and ice that reaches an ice jam site." The hydrologic aspects of ice jams have not been heavily researched. In the past two decades there have been consider- able advances in understanding the mech- anics and hydraulics of floating ice jams, mainly clarifying and applying existing theories to solving engineering problems (Pariset et al., 1966, Uzuner and Kennedy, 1976, Calkins, 1983, Beltaos, 1983). Numerical modeling of the ice processes and hydraulics has been expanded to offer greatly improved predictions of the water levels associated with ice covers and ice jams (Pashke and Schoch, 1985, Shen and Yappa, 1984, Calkins, 1984, Petryk, 1981). However, the significance of the hydro- logic input to the ice jam has been neg- lected. For example, the prediction of the hydrograph and the associated water temperature at a certain location from rain on frozen ground or rain on snow or snowmelt is an important feature that needs attention. Anderson and Neuman (1984) recognized the importance of frozen ground on the runoff processes and pro- posed a frost index correction to the National Weather Service flood prediction model. This is a start in the process but, as they indicate, a more physically based model is necessary before signifi- cant gains can be made. Hydrologic models that can predict both the flow rate and the temperature during the winter season have not been developed because of a lack of understanding of some of the basic hydrologic processes. POTENTIAL RESEARCH OPPORTUNITIES Several areas have been identified that are deficient in knowledge or lack *U.S. Army Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire 03755-1290. 603 sufficient accuracy in the data when deal- ing with ice jams. These areas are flow discharge measurements, flood frequency analysis of flows and stages, watershed cooling and freeze-up flow prediction, natural river thermal regimes in winter, ice discharge, and snowmelt runoff predic- tions. The latter has received much attention from a mass prediction view- point, but very little has been done on the quantity of heat available during these events. For example, melting of the ice in a river occurs every year and is often the only relief from ice jam flooding, yet little is known on the tem- perature distribution and its timing dur- ing the runoff period. Flow measurements and winter records There are ongoing programs within the u.s. and Canadian Water Survey Agencies that are addressing the issue of poor records for the winter season. At a re- cent conference in Michigan, researchers and engineers presented their work on how flow measurements were conducted under the ice and how the stage records were adjust- ed to account for the ice cover. A major difficulty in conducting flow discharge measurements under the ice cover has been the uncertainty of whether the current meter was functioning properly due to ice accumulation on the cups. Vane current meters are now being used with only minimal problems. In some of the frazil-laden rivers, finding the flow zone was a major problem; simply putting down 20 holes evenly spaced across the river does not adequately find the flow area. An area that still needs research is in the conversion of river stage data to flow discharge. Most countries adjust their records in a similar manner taking into account other basin flow records, air temperature, precipitation, and usual- ly one discharge reading during the win- ter. The actual details of correcting to the proper stage will not be given. The flow records during freeze-up are probably more accurate than those during break-up because of the non-dynamic conditions of the ice and water and the noticeable trend of declining discharge as the winter sea- son begins. 604 B "I } R2 = h2 /2 h 1 = xh 2 h =h +t+h X. I 2 Rl =hi nb=ni=nt Figure 1. An example of flow occurring beneath and over the ice sheet. Flow records during break-up can be improved by accounting for the mechanics of the ice cover and ice jam conditions. For smaller rivers the ice cover prior to break-up may flood with water over the top. Figure 1 is an example of flow occurring beneath and over the ice sheet. The flow hydraulics can easily be written for both areas, as uniform flow can still be assumed in both zones and there is no pressure flow in the lower zone. The Manning equation is applied to the two flow areas. For ease in presenta- tion, the bed and ice cover roughnesses (both top and bottom) are assumed the same, ni = nb = nt• The river with a uniform slope S is sufficiently wide to assume that the wetted perimeter is nearly equal to the river width B. The flow depth over the top of the ice is repre- sented as some fraction (x) of the flow depth below the ice cover, h 1 = xh 2 • For the flow over and under a horizontal ice cover, the discharge is = B sl/2h25/3 5/3 Q ~-(x + 0.63). A potential error could arise when calculating the discharge if the flow over the top of the ice is ignored, and this is computed on the basis of the ratio Q /Q where Qc is the total flow (under an~ u over the ice) and Qu is the flow under the ice at the same stage. The flow under the ice cover is 0 1.0 Figure 2. Error in actual discharge by not accounting for flow over the top of the ice. Bs112 hS/3 Q = 0.63 -- u ~ t where the ice cover thickness t is assumed to be negligible in the definition of the total stage, ht = h 1 + h2 + t. By plotting Q /Q versus h 1/h in Figure 2, c u t the error in actual discharge is seen. For example, if the flow depth over the ice cover is only one half the total depth, the calculated discharge, if one assumes the cover was fully floating, is about 20% low. When the ratio h 1/ht ) 0.70, then the flow discharge will be greater than that assumed for the floating cover. The ice sheet will not remain hori- zontal but will deform due to the buoy- ancy. It has been assumed for this ex- ample that a triangular shape will repre- sent this deformation and that the ice sheet bends at the midpoint of the channel and rises with the water level (Fig. 3). The combined flow discharge is now BS 1/2 h 5/3 2/3 Cf-) [ 2 ( c 1 + ~) o + ~) ) 605 If the same ratio of flows Qc/Qu is developed for the triangular-shaped ice cover, the relationship improves slightly (Fig. 2). The triangular-shaped formula- tion would improve the estimate for the discharge up to h 1/ht = o.s, but there- after the use of a triangular representa- tion will always underestimate the flow. A time-dependency creep model could be developed to predict this shape, which would be more like that shown in Figure 4. Realistically, the error in calcu- lating the discharge is probably between these two curves up to h 1/ht = 0.5 for a finite portion of time, and this is why a time-dependency model would be more use- ful. Beyond h 1/bt > 0.5 the discharge computation should depend upon the ice sheet shape in the channel. Field observations by the writer in New England have indicated that when the width is in excess of 100 m, no signifi- cant water flow over the ice exists. There is a critical river width that is related to the stiffness (characteristic length) of the ice sheet, which would be a function of the ice thickness and stage rise. Figure 3. Triangular representation of ice cover. Figure 4. A more realistic representa- tion of the ice shape in a river. .. "' 2 !/) 0 Max. Ice Jam Thickness Time Figure 5. A typical break-up hydrograph. During ice jam conditions, the flow estimates should be checked versus the flow computations one would get using the knowledge from the equilibrium theory for ice jam thicknesses (Beltaos, 1983, Calkins, 1983). At this time one would still have to rely on the judgement of the individual reviewing the stage records to determine when the ice cover went from a single-layered sheet to a multi-layered ice jam, which is not hard to determine from the stage records. Further research in the analysis of break-up hydrographs would help selecting the times when break- up and ice jam conditions were present. Beltaos (1984) has taken the records from one hydrometric station to evaluate the river ice break-up using the stage records for the winter period and a simple thermal energy factor. This type of analysis needs to be extended to many more sites for further evaluation. In Figure 5 is a typical break-up hydrograph where the sharp decline represents the initial break-up (water and ice moving out of storage), the next steep vertical rise in- dicates that the ice jam has stalled or is moving very slowly (positive wave moving upstream), and the rapid decline again in- dicates that the jam has moved downstream (water and ice coming out of storage). River water surface gradients during this period will be very steep and a single looped rating curve is probably not appli- cable. Frequency Analysis The frequency analysis of maximum annual winter stage and winter discharge records is becoming recognized as impor- tant for studies such as ice jam flooding, 606 hydro-electric winter flows or ice con- trol, water supply regulation, etc. The population of peak flows occurring in the ice-covered and non-ice-covered seasons can be seperated into "winter" and "sum- mer" events and treated as independent sources because the processes for peak open channel floods are hydrologically based without the influence of ice on the river or runoff from snowmelt or frozen ground conditions. The winter season is defined as beginning with the day ice first forms on the river to the day of ice out. Distinctions of flow periods within the winter season are also recognized, e.g. the freeze-up versus the break-up period. Techniques for analyzing the stage and discharge frequencies have been presented in several recent papers (Gerard and Karpuk, 1978, Gerard and Calkins, 1984). Research is needed to determine the type of distribution the winter flows follow. Can the distributions be region- ally characterized? Another area that needs attention is the proper use of limited data bases and the kind of confi- dence limits or risks that are associated with these data sets. Freeze-up flow prediction and watershed cooling The prediction of flow during the freeze-up period is critical when trying to assess an ice jam flooding problem that occurs as a result of the excessive ice production. This type of flooding problem usually occurs when there has been suffi- cient cold weather to have frozen ground and little or no snow on the ground. A warm front then moves through the water- shed bringing rain on frozen ground. The warm front is then immediately followed by a cold front that cools the river water to its freezing point, and ice generation be- gins with the river at relatively high flows for that time of year. The prediction of runoff at the be- ginning of the winter has received little attention. Predicting the water tempera- ture of the runoff requires knowledge of the effect of frozen ground on basin run- off and the heat loss experienced. There are algorithms for use with various hydro- logic models that incorporate the effect of the frozen ground (Anderson and Neuman, 1984) on the response of watershed runoff, but the heat flow associated with the run- off has not been evaluated. This writer has not been able to lo- cate many references that address the cooling of watersheds and the impact it has on the runoff. Groundwater and other shallow source zones are under the influ- ence of the cold front that penetrates into the soil column, but how these sources influence the river water tempera- ture during freeze-up has not been quanti- fied. There have been some observations on river water temperature and correspond- ing groundwater temperatures in nearby wells, and the definite lag in the ground water temperatures versus the river water is clearly evident. River thermal regimes The cooling and heating of river flows has been studied for many years in relation to excess thermal energy supplied from such users as nuclear power plants, sewage treatment outflows and heavy indus- try. The natural cooling regimes in riv- ers have not been studied in detail be- cause of the complexity of this process. Shen and Yapa (1984) have used a convec- tion-diffusion model for the cooling of the river flow in the St. Lawrence River. At this time the model does not account for lateral inflow and its temperature. The cooling of the river water to its freezing temperature is important for pre- dicting the onset of ice production and the subsequent ice cover formation. In assessing ice jam flooding, the solution that is often reached, because no other option is available, is to let the jam melt and the water levels subside. The question is, how long does the jam take to melt? If the water temperature and flow discharge are known or can be predicted, then the available heat to melt the ice jam can be computed, as well as the time. Water temperature immediately following break-up is extremely important. Parkin- son (1982) reports that the water tempera- ture at the leading edge of the break-up ice jam on the Mackenzie River was in ex- cess of +8°C. Which is the major heat 607 source: short wave solar radiation, sensible heat, groundwater contribution, tributary flow or all of the above? Pang- burn (personal communication) has recently measured the water temperature during the 1986 snowmelt season at the Sleepers River Watershed in Danville, Vermont. There is a tremendous shortage of data on the temperature regimes of natural rivers immediately prior to freeze-up or just after break-up. With the use of pre- cision thermistors, accuracies of 0.05°C are easily attainable and are necessary for research in this area. Portable water temperature instrumentation with data log- ging capability is now available. Ice Discharge One of the unknown quantities of riv- er flow during freeze-up is the ice dis- charge, but it is an extremely important factor when trying to predict the water levels and the ice cover progression rates. Very little field data have been collected and limited analytical develop- ment on the heat loss at the water/air interface has been performed. A few stud- ies have been undertaken to understand how surface and frazil ice is incorporated in- to the shore ice and anchor ice. Attempts have been made to measure the ice dis- charge on the Susitna River (personal ob- servations, Carl Schoch, R&M Consultants), but conclusive results have not been achieved as yet when compared with analyt- ical computations and field measurements of ice cover progression rates. Another important aspect of the river ice discharge is the role it plays in de- termining the locations of the initial ice bridges that start the progression of the ice cover. Predicting the first ice bridge and its location is important and significant to any modeling effort when one is attempting to change the hydraulic regime in some manner, i.e. increased flows or channel modification. An equal- ly important consideration is how much ice can be expected to reach a jam site during break-up. The first step in improving our knowledge is to develop field equipment and techniques for measuring the freeze-up ice discharge. Sampling and measuring the collected ice will be the problem. Is a full depth sampling apparatus necessary or does only the upper few centimeters of the flow depth need to be considered? How many samples in a cross section are needed? Is using a calorimeter the only method that ensures that trapped water around the ice crystals is not included in the calculation? How will the ice that is trapped in the shore ice or anchor ice be accounted for in the heat budget approach to calculating the total ice pro- duced? As one can see, there are many questions to answer. Snowmelt runoff prediction This paper will not go into much de- tail on this subject, but the reader is referred to recent conferences and sympo- sia that have focused specifically on the subject material, for example Unesco-WMO- IAHS 1972, CRREL 1978, the Northern Re- search Basins Symposiums, the Eastern and Western Snow Conferences, etc. and other journal articles. The one area that needs consideration is the prediction of water temperature along with the flow discharge, and this has been addressed in another section. SUMMARY AND CONCLUSIONS The importance and significance of hydrologic information for the analysis of ice jams has been emphasized. Areas of research to strengthen the quality of the hydrologic input have been suggested. These major areas are: 1. Discharge calculations during the ice cover break-up and ice jam periods based on the mechanics of the ice cover and equilibrium thickness of ice jams. 2. Statistical distributions for winter peak flows and enhanced analytical techniques for sparsely populated sam- ples. 3. Runoff prediction for freeze-up periods where the impact of frozen ground is important. 608 4. Field data on the water tempera- ture of natural streams and analytical methods for predicting the cooling of water and melting of the river ice in these streams. 5. Field measurements on the quant- ity of ice being generated and transported in rivers, both at freeze-up and break-up. REFERENCES Anderson, E.A. and Neuman P.J., 1984. In- clusion of frozen effects in a flood forecasting model. Proceedings of the Fifth Northern Research Basins Symposi- um. The role of snow and ice in North- ern Basin Hydrology, Viermak, Finland, March 19-23, 1984. Beltaos, s., 1983. River ice jam- Theory, case studies and applications. Journal of Hydraulic Engineering, ASCE, Vol. 109, No. 10, PP· 1338-1359. Beltaos, s., 1984. Study of river ice break-up using hydrometric station records. Proceedings of the Workshop on the Hydraulics of River Ice, June 20-21, Fredericton, N.B., pp. 41-64. Calkins, D.J., 1983. Ice jams in shallow rivers with floodplain flow. Canadian Journal of Civil Engineering, Vol. 10, No. 3, pp. 538-548. Calkins, D.J., 1984. Numerical simulation of freeze-up on the Ottauquechee River. Proceedings of the Workshop on the Hydraulics of River Ice, June 20-21, Fredericton, N.B., pp. 247-278. Gerard, R. and E.w. Karpuk, 1978. Prob- ability analysis of historical flood data. ASCE Journal of Hydraulic Divi- sion, Vol. 105, No. HY9, Sept., PP• 1153-1165. Gerard, R. and D.J, Calkins, 1984. Ice Related Flood Frequency Analysis: Appli- cation of Analytical Techniques. Pro- ceedings of Cold Regions Engineering Specialty Conference, CSCE, Montreal, pp • 85-101. Pariset, E., R. Hauser and A. Gagnon, 1966. Formation of ice covers and ice jam in rivers. ASCE, Journal of the Hydraulics Division, Vol. 92, No. HY6, PP· 1-24. Parkinson, F.E., 1982. Water temperature observations during break-up on the Liard-Mackenzie River System. Proceed- ings of the Workshop on Hydraulics of Ice Covered Rivers. June 1-2, Edmonton, Alberta, pp. 252-265. Paskhe, N.w. and c. Schoch, 1985. Simula- tion of river ice processes on Alaska's Susitna River. Paper presented at Waterpower 85, Sept. 25-27, 1985, Las Vegas, Nevada. Proceedings to be pub- lished by ASCE. Petryk, s., 1981. Numerical modeling and predictability of ice regime in rivers. 609 Proc. IAHR Int'l Symp. on Ice, Quebec, vol. 1, pp. 426-436. Shen, H.T. and D.D. Yapa, 1984. Computer simulation of ice cover formation in the upper St. Lawrence River. Proceedings, Workshop on Hydraulics of River Ice, NRCC subcommittee on Hydraulics of Ice Covered Rivers, June 20-21, Fredericton, N.B., PP• 227-245. Smith, K., 1974. Water temperature varia- tions within a major river system. Nordic Hydrology, vol. 6, pp. 155-169. Uzuner, M.S. and J.F. Kennedy, 1976. Theoretical model of river ice jams. ASCE, Journal of the Hydraulics Divi- sion, Vol. 102, No. 9, pp. 1365-1383. AUTHOR INDEX Alger, George R. . . . . . . . . . . . . . . . . . . . . . . . . 27 5 Allen, Milan W .......................... 531 Ashton, William S .................... 415, 501 Asvall, Randi Pytte ...................... 593 Benson, C. S ........................ 101, 471 Berg, Neil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 Bergman, James A ....................... 367 Bergstrom, Sten ......................... 461 Bjerklie, David . . . . . . . . . . . . . . . . . . . . . . . . . . 345 Boone, Richard L. . . . . . . . . . . . . . . . . . . . . . . . 263 Bowling, S. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 71 Bradley, N. Elizabeth ..................... 407 Braithwaite, Roger J ...................... 485 Bredthauer, Stephen R ................ 415, 573 Burkett, R. D ........................... 179 Calkins, Darryl .......................... 603 Campana, Michael E. . . . . . . . . . . . . . . . . . . . . . 263 Carlson, Robert . . . . . . . . . . . . . . . . . . . . . . . . 345 Chang, Alfred T. C. . . . . . . . . . . . . . . . . . . . . . . 521 Chapman, David L. . . . . . . . . . . . . . . . . . . . . . . 491 Clark, Michael A. . . . . . . . . . . . . . . . . . . . . . . . . 113 Clarke, Theodore S. . . . . . . . . . . . . . . . . . . . . . . 329 Cobb, Ernest D. . . . . . . . . . . . . . . . . . . . . . . . . . 135 Coffin, Jeffrey H. . . . . . . . . . . . . . . . . . . . . . . . 501 Coleman, H. Wayne ................... 63, 557 Cooley, Keith R ......................... 439 Cooper, David J .......................... 93 Dean, K. G. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339 D'Souza, Giles . . . . . . . . . . . . . . . . . . . . . . . . . . 113 Edmundson, Jim A. . . . . . . . . . . . . . . . . . . . . . . 179 Edmundson, John M. . . . . . . . . . . . . . . . . . . . . . 179 Farnes, P. E. . ........................... 13 Feinstein, Joel I. . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Foster, James L. . . . . . . . . . . . . . . . . . . . . . . . . 521 Futrell II, James C. . . . . . . . . . . . . . . . . . . . . . . 131 Gemperline, Eugene J ................. 3, 6.3, 73 Gieck, Robert E., Jr ...................... 283 Gosink, J. P ......................... 31, 471 Granger, R. J. . . . . . . . . . . . . . . . . . . . . . . . . . . 427 Gray, D. M. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427 Greenlee, D. L. . ........................ 157 Gross, Harry ............................ 121 Hall, Dorothy K. . . . . . . . . . . . . . . . . . . . . . . . . 521 Hamblin, P. F ........................... 167 611 Harleman, Donald R. F .................... 39 Harrison, William D .................. 329, 471 Ishikawa, N. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305 Janowicz, J. Richard ..................... 313 Johnson, Douglas . . . . . . . . . . . . . . . . . . . . . . . . 329 Jones, K. C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Kane, Douglas L ..................... 283, 321 Kattelmann, Richard C ................ 359, 377 Kettle, Dr. Roger J. . . . . . . . . . . . . . . . . . . . . . . 113 Koenings, J. P. . . . . . . . . . . . . . . . . . . . . . . . . . . 179 Kobayashi, D. . . . . . . . . . . . . . . . . . . . . . . . . . . 297 Kojima, K. . . . . . . . . . . . . . . . . . . . . . . . . . 297, 305 Kraeger-Rovey, Catherine . . . . . . . . . . . . . . . . . . 93 Kyle, Gary B. . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 Kuusisto, Esko . . . . . . . . . . . . . . . . . . . . . . . . . . 397 Lal, A. M. Wasantha . . . . . . . . . . . . . . . . . . . . . 583 LaPerriere, Jacqueline D ................ 31, 143 Landine, P. G ........................... 427 Latkovich, Vito J. . . . . . . . . . . . . . . . . . . . . . . . 131 Lindsay, Robert E. . . . . . . . . . . . . . . . . . . . . . . 221 Louie, David S. . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Marron, J. K ............................. 13 Marsh, P ................................ 23 Martin, Donald C ........................ 143 Mayo, Lawrence R ................... 471, 509 McGurk, Bruce J ......................... 359 Miller, Woodruff . . . . . . . . . . . . . . . . . . . . . . . . 541 Mosher, Frederick R ...................... 531 Motoyama, H. . . . . . . . . . . . . . . . . . . . . . . 297, 305 Munter, James A ......................... 245 Olesen, Ole B. . . . . . . . . . . . . . . . . . . . . . . . . . . 485 Panfilova, V. K .......................... 293 Pansic, Nicholas . . . . . . . . . . . . . . . . . . . . . . . . . 221 Parks, Bruce . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 Paschke, Ned W. . . . . . . . . . . . . . . . . . . . . . . . . 557 Petrik, William A ........................ 253 Pott, David B. . . . . . . . . . . . . . . . . . . . . . . . . . . 221 Predmore, Steven R ...................... 565 Prowse, Terry D. . . . . . . . . . . . . . . . . . . . . . . . . 121 Rhodes, Jonathan J. . .................... 157 Robinson, David A. . . . . . . . . . . . . . . . . . . . . . . 54 7 Rothwell, R. L. . . . . . . . . . . . . . . . . . . . . . . . . . 231 Rovey, Edward W ......................... 93 Rowe, Timothy G. . . . . . . . . . . . . . . . . . . . . . . . 213 Rundquist, Larry A. . . . . . . . . . . . . . . . . . . . . . 407 Sand, Knut . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321 Santeford, Henry S ....................... 275 Schmidt, R. A ........................... 355 Schoch, G. Carl . . . . . . . . . . . . . . . . . . . . . . . . . 573 Shafer, B. A. . ....•...................... 13 Shen, Hung Tao . . . . . . . . . . . . . . . . . . . . . . . . . 583 Sherstone, David A. . . . . . . . . . . . . . . . . . . . . . . 121 Skau, C. M ............................. 157 Slaughter, C. W .......................... 101 Sturges, David L ...................... 53, 387 Swanson, L. E. . . . . . . . . . . . . . . . . . . . . . . . . . . 231 Tabler, Ronald D ......................... 53 Tesche, T. W ............................ 449 Theurer, F. D ............................ 13 Trabant, D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 71 Wege, Russell E. . . . . . . . . . . . . . . . . . . . . . . . . . . 87 Whitfield, Paul H. . ...................... 149 Whitley, W. G ........................... 149 Wei, C. Y. . ............................ 167 Woo, Sheri . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 Woods, Paul F ....................... 195, 213 Wu, Yaohuang . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Zuzel, John F ........................... 237 612