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<-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