Loading...
HomeMy WebLinkAboutAPA2832S L *-i*a.. ^- sc, -,hi rr re,** a,$*"$ *wa&-'w*2-mmwmamm%=*mm--- *~~~~~~~m~d~7~u~*M~-*w~ ~~~~~~~-"~rx1*-IIUI~aurW~="l~ f --Ms--M.-*- """" w-mw""pv -w-- I_ mppm-v-- * . -+ i 1 i fgt 1" 0 "W -. a ! , && Gy D 1 1% A P l a* f* P~~JEWAL ENERGY REGULATORY 68Mh418SWN PR03EGT Na. % 3 $4 STREAMFLOW FORECAST iNG f Y ST'UBY Prap%~*@d BY i+.iVDE>( CORPORATION FINAL REPORT QAQ@~ c@@tga~e HblRZAaE@ABGO ~NE "$435 SUSlf NA JOlNT VENTURE Docmsnt No1 2832 SUSITNW HPDRBELECTWHC PROJECT STREAMFLQH FQRECASTIBG FEASIBXEITP STUDY Report by Hydex Corporation Eugene L. Peck Thoma% N, Keefer Under Contract ta Harza-Eba sco Sus i t na Jo int Venture Prepared for Alaska Power Authority t..- 4 *< rq 3 .L.i~ objec"ii7e of the Susi trla Hydroelectric Project is to e <~~~c~r.:~ze the iise of water for power generation while meeting environmental constraints, suck as requirements for fisheries and >~ildlife, The purpose of this study is to provide information on hyd-o? ogical forecasting nrtetlnods (including basic liata reqcire- fner?ts) to assist the Alaska Power Authority in devc:loping an eiFjiectj - "j,, water supply foxezasting program for the 2rajecte Only about half of the annual runoff from the Susitna Basin origina':es from snowfall during the winter months (October-April) and acurate forecasts of this snowmelt contributiean to tile sessonal runoff are possible. The question of how accurately icunoff from summer (Hay-September) precipitation can be forecasted is more difficult, out foracasts of runoff for two to three weeks in advance are possible. For accurate long-term forecasts of the total seasonal runoff, it is necessary to have a knowledge of; (a) the amount of runoff that will be contributed from rainfall and snowfall (including glacier ice) that have occurred in the basin, (b) short-per iod weather forecasts, [c) current seasonal trends of precipitation and temperature, and (d) statistical indices of climatological cond itions for the rest of the forecast period. An operational forecast program can be developed that will provide short and long-term forecasts of runoff of sufficient accuracy to successfully operate the Susitna Hydroelectric Project and achieve the objective stated above. This will require i:he collection of additional information on the precipi tatinn regime sf the basin and application of hydralogical models, To cost effectively accomplish the collection of required basic data and ts develop an operational forecasting service for the project, the follow~ny recommendations should be followed. 2. Basic Data Collection Program A. Findings The current data collection program does not provide sufficient inEormatian to support either the development nor the operation of an hydrological foreczsting service. 13 ,, ~ecsrz~~qendat ions In order to have sufficient information for c?evelopn:ent of an o;jsrz"Lional forecasting service "ie following actions should be taken during the period until the project becomes operational; (1) Discontinue measuring at meteorological stations (except at Matarla base camp and for other specialized purposes) all. meteorologicaL variables except preci- pitation, teinperature and wind movement-, , (2) Operate and collect in real time measure- ments from 12 meteorological stations shown on Figure 2, page 24, (3) Continue to coliect snow survey informa- tion (in cooperation with SCS) for sites shown on Figure 3, page 26, (4) Continue to collect streamf low records from the three 'JSGS gaging stations, and (5) Summarize and archive all basic data collected fsr the forecast program, When the project becomes operational do the follo~.iing; (1) Collect in realtime streamflow, reservoir stage and information on reservoir re- lease sf water, (2) Continue collection in real time measure- ments from the 12 meteorological stations, and (3) Continue collection of snow survey measure- ments from those sites whose data are found to be required for the operational forecast pragzam* It is recommended that data collection be accomplished usin52 a satellite telemetering system. 3. Development af Operational Forecasting Service PA. Even with the additiorlal basic data proposed above, it would not be possible, using presently available niodels and imataloij ica1 informat ion, to forecast the runoff for the Susitna Basin with sufficient accuracy. Additional studies need to be accomplished to im2rove the utility of models that arc currently used for forecasting runoff: from other glacier izec-i basins in Alaska. Actions ta be taken incl-ude; ."'. CI in~aisological Stud ies (I) Develop seasonal isohyetal (aver3ge ppreci- pitati011 maps) for the Susitna aasin. (2) Develop climatological statistics of Play- September weather conditons for Susitna Basin required for ex tending short-term supply forecasts to seasonal forecast (semi-conceptual model) . B, Model Studies (I) Arrange for application of hydrological forecasting models to basin as recom- mended in Chapter 3. Models to be con- side~ed for use are: (a) Wnder son NWS snomeE t: model (b] Quasi-conceptual seasonal runoff model (e) NWSRFS and the SSARR soil moisture accounting models Other actions to be considered in the future for improving the forecasting service include: (1) Adjusting precipitation records (espe- cially snowfall) to reduce variability and bias introduced by wind sction on gage catch, (2) Incorporating satellite techniques for enhancing the knowledge of the magnitude and distr ibutiomn of summer: thunderstsrm precipitation, (3) Using aerial gamma radiation surveys to obtain better knowledge of the areal distribution and magnitude of the snow cover on non-glacierized arsas, (4) Using meteorological i~formation (such as upper air measurements of moisture and wind) to improve knowledge of precipita- tion over the high areas of the basin, and (5) Obtaining reliable measurements of the axes% extant and af the seasonal water balance af the primary glaciers in the bash~.rr, TABLE Oi7 CON2'ENTS r>;F"" "rn' .-JLLj~aVE SUMMARY,...,,,., ,.,.e,,-,,,.,,,,,,,,s,,s,,,,eaosei 1, QBJE@TIVE,~,eeo*~~~oslfo~kyIes~biaosocpoaia~~~aoe~*~~~- 2, BASXC DATA COLLECTION PROGRAF4, ts e e e e e a 0 e e 9 Cs e 14 I 3, DEVELOPMENT OF OPERATIONAL FORECASTING * I SERVICEeaoe+eetr;ecalaaaa~cseeh-ea,eoiJo~adevaaeazsaso~~ 4, OTHER CONSIDERATIONS, , , , , 8 a w1 i&i ?P ea 18 "4 ii -e" CHAPTER 1, PNTRODLT,FGTION,,,..,,..,,.,,,,,,,.,.,,.,,,,.~e,..~. 1, BACKGROUND, , , . , , , . . . , . . , , ,, , , . . , , . , . , , , . , , , , , , b 2, PURPOSE AND SC~IPE,~.,.,,,,.,,,,,,,,..,,~~.,~~~ CHnFTER 2. DATA REQUIREMENTS.. . . . . . iQ qb %11 rn 4P B) QP a i&i iPd bB , B, I a , Ib a a 4 2, RESULTS OF FIELD TRIP TO ACA%KA,,..,.......,.d 2, ACCURACY AND REPRESENTATIVENESS OF aASH@ DW'~A~s+rn890eBC~.0~e98b.~0BI~~~~0)(h8tIs~*IIa41)e~~~e~6 3, EVALUATION OF EXISTING DWT8"A CBLIL8ECTIBN PROGRWMB6m88e~QaiQOgIPIB~~~e4Bb~~gCB)Bdl,D~OOs~~~ 4. DATA REQUIREMENTS FOR HYDROLOGIC FORECASTIMG,,D,,,,,~~rsC69,(g,Q~$,~,or?rO~r)~'6F~Is~~~~f~ 5, CLIMATOLOGICAL DATA REQUIREMENTS FOR MODEL CA&&BAATfQNaeeese,~esrB)eaill.Bs~*eaoIea9po1eIj*I1"3" 6, OPERATIONAL DATA RE~UIREMEMTS,,,,,,,(iiUeO(irsbsaI~8 7, RECOiqMEMDATf ONS FOR CHANGES IN DATA COLLECTION PR6GRAMrB)e*eeeea*eaaa~Vre.*e~r~e~22 8, OTHER RE@OMMENDATPONS,,e,rB..iJiOIRBOeD8~(IIaDOQ0~.&,O28 CHAPTER 3, MODEL TECHNOLOGUeaea,*,,e~pIoLa~ceaa,aeeeaa~e01:*p1a~O L, FACTORS FOR SELECTING FORECAST MODELS* B, a *. :30 2, MODEL WLTERMATHVES,,,I,,,Qmr)Ls~B1,~s~eCaD,~~dXU~37 3, FORECAST ERRORS,,~,ge.~,B,..,ssQDO~Osdee.t(b~OIR~TO 4, MODEL REGOMMEMDATIONS,a,Ba~iQ*.~p~e91a1P,ussdiC.El.S3 CHAPTER 4, OPERATIONAL CONSEDEWATIONS,,,,,(ZQ(D,PJr,09$IBe.*Qd55 l, INTRODUGTIQN, . , . , , . . . ., . , , . , . , , . , ,. . , . . , . . , .!55 2, OPERATION OF TEIE FORECAST MODEL,, , , , , , , , , , , ,55 3, DATW GOGLECTIBN RNB REEAUeea.a,~s~,,,,,0re4L~!55 4, RECOMMENDATION ~~,,~~**~~~~~m~a~*~4r~Saayo~eaVra4S3 5, ALTERNATIVES FOR TEiE SUSITWW BASIN,, , , , , , , , ,58 5, COOPERATION ON DATW RECEPTIBN,,,,..,,,,e,,*06:3 7, PURCHASE OF DEDICATED GROUND STArrC"ISN, .. . . . , .63 BUDGET iCST&MWTION FICISRESu BL I$ + * 0) BL ,#. . rp, k. q, ap 64 9, ADDXTIONAL RECOMMEMDATIBNS,*e,~e~*$ieweaspJ~nrg,6~$ FfAPS OF MONTHLY PRECIPITATION, SUSTTNA RIVER BASIN PIaOTS OF DAILY DISCHARGE AND DAILY BRECIEK'3TArEC"ION, SUSITNA RIVER BASIN RESULTS QF MULTIPLE REGRESSION ANALYSTS, SUSITMA RIVER BASIN AT DENBEI, ALASKA EXPLWWATORY INFORMATIONd DFFPNfTfONS OF STATES WMD PARAMETERS, AND SCHEMATIC BXAGWAHS FQK CONCEPTUAL HYDROLOGICAL MQQECS R & M CONSULTANTS LETTER OF ILWRCH 8, 2985, The S&'isitna Hydroelectric Project includes the construction of two tla~i~s, Watana and Devil Canyon, on the main channel of the Sustina River, The 8atana damsite's drainage area is 5,180 square mi?zs, the Devil Canyonss, 5,810. Tkie project is designed to p?.ovide eltzctriclty to the Railbelt area of Alaska Ear well iflto the rlexc zrzrrtury and to have Elow regimes that will minimize impact on fisheries and other environmental resources, The prc3posed Watana Reservoir will have 9.7 million acre feet of storac;t? of which 3.7 million will be active storage, The Devil Can:ron Reservoir will not have any significant amount of active storage, Since power requirements are greatest during the winter, tl~e psoject is designed to have as much water in storage as possrble by September of each year. Durint~ visits by Eugene L. Peck and Thomas N. Keefer to Anchorage, Alaska, and Chicago, Illinois, in December 1984, discussi.o~2s were held with representatives of Warza-Ebascu and the Alask.-i Power Authority (Mr. Eric A. Mari-hegiani). The following :nformation perzinent to hydrological forecasting for the pxoje#:t: was received. Based on the discussions held in December 1984, the following bydrologir:al forecasts are required for the operation af the project: t 1) Forecasts of seasonal May-September inflow to Matana Reserve ir (long-term forecasts) , -2) Forecasts of inflow to Flatana Reservoir for two-week periods (short-term forecasts) . The i$~ng-term seasanal forecasts are necessary to schedule releases i'rom the reservoirs to ensure the maximum water storage hy S~zptemkjer. These flarecasts are required at semi-monthly jntervals starting on April 1 but possibly as early as January L of each year. Each forecast should be a probability forecast for the May-S~ptember runoff (or fox forecasts after May 1. for tile period frcrn the date of the forecast through September), These forecasts, used in canjunction with standard reservoir operational. ru1.e c~rves, will percmit the reservair operators k.3 plan for having maxisum water storage in the Watana Reservoir prior to the winter. The $tort-term forecasts (inflow to Watana Reser'~oir for the next two wgzeks] are rctquired daily from May through September, The Susi tna Bydroelectxic Project includes the construction of two dams, Watana and Devil Canyon, on the main channel of the Sustina River, The Watana damsite's drainage area is 5,180 square ritiles, the Devil Canyon"s, 5,810. The project is designed to provide electricity to the Railbelt area of Alaska for well into the next century and to have flow regimes that will mini-mize impact on fisheries and other environmental resources, The? proposed Watana Reservoir will have 9.7 million acre Eoet of storage of which 3.7 million will be active storage. The Devil Canyon Reservoir will not have any significant amount of active storage. Since pcwer requirements are greatest during the winter, the project is designed to have as much water in storage as possible by September of each year, During visits by Eugene i. Peck and Thomas N. Keefer to Anchorage, Alaska, and Chicago, Illinois, in December 1904, discussions were held with representatives of Warza-Gbasco and the Alaska Pc~der Authority (Mr. Eric A. Marchegiani). The following informat ion pertinent t3 hydrological forecasting for the project was received. Based on the discussions held in December 1984, the following hydroLogical forecasts are required for the operation of the project: (1) Forecasts of seasonal May-September inflow to Watana Reservoir (long-term forecasts), (2) Forecasts of inflow to Watana Reservoir for two-week period;^ (short-term forecasts] . The long-term seasonal forecasts are necessary to schedule releases from the reservairs to ensure the maximum water storage by September. These forecasts are r.equi red at semi-monthly intervals starting on April I b~t possibly as early as January 1 of each year. Each forecast should be a probability farecast for the May-September runoff (or for forecasts after May I Eor the period from the date of tl7e foreeast througl~ September). These forecasts, used in conjurlct ion with standard reservoir operational rule curves, will permit the reservoir operators to plan for naving maxiinurn water storage in the Watana Reservoir prior to the winter. The short-term forecasts (inflow to Mc~tana Reservoir for the ne,:'r. two weeks) are reyuir~d Slai1.y from May throucjh Septemis6:r. ,z . st L;z::ei.asw swill ~rovide "ihe necessary infc?xn;?tion "c \jdj k- b; P_j % - sca!:onal inf Itow Eorecasts, to prevent enc~:oachme!~i 3r. dam " - "" "*-** p> - "I ..... 1 udrd req:li~;emeflirs - and to sa"iisfy doxnstzeaa i- iver *Fp-iX '1 - . C~L~J remenks v~i'ch a mirlimum reduction in storage, --a Ij-h 4 e k~:, 2:p(3t se and Scope a1:.. " AS stuljy is to provide information on hydralog ical forecast r:?4?:5i-,ods (and on basic data requirements) that will help the ,TI% l z s ha *.- Power 8ulii~;ri ty develop an effective farecast inq service - ',- ? 4- -&:GC will ensure that the project's objectives are achieved (i. 6% ,-- -- to maximize the use of the reservoir" inflows for energy genexati~r~ and to provide flow mregimes in the reach of the river belo%> the reservoirs that will minimize the impact on fisheries arid sther environmental ~~SOUTC~S) . The scope of the study is: (I) Ta review sets of alternative forecast syst~ms that could be used for operational forecasti~g fog the project; (2) To evaluate each method's ability to provide the required type, accuracy, frequeczy and lead time of fsreeasts; and (3) To p~ovide information on the basic data re- quirements associated with each alternative forecast system. The approach to the study is: (If To rneet with Harza-Ebasco and Alaska Power Authority personnel to more fully understand the needs for the operational forecasting service; (2) To meet with state, federal and other agencies in Alaska to evaluate the benefits that can be achieved from cooperation with such agencies for developing and operating a forecast service; (3) To review the methodologies and the basic data requirements for a forecasting service; (4) To review specific forecasticg methods and evaluate their usefulness far the project; (5) To provide a list of alternative sets of forz- cast methods together with information on the accuracy, sensitivity and data requiremenks for tzheir op~hrational use; and (6) To present a draft of z final report and to meet wi?h Hazzil-Ebrtsco ofEi:~i~Ls to revie?w the draft and sublnit a final xeport, *- U A j: scuss jon of the basic data needs for preparirrig 5'-or" ihfqid 1 sr.g-- term hydrological forecasts for the Susi tima Bas! n and r;:;eci f ic rceomn~endations for cllanges in the existing data collection network are presented in Chapter 2, Forecast model technology is discussed in Chapter 3, with reviek~s of alternative corecasting methods. Enformation aad recommendations for an cpe;:ating hydrological forecasting service for the project and (3 d.LuLussions ; c s4 on collection and psocessing 3f dat.a, computer and manpqwer requirements and cost factors are presented in Chapter 4, The executive summary foilowing the Table of Contents contains recommendations for collecting and processing basrc data foe the operational forecasting service for the Susi tna Hydroelectric Project. Durirlg the week of December 10, 1984, Eugene L, Peck and 'i'homas N. Keefer visited Anchorage and disccssed the preserlt da. ta ccllection networ;~ of the Susitna River Basin with many agencies anCi a~ganizations in Alaska. They also surveyed much of the upper Susi tna Basin by he! icopter. Considerable information was furnished by Harza-Ebasco prior to the field trip. Jeffrey He Coffin, H & M Consultants, Xnc., furnished additional information with his letter of December 19, 1984. Hydex personnel were impressed with the attention given to Cne develapment of the data network for the basin. The cooperative attitude expressed by federal and state repre- s~:ntatbves, by R & M Consultants and by personnel of Harza-Ebasco was appreciated. The major meetings held during the visit in Alaska with brief comments on items discussed are listed below: General overview of the project and the cooperating agencies was given by E. J, Gemperline al!d Khalid Jawed. Other Harza- Ebasco personnel participated. The primary discussion items cere forecast needs, wildlife habitat and other environmental requirements. Wayne Dyok provided information on power requizements and the reservoir levels that can be maintained at different times of the year. Eric Marckegiani, Alaska Power Authority, provided specific information on forecast needs. It was clear that the primary forecast need is for accurate estimates of the seasonal inflow from early in the calendar year through August 1. A secondary need is for short-term wa*ter supply forecasts (inflow to reservoir for next two weeks) from May through September. No specific need is foreseen at the present time for river fo~ecasts for ti^^ river below the Devil Canyon Reservoir, It was felt that the operation of this portion of the river could be accomplished by krlowledcje aE reservoir f ei.ea~e~* M Consultants, Jeff Coffin and Steve Bredthauer advised us of their data collection program and provided information on R & Mgs overall activities, They have considerable first: haild knowledgo of 'rile Susitna River Rasirl and the dai:a collection program, Clagett provided complete records of ail $no;; survey %-:a~--.rds L b % s-~ "ihat haye been collected by the SCS in A1.a~ki.i~ Geor~e is ;;:.lowLedgeable of snow conditions in Alaska and is very wil1irr.j tJ t,LId cooperate. George classif ik2d each snow survey course in "LA- Susi tna Basin on the basis of representativenes~:. D, Alaska State Climatologist. Jim Wise, Artic Environmental Information and Datz Center, supplied a photocopy of portions of an annual precipitation map fo -6- L Alaska that covers the Susitna River Basin, The office has printouts of all climatological records that have been collected by the National Climatic Center in Asheville, N. G, , an6 provided copies of records for some stations that operated for a short period of time in the Susitna Basin. E, U, S. Geological Survey. Raymond S. Geozge, Larry Leveen and others discussed lnt-eeent creamfl flaw measuring stations and the proposal to establish a station upstream of the high water level of the major reservcir, F, Nat ianal Weather Service, The hydrologist in charge, Jerry NibLer, and members of his staff and the meteorologist in charge, Edward Diemer, provided several reports and personal information on precipitation, snowfall and thunderstorm climatology of Alaska. These reparts will be of value for understanding the precipitation regime of klaska. Jerry also provided complete sets of parameters for the eleven basins in Alaska that have been calibrated using the Platianal Weather Service River Forecast System's (WSRFS) canueptual model. C, University of Alaska. Will Harrison, glaciologist, Geophysical Institue of the University of Alaska, presey-d an interesting discussion on glaciers and their relation to streamflow and sediment discharge in the Susltna Basin. He disccssed the effects of surges in the discharge of rufiaff from the qiaciers and the resulting very large changes in the sediment discharge. Fie supplied consi- derable information on the lo-.lg term changes in the glaciers thai-. indicate ttle glaciers have, cn the average, been wasting for th.2 last 30 years and therefore 3 substarltial portion of the streamflaw of the Susitna River has originated from this source:. From his observations of thf2 glaciers and snow cover on the glaciers d~ring the past three years he llas computed the wlnter and summer balance (loss o: gain af the glacier during the season) and the annual bal anee. T';.iese estimates ind cake that some of the glaciers in t31e Susitna Basin have grown while others appear to have wastled during the past three years. rn . ;.; ie ,.. information f~rn-~shed by Dr, Harrison, especially cn !:hi? pc::csntage of the stream discharge for specific sub basin:^ tha T: = a p:zr.iglniite from the glaciated areas, will help in evaluating i.-izr iaus r~odel s for forecasting snomel t runoff, iiowever , GUT upinzorr, a,:3 one that is apparently shared by Dr, Harrison, is .chat his observations are more impoetant fax estimati.ng the r) lr" sxreuir of glacier contribution to the long-term (33-year) strearnflow "clan for ilse in forecastin9 water supply during a specific seasan. In establishing a basic data network for a hydrological Corecasting service several factors are importznt. For precip- itation and sncw cover measurements two important factors are sometimes not given the attention they deser-~e for the data to be of maximum value, The first, and most critical of these factors, 4s the effect of wind (gage exposure) in reducing the gage catch and causing variation in the measurements relative to the true pr~cipitatlon at the si%* The second factor is the need to have the site "exposed" to the general paths of storms so that the records represent the average true precipitation for the general area {representativeness). Information on these two factors are presented in tte follvwing sections. A, Gage Exposure Hydrologists have long recognized that deficiencies ex ist in precipitation measurements, especially for snow fall. Errors in precipitation data account for a Large portion of the inaccuracies in precipitation-runoff relationships and in the simclaeted streamflow from operatianal hydrological modelas, Our inability to evali,,ate adequately the areal average water equivalent of the snow cover is the greatest limitation on inlproving the re1 iability of silomelt forecasting (1) . Snowfall is the most difficult form of precipitation to measure. Most precipitation gages have an orifice that is exposed in a horizontal plane for intercepting failing precipitation. The effect of wind on a gage is the largest: cause of measunement. error (2). Wilson (3) demonstrated that snowfall and snow cover measurements that were most representative of the actual snowfall had the highest correlation with streamflow measurements of small basins in the high Sierra Mountains, Brown and Peck (4) showed that the reliability of a precipitation measurement (snowfall) was related to how well the site of the gage was protected from wind movement. Lai~~on and Peck (5) showed that; the accuracy of conceptual models for simulating s~omel t streamflow was enhanced when precipi tation measurements are adjusted far bias in catch due to %find action on the precipi.tation gage. Figure I, from Larson and Peck (5) , presents a summary of gage catch deficiencies that have been observed in measuring rainzall arid snowfall jn relation to the average irlind measlired at tkje gage or i Eice, r*- * L" i2 2 precipitation measurement problez h%:s receivel consi- ciitabi~ attention by international melreoroLsgical cind hyd~o- li~j ica2 organizations, Dux ing the 1912 World ~etearolog ic.1 Ora-ri%n-izetionn symposium on Distzibution of Precipitation in v m >!i.iini:alnaus Areas he1.d in Geilo, Norway, "ce pprbblsrns uf w* - i.tescsl- ing precipitation in mountainous areas were discussed (61 . 1'fB ;.- -.tie conclusions stated that the mast xeP iable and consistent: i;?.i.ascremen"i those obtained at sites th~t are well protected from adverse wi 3 action. It was also repo~ted that adjusting preci;>itation records using wind measurements was a good method t Lu sr reduce the bias and variability induced by wind on the 6, OPX fxee,, F*?easurements of the average water equivalent of the :snow cave*: may also be greatly affected by wind. Snoi~fall on the site may be blown from the location by very low (less than 5 mph) rgind speeds, Sites located at the edge of forests or other ccver may col lect amounts of snow greater than the average snowf all. ii. Representativeness of Precipitation and Snow Cover. Heasurements, AL?:hough it may seem obvious, it is not always recognitzed that :cipi tation (snowfall) and snow survey measu~ements'~ are most valuable for forecasting snomel t runoff whon the k measurements are representative of the true snowfall at the site, Ideal exposure for observational sites for other meteoroloyi- cal parameters such as wind, temperature, humidity and evaporation are different than those for precipitation. For these measurements the best sites are those./that are sanewhat open to wind movement. f B For snowfall and snow survey measurements the gaging sites should be away from the immediate influence of trees, buildings and water bodies and in such a position as to afford a fair representation r E surrounding condtions. A station should not be ~ited upon, or close to, steep slopes, ridges, cliffs or k~ollows. Pol- some purposes (e. g., modeling crop conditions) the station location should be situtated to lneasure the conditions in the agricultural field. However, for use in modeling rjver basin eond itians the site location should represent average conditions i rr t:he basin rather than those of Local conditions. The Location af the site with respect to the general direction of storm movement during periods of precipitaticn is an important factor as to how well the preci~i tation at the site eelates to tbc average precipitation in tile gcirreral area. En this report this is referred to as the repres~ntstiveness of the si- te. Dased on experience in developing procethure sfor forecasting water supply for the mountainous areas of the western United States, Lo{::,tinq gages a"ci;ites that axe '%exposedw ta the gentzral direction of stab-m..; is an irnpor:tant factor for the data t.ci bf: of mos";\ri~lue, RPCXF~?~ @ark b)Oi;$CIe:". in dcvslopinq 8 +. s ,as ,- ,. . for nehwork design in mountainous zreas hzz also \. 'a, Iemonstraf.ed that a measure of how well a site is "e::~~~~ed'"-*~ ks,3 {jt5:?e1:31 storm movement is an important parameter fcr *\ -!ldez.s"ianding the relation of average precipitatl-on at one Z r~cation with average values at other locations, h \ In summaryi for pxecipitation or snow survey information to \-, be cf most value for forecasting the seasonal sncwmelk i:unoff in mountainous areas, it is important to know the relation af the rrecipitation catch to the true precipitation and how w~iL the s; Le is '"exposedw ((representativeness) to the general d irect ion of ansv~rnent of ~'k~rms, For high elevation basins like the Susitna there may not be zany sites that provide proteetion from wirld and are also wel.1. F W. exposed" to the general direction of storm movement. In the f~ll owing discussians the need for protected and re2resentat ive sites k~here precipitation and snow survey measurements are made is a major consideration. Consideration is also given to the techniques for adjusting precipitation measurements to correct for wind effects and to obtaining enhanced knowledge of precipitation at high elevations by using meteorological parameters to extrapolate measuremtnts observed at lower elevations. The advantages of remote sensing are also presented. ~ 3, EVALtlATPON OF EiXXSTPMG DATA COCLECTZOH FROG The data requirements for the forecasting service depends to same extent sn the models selected, Haw wcZl the data eollecyt2d meets the forecasting program needs depends upon the network design and on the accuracy and representativeness of the sbsgrvations, ~ A. Data Requirements for Long-term Seao~al Forecast iny Data requirements far preparing long- term seasonal water supply forecasts may be different than those for short-term forecasts during the May-September period, For seasonal farecasts issued on or before May L a knowledqe of the winter precipitatia~ (and/or water equivalent of the snow cover) Is required. Comments on the usefulness of the measurements fccs the established meteorolag ical stat ians in tl.15 basin for lt~.:~q~-- term seasonal foreeast ing are given below. (1) Air Temperature. Ternperatur~ measurernoilks are rjot specifically required for most forecast mode2.s unless there are periods of winter snowmelt, Tem- perature measurements have been used along with wind movement for adjusting precipitation records for deficiencies in gage catch (7), Obse~vatisns now xecerded in the basj-n axe $!YE$ maximum, minimum and mean air temperature in deg.:ses @czJ-sius, (2) Wind Measurements. Like temperature, wind iileasure-. ments are not required for forecasting long-term seasonal runsf f, Wind measu~ements are used to adjust precipitation records for def ieiencies in gage catch and ta estimate precipitation at high elevations using meteorological models, Six daily values for wind are recarded at the meteorological stations: average wind speed in m/s, maximum gust speed in m/s, direction of maximum gust in degreesl zssuitant wind direction in degrees, resultant wind speed in m/s, and pre- vailing wind direction to 16 points of the compass, (3) Moisture (humidity and dew point] Measurements. Generally models that forecast seasonal water supply generally do not require input of moisture measurements. Measuremenrs of moisture could have some value for computing sublimation in snowmelt models. Daily values of the mean daily relatire humidity in percent and the mean daily dew point in degree Celsius are recorded at the meteorolog icaL stations, '. -. (4) Precipitation. Precipitation, along with measure- ments of the water equivalent of the snow cover, is \ \ the most important meteorolog ical input for models '\ forecasting seasonal water su~~ply during the winter ,a \ Preci.pitation is generally not measured at the meteorolog isal stations during the winter \. aycept oceasi.onally for the month of April, ~iecipitation measurements are available from the Wyoming gage at the Wa tana rrleteorolog ical s"lation. (5) Radiatiqn Measurements. Measurements of sal.ar (short wave) and long wave 2:adlation arc? not generally required for seasonal wcl ter supp'ly forecasts m~de during the winter. The nftec2 for rad iatioo measuretnenksr especially dilr ing the winter, is not clear, division of the Naticna!- Weather Service, NORA, h:a3 19,ade no evaluation of the accuracy or reliability of this instrument. The following facts sake the ~ePisbility of the recorded data during the winter months of questionable value: (a) The information on the pyxanomewer states that the cgsine response uE the instrument is true over a range of 70 degrees, it is not clear if this is 35 degree each side of vertical or 70 degrees from the vertical. In any event, during the winter months the sun is at a very low angle and the response of the instrument to the solar radiation would not be accurate, (b) Most of the instruments observed during the field survey were covered with snow and ice. This prabably occurs ofter~ duxing thd winter esp2cially following periods of srlowfaLl and the sensors remain in this condition until the snow and ice are blown off , melted off, or are removed duuing an inspection visit. The brouchure on the instrument furnish- ed by R & M Consultants provides some information on the effects of temperature on the change in output for constant radiant input. However, this information is furnish- ed relative to 28 degrees Celsius and only dawn to O degrees Celsius. How the instru- ment responds to the very cold temperatures of Alaska is not known, (6) Snow Course Meesurrmsnts. The depths and water equivalent af the snow cover are read near five stakes that are installed in the ground along a line extending north from the instrument shelter at some of the meteorological stations. Because these si tes are generally open to wind movement, the depth of the snow cover is primarily limited to the height af the stubble along the snow course. Except for those stations which have natural protection from wind movement: (such as at the Sherman station) these measurements are of no value to the water supply forecasting program. Sojne aflotl courses estatlisi-ngd by Soil Conservation Service and by & M Consultants ha.. aerial markers tilat are read frrtrr. :;e'licapters or light aircraft, This techciaae bas proven success-* ful ic other high elevation snow areas where same water equivalent measurements for computing scow density are available at similarly located sta- tions. The value of the readings is related to tile windiness (exposure) of the aerial marker snow stsite site. If the immediate area is protected from iqind movement, such as by trees or by terrain featcres, the readings can be of considerable value fur seasonal water supply forecasting, B, Results of Field Survey. Comments on exposure of precipitation and snow survey measurements sites visited during the field trip on December 13, 1984, follow. The exposure classification for the immediate area of the gage or snow course is based on the system developed by ?xuwn arid Peck (4). (I) Watana Camp Meteorological Station. Exposure. The site is open and is classified as a windy site. There is a slight rise in the terrain ta the north of the measurement site. The location is exposed to the west and the site is cansidered as representative for storms in the lower portion of the basin, Pzecipi tation. A Wyoming precipitation gage has been installed. However, the maintenance of gage is poor. The wind vanes dip in the middle of each section and the northern seetion was disconnected at top and hanging down. The precipi- tation gage catch in properly constructed Wyoming gages is considered to be more consistent (approx- imately PO percent of true value) than a gage in a windy location without such protection, Snow cover survey. Records of water equivalent or of snow depths Erom the five-stake snow course at the station are of little ox no valve far a water supply forecasting program. (2) Marlahan Flat Snow Co~~rse Station (SCS) Exposure. The present sic@ of che snow course is in trees southeast af lalre, The classif icatioil of the exposure sf the site is rated as fair1.y wstl protect;ed t:o moderiltely windy, Prior to 1983 the SCF sllow course ~23s located in i mucir inove ogotin site arid khprcforr rec(~r(js for chi? f;r~io periods arc? not $+* p am* f- . fSieasurerqen,s 'nave bg>e!l zade both sj-tes, ::II published recaras arc C L Erom a-e -: 4"Q The SC.S plans to publish22 sa~ustec! va?;cs for the entire period when a sufficien;; overlap of data is availalzie. Snow appezrance in 3x23 during the field sarvey indicated that the wind movement at the Monahan survey site is much less than 5 miles to the south near the De~iali Highway. The station location is representative for the central portion of the upper Susitna basin 13) Susi tna Glacier Meteorological Station Exposure, The measuremr:nt site is sn a ridge be- tweerr two Eoarks of the Susitna Gl.acier, 'The primary drainage wind movement is below the eleva- tian of the station. Rcrwevez, during storms, the site could be subject to strong wind movement. Therefore the exposure of the gage site is classi- fied as windy. Snow course measurements at the site would be poor. The site is open towaxds the southwest. The site is considered good for representing the upper portion of the basin. In factr true precipitation os adjusted precipitation values from the site would be nearly ideal for lrepxesenting the high elevation area of the basin. (4) Caxibou Snow Course Exposure. The site is not protected from wind accompanying storms but is protected from genexai winds, especially the major drainage winds in the area, The site is classified as fairly well pro- tected to windy. The measurement site is located in a saddle at a hig:~ elevation south of the Susitrta Glacier. The snow course information is probably as good as one could obtain from a site at: a high elevation in the basin. The site is opened to the soutl'lwest and measurements should be representative of snowfall at high elevations in the basin. (5) Dcaal i Mcteorologicar Station Exposure, The gage site is completely open with no protection from wind. Due to this and the fact that the site is subject to strung drainage winds, the gage site is classified ss vsxy windy. The precipitation gage was in need of rerdair and was also unlcvc+1, Tila prc~ipi"c:tbni? gage in use does not have sufficient strength fox the winds experienced at this site, Snow course data observed at five surface stakes e3rtcnding north from "Le instrument shelter are considered useless. A location a half mile to the northeast of the present site would place the gage in trees and out of the path of the strong drainage winds that originate in the glaciers upstream from the station. Relocation of the gage would reduce the bias and variability of the precipitation mea- surements in relation to true precipitation. The location of the present site (or the proposed one) is considered ta be such that true information on the alztual precipitation at the site would be representative of the precipitation at the middle elevations of the basin. (6) Tyone Meteorological Station (Discontinued) Exposure,. The site is in thinly scattered trees near the Tyone River. No strong wind movement was evident from snow patterns. The gage ~ite is classified as fairly well protected to moderately windy. The measurements would be considered as repre- sentative of a large portion of the lower drainages of the Tyone and Osbetna River Basins. (7) Kosina Meteorological Station Exposure. The location of the gage is open with little or no pl-otection by trees. The appearance of snow in the area of the gage indicates fairly strong wind action. The gage site is classified as windy. A location to the north at lowar elevation in trees would provide a much better measurement of precipitation and snow cover. Accurate measurements at this site would be representative of a large area of the middle Susitna Basin around the proposed Watana Reservoir. (8) Devil Canyon Meteorological Station Exposure. The station is Located on a small ridge with protection from wind movement afforded by the general terrain. The gage can be affected by wind during snowfall. Although the exposure of the present gage site is fairly good (classified as fairly well protected) relocating the precipitation gage just to the north off the ridge an^? in trees would enhance the protection for the gage (to well protected) . The site is considered to be representative cf a large portion of the lower middle basin. (9) Sherman F'ieteorolog ical Station Exposure. The gage can be affected only by strong downdrafts during storms. The site is classified as well protected. In fact, because of the good proteetion the srlow builds up on the wind sensors, During the field visit the wind sensors were com- pletely inoperative because of snow and ice accumulation. Snow survey measurements at the site are considered to be very good. The station location with respect to general storms makes the site very good for representing precipi- tation and/or snow cover for the lower portions of the basin near the village of Gold Creek, 6, FieEd Bbservatisn of Wind Movement in Basin, The light snow cover that existed over the basin during the aerial suPvey on Decembes 13, 1484, pxovided an unusual opportunity to observe the general wind movement patterns of the basin, The snow cover for regions with light wind movement was swath and unbroken, Ir1 other areas the snow was blown into ridges or showed other evidence of wind action. The area having the most effect of wind action was approximately 10 miles west of DenaLi near Butte take.. Strong wind effects were seen south of Butte Lake as far as Deadman Lake. Flying north from Butte Lake over the Denali Highway towards Monahan Flat the wind effects became less observable. There was little evidence of strang wj.nci mavernest in the imimediate area sf Mornahart. Plat, Drainage winds originate over the Susitna, West Fork Susitna and East Fork glaciers and flow dcwnhill towards the confluence sf" the West and EasP: Forks af the Susi tnla River, The wind f'law tends to split near DenaLi with a major flow westward towards Butte Lake and the rest moving down the rnain Susitna River ck~annel, D. Evaluation of Snow Course Information The locrati~n sf all snow course sites (SC*- -nd R & Mj nlc4t visited during the field survey were located on O. S. Geological Survey maps and the terrain features fur each site were reviewed. Tkrf- stat$-ons have been well located and arc considered to be ri7cesentative of the snow cover in the general, area of each sitelr The following evaluation of exposures Eoz snow survey courses in the Susitna Basin was pravided by George Clagett, SCS snow survey supcrvisnx, based on iiis personal experience: Classification of Snow Course Exposures by George Clagett Snow Co~anse Classification Cathedral Creek (new) GPcarwater Lake [reactivated) Devils Canyon (aerial marker in trees) Fog Lakes Worsepasture Pass [aerial markey) Jatu Pass Lake Louise Malemute 2 (new) Plonahan Flat (and precip gage) Monsoon Lake (new) Square Lake Tyone River (reactivated) PJest Pork Glacier Watana Camp (Wyoming gage) pss r good fa FP fair fair exeellent -? fair ? exeelient good excellent 4, Data Requirements for Hydrological Forecasting. The data used in hydrolcgical forecasts can be divided inko two groups: the material required to develop the forecasting method (cal ihration or parameter estimation) and the informat ion needed to operate the forecast (operation). Data requirements depend upon the forecast method used, the time period of the forecas"cand the hydrological characteristics of the basin. Factors affecting the decision to use different hydrological forecasting models are discussed in Chapter 3, Model Technology, Although available data may restrict the actual choice of forecasting methods? the recommendations for changes in the network presented in this chapter assume that the project will support a newwork that will meet lcealistic requirements and that a physically based (conceptual) model will be an alternative (at least for headwater areas). Cal ibration of a model requires conventional time-ser ies o f hydrological information e p~ecipitation, temperature and streamflow) as well as information on constant basin and river characteristics, such as subcatchment areas, area of woodland, soiltype, channel dimensions and slopes, and, in the case of the Susitna Basin, on glaciers. For operational forecasting, data requirements include the hydrometeorological data specified by the forecastiilg scheme to characterize the state of the catchment irnmed iately befare the issue of the forecast and may include a measurement of the farecast elelnent itsel f for monitoring the forecast perform;lnce updating the farecast mt~ciel, anly the hydrometeorolog ical data requized to cal ihrate nd c]?s@rz k2 7-3 forecast models are discussed in this chapter, aecat~se he unraalkial climatic condi tiana in the SusHtna Wil~ex Bacil, special attention has been given to problems cf collecting accurate and representative measurements and to evaluating the e: ist ing collection network. P" 2 e3 ClimataPogical Data Reqni~ements for Model Calibration A primary data nced for calibration of a forecast model is time series of precipitation, water equivalent of the snow cover, s,ir temperature and streamflow. Care must be taken to insure tha3:here is not a bias between the data used to develop the forecast procedure and the data used for operational forecastiipigw Consiskeney of the records is as important for calibration as having records of sufficient length of the required data. The application af any hydrologic model to a basin requires exper lence and hydralog ic expertise for best results. W statistical model (commonly referred to as a black-box model) can be adopted to a basin without considerable knowledge of the hydrological characteristics of the watershed. The overall operational accuracy of a black-box model may not greatly differ whether applied by someone with experience or by one without much experience. However, E02: application of a conceptual model the operational accuracy from an expe::ienced hydrologist and f ram one without much experience can be greatly different. To best adapt a conceptual model to a basin, the modeler should understand the mathematical representation of the model and how this relates to the real world. In applying a conceptual hydrologic model to a basin the model representation must relate? to the xeal world conditioils. Thus, the components of the basic vatex balance equation Discharge = Precipitation minus Basin Lasses plus Change in Storage (1) must be xealistic. If this condition is not met: the calibration fpalrameter estimation) may be biased and the model, while demovstratitag a goad relationship for the calibration periad, migilt not be of much value for conditions not covered by the calibration period. This has been demonstrated in the paper "Advantages of Conceptual Models far Northern Research Basins Studies" by Eugene L. Peck and Thomas K. Carroll (8). This paper axso demonstrated that additional information not used in the calibration can not be used at a later time to improve the mode19s aecuzacy if the calibration is no: in line with real world cand it ions, To insure that the caxibratian af a nodeX in mauntainaus areas is irn $ir..ne with real. mx$d ce3r74dit ions IL;he model~r rnust have 3 5;rjad !;nowledye of the basin" average pl-ecipitatzicn. Tkij-c3 requirm.is accurate maps ( isohyetal) of ';he seasonal precipi k; z.i.03 A- - Lg-hat E~FS mast xelated with stxeamfffsw, At the present time ;here t3 'a"l 3- ~tjree 1' isohyetal maps of annual preeipi tation for the Susi tiia _11-0 + @:iver Basin, These were prepared by N'AA in @arch 1974 (9) by the SCS in August 1981 (10) 2nd by the State Climatologist for Alaska (11). As far as can be determined ,, the maps were c'reyeloped using only station means for %$hatever periods of r;.cord were available -- not specifically taking into account "iopocjraphic effects and probably not making maximum use of snow A good method for developing isohyetal maps in cold cl.imat:2 mountainous areas is the anomaly technique developed by Peck and Brown (12). In this technique the authors develop maps foe seasonal. periods during which storm paths are fairly consis~ent and siqqificant relationships can be established between average preciplcation valuer: and topographic features. These rela- tionships are used in preparing the seasonal isohyetal maps, These seasonal maps are graphically added to obtain annual maps, The technique has provisions for using snow cover measurements to provide additional estimates of winter season precipitation. Relationships between annual precipitation values and topographic characteristics are not generally adequate for preparing annual maps directly. A substantial portion of the May-SeptemGer streamflow in the Susitna River Basin results from precipitation that occurs during the winter months of October-April. The lnost useful application of any conceptual (and even statistical) model for forecasting the long-term water supply runoff for the basin could be achieved only with the use of an accurate and representative isohyetal map for the October-April winter season. Fleasurements of year-round precipitation are very Ximite6 for the Susitna Basin. The actual calibration of any model to the basin should be delayed until this lack of data is improved. As discussed in Chapter 3, the period of time that records are required for calibration depends to some extent on the range af conditions that are covered by the records available for use in calibration. Most ideally, as wide a range (from very high precipitation years to years with very Light precipitation) should be available. If a wide range of conditions does exist, four to five years of records may be sufficient to obtain a good calibration far a conceptual model, as many as 20 to 30 years of records may be required for application of a statistical (black bax) model. 6, Operational Data Requirements The primary concern ir~ forecasting with any model is the uncertainty of the forecasts, Measurement errors, model errors and the natural varriability o: meteorological inputs are causes nf uncertainty in the forecasts. Methods exist to evaluate the accu:acy of hydrolay ical instrumentation, to quantify the natural :-?yc?rolog ica? va-iabil ity of meteoroloy ical inpets ar:d to asse: c: ,- 'L iitr accuracy of the hydolog ical models by empir ical iy compar ing s i-nj-18ak:ed xest~lts with observed data, ii major source of uncertainty for forecasts whose lead tine i time f rorrl forecast issuance to observed flows) is greater than f~ bga .. ... tirne of concentration (time from occurrence of rainfall to observed flows) is the uncertainty in future weather conditions, especially the occurrence of precipitation. Thus an early seasonal forecast issued on February I has a high degree of uncertainty since precipitation for a considerable portion of the season has not been observed, Since the Susitna Hydraelectrical Project forecasting need is primarill for short- and long-term water supply forecasts from May tinrough September the primary data requir~ment is for measurements that adequately describe the magnitude and spatial distribution of precipitation (especially snowfall) in the basin. Accurate and representative measurements of the snow cover as of April 1 have been found to be very useful in describing the average winten: snowfall and for forecasting the runoff that results from that snow cover in basins where there is little snowmelt prior to April 1. This is in part due to the fact that the snow cover as of April 1 is generally near or at the maximum water equivalent of the snow cover for the season. Far forecasting seasonal water supply runoff on dates other than April L (using the April snow course data) , accurate and representative measurements of precipitation (rain and snowfall) are required. Therefore, for seasonal forecasts prior ta April 1, fur seasonal forecasts after April 1, and for short-term water supply forecasts, precipitation measurements are essential for an operational forecast program for the Susitna River Basin. Data required to support a hydrological forecast program in the Susitna River Basin are discussed by data type in the following paragraphs: a, Streamf low. Streamflow measurelnents are made by the U. S. Geological Sgrvey. Ideally, observations of river stages (for conversion into discharge values) should be available in real time for aLZ forecast points. b, Reservoir Storage, To compute reservoir inflow, lake stage values (for converting into reservoir storage) should be available in real time for 311 reservoirs far which inflow water supply forecasts are to be issued, c rS Reservoir Releases. As for reservoir storage, real time information on the amount of water released from reservoirs should be available for all reservoirs Eor which forecasts are prepared, The formation would also be required to torecast rivei- csnddtions belaw the reservoirs fo~ environmental and river operations, d, Snow Cover. The primary snow measurements re- quired are water equivalent for selected snow courses. The Soil. Conservation Service (SCS) has recornended procedures for establ isbing and oper- ating snow survey stations. Measurements of the water equivalent of the snow cover should be available either in real time or at set times durinc the season, The SCS also establishes snaw markers for reading the depth of the snow cover from aircraft. Some of these are in use in the Susitna Basin, The National Weather Service has an operational program to measure the average water equivalent of the snow cover by means of aerial gamma radiation surveys (13). These measurements provide average values of the water equivalent (or measure the sail moisture) of the snow cover for an area about 1500 feet wide and from 6 to 10 miles in length compared to the small area of a single snow course, Such measurements wsuPd be of value fox the Susitna Rivea Basin, e, rive^ Ice, Measurenents aE ice covex sn rivers is not critical for the water supply forecasting program. They are of value for other hydrological or environmental purposes. f * Glacier Measurements, Information on magnitude of snow and ice making up the glaciers in the Susitna Basin would be of potential value for use in the water supply forecasting program. Knowledge of the areal extent is an important factor that could be obtained Esom aerial (or satellite) photographs prior to the beginning nf snowfall in each season. Other factors that could be sf value include periodic measurements of the albedo of the snow cover on tho glQciers, measurements to determine the winter halance (what comes in during the winter period) and the summer balance (what snow and ice melts during the summer) and measurements to determinec'furing a seasan if a glacier is accumulating additional ice or is wasting. Except for measurements of the areal extent of the glaciers prior to winter and possibly remote sensing measurements sf the albedo of the snow GBV~~~ no othez i~easGrements listed above are presently available or useful for operational forecasting. Techniques for handling snow and ice melt from glaciers in an operational forecasting program will be covered in Chapter 3, Model Technology. r1 Trecipi3:ation. Measurements of preeip~ kai;ic5 slid snow cover are "the primary inputs far models that forecast water supply for basins such as the Scsitna River Basin, For seasunal forecasts ~SSC~P~ mi3fithly I oniy measuremsnts of monthly precipitation and/er the water equivalent of the snew cover may be suf t &cienZ to provide generalized outLooks prior to beginning of the snamtei t season, Short-tee~n forecasts of water supply (i.e. inflow to a major reservoir for the next two weeks) require daily measurements of precipitation (sncwfall and rain- fall), air temperature and forecasts of future precipitation and temperature conditions, The same measlrremer~ts would also be required to piaduee revised forecasts of the subseqcient resrcxvair inficw fsx the rest of the season, Meteorological. Various hydrological forecast models require metecrolog ical measurements of temperaturer wind, radiation and evaporation, 1x1 add ition, knowledge of meteorological cond i tions (i.e., parameters relating to precipitation and storm wind direction) can be used to improve the knowledge sf the magnitude and variation of precipitation in a basin. Remote sensing measure- ments of precipitation, snow cover and soil moisture may also be used as inputs or for updating some hydrological forecast models. Cornments of need far aeteoroLogical measurements are given below: (1) Temperature. Measurements of air temper- ature are required in most: snowmelt forecast models. Air temperature is also used in many conceptual hydrolog ical model s to est imaize the form of the precipitation (sol id or liquid). It is ass~med that daily temper- ature measurements (maximum and minimua temperature) will be required for the recommended opera tionaX pragram. Evaporation. Estimates of the rate uE daily evaporation in a basir, are not: generally required for season31 water supply fore- casting but ~k3y be used for short-term forecasting, Evaporation estimates can be based on pan measurements or derived from met~oralog ical rneasuren~ents of wind, tem- perature and some measrJre of incoming solar radiation (i.e,, such 21s percentage of cloud iness) , (3) Wind. Wind measurements $ie rhak used direce;ly ill most hydroioqical models, Mow- Ever, wind ar;asurert~ents car1 be asefu"!* in adjusting precipi tat ion m@F~surements to account for adverse wind a~tion that redu<zes the precipitation catch, Wind measurerc;entr may also be used in @$timating evaporat*inn and for computing wirrd surges ir, reservoi~:~. (4) Radiation, Some sncG me1 t raadela use radiation as an inpbc, ilowever, based on experience of tha National Weather Sf2rvjce River Forecasting Service, it is very difficult to obtain and ta tfiaintain can- sistency in radiation measurements far operational. forecasting, There 73rz alter- native methods for obtaining estimates of radiation. Unless there is a clear need for radiation measu:rements, their us~i should be avoided, (5) Meteorological Pclrameterr;. Upper air parameters can be used to enhance the knowledge of the magnitude and distribu- tian of precipitation in relatian to topo- graphy. Such parameters call be used for this purpose in conjunction with point precipitation measurements or with infor- mation obtained by remote sensing (e,g,, radar ox satellite sbservakisns) , endatisns for Changes in Data Callection Proyraa R & M Consultant Stephen Bredthauer's letter cE March 8, 1985, (Appendix El comments on the preliminacy reeommendat ions for changes in the location and operation of the meteorological statiorls in the basin have been considered in the followincj xeecmmendatisns: Caribou Snaw Survey site We agree with Bredthauer" srecommendation to operate a meteorolegical station at the Caribou snow survey si be, However, it is believed that May-September precipitntbon measurements at a site with similar exposure as that for the Susitna Glacier station would be very valuable for the prajeck, Tyone site c r We agree with relocating tn a site near 2 lake, ~f ~~~gerl $$ fl we feel the station should not be moved tocj far since its location fills a major. gap in ci-ie prapased ne::gqcrk. A. Changes in Winter Measureaient Program rq ii., $+-< a? location of the meteorolagical and snow survey (snow a"b .ak~+*~d.s2s A a * Ye sr:d aerial markers) statiofis are well selected tcl ~~,*cp:~*::~sent the var igus areas of the Susitna River Basinqb ~-~e~~~~~~ tbs aa3ata colfection program does not provide sufficient accuzacy i*P u, 3M quantity of information on the precipitation for khe entire area. As indicated in the previous sections same of the d3t.a beir:g cn3,Iected during the winter months are Rat required far $zu;?pnr.:t- of a water supply farecast ing program. Recommends t ions X LOX in the data collection program during the wrntcr mt3nths are: (9) Discontinue a11 measurements at meteorological. stalkions except precipitation, wind movement atid =i i Y ti2mpt2rature at: observation time. (2) Incraase the number of stations at which met- eoralagical data are collected during the win!:#er to those shown an Figure 2 and as indicated 43 TCobie 2, (3) Instail equipment for telemeteriny metearalogieab msasuzements% (4) Qevelap or purchase (see chapter 4) computer and software capability for collecting and processinrj the meteorological measurements azd to monitor and evaluate the data in real tia~e, (5) Maintain March I, April I and May 1 snow surveys at snow couxses and of aerial rna~icezs Ear the stations shown on Figure 3 and as listed in Table 3. Possible improvements in locations oE snow survey stations should he discussed with George Clagett, SCS snow survey supervisor, based on his evaluation of tht2 sdtes as listed in Table k, TI73 number of snow survey stations required far ti-re operational Eorecast ing may be reduced when actual forecast procedures are developed. fbos ina &fa t a %?a Devil Canyon 7:f~nahan Flats a r p+-R I t Discontinue, ~e a precxpx ta- tion station to operate Slay-Sep with similar exposure and distance a X to high mountains and not requlrlng aver-glac ier serv ice f l ight s. Relocate to protected site (precip itation, temperature and \*rind n?oxTe- ment) Reactivate at protected site (pre- cipitation,, temperature and s~ind movement) Relocate to protected site (precip- itation only) Rehabil i tate !?iyorning gage (~3reu.p- itation, temperature and wind move:- ent for water supply forecasting, other variables as required) Re?iocate to Nigh Lake (precipi t;-i-- tion only) Establish at SCS location (precip- itation, and temperarturc and w i nd movement) Esstablish at snow survey site (preeipi tation, temperature and wind) Consider installation aE Wyoming shielded gage. Establish a precipitation gage only in protected site with exposure to southwest and r:ot requiring over-glacier service flights for May-Sep period only. Establish at snaw cause lacatla~s in protected site if passiblc ar consider Wyoming Sage install- at ion, (preeipi tati<,n, tempera- t:a~:e atr~ti ib~i;*ind rnavej~er~t) Xs~k6: Ebauiwsle EstakjZ ish at snaw course Sc~catiaws in prarected site if possible (precipitation only) Stgw taite E~;t~i'k~~?IF,fst') at ssr-eow cejurse? lZ,caeat:io~m in praccrted site if possible (prcc i pi t;t t ion crn! y) . Sheg~~an Gt:a% ion ~S~~;.~C:IOO~*~.~~IPU~~ if tak>c+ded tor other purposes, Fiaore d 3 Snow Survey and Meteorolggica: Stations, Present 235 Pr~)pr,~~d ; Susitna Ri.*-- VGJ, Bt%~in --ulx, ------- --par*ri .a~rrl~~mv*mmw -mw-- *-m--- *----*- Snow Survey Courses West Fork Gbaeier Caribou Valdez Creek BsaPder Naxth Susitna Gbacie~r Main Fsxk East Fs~k %usjA ta Gf a@ ier Pyramid Jatu Bass Monahan Flat Butte Cxeek Cathedral Creek Clearwater Lake Tyone Rivez Lake Louise Horsepasture Pass Square Lake Fog Cakes Devil Canyon -- a, h,?-omnended ehailqrs in datz ccailectisn progra&: dur Zumrner months (?lay-September) . 'T'he fo~ec~ssting prugraro will require dai;y streamflow ii.Eormation for the Susitna River at Denali, the NacCaren River ;;ear Paxson anti the station to be establishzd in lieu of the +2i3sib kna River ilt. Cantwell. Recorninendat ions Ear tho- sclminer o;>cia kions for streamf low gaging stations are; (1) Ii~stall equipment at the three U, S, GeoLogicaI Suvey stream gaging stations listed above for telemeter ing river stages. When reservoirs are in operation, information an reservoirs stages and reservoirs releases will also need to be telemetexed, (2) Maintain the meteorological stations istnd xn Table (2). (3) Develop computer and software capability for collecting and processing the streamfLow and meteorol.oy ical data and to monitor and evaluate the data in real time, (4) Continue pan evaporation measurements at the Watana Camp, The equik2ment consists of a Class A point mear;jirenent evaporstion pan equipped with %a teil temperature sensors. A pan anemometer should be installed. The site should be an opea site and as free as 2assibie Eronr influence of wet areas, (5) Coalecfr at the Watana Camp station (and possibly at other sites arn%,nd the Watzna Reseirvoir) additional metearoiog isax measurementc required for other purpases. These may iaclude long and shcrrt wavlf* rahfi.stion measuzerirenta, humidity me~m~re~enks an23 aBdcKiti~naX wind mezls\,aretner:t%, insure that the fo1:ecasting proyra:n will. upr2rate at maximum eff icierlcy when the reslarv\2ir is complet~>f3, the "ci13wing actisns m:lrit be taken in the near :iuture tto have ariey~.~;it~ informatian fur cal ibratio~~ of the foramst models. The da td collect j.013 prograris xecommended in {:he pre-v iot.is sec"kon shuuLd be commenced as :ioc317 as pract iaablc , 8, Preparation of Seasonal Xsohyetal Maps Isohyetal maps (~ctober-Ap~~il, May-September and annual) should be developed for the Susitna Basin. The aaps should be based on physiographic relations and make maximum use of the large amount of information available on precipitation patterns for Alaska, This includes studies on meteorolog icaL patterns asso; late6 with snowfall (La) , studies on frequency of occurrence of thunderstorms (15) and studies on storm movements (16). C, Techniques for Adjusting Snowfall Measusements 'Techniques to adjust precipitation (snowfall) measurements using wind movement measurements to improve their accuracy and representativeness for use in the hydrological farecastinq models (as discussed in seetion 2 of this chapter) should be developed so that the techniques can be applied prior to calibration of the %208eI* s 0 9. Future Items for Consideration Depending upon the success of the calibration and operational use of the hydrolog isal forecasting models, othfsr actions may be considered in the future. These may include reviewing the use of satellite information to improve knowleclge of the temporal and spatial distribution of precipitation, the use of aerial gamma radiation surveys to measure water egirivalent of the snow cover over non glacierized areas and the use of meteorological models to improve the knowledge of precipitation at khe higher elevations of the basin. CHAPTER 3 ibfany excellent hydrometeorological reports on the Susi tna Basin have been furnished by Warza-Ebasco and no effort is made ZO sunmar ize the findings of these reports. Xowever, there are several factors that have an important bearing on the 1- irldroloyical . . regime of the basin that need to be given special consideration in selecting models for forecasting the short- and Song-term streamflow for the basin. A Data Limitations A primary consideration as to the type of model that can be used is the adequacy and availability of basic data for applying the model ta a basii. Th3 only hydrometeorological basic data tiohat has been recorded continuously in the basin for a fairly long period of time have been streamflow measurements by the U. S, Gealacjical Survey and snow survey measurczments by the Soil Consexwat-.idn Service. Records of other hydrometeorolog ical parameters (precipitation, temperature, wind, radation, etc.) have been colLceted by B & M Consultants in the basin since 1980 and are discussed in the Chapter 2. One result of a sparse data base is a limitation on the use of statistical models for forecasting seasonal runoff. Correlations between long-record, consistent precipitation and/or snow course obacr~~ations and streamflow records are required for t"n deeveLopment of thz.c;e procedures. The only records in the Susitna Basin of sufficeat length for developing a statistical relationship with the streamflow records are snow course me23surements. Even these are l imi ted in the ir usefulness because sorrle of the snow corjrses are subject to considerable wind actic~rl and sorne have been moved and do not pr~vitle a consistetnt index to the average snow caver, Correlations between April 1 snow course records and May- September runoEf for subdrainage areas in the Susitna Basin indicate that less than half of the variabil.ity (coefficient of 2 deteri%ination, r jess than 0.50) of the runoff is attributed to tl-rc winter snowfall. Marchegiani (171 found a coefficient of deterntination of O,6Z between a May 1 inclex (weighted April water equivalent from four snow course in the basin and the April precipi tat~ion at Gulkana, Alaska) and the April-Septembex runoff fur the Susi tna River at Gold Creek, Alaska, Since precipitation has been measured far only a few yeasx% in the basin it is difficult: to develop a statistical index procedure using ex ist-.i :jq records that would be an improveinent over that fuund by ~archegiani.. It ~;viclunt: from the low April b watc-r g+j .-<~ivalents A+-s u ., observred at most SCS snow cdurses and fronz the Lsi: e~efficients of determinatiun reported above that a Large portion I-X + che total runoff from the basin resalts from precipitation chr in9 the months of May through September, B, Glaciers Infosmation on glaciers in the basin is limited to a short 1 -a? p~eziod of time and no consistent: periodical measurements of the areal extent sf the glaciers, of the depth of finn ice ar of the perennial snow aecumulaticil on the glaciers are available. William Harrison of the University of Alaska and R & M Consula~ts (18) hajie determined that wasting of the glaciers during %he past 30 years was a cocsiderablk percentage of the average runoff du-incj that period. Recent calculations by Harrison (private communication) indicate that 32 percent of the runoff of the Sinsitna River at Denali is from summer melt: from the glaciers and that approximately 11 percent of the flow at Watana damsi te is from summer melt from glaciers. Larry Mayo, U. S. G~ological Survey, Fairbanks, Alaska, furnished the schematic diagram shown in Figure 5 depicting the relative values of the annual precipitation and runoff at various elevations for the Yukon Tanana Uplands and the Alaska Range are2 east and southeast of Fairbanks (private communication) . The upper line on Figure 4 indicates the total precipitation wbi-ch is divided into the arrlount that is from rainfall, snowfall and the amount that is perennial snow accumulation, The lower dashed line indicates the total water losses fxom the basin whieh is divided into the evaporation loss, snowmelt runoff, rain sunoff and the perennial ice ablation. Although the relative amounts of the various portions of the precipitation and runoff would be different for the Susitna Basin, the diagram provides information thzt is important ta consider in selecting or developing a rnodel for forecasting the runoff from the glaciated portion of the basin, The information on Figure 4 indicates that an the average there is no runoff from rain on the glacier at elevations above 2100 meters. Mayo has stated that rain water above this elevation is retained in the snow cover of the glacier and becames part of the perennial snow cover. This snow cover either csntxibutes .$o the rrunogf at a later time as snsmelt ablation at: becomes part of the firn ice of the glacier. C, Climatic Regime The drainage area of the Susitna Basin is subject to considerable difference in climatic r=onditi9=ssns, The eastelrm portion of the drainage area (lower elevations of thr E-lacLare-n River basin and the Tyone River basir?) are subject to a ~ontincn~ial climate as is the upper portion of the Copper Rivcr Basin (repesented hy the Gulkana Meteoralog ical Stat ion nf the National Weather Service), The cl irnat~? of the '1~:d~'r rt?;iehe:; of METERS EtEVAf ION with eleua- Figure 4. Distribution of prccipi tat ion anrl runocf =.- tion - Yukon-Tanana Dplands cast and southeast of !?a i rbanks, nlaska (L, Pfnyo) the Susitna River Basin west of the Talkeetna Mountains can be eiassli. fed as mnsd ifid mar itirne f represented by the Talkeetna t*?ereorolog ical Station of the National Weather Service) . Most of the drainage area of the Susitna River Basin above the Watana damsite is in a transition zone which is subject to both modified maritime and continental climatic cand i"%i~ns, The precipitation data that has been collected by R & M GsnsuBtants f~sm 1980 ts 1884 were reviewed $a evaluate the effects of the varying climatic conditions over the basin. Maps showing monthly val.ues of recorded precipitation for stztions located i:~ and near the upper Susitna River Basin are ?resented in Appendix R. R and M Consultants have indicated that measurements from the meteorological stat ions operated in the ba~in may not include all snowfall. In general, it appears that precipitation in October and April (the only maps available for the k~ir~ter months) is in line with the amounts that occur over the Cooper Ri.ver Basin and is representative of continental. c% imate eand itians, During the summer, the precipitation over the upper Susitna River Basin ia mare representative of a maritime climate. The Large amounts of precipitation recorded during some July and August months at the Susitna Glacier station (Maps A-4, A-6 and A-7) are a result of the inflow of moisture from the Gulf of Alaska, Studies sf thunderstorm activity in Alaska as discussed in Chapter 2 (15) ~Lso show that this area of the basin is snbjeet: to considerable thunderstorm activity in July, August and sometimes September, The thunderstorms foxm over the southern and western Talkeetna ~ountains and move to the norlheast, It is evident that a large percentage of the upper Susitna Basin, especial-ly the primary and west forks of the Susitna River, recieve a major porti.on of their annual precipitation during these summer months. Studies on the Gulkana Glacier east of the Susitna Basin have shown that fairly heavy preciptiation does occur on this basin during the late summer months. However, the upper elevations of the Susitna Basin may receive more surnmer thunderstorm precipitation than does the upper elevations of the Gulkana Glacier. One reason is that the high mountains of the RLaskan Range are oriented as to receive maximum precipi tat iorl Peom thunders"corms that move frwm the southwest and that originate over the Talkeetna Mountains. The mountains above the GuLkana Basin are also similarly oriented, Novever, the overall summer thunderstorm activity in this avea is probably somewhat less since the mountains where the thunderstorms originate arE I.nwer in elevation and are also more glaciexized, ' "*, Ronof f and Precipi tation-Runof f Analyses (1) Runoff Analyses. Daily discharge records from streamgag ing stat ions in the t~g~pei Susitna Basins were plotted for the summer months for the years 1981, 1982, 1983 and 1984. Daily precipitation as recorded elr the Paxson and the Talkeetna weather stations operated for and by the National Weather Service were also plotted for the same periods as the daily discharges. For each year the following plots are presented in Appendix B: Susitna River near Dendi, Alaska 2, Placlaren Mver near Paxson, Alaska 3, Susitna River near Cantwell by solid line and combined daily flow for Denali plcs Baxson by dashed line 4, Daily precipitation for Talkeetna and Paxson A subjective review of these plots for the Susitna River at Denali an2 of the temperature variations that occurcd each year provide a good view of the runoff characteristics of this basic. During the months of May and June rises in air tempe.:ature result in an increase in the snow and ice melt with only a small lag between the rise in temperature and a rise in the discharge, Following heavy rain durinq the later months of July, August and September there are also fairly rapid responses in the discharge. Far the Maclaren River near Faxson, increases in discharge resulting from increases in air temperature during May and June appear to be less noticable and generally start later. The rises resulting from rainfall during the latter part of the summer are i3sre in line with those for the Susf tna abwe Dena%i, RevSew sf the combined plots (Cantwell and Denali plus Paxson) sho\qs that the contribution from the ifitervening area below the upper gaging stations and above the Cantwell gaging station (indicated by the spacing between the dashed and solid lines) during May and June varies*conslderably in relation to the amount of runoff from the glacierized areas above the upstream gaging stations. Semilog plots of the daily discharges for May-September for 1981-E984 for "Lbe DenaSi and Paxsom streamflow stations and far the runoff contributed by the intervening area are presented in Apendix B (Figures B17 through 8-28). Review of these plots for May and June each year shows that the discharge peaks resulting from snowmelt at the Denali and Paxson gaging stations are fairly well correlated with possibly a longer delay of one to two days in the time to peak for the Uenali station hydrograph. iin intrrxestinq fact observed from the Cenal i and Pa:;sor; i>?ats is the gradual decay in the flows during August and S~~ptenber (day 93 fro day 153). The decay in runoff xesults f rain she deexease in the areal extent af kke snow cover, Peaks in the runoff resulting from nain are superimposed on the decay curve. 2-'" - ' ~'nss is sn indication that a snomelt madel that aceaunts for variability in the axe31 extent of the snow cover should model the basin snown~elt very well. (2) Precipitation-rbnoff relationships Studies on Probable Max imum Prec ipi ti2 t ion for the Sus i tna giver Basin were received f xom Haarrza-Ebasces (19 and 20) . These reports contained results of studies on unit hydrographs for subbasins of tile area and some application of the HEC-1 rainfall- runoff procedure. In the report prepared by Acres (20) the SSARR (Streamflow Synthesis and Reservoir Regulation) model developed by ?:he U. S. Army Corps of Engineers was adapted to the entire basin above Gold Creek for part of the years of 1964, 1971 and 1932, The above applications of precipitation-runoff models to the basira are sufficient ts demonstrate that the basin 'can be reasonab:ly fitted with models. The accuracy of the relationships was limited because of the lack of precipitation and temperature data fxsm within the basin, although it was not planned to fit any model during this study, in crder to better understand precipitation-runoff sebationships for the area and to obtain a knowledge of the relativc importance of different basic data, some statistical ,analyses were performed. These included correlations cf snow course, p4:ecipi tation and temperature data with streangag ing zeco~ds for the Srasitrsa River at Denali, the Maezlaresl Riven at Paxsan, the Susitna River near Cantwell and for the lower intervening area of the basin (the area above Cantwell gaging station excluding the area above the Denali and Paxson stations) . The August ~~noff of the Susitna River at Cantwell has a 0,77 correlation with the combined runoff observed during August at the Denali and Paxson streamgaging stations. Using the precipitation recorded during August at Talk ree$_~~a as an additional variable increases this correlation to 0,84 even though Talkeetna is over one hundred miles away. It is also interesting that the July-September TaLkcetna precipitation has a corcelation of 0.77 with the July-Sept~mbsr runoff from the lower intervening area above Cantwell. Summer runoff Ear the glacierized basins is not as well rt3lated ta precipitation observations. The best single cor- rglation with the May-September runoff of the Susi ~na River at Denali is with the April 1 snow survey data observed at Monahnn Flats (r = 0.60 or in coefficient at determjnation of only 0.36). This indicates that the winter precipitation accounts f~r 8% b:ilv a small. f;orfion of the Hay-Septembex runoff of the Susitna River above Denali, Other factors that influence how much runof i %:ill occur dillring the May-September period for this basin are: T, Amount of energy received in the basin duririg the summer months 2. The sun-iiier precipitation over the basin 3, The amount of snow carryover in the basin A multiple correlation analysis was made to determine if the effect sf each of the above factors could be faund using the ax4faialahle data base. Monthly departures from the average temperatures at the Talkeetna meteorological station for five inonths [14ay through September) were summed each year for an index 02 the amount of energy received during the runoff period. The May-September precipitation at Talkeetna was used as an index of the summer precipitation in the basin. The May-September precipitation of the previcos year was used as an index af the amobnt of snow that was carried over from the previous year. Most of this is seen as patches of sncw on non-glacierized areas at higher elevations prior to the snow cover of the current year. The data used in the correlation anlysis and the statistical results are shown in Appendix C. The final correlation using all four variables was 0,.6 1 and the separate contribution ind icated for each of the four variables appears to be significznt from the t and F test statistics, Although the correlation is not high, tne results are very good sir3ce the analysis was made using only seasonal values for data t-eeo~ded a very long distance from the basin, The results suppart the hypotheses that the winter precipitation for the Susitna basin above DenaBi accounts for abcut half of the total ranoff from the basin during the period from May to September. A trend analysis was made on the basic data. This gave a carwelation of 0.52 between the May-September runoff and time (year to 2 digits)) with a trend towards increasing amounts. This could be related to changes in the discharge measurements, changes in the amount of wasting of the glacier, a real trend towards wetter years or a combination of these factors. During meetings with the various scientists in Anchorage an interesting statement was made by William Harrisan. It was his observation that during years with very heavy precipitation there seemed to be less runoff than would be expected for *e Smsitn? River at Detilali, Pje also stated that there seemed ts b? a negative correlation between summer precipitation and the May- September runoff. Al thc ugh no conclusive ev idence has been develop--d to prove Harr ir;on8s statements, the resul ts of the correlation studies using monthly means values seein to substantiate to some degree his observations. In addition "'9 "^ k8tecipitati,sn recoxds cjbserved in the Power Susikwa Wasin ds carrelate negatriv4%ly ufith the May-September runoff of ",he Susitna :?ivef at Denali,. There is na way to prove the interactive role played by ~jrecipitation and temperatune in the Susi tna Rivez basin above Denzbl i without detailed measurementti from the basin. However, assuming that the zesults of the corxelations analyses made using zlvailahle monthly values are related to the true cond it ions, some observations can be made. When heavy precipitation occurs during the sunlmer nronths the cloud eover is much greater anfj the amount: of energy reaching the basin is reduced. Although the total runoff from the basin may he greater than it would have been with Less sumR7ex precipitation, he amount added to the perennial snow covey: above elevation 2180 cnuld be greater than the amQunt ablated. Thus the aniount of snow added to the perennial snow cover (iind on the notl-glacie.:~zed areas) during those years with heavy summer precipitation (and consequentlv proportional less snowmelt) should be much greater than durin9 years with light summer p~eeipitation. This supports the use of the previous May- September pzecipiati~n to account far the apparent increase in runoff during the current year. T1ne method selected for forecasting the May-September runoff of the St~sitna River at Denali and the Maclaren Rivet. near 2axson shot~ld take into account the importance of the summer precipitation and the carryover effects. If accurate and consistent measurements of the perennial snow cover on the glaciers and of the caxryover snow -over of the non-qlacierized areas of the basir: were available they could be used to account for this factor. Howevex, such measurements are difficult to obta i il.. Without such ~~sasurements, techniques can be develci?ed to hzlndle this effect for forecasting the seasonal water supply. This can be done by finding a variable, such 2s the previous yay- September precipitation, that is an index to the amount of carryover and correlating this variable with the errors in the farecas t procedure for the seasonal runoff, The factgsrs discussed in the previous section are the prim;iry ones to be considered in selecting the models for cse in the forecasting program for the Susi tna River Hydrnelectx ic Project.. Anokher important factor for consider.3t ion is the abil ity of ttlc. r~iodel to be updated using real time streamflod mea!;uremenfFs and/or otrher hydrorncteorol og ical measuremerrts tld,at zay become available in real time in the future. I. P* " ine first part of this section consists of a review of the modeis that could be used for forecasting short- and long-terxri xunof f in the basin, This inclt~des disci1ssi~ns on acctlxacy af 4- ** ,fie forecasts far the various models and an the seqsitivi ty cf the dodels to the basic data input. The tern '"modelling of hydrolag ical systemsqp generally means % 1- .4~ a appL ication af mathemat leal and logical expresrjiuns wjhich define the quantitative relationships between the flow chaxac-teristies (output) ar,d the flow-forming factors f input] . This is a very general definition which covers an entire spectrum of approaches. At one extreme are the purely empirical 4 statistical "black boxwP techniques that make no atteinpt to modelf the intorr~al structlare and response of i.he catchmerat: system. At the other extreme are techniques involving complex systems of equations based on physical laws and concepts, the so-called conceptual models, Bath %!~e statistical $Pack bax and the conceptual models axe deterministic models and any classification of a partic~l~ar model to one cl,ass or the other forces a decision on the degree of empiricism. Xn this review the terms black box and conceptual will be used to classify the models, Most deterministic models musue calibrated ts a basin using past records of hydrometeorological variables to determine a set of paramters to fit the model to a particular basin. In general, black bax models require a Longer length of record to determine the ~iost suitable parameter set. Por conceptual models some of the parameters can be determined directly from physical eharactcristics of the basin fie soil type] and from ana1yst.s of streamflow recolds (i.e., coefficients for ground water d ischarge) . Because the Susitna basin has a large percentage of runoff ~esulting from melting of snow and glacier ice, special attention is gj.ven to review of approaches for forecasting the melting of SI:OW and ice* Water supply forecasts for basins with considerable snow cover (and glaciers) are made by three basic techniques: a, SnowmeX t forecasts, Conceptual models c. Time series analysis. The modc1.s for forecasting the snowmelt will be considered first, follow~d by n review of conceptual models for forecasting the entire runoff from a basin (complete or in conjunction with a snowmel t model) . Many snowmel t model.s, use a simple degree day approach as an index of snowmelt, lqBDELS FOR FORECASTING SNOWMELT AND RUNOFF FROM CLACXERPED BASINS VAL Kss\ovalov 621, 22) an6 ethers in the USSR have -i~:vc?oped tec,?niques for forecesting snowmelt and glacier melt, - 2'ar r~ountainous areas, most of the models compute snowmelt by ale!i;lt ion zones and require eoi3siderable input of snow IneasGrements, FOP glaeiei meit forecasting, Konovalav (21) has discussed the rel~tive importance of knowing the ratio of the acc~niulation and ablation areas of the glacier. En addition, he has stressed the need for computing the sepatate estimates of the voiu~~s of melting of pure ice, ice under the moraine, old firn and a*.intem: and surmer snokf making up the total volume of melting from a glacier region. The data requirements for these ;%ppra.lches s.re deemed to be mote than necessary for the Susitna ??rojece, ~ (2) Models for Predicting Runoff from Glacierized Basins Wodrew Fountain (U. S. Geological Survey, Tacoma, Washington) ktas ptovided a copy of a chapter (23) prepared by him and Werldell J Tangborn (WyMET Company, Seattle, Washing ton) for an ~npublished report by the Working Group an Snow and ice Hydr~l~gy of Glacierized Basins, International Commission on Snow and Ice, International Association of Hydrological Sciences. This report, "Contemporary Techniques for Predicting Runoff from Glacierized Basins," provides comments on ,he advantages and disadvantages of several models and includes a table summarizing the main features of each mondel, Most of the models reviewed in the report are nesearch models that have been applied to only a single basin. Exceptions t-.o this are the model by E. A. Anderson (National Weather Service, Si lver Springs, Maryland) which has been appl ied operationally by the national forecasting services in the United States and Canada (including 11 basins in Alaska) and the Tangborn model which was reported as being applied to three hissins in Canada, (3) Eric Anderson Model {NWSFRS Snowmelt Model) The Sno9~ Accumu%ation and Ahlation MoSel of the National Weather Service (NWS) has been in operational use by the NWS since 1913 (21), Nibler, National Weather Service River Forecast Center, Anchor3gc, Alaska, has modified the model to include the presence of glsciers in the basin. T$& snowpack depletion curve (area of snowc3ver as a function of znowpack water equivalent) was altered so the minimum snowcovered area is equal to the basin's glacie~ ized area. This area is then assigned a snowpack thickness greater than what cou1.d be lost by summer ablation. Nibler reports that this modification $$arks satisEactori2y fiur estimates of seasonal runoff f ro~n glacier izeci basins, bii t it is d5sappointini~ for daily forecasts. Calculated runoff has a very quick and "flashy" response relative to the more slowly rzsponding cbsenved runoff, which he suggests is caused by neglectang storage faeturs of the glaciers. Parameter values far eleven basins that are forecasted operationally by the NWS in alaska have been furnished by Gerald NibLes {private c3nrmunication), For most of these basins the model has been appa ied with two zones (based on elevation) , Although the NWS Anderson snowmelt model has generally satisfactory zesults, certain deficiencies do exist. Insuffi- cient knowledge of the actual occurrence (and of the noriBa1 precipitation during the various seasons of the year) of precipitation at higher elevations c%n result in runoff volume errurs. There is also a problem in cia-sifying precipitation as to rain or snow based solely on rnax~mv3-minimum temperatures. This is mainly a peoblem in basins where both rain and snow freqiaently occur during the same event. The above are deficiencies in basic data and not: in the structure of the model. and wo d be applicable to all models. The Anderson Model uses temperature as the sole index to - sttowme1.t but nat in a degree day approach. Under certain temperature conditions the use of air temperature as the sole inde:r of snowmelt h2s proven to be inadequate. This can occur under clear skies with abnormally cold temperatures (under- predicts), under very warm temperatures with little or no wind (ovexpredicts) and with very high dew points and high winds (underp~edicts). The latter two conditions would not be a factor in thrt Susitna Basin. The first condition should not be a major arie because of the high amount of cloudiness that occurs during snowmelt period in the basin. In any case, these conditions affect th~ timing and magnitude of the peak runoff more than the total volume for a period. The Anderson model was reviewed in a report prepared by the authors for the National Aeronautics and Space Aministation (NASA:, a diagram from that report (25) illustrating the model is incl~ded in Appendix D (Figure D-1) . Also included in Appendix D are listing of the states and pauam~t~rs for the model (Table D-l and D-2). Informatian on the schen1;tic diagram and on the definitions of states and parramckers as used in the NASA report is included in Appendix D. The paxameters foi the Anderson model have been found to be related to cl imatic and physiographic characteristics and reasonable initial parameter values can be obtained from a kllowledge of typical wand i t ions over a watershed. A discussion nn the possible appl ication of hi s model to the Susitna Basin was he1.d with Eric Andersan, Anderson believes that fc~r ylacifzrlzed biisin~: "the daily forecast values of runcff wauld have mane vaxiance than obse~ved values but that this c.Efeci: would tend to average out for longer pelricds [such as for b7i t:k~a-week fore~ast) He also staed that reb% iable seasanal :naps 05 23:ecipi tation would be required to properly calibrate the mo2el. He agreed with Gerald Nibler that a separate zone should he ineluded and assigned a large water equivalent value for the glacierized area of the basin. With an areal depletion curve as used in his model, he believes that you would not need many zones (probably 2 but a maximum of 3). For models not having an areal depletion curve many zones would be required as is done in the USSR, Anderson furnished an unpubl ished report summarizing the findings of the recent comparison study on snowrnelt models discussed in the next section, Anderson has fitted his mqdel in Al?ska. An interesting result is the difference in the shape of his curves for distributing melt factors through the year as shown in Figure 5. In the contiguous United States the curve is a sine wave; in Alaska the curve stays at a minimum until it starts to rise sharply during April. Anderson attributes this to the differences in length of sunlight and is a reason why the streams in Alaska do not: have a significant increase in snowmelt runoff until after May 1 as may be seen in the report on Daily Flow Statistics of Alaskan Streams by Chapman of the NWS (26). (4) WE40 Comparison of Snowmelt Models During the period 1978-1983 the World Meteorological Organization (WMO) carried out an international comparison of canceptuai models of snowmelt runoff (27). 'The final report of this study has not beec released by WMO but comparison figures have beer: furnished to the participants and were reviewed in Anderson's office. Three models developed in the Uni ted States were included in the WMO comparison study and the calibration used in the comparison study were those fitted to eackl basin by the authors of the models, The unpublished paper furnished by Anderson was p1:epared for WHO by two Canadiazs (28). The statistics for the results of the cakibra"%;":on and verification runs far the "ezhsee Ui, S, msdels are summarized below and are considered to be more represents t ive than most comparison studies where the models are often calibrated by hydrologists who may not be experienced in cal ibrating the ~aodels compared. The U. S. models included in the WMO comparison study were: a, MWS Anderson Model b. The Precipitation-runef f Modeling System (PRHS) of the U. S. Geological Survey c, The Streamflow Synthesis and Reservoir Regulation Modcb (SSRRR] of the O, S, army Corps of Engineers 4 1 $8 & # * gj Figure 5 Seasanal variation in melt fitctors used durir.8 non-rain periods. Th@ statistical vaf ues aQ the simulated and obse~ved s::reamflow were compared in the unpublished study for ~:.gi?i lficanee using four statistical vau iables, A valtle fox each 02 the four statisticd vasiabfes was computed fur the r:zi ilsrat ion-complete year, cal ibration-snobnneltr period , vex i f ica- <:ion-complete yeax, and vex ification-snotg~nel t period) fonr each f the three river basins that were calibrated by all three of the 0. 3, models. This provided a total. of 48 cases. For each case, each model's results were compared with the results r" the model having the best statistical value f;r that case. 3 ?hat wore found to be as significant as the best model for that c ;e (withit? a 95 percent confidence interval) were considered to be as good as the model that had the best statistical value and were considered as a tie for first place. Tables showing the ranking of the models for each of the 48 cases were presented in the unpublished report. The following indicates how often each of the three U. S, models were ranked in first place; NWS Ranked first in all 48 cases PRMS Ranked first in 22 cases SSARR Rarrslsed first in 21 cases The Anderson model is the most physically based snowmelt model in use today and the results of the test program indieate its accuracy and reliabity for forecasting basins in different cl imates, The model was developed f ram physical concepts using detailed measurements of all measurable variabies yet was designed to use only standard measurements of preeipi ta tion and tentirperakure. B, Conceptual Soil Moisture Accounting Models Five conceptual soil moisture accounting models were selected to be considered as alternatives for the operational system for the Susi tna Hydroelectric Project. These are: Tangborn Model (HyMet) National Weather .C,irvice River Forecast System (NWSRFS) Precipi tat ion-Runof f Madel ing System (PRMS) Stanf srd Watershed Model (SiqM) Streamfow Synthesis and Reservoir Regulation (SSARR) Three of the above models we:re included in the review oE lnodels completed by Hyclex for NASA (25) and by the WMO (30) . This study provided a yaod methad to study and canpare models. Since informatian on three of the models is available in the form vlse+l in the NASA study it was felt it would be beneficial to have all mode3.s in !:he s2r9ae format, The at~thors 0% $ha two models tnat wexe nat included in th~ i4aSA study were contacted to obtain more complete information cn the podels and their structure. Wendell V. Tangborn (HyMet Company, Seattle, Washing ton) furnished infoxmation on the HyMet iaodei and George PI, Ceavesley (U. S. Geological Survey, Denver, Coio~ado) on the PRMS model. Schematic diagrams and gefinitions of states and parameters for the two models were prepared and* submitted to Tangborn and Leavesley for review. The schematic diagrams and lis- sf definitions for all five models are presented in Appendix D along with the same material for the Anderssn Snowlelt Model, k glossary of terms used in review of the models for NASA and a Legend for the seheniatic diagrams as ori2inally published in ,the NASA xeport (25) are reproduced in Appendix D. By developing the diagrams and studying the states and parameters used in the models, a reasonable understanding of the structure and operation of each model can be achieved. Based on this knowledge, and on our experience, comments on the models for possible selection for the project are given below. (1) HyMet Model (Diagram D-2, Tables D-3 and D-4) (a) Data Requirements The aperation of the model requires only the standard measurements of precipitation, temperature, (b) Field Experience The HyMet model is being used operationally for seasonal forecasting in the Pacific Northwest and in Central Wtiasna, As far as can be determined there is no complete documentation for calibration or operational use af the model, Short term streamflaw forecasts sf 1-3 days duration require weather forecasts of preci- pi tation arid temperature. Seasonal forecasts are based on regressing indices of the total water in storage (in snow cover, soil, surface and groundwater] in the basin on the day of ti7e forezast with subsequent runoff for a specified ti~ne period. The model has not been included in any af the WMO comparison tests and no results of comparison with other models have been found. The only available results arc those published by the author with little or ns supporti\lg inEormation for evalulating the model perfor-- n3:aan"ree, [d) Use in Snow and Glacierized areas It is evident from the Diagram in Figure 0-2 that the original Tangboxn model has been adapted 'for Lase in glacierized areas. Tangborn furnished 2 paper "Prediction of Glacier Deri~ed Runof: for Hydroelectric Development" (29) which presents the basis for the glacier protion of the model. The paper reports on the analyses made fox two small basins in Bnitisb Columbia, Canada. T12e assumption is ~.zde that englacial storage of water (from ablation of snow, firn and ice, and precipitation) occurs during the period from November L to July 14 and that outflow from this stored watex islt%ceurs from July 15 to October 31. From the information furnished on the model it was not clear exactly how the model handles the states in the model for firn ice and glacier net balance. Bscause the precipitation input to the glaciers in the Susitna Basin is considerably different than for glaciers along the west coast of ?J~r%h Werica (Less than haif of the inj?ut during the winter compared to a very large percentage for glaciers along the west coast of the United States), it is not certain if the adjustment technique for englacial storage as used in the paper would be of direct value for the Susitna Basin, For glacierized areas the HyMet model uses up ts five elevation, zones and snaw accumulation for each altitude is determined using standard precipitation measurements from one ar two selected weather stations, The snaw, firn and ice melt axe calculated from the mean ai~ temperature and the range of daily temperature, (el Updating A technique using tht. error far a short-term forecast as an index to the exrar in the seasonal forecast is used for ad just ing the seasonal forecast an the basis of comparison of simulated and observed flows. For example, a seasonal forecast prepared as of May 1 uses the exrosr in the simulated versus okserved flows during May LJ~ revising the seasonal forecast. This has meare vaLilae for basins where the major. portion of the input for the s9asonal flow occurs prior to the date of the forecast. It is no% ezrtain how it would wo;:k fox the Sasi tna Basin, The updating prucedure has not been incorporated into the shor t- term forecasting method (311, The soil mositure accounting structure of the msdel is sirnilax ts other madels and it is clear that the model was not developed with the concept of having states of tbe model to re? ate directly to measureable hydrolog j>cal variables in the real woxld, The~e is no indication that techniques have been developed that would objectively use remote measurements (i, e., water equivalent of the snow cover) for updating the model, (f) Operational Factors Unless the model was applied by the HyMet Company the lack of documentatioa for the model would be a serious Iimitatisn, Thexe is insufficient information ta determine how the madel would be used fox short and long-term forecasting. (2) NWSRFS Model (Diagram D-3, Tables D-5 and D-6) (a) Data requirements. The operation of the model requires only G tandaxd measurements. jb) Field Experience The NWSRFS is used operationally throughout the United States and Canada. There is complete documentation for the NWS River Forecast System and for each sf the models used, A report an techniques for catchment mo3el ing and initial parameter estimation for the NWSRFS soil moistare accounting model was published by Peck (3ZL The MWSHFS was i~cluded in the WM8 comparison study on soil moisture accounting models and although no formal comparison was roported on the results, a review of the published ecmpariaon indicates that the model performed as well or better than any of the models tested. The Hydrolog ic Research Labor3tory of t,he i\iWS did an extensive testirlg program in 1975. Three models were comk3ared with a teen prepared onlk- for prediction use where ax1 data for one year (or a long period of time) is imput at one time. There is nc operational model fox use in Es~ecastlng but Leavesley states that they plan to prepare one. The only known evdluation of the model that has been published is that reported above from the WMO snowmel t compar ison study, (d) Use in Snow and Glacierized Areas The model has a component for handling sno~xnel t runoff as may be seen on the schematic diagram (Figure D-4) 'and it is based on energy balance equations. The model has not been adapted for glacier areas. (e) Updating Since there is no forecasting mode there axe no updating techniques. (E) Operation Use The model is well. dacaniqsented (3"$ ). (4) SWN Model (Diagram D-5, Tables D-9 and D-10) (3) Data Requirements Same as far NWSRFS model /b) Field Experience The model has been used throughout the world by the autk~xs but it is not certain where the model is presently being used. The SWM was included in the test cs~lducted by the NWS Hydroingic Research L2')oratory and in the WMO comparison tests. In both studies the results were very camparable to those for the NWSRFS, Uf3e in Snow and Glacier ized Areas TI"$ model has been used in ~120~ melt areas but no information is ava'lable for glacier ized areas, 'S"k~e i."sndcnr~ion mode% is used for snowme1 t model irig , [e) Updating The authors have updating tect:rliques us in2 observed streamflow but no speci f it informat ior; is available, (f) Opexat ional Factors Similar to those indicated for the NWSRFS, (5) SSARR Model (Diagram D-6, Tables D-TI and D-12) (a) Data Requirements St.awdard measurements, (b) Field Experience The SSARR model has been ixsed throughout the world and is used ~peratj~onally by the joink WS-Corps of Engineers river forecast center in Portland, Oxegon. Excellent documentat ion exiets foe the model, The SSARR model is primarily a river routing and reservoir regulation model. The components for its soil moisture accounting and sno~melt components are not as conceptual as for the other models considered. Tests conducted by the NWS and the WMO (on both the soil moisture aceaunt ing an3 the snowmel t compar isons) have shown the model to have more limitations in forecasting for individual basins. (d) Use in Snow and Glacierized Areas The model has been used in many snowmel t areas. There is no component for accounting for glaciers, The model has been used ta calibrate the Susitna River Basin for the months of August and September using standard tables by Bredthauer of R and M Consultants. Some subjective methods for updating the model operat,iooally have been used. However, the structure cf the model (large use nf tables) makes it very d i ff icul t to develop procedures for objective updating the model, Rnalysis by Hydex for NASA has detertniner3 that the use of (1") Operational Factors There are no known problems Eor using the SS?lRR model operationally. Knowledge of the sour-es of error that would occur in forecasting short- and long-term runoff of the Susikna River i:: ?:in important factor for determinin, the forecast models that sl3aukd be used, Possible sources sf foxecast errors are shown by the schembtic diagram in Figure 6; rbese are model, basic data and cl ima tolog ical, !%ladel erroar results from the fae"i:hat the madel (and its parameterization) does not accurately represent the real world, The magnitude of the model error in forecasts of short- and long- tern seasonal runoff would be apgroximately the same for either a statistical or a conceptual approach, The accuracy of a seasonal forecast for water supply depends on the ability to knew the true basin averages of hydro~et~orologid parameters used in the model. Even if a perfect model and a perfect set of parameters were available, inadequacies in basic data would introduce i.al:ge errors in thc. forecasts of the May-September runoff far the Susi tna River Basin. The basic data error For the winter precipitation (due to fairly good correlations agnong the April I snaw course measurements) is probably less t.han the data error fop forecasts du2: ing the summer thunderstorm period. Improvement in the knowledge of the actual precipitation by proper locating gages, by adjusting precipitation measurements for gage catch defi- ciencies, and by improvements in the network (additional stations and representative exposures as recommended above) all help to reduce the basic error component of the total forecast error. The climatological poxtian of the forecast error results from futurc weather conditions which are unknown at the time the forecast is issued. The percentage of the total error due to cl imatolog ical error for forecasts for the Susi tna Basin above ~enali is directly related to the percentage of the total input that has occurred up to *.he date of the foreczst. Since this is approximately less than &~alf of the total seasonal input as of April 1, the climatoloai--1 error is probably much greater for the Susitna River Basin ct~an for basins where the percentage of the seasonal input is much larger prior to April I. Climatological Error Peb Mar pigure 6. Schematic diagram of sources of forecsst errors Jan Feb ~iglrre 6. Schematic diagram of soirrces of forecast e+rorr; Sep It is not possible to determine the relative errors t cil imatolog ical , basic data and model.) that c~ou'id be assoc ia ted .;it23 -eacn model without completely applying the models to the S7.sicna River Basin. However, the climatological error should be approximately the same for each model used. The basic data errsr ~i5sociated with the use of the five models reviewed is prbmar lly dependent upon the basic data and only marginally on bow the data axe p~seessed in the saodel's, It is assumed that the basic data error and the model erxon may be approximately the same fur dz A-srecasts prepared using the PRMS, SWM, HyMet or NWSRFS models .A somewhat greater for the SSARR model, The actual error of the forecasts finally developed for the project will depend much more on other factors than an the sePeel:isn of the model or models to be used,, Basic data error ~esul t:i ng fzom inconsistency {change in time or l neonsistencies in the data used for calibration and for operational forecasting) can be very great. The successful application (and minimum forecast error) using any model depends to a larye extent on the experience of the modeler and on the characteristics of t:?e basin, One advantage of using a conceptual model over a black box regression model is the increased ability to forecast extreme (dry or wet) conditions that were nct experienced in the set of data used for calibration of the model. For a conceptual model to realize this advantage the input data for the n~~del should be as representative of "truew values as possible. For calibration an aceurate knowledge af cl imatolog ical averages of the magnitude and distribution of the seasonal precipikation over the basin is requi red (seasonal isohyet31 maps) . For operational use information on the climatol~gical averages of the possible weather conditions for the rest of the season are required (statistical indicies of rlimatological conditions]. The only way to reduce the climatological error of forecast is by using forecasts (that have information content) of future weather condl tions. Availability of short- and long-range weather forecasts and monthly outlooks have been summarized by Nibler (38). Forecasts of temperature and quantitative preci- pitat ion forecasts (QPF) are available from the National ~eteorolog ical Center, National Weather Service, Sui tland , Maryland, and are received directly by the NWS off ices in Anchorage. Specif ie forecasts of precipitation and temperature conditiorls are issued for 12-hour periods for the next 48 hours. Other prognostic information issued for op to 72-hours periods and for 6-20 days ahead can be used to derive forecasts for precipitation and temperature for up to 10 days in the future. In addition, information on expected average monthly values is avsilf3b2e frorn pub1 ished monthly outlooks for Alaska, Sevr?ral individuals with whom the study has betn disct~ssed '3 3 . ."A indicated that forecasting the snow and ice meit from the 7.i-zcierized portions of the basin ~~olnld be difficult, iiowever, ba:?d on the review uf the hydrometeorological data for the basin and thri results of previous model studies, it is bel ieved tha: the Ariderson model will provide the accuracy and relitibility fa; suc:n forecasting when used with available updating procedures 2nd real-tjme measurements af the streamflow. Thus, the Anderson i~1odil.1 is recommended for use in forecasting sncw and ice mnir for ?he project. The Xack sf accurate and consistent data Ear a number of years (except for some snow survl2y and streamflow data) makes i E difficult, if not impossible, to use regression type "black-hcsw approaches for forecasting the long-term water supply runoff. J*ssuming a real ist ic knowledge of the magnitude and variability of seasonal precipitation is provided by the propn~ed basic data coLlection program, simple quasi-conceptual models can be usg:;d ta prepare Long-term farecasts of the water supply runoff, Forecasts of the long-term water supply runoff can als43 be prepared by extending the forecast periods for the conceptual models discussed in this report. This zan be done by initializing the model as of the date of the forecast by using measurements of the water in storage (in the snow cover and englacial) , by using forecast weather conditions for the rleair filirurcl and hy applying Extended Stre3mfEaw Prediction (ESP) procedurks used by the National Weather Sevice (39) far time periods fax which weather fo~ecasts are available, The long-term water supply forecasts can be issued in terrns of conditional probability that will provide considerably more information for the reservoir operations than a simple volumetric sr'orecas t. For example, probabil ity distr ibtltion relations can he produced providing the expected probability oE occurrence of thz ral7ge of water supply that could be expected. Such information is valuable for reservoir operational decision making. The time and expense to develop a data base and process all data for use in forecast modeling is considerably greater than the time required to test and ,evaluate the use aE different: Fsreeast models, "Peerefcare, it is ri-acornmended tbjat mo&e thaa3 arcre lnodel be tested for the soil moisture 2ccounting and the long- term forecast models for the project. This is considered viable in view of the tart that the actual operation of the reservoirs will not- commence for several yEars, Re<-ommendations for mode-1s that should be investigated for use in the project arc.: A, For short-tern forecasting: (1) NWS A~--aderssa %now ablation and k~t:cusnul-c~t ion itaodcl for forecasting snob] anrl rjliicicr mc!l? fur kt-!F. bnsir-i. (2) Test the SSARR and kl\qzj models for soil :~iojst~jre accollnting and use as a routing model. R, Pox long- te~m foretasting,; t,*j Use the conceptual ionecast model selected for issuing short-term water supply runoff to forecast ~sn~-tkrm (from date of forecast to September 30) wzter supply using weather forecasts and E-c teilded Streamf low Prediction procedures. (2) For early season forecasts [prior to May 1) def~el op a simple quasi-conceptual yodel :. ssed on wiitershed conditions as cf April 1 si-3:~ sta- istic:% of climatological conditions dux ~ilg the ilay-September per iod. The seetion includes a shost summary of current data ~3llecti.on snd relay methods, a discussion of requirements for operational forecasting, several alternative communications schemes, and rece0rnmenda"L-,~r1~~ A. Existing Data ColLection Methods The existing data collection methods are well summarized in previous portions of this report, At the present time all data from the Susitna Basin is recovered by helicopter and to a limited extent by ground transportation after having been recorded by automated weather or stream gaging stations. Data from the basin are thus available at approximately monthly intervals. The time delay from collection until the data arc plzoeessed into useful farm can run from one to seven weeks, B. Data Callection Requirements for Operational Modeling Discussions held during the field trip to Alaska indicate that operational forecasts will be required as often as daily dueing the summer months. The forecast tilne interval dictates the minimum time which can elapse between recording of data in the field and recovery and processing. Models can obviously be run at any time, However, if the sophistication of rainfall-runoff models is to be used to maximum advantage (i.e., if short time interval runoff from thunder starms is to be modeled) then data must be available at short intervals and soon after the events occur. Hourly 2ata in real or near-real time would not be an unreasonable requirement, C. Alternative Data Gathering Systems (I) Con"einue Present Methods One alternative which must be given considerati3n is tc continue data gathering just as it is done now. This rtay, i? fact, be a satisfactory alternative during the winter months where forecasts will not be made as often. During the summer the frequency of site visits could be increased to weekly or biweekly. Fuxther automation of the data handlintj process could take place so that data can proceed directly frovt the field tapes in a farm suitable for input to the models. Current methods can not provide the timely information needed far rainfall/runoff modeling in the late summer period. (2) Automated Telemetry Systems There are several types of automated telemetry systeias available for the Susitnn Basin Project, Each system requires field processing and teletnetry klardwarc, a communications link or I = : inlis, central receiving equipaent, and data processing ey9l i pne~, (39 Field Site Hardware Most current automated telemetry systems make use of field hardware called "Data Col lection Platforms" (DCPs) . DCFs incorporate microprocessors, optional tapes or other form of storL3ge, and one or more telemetrl- modules. The telemetry modules can be line of sight radio transmitters, satellite transmitters, or telephone modems, some even including voice capability. The meteor burst system (to be discussed later) uses remote data terminals that operate only with that system. DCP's are designed tc interface with a wide variety of strea~u gaging and meteorological equipment. They are widely used by the U.S. Geological Survey, Corps of ~n~ineerk and other agencies, They are normally small in size and powered by 12-volt batteries charged by solar pan5ls. Currently there are three U,f. firms actively manufacturing DCPs. They are Sutron Corporaticn, Nandar Corporation, and Synergetic3 Corporation. All of the firms have sizeable numbers of platforms in service and all offer similar capabilities i ~4 6istinctly different pae:kages, Considerable experience by government agencies indicates t:22t the use of DCPs and real-time data obviates the need for an- site recording. This is contrary to the suggestion by R&M Concultants (W&M review memo dated March 8, 2985) that additional mechanical recorders be added at each site. The lemetry equipment has been found to be more reliable than on-site reccrd ing instruments. X t makes l ittle sense to make dupl icste recordings of had data on-site. No goad data is no good data whether recorded on-site 0s: telemetered to some centnak Xscatics~ and recordcd there* Not only that, but the adds an the central site computer functioning properly are much higher than for any field recorder, The less moving parts in the field, the better, However, there would be some value in retaining on site precipitation recorder during the period when data are being collected for model cal ibration purposes. (4) Central Receiving Stations Central receiving equipment is specific to the communictions link used. Cine of sight radio systems require an antenna, (usually a whip or Yagi mounted on a tower) a receiver, and an RF' modem (modulator-demodui-ator) for data recovery in computer compatible form, Satellite systems also require an antenna, but this must be a 3 to 5 meter dish permanantly focused on the satellj te beitlg used, A receiver, demodulators, and inul- t iplexiny kquipment are also required to reduce incoming data to computer compatible form, Telephone systems require only a roodern to receive incoming cal.1~ and interface with a computer. Enormous variatians are poscible in central site data processing i 1 Opt ions vary from desktop personal compilters ko s s ri- computeus us capable of not only receiving the data butt pA-?cessinq i t in real-time and updating the inod+zls. - r Fk,i ke:xnati~~~s EQK the Susitna Basiw At the current time the writers believe that there are tkt:es alternatives fax !:he SusJ tna Basi n: (1) A combined line-af-sight radio (with or without voice capability) and telephone systgrn d, (2) A satellite system with a variety of central site options, and (3) A Meteor Burst system. The alternatives are discussed below, A. Line-of-Sight and Telephone System The wide open nature of the Susitna Basin makes a line- of-sight system reasonably attractive. It should be possibls tto relay data fro21 virtually any part of the basin to a central locntio~? such as Watana using a minimum number of repeaters* Two or three should be sufficient. The systerct call be designed strictly for data purposes or it may be ccrrnbined with a s.aice network for use in managing the reservoirs, A small minicomputer at Watana can save the data and relay it to Anchorage on a daily basis by telephone. The assumption being made, of course, is when the reservoirs are built that phone service will foilow. If such is not the case then line-of-sight is probably not a practical alternative. Discussions during the field trip indicate that thunderstorms are a frequent acculrxrence in the Scrsikrra basin, Thunderstorms are the most frequent cause of faiaure in line-of- sight systems. Lightning strikes near repeaters nearly always cause failures, Any system should probably be designed with dual daka paths (multiple repeaters) to minimize data loss and allow for tirne to repair. Hot standby repeaters could also be considered. Hat: standby uses duplicate transmi t/receive equip- ment with a spare ready to take over the minute the aperational unit fails. The cost for hot standby is fairly substantial since ik doubles the amount of equipment at each repeater. It: is possible to design a line-of-sight system using polling, wherein the cent:rol sight queries the field sites for data, This requires more complex equipment on both ends 2nd is rlot rec~3nmtznded. DCPs are easily capable of determining when data should be sent and with great reliability. Do Satzellite Data Relay System A17 attrackive way to recover diita from the Suslhna Basin is by %;cans of the GOES (Geostationary, Operational Environment81 Si'telliete) system. The GOES system, owned by tqY'OAA, is the szme satellites wh'ch provide the satellite wesLher pictures seen on tel evision newscasts. The Data Collection Systen~ (DCS) on the sate1 1 ites provide 256 channels far environmental data trai~smission. The two U, S. GOES satellites provide sufficient aeegrayhic coverage to send data from Florida to Alaska with no -i ~n'iermediafe repeaters. GOES is truly the "ultimate" line-of- aiqht repeater, The GOES satellite data collection system has been in use for many years, Tbe U. S, Bureau of Reclamation operates a network of over 150 DCP" in the upper Snake River Basin for which part of the area has cold temperature extremes similar to those experienced in the Suisitna River Basin (near Yellowstone Park Wyoming). The contractor in charge of the system maintenance is paid based on the percentage of data received (percentage of 15 minutes readings from many types of sensors). The Eormula is such ?hat the contractor receives full payment for data reception of 95 percent or greater, 90 percent payment for 90 to 95 percent receptios and so on until no payment is received If reception falls below 50 percent, Ta date the contractor has xJ.ceived LOO percent payment over the life of the contract. The U. S, Bureau of Reclamation officials in charge of the project using the GOES system is Dan Lute, DSBOR, Box 043, 500 West Fort Street, Boise, Idaho, 83724, telephone 208-334-1976. A newly installed GOES collection system at high elevations along the Ro-ky Mountains in Colorado for the Office of the Colorado State Engineer has been demonstrating similar performance as that for the USBOR in Idaho after initial instal latiox1 bugs were worlted out. The "hid method for relaying data from the Susitna River Basin is the Meteor Burst or Meteor Trail technique. Meteor Barst conu~unications make use of ionization trails in the atmasphere to reflect high fcequency (HF) radio signals. The trails take the place of the satellite in the GOES system or the repeaters in the l ine-of-si te systems. Meteor Burst systems are, by nature, more complex than satellite or line of site systems. First, the systen is two-way. A base station sends aut signals to one or more remotes requesting them to report stored data. The remotes then send out data signals, The requirement for a polling scheduler and 017- site receiver^ more than doubles the cost of the hardware over a one-way system. Second, the meteor gurst field sites require 2 to 3 times the average power to operate as a satellite or other one-way system, The receiver at the site must be sn at all. times d 3 20 5 times the transmitter power (25 to 40 :~at*ks) is d 1-0 bounce signals off a :neteor trail as compared to a sateiiite (8 watts) or line-of-site repeater (2 to 4 rqatts), ""- kJ::gresed in a different way, that battery will. last 3 to six d- " ~~rnes as long as the same battery powering a meteor burst station .hen both transmit the same message length, Finally, meteor Djrrst hardware 3re larger in physical size tnan satellie or Line- f-.- s; te systems. The transmissian units are large^, and much la3:ger solar panels and more batteries are required, Meteor burst's one true advaz~tage is the preserrc . of 2 :;.cciver at the remote sites. It is possible to send and receive text messages at the remotes which may offer some further safety for field crews in remote lacatins, The basic cost of a meteor burst remote station is roughly twice the cost of a satellite DCP, exelusive of the accessories and iilstallatian. Accessory cost is also higher with antennas costing 2 to 3 times as much as GOES ($500 to $700 compared to S250B er line-of*site, At a meeting with Narza-Ebasco and Alaska Power Authority officials in Anchorage, Alaska, on May 30, 1905, George Clagget, SCS, Anchorage Alaska, indicated that considexable improvement has heen experienced in the operation of the SNOTEL data collection program (meteor burst system] of the SCS. However, CLagget did not have detailed information on the performance of their system. "ontact was made with the SCS offices in Washington, D. C. and in Portland, Oregon to obtain first hand information. Art Crook, Water Supply Forecasting Staff, Soil Consexvat ion Service , Portland, Orgeon (503-221-28~3) supervises the SNOTEL program for the SCS and provided information on the syst:,em, Figure 7 and Table 4 were provided by Crook and provides specific information on the SNOTEL system performance (for data collected by two master stations, one in Boise, Idaho, and one in Ogden, Utah) from February 1981 to May 1985. The sclid line on Figure 7 represents the pe,centage (by a four-month moving average) of sites that successfully responded during the once-a- day colletion period (from 0500 to 0800 local time). The dashed line represents what the SCS believes the collection percentage would have been assuming that the collection would have been sccompl ished with present day software (5 percent improvement) and without a master station failure during the 1983-84 winter. D, Comparison of Meteor Rurst and GOES Satellite Systems As indicated in tt3e abave diesussion a radio line-of-site system is not recommended primarily because of the need for repeater stations and the thunderstorm activity of the area. There are many factors to be consideved in making a decisisn between Meteor Burst and the GOES Satellite system. Cost of the initial hardware and accessories (iracbudirrg antennas) 3"s much OBSERVE ',hi) JUSTF~P' -- LI OBSEj;VED - -- ADJUST. DATA LOSS ----I( ADDNL. ADJUST --..-=--. PREDICTED Table 4 nonthly Wve~age Percentage oE Sites Repo~ting Daily Miomial Pa:LLs dazauary Febr~ary p@< a c pfi April $4" i 4ay June July August Segtenben October kJavember December less far the GOES system. Installation costs would be very eompa~able. Maintenance costs would also favor the GOES system since the power requirements for the Meteor Burst system is much qxeater, The reliability of the two systems for once a day collection of reports is assumed tc be about the same based on recent performance information. However, the reliability of ebtaining 15 minutes or hourly readings far an entire day would be much greater using the GOES system. The two way communication potential of the Meteor Burst system was indicated as a plus by the federal agencies in Alaska that attended the Play 30, 1985, meeting. However, this would be of advantage only during the maintenance visits, For the Bureau of Reclamation net~rork for the Snake River Basin this is often only once a year. The GOES system does allow for text ("related textN) aqd emergency messages to be sent from the DCP. In fact, NOAA, owner of the GOES satellite, currently operates a program called SEAS (Shipboard Environmental data Acquisi t ion System) which prefaces each data transmission with text detail ing weather and ship locations. The message are keyed in by the shipboard operators. If it is truly important to have the meteor burst 2- way messages it may be best to buy a single transmittor for the Meteor Burst system for use by the servicing field crews to carry and operate it in cooperation with the SCS ineteor burst system. When operation of the hydroelectric system commences there will undoubtedly be a greater need for data for short periods of & * &nrne, i, e,, for information on resexvair releases and dawnriver streamflow to ensure that envirornmental requirements below the reservoirs are satisfie$. Th2 added reliability of the GOES system for collecting 15 minute or hourly data for the entire day would be of vzlue, based on the abo~~e cons idexat ions, it is reccmmended that "; '";? GOES satellite system be investigated for ilse in ti?e prt~ject~ " )"* ' : 81% .r?~ould include an ansite field survey to ensure thar ail :;r?lected sites requiring collection of data can be seen by the $:x;3 $:Q 1 1 i ke, 6. Coeperration on Data ~eceptian A Large number of firms and agencies awn and maintain GOES !+.a -. b, :a receiving stations. The U. S. Geological. Survey, for tixampie, operates a portion of its own gaging network by means of a receiving station in Anchorage. Any such site can collect the nccessasy data and hold it for telephone transfee to the comput<+r chti:re the models will be run. The geographic coverage of GOES is emphasized by the fact that nydex Corporation using the Sutron Csrpoxation downlink in Herndon, Virginia could easily receive the data, xun the models, and return the answers to Anchorage. In the initial stages, however, the U.S.G.S. site will certainly be attractive, A satellite system using the U.S.G.S. site would require only an investment in DCPs and sufficient telephone equipment and software to take the data from the ground station to the computsr isihere the models will be run. Larger, moxe nlodern ground stations in t,he $50,000 to $100,000 range conre with complete data base software and model inter faces. This rnore sophisticated system should be considered if caoperatior~ with II. S.G.5. is not desirable, I, Prrrehase of ddicated ground station Purchase of a dedicated ground station will have to be weighed on the basis of cost effectiveness. Such stations are avairable with a wide range of capabilities and corresponding wide range in prices, There isre rur-rent%y sszy ewe manc2ac%urer% sf qrauz2-& stations in the U,S. - Sutron Carpazstion sncZ Synsrgetics Corpauation. Sutron is the newes of the two coxparations and offers mare data handling capability and smaller, fully digital electronics, Synergetics has rnore stations in place, The markci- is highly competitive, The simplest receiving site from either manufacturer costs approximately $27,000. Far that ?rice the user receives an antenna, frequency downconverter, receij:sr, cab1 ing , and demodulators for monitoring the satel l i tc chann6sks. Also included is a pr:rsonal eompiiter to select channels nn ;1 t ivic t;chedule loasis and to store and print: out: data. Stlch botton. end receis~e sites are minimally useful and are usually ussd as "frot*ttr essds" for h~rgera sys%e~%, To increase the receiving site ~apa'uj.1 ity is primarily .1 matter of add incj computer capacity, H<~nufacturers 0t:fer 2 :q;i;ased --* d_i processor and disk space in nouyhly 510,000 doila: ii;cremcnts, realjstically useful receive site with a cc.tnpu?r.er *. czgJa122e 0: runllinq the real-time m~d~ls and a goad data base i*.lLl cLc;st approx inately $60,000 without such extras as an ~i>i*Ote~ruptible power supply or installation. Additional maney Tg: ..:zA - d- *- be ccrrsidered for data handling software. Sutron offers a 1 icense data base package designed to provide real-time data ** -8 :r:spiays and to interface with models. Xt cilrrntly 1 ists for 525,000. This is approximately the cost for single-purpose cus"im software if a "hare bones" receiving site is interfaced tra a!: EX lsting computer. 8. Kgudget Estimation Figures Detailed budgeting at this time is not practical, However, cst inates of telemetry system costs axe needed for planning porposes. The following are reasonable price estimates in 1985 dQ$lans Ear V~P~QK~IS items wkPch would be ;.meelled far a satefli te systd:?rn: a Data coLlection platforms - satellite $4,500 ea, (includes antenna, solar panels cables, batteries and insulated csrxtainer) , arad o Minimum satellite grcund station for interface to existing computer $ Snelude% S2S,OOaS custom softwar@ , ar a Sakellite ground station capable of running real-time models as well as receiving data (includes data base saftware) ar 9, Add i kisnal Reeoamendatians If the Susitna Hydroelectric Project will have a line-of- sight cammunications system installed for: other seasons it: would msjce sense to multiplex in the data gathering and combine the maintenance. However, if no such system is planned a satellite system as recommended above seems highly atfractive, If suff icianl: funds are available a gcourld station ::~~ould be puxchased. The USGS site is fajrly old technology and the data handling software for it is primitive by today's standards, A new receive site could be purchased with an-line quality aontral, ;! camp'lcte data hase, and sljfficient capability to run all 1:lie ~OXEC:SS?S* rn the mean time, it wot~ld make a great deal of sense to buy UCPs and place them at the existing data gathering network sitt~s. Data shouLd be manitored daily and as-needed maintenance under takex-i. By usj.ilig UCPs and the USGS grourld station it is gmnsslble to knai.4 ilnmed late1 y when px~b;lre:ns OCCUI- and cnfould - - cJ~..:s!.~c.~sE,1y up the rtzl iability ~"ata ret.rieval, if DCPs jre s. ~~ed i; will be possible to get rid of t'-9 on-si:e recorders, S:IC~ t;lcire wculd axso up the reliability of the sibs. ALL dat~, r<:&ccrding ;*.auld take place in data fifes on the computer that P pcrz *j ~ps - *- & the data, The conside~able experience of R&M Consultants in the - *.l -2 s, 9 :gi.-e~iiliatior: artd maintenance of the currznt system rgould seer3 :a 4 -* ?~io~cate them as a logical choice to install and mainkain a satel'{ ice system. They have accurately defined the cost of s*aa-j !i,.c_i.;ea - :instruments a and varying kinds of maintenance schedules, iwy P ?:?ray w + --, could easily be trained in the installation and setup of r$6, u?, LAq* P 5 * 1. E. L, Peck 1972, Snow Maasure~srat Predicamen.%;, Wantei- !?e5wr~e5 RQS~B~C~, Val, 8, N0, Feb, A972 2, C, E, lsraalsen 1967, R~liability ~4' Can-Pyps Precl- pitaef cn Gage Heaeurenents, Techn Repu 2, Utah State Water Re$. Lab., Utah State U,, Lerqan, Utah, Jetiy 3 U, T, Wilson 1954, Rnalysis of Winter Pre~~p~~tatlan in *$he &oapesa%ive Snow 11% tss$%ga$iona, Manth1 y Weac Re\#- 82, pp, f83-%99, July 4, M. 5, Braun and E, &, Peck 1962, Reliability a+' P~eci- pita'kion MeasurnenSs as Related ta Expasure, Sour. af Rppl. Metea~s, J143e 5, &* Larson and E, h, Peek %974, Rewsaey of Preeipitataan Msa~ursmen+% for Hydro1 sgic Hodel ing, Water Retiaurcea Research, Vol, 18, NQ. 4, kguat 6, E. L, Peck 1973, Discussion sf Problems in Measuring Prccl pi tat ion ir~ F"lraun*%;ainous Areas, Symp~sium art Oistrf bu%icbn of Pracipi-katian in Motantairlous Areas, EeiPo, Yoruay, Nav, I, WMBIBMM, Ma, 32S, Genava, Swi tzarland 7.. Ea G. Bsgdaaova 1968, Es$imat~ of tf-se RelixabiEity of the Charactariskica of $h~ Shortage in Solid Preci- pitation he to Wind, Sor. Hydral. S~&ae%ed Papers, Esagl. fsansf,, Ma* 2 8, El L, Peck and T. R. Carfmal1 1979, Advantages af Coxlceptue% Madal a for Mlsr*.ft;t"aarrn Research Bafaan 5tridsz?s, lhdrd Symp~si~m EHP Rege Idorkrng Group 017 Marl;hef-n Rt3aearplh Basins, hebe&: Crty, Canada, Jurle 11-15, 9, G,, 0, Klilday 1974, Mea8.o M~nChly and Rnneral Prsei-. pitatxan, Alaska, WAR Vaeh, Hema, NU5 AR-18, Anchr~raqc, Alaska Mar, 18. 5C5 1991, Maarr Al.snuaX Precip~Gat~~n, Susltria Rrver- Baszn, filaskt3, Prepared by SCS WSTC Casts Unx t , H'7-M-24165-1, X3u~ust 11. J. W~ss $9Q49., Annual f"%reorpafat~or-r Map frlr- HI,-s5shti rdrrpu b2 l sfled, Per sasial Crrr*reuporrdcncu 12 El L, PCC~ AI?~ M, r7, 8r~~9"t 1962, Ai7 A~praach ti) the Dcveiaj~~~aaat a$" S3ahyatal Maps for Hsur~tal rraus fire~95 T, R, Gar~aIl and K. 6. Vadnais f 980, Operatione31 fiarborne Heaauranen-8; sf Snaw Wakes Eqi~ i val ent Uai ng Natural Terr~strial Gam~a Radiatzon, Prac, 48th RnrlediaP Weatern Snow Cansf,, taramis, Wya,, fipril D, &, Fineh and J, P Ualker 6979, Snouf a1 l at @nchorage, Alaska Rssociated with Cold Advecti an, NOBa reek, Mema* NWS AR-25, Oct, D, Blanche* 1983, Chugach Mat ional E~visonmental fit las, Chugach Natianal Fares$, Alaskan Ragisn, Rep, NQ. 124, Ee A* Ma~chegiani 1385, A Relalianship Between Snow Cwrse Infsrmatfe~~.r and R*lnguiB.'f, Unwblisk-sed Paper. R & M Can~ul$ants and U, Q, Hz3rsisan 1381, elaska Pawer RuPharity Susitna WydraaekacSr~,c Prsj ect , Task 3 - Hydrslagy G1aci er Skudi es RG:~O~B prepared far Acres American Xncarporated, bffafa, New Yark, Hasza-Ebascs Susi %na Saint Venture 1984, Watsna and Devil Canyen Sites Prabeble Maxxwn Flaod, Draf d Regasrt, Docamsnt No, 457 Jan. Rcres American Inc, 698B. Appendix AZ, Watana and Devil. Canyon Sites P~cubable Fvtexlb~gum FIOC)~, Subf ask 3,85 a$* G. Kaolovalov 2988, i"E@l%lrag! and Runoff in G~acaer Basin%, Ssviet Hydrology, Selected Papers, Val 19, Na, % V, G6 P(onsve1av H98B. Madel sf fPle Hovement aC *the Snow Line an a Gaaciss =ring $he flblatzan Period, SovxeP Wydrslagy, Sele~ted Papers, 901. 19, No, Z A. 6, Fountain ab-~d W, kd, Tangtjharrr 1985, Cantampssrary Tacfrrrnwas Csr Pr~dlctiprg hra~~ff from Glacleri~cd Bas] na, esnptdhliahed by Working Grcaup on Snow and Ice Hydrningy of Glacieri ncd Basins, international Eanm an Snaw and X ce, T AHS ( pri vata csmmu ral e:a%nat.1) E. W. Rr-icdersan 1973* %dat;rr~na,% WeatP-\er* Sera vt:e Rn ver Forecast System - Snuu Accanmulatzon and Ahll3tlurr ?.ladel* NQfiB Tec5-I, Hexuao. NUS Hydf-0- %'7s Snlbver S~rl rng, Rd. N(3\ . E. I,, Peck, 8, S, FBcQurvegt, 'F, N, Muefer, E, t-4- LIcxE-tr~ciarrr, and J, I,, E~aksruag 1,981. Revrcw of b4ydr-alagic tq'%ocJ;:ls Ftti. Evaluad lt mg Uss a43~ar~ro-t;~ Sen42 rag Capabr 1 n t A ns, tJHSi"i CR 1666'74 Mar 8-s i c. D, L, Chapman 1902. Daily Flow Statistics of Alaskan Streams, NOAA Tech. Memo. NWS AR-35, Anchorage, Alaska, oc t 2'*7, PJMO, Unpubl ishedl rep~xt on Intercompar ison of Snowmel t Model (private communication f lorn E. A. Anderson) 20, G. C-vadias and 6;. Morin 1985. Approximate Confidence Intervals "~r Numerical Verification Critexia used in Hydrologic, . Models. Applicatipon to the WMO Inter- comparison of Conceptual Models of Snowmelt Runoff. Unpublished report (private communication) Jan 29, W, V. Tangborn 1984. Prediction of Glacier Derived Runoff for Hyd-qelectric Development, Geografiska, 668, 3 30, WMO 1975. Intercompar ison of Conceptual Models used in Operational Hydrological Forecasting, Operational Hydrology Report Ha. 7, WMO - No. 429, Geneva, Swi tzexland 31. W. V. Tangborn 1977. Application of a New Hydro- meteorological Streamflow Prediction Model, Western Snow CanE, Apr 32. E, L. Peck 1976. Catchment Modeling and Initital Parameter Estimaeion for the National Weather Service River Forecast System, NOAA Tech. Memo. NWS Hydro-31, Jun 33. E. L. Peck, E. R. Johnson, K. M. Krouse, T. R. Cirroll and J. C. Schaake, Jr 1980. Hydrological Update Tech- niques rJsed by the US National Weather Servi.cer Proceedings of the Oxford Symposium, IAHS-AISH Publ. No 129 Rpn 34, W, T, Sittner and K, M, Krsuse E979, Improvement sf Hydrologic Simulation by Utilizing Observed Discharge as an Indirect Input (Computed Hydrograph Adjustment Technique - CHAT), NOAA Tech. Memo. NWS Hydro-38, Feb 35. T. R. Carroll and K. G. Vadnais 1980, Operational &AT- borne-Measurement of Snow Water Equivalent Using Na'rual Terrestxial Gamma Radiation, Western Snow Conf, Apr 36, E. R, Johnson, E. L. Peck and T, N. Keefer 1983. Creating a Bridge Between Hcmote Sensing and Hydrologic Models, NASA CR 170517 Jan 37. G. He Leavcsley, N. W. Lichty, 8. M. Txoutman and L. G. Saindon 1983. Precipitation-Runof f Modcling System: User's Manual, Water-Resources Investigations Report @,-4238, U. S. Geoloy ical Surveyr Denver, Cola. 38, 6. J. Nibler. Information Content of River Forecasts, Unpublished Paper (private communication] 7 Q r-3 id. L. W. Twedt, J. C. Ssbaake, Jr,, and E, L. Peck 1977. National Weather Service Extended Streamflow Fred iction, We.:teprn Snow Conf. Apr APPENDICES Appendix A Maps of Monthly Precipitation, Susitna River Basin (Figures A-l to A-9) Appendix B Plots of Daily Discharge and Daily Pzecipitation, Susitna River Basin (Figures B-l to 8-16) and Semilog plots of Daily Discharge, Susitna River Basin (Figures B-17 to B-28) Appcndix C Results of Multiple Regression Analysis, Susitna River at Denali, Alaska Appendix B Explanatory Information (Figure D) , Definitions of States and Parameters (Tables D-l to D-12), and Schematic Diagrams for Conceptual Hydrological Models (Figures D-l to D-12) Appendix E R and M Consultants letter of March 8, 1985 s bpbds Marawnan 8 - Pax?S9T?i " Coppr Canter C)Ld Edgerton Figure A-.I Precipitation April 1983 "$%rims Camp B Figure A-2 Precipitation May 1982 Figure A-3 Precipitation June 1984 ~i.gbre A-4 l>reci.p~talrion July 1.981. Center Edgar tan Tg.i&w Camp 0 V; Galkana L / lcb -- o Cetpwr Canter 8 :, @&- - - - 1.7+* @ - *r*a Glerutallen x&% 01x3 %:dqer%an Figure A-5 Precipitation July 1983 Bek*ll.y Par* 0 . wrly 1, SBS~&%$ Re+ P6,ta mw%& 2, Bkgaxen 8, w, Paxwn 3*).~3 x$l Figure A-6 Precipi.tat ion August 1931 Figure A-7 Precipitation August 1983 Sw$&qaging Smasns . waxy 1, gasi-a B, are Wnalf esnley Park dg @ 2, Hemen R. ~ar, ~X~IB . . 3, SUS%.%~~ 8, AZP. an-*%% Figure A-8 Precipitation September 1983 Figure A-9 Precipitation October 1983 DISCHARGE DENALI 1981 DAY OF MAY-SEPT, PERIOD Figure 8-1 Figure B-2 DAY OF MAY-SEPT. PERIOD Figure 13-3 DAY OF MAY-SEPT, PERIOD Figure B-4 HM&&&ddddM4v4 DAY OF k4AY-SEPT, PERIOD Figure B-5 DISCHARGE DATA PMmN IQB2 Figure 13-6 DAY OF bjAY--SEPT, PERIOD DISCHARGE COMPARISON I982 Figure B-7 DAY OF MAY-SEPV, PERIOD 1982 PRECIPITATION DATA Figure B-8 DAY OF MAY-SEPT. PERIOD Figure B-9 DISCHARGE 25888 248&16 23888 223@@ 21@88 28888 19888 188168 1 "%go A 16888 " fS&1PIB % " 14808 w w f 308651 12888 < $.arm@ nr~l~B g leagg &-@ 0 9888 8888 aBaB 6888 58B8 4BBB 3888 28@B l BBB B ~~~~~MB~3mMmu3mmBmmMmb3~Lnmmmmmm~b2mm ~~~~mmtfmm~~bRmmrn~~~~-NN~~~W~~ MMM~~~WVMM~~M Figure B-i~ DAMOF MAYwSE$$, PER I OD DISCHARGE COMPARISON 1983 pigwe B-1.1- DAY OF HAY-SEPT. PERIOD DISCHARGE I1 DAY OF MAY-SEPT. PERIOD Figure E.-13 DISCHARGE Figure 13-14 DISCHARGE COMPARISON 1984 Figure B-16 F-4 Figure B-18 -- DAY OF MAY-SEPB. PERIOD * .-i Figure B-19 6 Figue B-20 DAY OF MAY-SEPY, PERIOD r----------b MAY-SEPf. PEW I08 ~;~rs-p..a n-39 8A't OF MAY-SEPT. PERIOD I i I I I --.J Figure B-24 -mr*.-nnux..rr..rr^r-o~~.w-.~-e -"w.bw*"--I.- ! -,,- L-a--.-"-LIA-- -- -k . .-.Y--n"-"-' -5-- ---,.---.""-'-- I 1 ! I I . * ---* I i -- ----7. -.- * ll-O-*..i.. -----"-.-'--------' 1 4 I 3 i... I 1 j -- t I I ! 1 i .. ..! . :-; ----.--p.------.~--.".----".. -po,p--P--P--P--- I , . C -- -.".-..-tpL ----- id --I--- ----."------.---- i I t 1 - I- j I - i "-.rrc-"..---.--."-""----.----L Figure 8-25 . - / - .... - I.. .- 1 B Figure B-26 DAY OF MAY-SEPT. PERIOD 1 -*-- - - ---- --- --- -- --- 4 - --I 1 - L ! ! -.- ---- _..d I + I 1 --_-- -? Iy.-------LI --- ---- -1- --- I , -~---I I 1 I i I I I 0 Figure 5-27 DAY OF MAY-SEPT, PERIOD - .. "-".- 1.. . [ ... -. I-" -..-.--.- 1 -,-: -.-,-.-.-. ,. .. 1 .-.-. -...- ...--.. Figure B-28 DAY OF MAY-SEPT. PERIOD PROBLEM NUMBER REPLACEMENT BND DEPENDENT VARIABLE 16 NOW 2 NUMBER 66 VARIABLES DELETED 1 VAR %ABLE5 DELETED. . . f AMALY%%G OF VARIAYCE FOR REGRESSION SOURCE OF V&WXATBBM B. F. SUM QF HE AN F GGUAREf BGUAAES VALUE DUE f8 REGRESSTON. 4 3388. 7030B 1397. %7&2f 8.07810 DEW r AT ro~ ~asur ~~bkkbki bk: : a 4 2426.41992 8~2.93836 TQTAC.. , ie 808~. as300 INTERCEPT [A VALUE! IS 89,04"i)27 -- - lea. baJsao COMPARE CHECK ON FINAL COEFFICIENT. .,... 0. $0182 , INCWEHENTS FBB INDEPENDENT VARIABLES Q CUMULATIVE REGRESSIONS +a WAR SABLE SWHS Of PROP. F VALUE @ SBD. ERROR SWHS OF PRDB. WAR. . P H&T l PL6 ND. MARE SQUARES VAR . EACH TERH a 6F ESPEHATE 58WCtRES . " R 60. VALUE R 3 x 4 ~~83. 37861 8. 36~49 0. 82448 17.314329 . a~ias.~legn o.=lisosc% 43.82448 8. &Ql24 4 X2 159&. 311928 8. 19432 8, 98847 8%. "31230 4432, s397&& 0. 93'981 30. $91026 0. $4333 9 X3 &SO. 371983 0.07997 3, 29333 33,94820 3092. &TAT8 0.83378 8. ;Pa7 Bb 0.7913& k3 x 4 466. a2869 s, oar 9s z.eb790 53. as137 ~SSB.~OSQB 0. &OO'P~ 8.87913 0. @a339 BR9PQRT16N OF VARIANCE 5PE- ClFttEBTDLBfiXTWACifAiOLES O.008LBQ . 1 Year Twa digits far year ( 3 962 = 62 1 iI 'sJ Hay-September runaf f @ S~b~itna RIVBP at Benali, 1,888 acre fee$ 3 X1 + Water ewivalent, Bp~il 1 MsnahanFla$ snou causse, 1bdB inch86 4 Xil Sum of nonthl y departures f ran average %snpera%~nre f6r Hw-Sept~rnb~r TaPkast na degree6 F b plus dB degrees F! 5 H3 May-September precipitation, Palkeatna, 1Bb8 iarsche~ 6 X4 Value of X3 far psavisus year Regressian equaf ion PLlVHC SBA ESC PARAPaq ETERS STATES Figure D LEGEND FOR SCHEWIA'TIC DIAGRAk4S OF MODELS Inouts A- The set of driving forces required periodically by the model, Co on examples are precipitation poken- tiaL evapotranspiration, and temperature. For most hydrologic models the inputs are all m&teorologic factors, but some require inpnta describing man's activities (cropping practices). The key phrase in the definition of the inputs of a model is "required periodically." If it is possible to run the model without providing a valve for a particular item, that item is not an input. Likewise, if the model can be run with a particular item provided only once or perhaps intermittently, that item is not an input. Some models, however, may have default values for certain inputs (e.g., precipitation is zero if not entered). Parameters The set of values that are changed to make a general hydrologic model apply to a particuJar location. Parameters ar cans%ant with time BP at most, vary only slightly with time as compared to inputs. /. States The set sf intsxnak model values sufficient to start the model. The states of the model completely define the past history of inputs. These are usually values of moistQre stored in various model components (e.g. upper zone tension water contents), indices to model status(e.g.,API), or computational carryover values (e.g.,the carryover values of a unit hydragraph opera- tion). In each time step of ope~ation~the model uses the initial values of the states along with parameters and inputs for that time step in order to compute the state for the next time step. OutpI-:ts Variables of interest that can be computed from knowl- edge of the states and inputs. Usual examples are streamflow and actual evapotranspiration. In many cases an output will be identical to some state of the model, but such .does not have to be the case. The mdel may produce an output that is of vital interest to the mode% user but 2s not necessary to the model computations. PRECIPITATION, MA $42 MI@ TBEBtqPERAVU Figure D-1NIMSRFS (ANDERSON) SMCBWMELT MODEL SCHEMATiC DlAGRAM PA TERS {DEFINITIONS) N\#Qi$SRFS SNOmELT MODEL AmAL - DEPLETION CURE - Curve that defines the areal extent of the snow cover as a functian sf how much of the original snow coverremains. It also implicitly accounts for the re- duction in the melt rake that OCCUXS with a decrease in the areal e:4tewt of SCF Constant amount of mehk that occurs at the snow-soil inkerface whenever snow is present, Base temperature for snowmelt computations during nonrain periods. Maximum melt factor during nonrain periods; asswed to occur c,n June 21, Minimum melt factor during nonrain periods; assumed to occur an Dece&er 21, The maximum negative melt factor. Percent (decimal) liquid water holding capacity; indicates the maximum amount of liquid water that can be held against gravity drainage in the snow cover. The temperature that delineates rain from snow, A multiplying factor that adjusts pre- cipitation data for gage catch deficiencies during periods of snowfall and implicitly accounts for net vapor transfer and inter- ception losses. At a point, it also implicitly accounts for gains or losses from drifting. The mean areal water-equivalent ~Sove which there is always 100 percent areal snow cover. Antecedent temperature index (range is Oel;TIPM<l. - - 01 . parameter The average wind funetion during rain-on- snow periods. Table D-2 STALTES (DEFINITIONS ) NWSWFS SINQWIIEET $3QBEL ~ NEGHS antecedent Tempcsrature Index; represents the temperature within the snow cover, EAGRQ and S together define the amount of excess liquii water in transit in the snowpack. The amount of liquid-water held against gravity drainage. The maximurn water-equival.ent that has occurred over the area since scow began to aecmulate, Heat Deficit; the amount 0% heat thak must be added to return thg snow cover to an fsothemal state at 0 C with the same liquidwater content as when the heat defi~ix was previously zero. S and LAGRO together define the amount of excess liquid water.in transit in the snow- pack. The areal water equivalent just prior to the new snowfall, The areal extent of SFPOW cover from the areal depletion curve just prior to the new snowfall, The amount of water equivalent above which 100 percent areal snow cover temporarily exists, Water equivalent of the solid portion of the snowpack. "These states are onby used when there is a new snowfall an a basin with a partial snowcover. Pl :lGblRE D-2. WYFIET MODEL XMEmTIC DP bGN&M HYDEX, April 1985 TROS T:x 1 K TIIAX S $3 N S?,X EFF TFF QAm SBMF S U B-I, U2IN UZOUT GW1 E GROUT GLW TF GEf N GLOUT TFF GL S PK THRESBOD TEWEUT'dRE FOR MIN OR SMOW %WESEBLD "$IEMBEMTgWE FOR WA2OMTXOH TBRESBOLD TEWEM-TUBE FOR ShOIvWELT m$SWX OF FIN TO SOIL MOISTVRE MOWT OF SN8 LT TO SOIL PiOXSTUWE VARIATION EN MELT UTE DUE TO SW ANGLE VMXBATXON ZN IELT 'UTE DUE TO U35ATION SNObXELT PRODUCED BY AIR TEWEUTUEE DEPLETTOM OF SNOW CWEREB AREA SUBLZMTIOM FRQH SN6WAa IRPUT TO UPPER ZONE STBMGE OUTP5T FROM WPEW ZQHE STOMGE INPUT TO GROUNDVATEM STOMGE B3TPUT FROM GROUNDWATER STOMGE GUCTER CWER FPACTPBE GUCIER $.EL$ DUE TO UBIATION INFLOW TO ENGUCXAL STBWGE OUTFLOW FRlIM ENGUCEU STOUGE GUCIAL MELT DUE TO AIR TEWEMTURE GLACIAL m%T FACTOR DUE TO QRIENTATXON 'ff"&le D.-4 BYWT SLMUUTION MODEL STATES SNACC SOIL MOISTURE SNOW CWERED AREA FUCTION GBAE GLACIER BALANCE PERVIOUS AREA INf ERf LOW Figure D-3RIWSRFS (SACRAMENTO) MODEL SCHEMATIC DIAGRAM That fraction sf LT-ke basin that becomes impervious as all tension water require- ments axe met, Maximum capacity of I~wer zone primary free water stswage, Lateral drainage rate of lower zone prirnary free water expressed as a fraction of contents per day, P/Iaximum capacity of lower zone supplemental f%ee water storage* Lateral drainage rate of lrwer zone supple- mental free water expressed as a fraction of contents per day. Maximum capacity of lower zone tension water. Fraction of ilnpervinus basin contiguous with strea*m channels, The percentage of percolation water that direct- ly enters the lower zone free water without a prior elaim by lower zone tension water. Fraction of lower zone free water not available for transpiration purposes (incapable of re- supplying lower zone tension water). An exponent determining the rate of change of the percolation rate as the lower zone deficiency ratio varies from 1 to 0 (1 = c~mpletely dry; 0 - lower zone storage completely full) Fraction of basin covered by riparian vegetation. The ratio of unobserved to observed baseflow. SIDE Maximum capacity of upper zone free water. Lateral drainage rate of upper zone free water expressed as a fraction sf contents per day. Msximum capacity upper zone tension water. W fraction used to define the proportional in- crease in percolation from saturated-to-dry lower zone soil moisture conditions. This parameter, when used with other parameters, indicates the maximum percolation rate possible when upper zone storages are full and the lower zone soil moisture is 100 percent deficient. % Table B-6 STATES (DEFINITTONS) NwSRFS MODEL Additional impervious area. Lower zone free primary water storage. IIZFSC ----- Lower zone free supplemental water storage. Lower zone tension water storage. Upper zone free water storage. UZTWC -- Upper zone tension water storage. Figure D*~P~s ElODEi SCIEEIATIC DIAGRAM 13;b-f Tmprature k10tq wait:? ppr-ipitation is snow ard above r*rhich it is rain (dqrees F or C), C2ECIg Convection-Codensastion eqery coefficient for months I-12 (cal/degree abve OOC) ~~~~~ r cover density for major vqetation far each (decimal prcent] Winter cc er density for major vqetation for each HRU (decimal percent) Air tanperature coefficient for ET emputation for mn~s 1-12 Air tgnperature coefficient for ET camputation fon each Impervious drainage area for each HRU (acres) Initial density of new-fallen snow fdshal prcent) average maximum density of snow pack (decimal pxcent) Ehissivity of air on days without precipitation maporation pan coefficient for montk~s 1-12 Evaporation loss from hpervious area for each (imkes) Free water holding capacity of snowpack (Decimal percent of snowpack water Coefficient to compute seepage from each ground- wate~ reservoir to a ground-water sink Vegetation cover t fcr each mU (14.. Bare, 1= Grasses, 22 Shurbs, 3= Trees) of rout iny for each surface reservoir (8= ReaPs; S= Linear) Julian date to start i.ooking for spring snow me$ t stage Julian date to force snowpack to spring snow me1 t stage Month that transpiration ends for each HRU Proprtion of rain in ~.ain~'sr?ok~ event &ave bfiniriich snow albedo is not reset for sraoclpack mglt stage Inte~eeption storage capacity of unit area af vwyetation for rain during stxmer priod, for each (inches) Interception storage capacity of unit area of vegetation for rain during winter perid for each (ivchesf Sepg~: rate frm each sU&~SUE~~C~ neservoir to ground water reservoir ( ir~hes/day) Miniman contributing area for surface runoff when IS=:= 0; Cwffirient in contrij2utirig area - soil moisture idex relation when $SSR1= 1 Maximm possible contributing area for surface runoff as proprtion of each HR&I Coefficient in surface runoff contributing area - soil misture idex relation Seepge rate frm soil moisture excess to each ground9later reservoiz ( inchesiday] Snofqck set tlment the cans tant Paximm available water holding capacity of soil profile for each ( inches) Interception storage capacity of unit area of vegetation for snow, for each HRU (inches, water ivalewt) Maximm daily srlo It infiltration capcity of soil profile at field capacity for each kPU ( inches j Storage values in ou"cflow/storage table for Puls routing (aS days) hpse rate for minimm daily temprature for months 1-32 (degrees C or F) Lapse rate for maxirr,un daily air tfmperature for months 1-12 (deqrees C or F) Trawm~issisw coefficient for shortwave radiacisn through vegetation canopy for each HRU P?cnl:l1 to bs~in ckclsi~ for start of trar~.- spisatiew fox each Tranpsiration switch for each .lhU (0 = v8y- etstion donoat; 1 = vegetation transpiring) Month that thurderstom :y[* events end Outflow-storage table values for Puls rrouting F4aximm1 air tanperature, kihich when exceedeo, forces precipitation to be all rain ,Ujusted snowpack water obsev~d SMW course data Rntlting coefficient for each grouMi~ater reservo i P Linear routing coefficient for ezch absurface rewrvoir Mon-l inear routing coefficient for each s surface resexvsir Surface storage reservoir linear routing coefficient for each reservoir Y - Intercept for velation &tween tmerature (XI and 1) degree day (Y) or 2) sky cover (Y) *en ==lor2 Slope fox relation between tanperahure (X) and 1) degree day (Y) or 2) sky cover (Y) when =POP% Maximm Rrcent of potential solar radiation (decimal) mximm value of SfR for each (in~l=~) Coefficient for routing water from each sub surface reservoir to groundwater reservoir Maximum retention storage on hyx3rvious area for each [ ineks) Cwfficient for routing water frm each sub- surface reservoir to groundwater reservoir Proportion of rain in rain/snow event above which snow albdo is not reset for snobpack arcmula- tion stage TABLE ~-8 STAT1:S (DEFINITIONS) PWMS MODEL IJET e- PICE PKDEF PSS RES SMAV Computed ALBEDO for each Hydrologic Response Unit [HRUl Density of snowpaek on each HRU Depth of snowpack on each NRU (inches] Fxee water eontent sf snow on eacl2 MRU (inches) Storage in each ground-water reservoir (acrec- inches) Total seepage to ground water sink for each ground-water reservoir (acre-inches) Snowpack temperature (Degrees C) , each HRU Potential Evapoxanspiration, computed by model ( inche,.s) Porticon of snowpack existing as ice on each HRU ( inches) Calozies required to bring pack to isothermal state, .each HRU Accumulated sum of net precipitation beginning on the first day of snowpack formation Water equivalent of snowpack on each HRU (Inches) Storage in upper part of soil profile where losses occur as evaporation and transpiration ( incf-a~zs) Stosraqje in each subsuaf ace reseHvsi H (acre- inches) Retention storage on imprevious area for each HRU (inches) Number of days since last snowfall on each HRU Daily available water in soil profile for each HRU (inches) Depth of new snow on each HRU (inches) Initial storage in each surface reservoir (CFS-Da ys) Intercept ion for each HRU (inches) P""a kable -D--9 p TERS (DEFINITIONS 18 STWFQRD WATERSHED MODEL 171 A -- Percent impervious area. CB -- -- Infiltration index, Interflow index, which determines the ratio of interflow to surface runoff. Maximum amount of interception storage. Ratio of total stream and lake area to the total watershed area, Daily interflow rececsion coefficient. Daily groundwater recession coefficient. Weighting factor to allow variable groundwater secession rates, Percent of watershed stxeam surfaces and riparian vegetation. Percent of groundwater recharge assigned to deep . percolation, Evaporation loss index for the lower zc-lne. Overland flow length. Manning's "n" for overland flow. Nominal lower zone storage, an index to the magnitude of lower zone capacity. Nominal upper Zone storage, an index to the magnitude of upper zone capacity. Overland flow slope. 2. -0 STA-TES (DEFINITIONS) STmFBRB PJATERSHE1D MODEL PV SRGX --* EPX Surface depth. Interflow storage. Active groundwate:~ starage. Groundwater inflow index, Upper zone storage. tower zone storage. Interception storage. UPPER ZONE LEGEND PW PERIOD kENOPH RGP QENERATED RUNOFF % ROS GENERATED RUNOFF, SURFACE W I RAINFALL . 1TENSlTY RS SURFACE db9N8FF WP WEIGHTED PRECIPIWTION GROUNDWATER LOWER ZOHE ROUTlNG PHASE ,p BASE FHOka TSBF, W STORAGE h TERS ~DEFINHTIBNS) SSARR HODEL BFL ETI -- KSS -- ROP TSS TSBF _yl__ Base flow infiltration limit, Base flow, percent. Evapotranspikaticn index. Percent effectiveness of EXI (function of rainfall intensity, HI] . Limiting subsurface infiltration rate. er of routing phases (surface flow) er of routing phases (subsurface flow) er of routing phases (baseflow). Runoff percent:. Surface runoff percent, function of WSlXS table- Time of storage; surface flow. Time of storage; subsurface flow (interflow). Time of storage; baseflow. Tab,Le D-1% STATES (DEFPMITIDNS) SSARR MODEL -a- - Sail Moisture Index. Base Flow Infiltration Index, Phzse storage (discharge or stage) for surface flow Phase storage (discharge) for subsurface d.%,ew* Phase storage (discharge) for baseflow. ~ Marsh 8, 1985 Warra-Ebasco Susitna Joint Venture 7 I "8% 'Xtreet Anchorage, Alaska 99501 ~ Attention: Dr. Larry Gilbertson R &M Nos. 45241 9 452443 ~ Re: Review of Hydex Streamflaw Forecasting Feasibility Draft Report We have examined Chapter 2 of Dr. Peck's draft report, dealing with data requirements for streamflow forecasting, and have several comments. These comments primarily address the major conclusions and recommenda- tions in the report. A few additional recommendations are given for your cosssideratiar~: 1 , The objective of the hydrological-meteoroDogicaI data-collectian network will be to obtain input data for a streamfiow forecasting system for the hydroelectric project. Other purposes of the data collection should be kept in mind, such as terrestrial game studies and impact monitoring. Many of the existing stations were established to support such efforts during the feasibility assessment. Future data require- ments will be oriented toward project operation and monitoring. 2. An additional application currently being made of' meteoroiogical data is air quality modelling for the Watana site. Required parameters there ate wind speed, wind direction, and standard deviation of the wind direction (known as sigma theta). The sigma theta has been measured since October 1954. At least one year of data is felt to be necessary for satisfactorily modelling of the air quality. 3. The network of existing and proposed met. (meteorological) stations in Figure 2 and Table 2 of the report recammends that data be col- lected at a total of 11 sites in the basin .aboxt 3evii Canyon. Five stations would be at or near existing siaL.sns, one would be re-established at a former meteorological station site, and the other five wouid be new stations established at existing snow course sites. Factors which we consider important in selecting specific station Iscations at each site are: a. Ability for ~eteorological data from the site to be represeritative of a large area. ~ b. Protection afforded by nattlral objects, such as trees. 3 14 a 8e x i3 .is Ebaseo Susitrta Jaif~t Venture $A;3di.i~k $y, 1985 ["3 --. # . 3 dye c. Relationship to location where long-term data at the site have previously been collected. d. Logistical advantages offered, such as accessibility by fixed-wing aircraft f~r winter maintenance and Eocatisn on an efficient transportation route relative to the other statioils to economize aircraft logistics. e. Availability of and access by a local observer to provide "con- trol" data in the event of a system failure. 4, The 97 total proposed sites, shown in Figure 2 of the report, fairly well cover the expanse of the upper basin. However, a few advan- tages may bc offered by shifts in some of the iocations. Considering each of the sites in numerical order, with report recommendations shown in parenthesis: 1) Susitna Glacier (retain existing site). Relocation to near a small lake, a Few miles southwest of the existing site at the existing "Caribou" snow marker location would permit access by fixed-wing aircraft and still be exposed to much of the "glacier" weather conditions prevalent at the higher elevations. 2) Denali (relocate to protected site). This sounds favorable, likely utilizing the vegetated area to the north of the present site. 3) Tyone [reactivate). While this would be a good location, satis- factory data for the area could probably be obtained from sites 9, 30 and 13 around it. 4) Kosina (ml~cate Ps prsteicted site). a mare favorable facation may be higher up in the Kosina Creek basin, where a greater percentage of the annual precipitation falls. As with site 1, above, a lake is present which would permit fixed-wing access. 5) Watana (rehabilitate Wyoming gage). Agree that this site's proximity to the base camp makes it favorable. Perhaps relocate to the north near a small lake and eatiiize a loeat observer. . 6) Devil Canyon (consider small relocation). Rather than simply relocating to the trees near the present site, a complete move to near a lake on the other site of the Sarsitsla River would offer fixed-wing access as well as natural wind protection. High Lake, which has a lodge and an airstrip, is a possible site, with potential for a local observer as well. ~ 7) Monahan Flat (establish at SCS location). Agree. 8) Cathedral hake (establish at snow course location in protected site if possible). Agree. ~~~ci;zl~-sxa'-Ek~asco Susitr~a JainC Venture ?;JaS*c13 8, 1985 $age 3 * 9) Clearwater Lake {estabiish at snow course location ir, protec"led site if possible). Agree. 13) I.ake Louise (establish at snow course location in protected site if possible). Agree. Use local observer, if possible. 41) Square Lake (establish at snow course location in protected site is possible). Agree. if site 3 is deemed superfluous, as discussed above. a new installa- tion could be made in the upper Tsusena Creek basin. There is a fairly large "hole" in the data network in this area. The site would ' represent the upper elevations of the Tsusena and Deadman Creek watersheds and to same extent Watana Crcek, each of which contrib- utes to one of the two reservoirs- I$ is alss ~10563 to the headwaters of Portage Creek, the largest tributary to the Middle Susitna River and an important salmcn stream. 5, A few additional sites, besides the 71 discussed above, may ofler advantages for other specific data-collection purposes. These ate locations downstream of the project and would thus not be directly applicable to forecasting of streamflow into the reservoirs. The sites aped suggested uses of the data are described below: 1) Middle Susitna River - Since mainstem Susitna flows in this reach, from Devil Canyon to Talkeetna, are of concern for fisheries habitat and spawning, knowledge of the meteorologic conditions influencing river inflow fron, below Devil Canyon Dam is important for forecasting and impact analysis. The ex isring station at Sherman could be continued, or a never station could be established at Gold Creek. Each site wouBd offer shelter by surrounding trees. Sherman would offer arl existing record which would be extended; Gold Creek would offer potentiti1 for a laca0 observer. 2) Lower Susitna River - Temperature conditions in the lower basin are of concern because of their effect on river freeze-up and ice conditions. Some data have been collected fcrr one season at Susitna River Mile 61. 7 his station could be continued if des i red. 6, Dr. Peck has noted that the important meteorological inputs for modelling and forecasting streamflow are precipitation, temperature, and wind speed. The cne most important is precipitation, both summer and winter, The mast efficient and economics! hydro-meteorologicaI system would consist of a series of weather stations at il~tervals throughout the watershed to record and u ltirnate- ly transmit representative data to a central location. The system shou!d provide the data for calibration and operation of the forecast- ing model, yet remain flexible enough to accommodate additional future requirements. By the time the hydroelectric project has become operational, installations would most likely be "ransmitting r-c.ala.time data to be utilized in the forecasting model to a central location. 7, The discussion of ~lhich parameters are necessary to measure at each site focused on precipitation, temperature, and wind speed, and min- imized the importance of other data currently reported: wind direc- tion, relative humidity, shortwave radiation, and longwave radiation. We fee! also that these latter variables would not be cri"kical for streamf low forecasting and could be dropped from the d3ta-collection program except possibly for continued efforts during the summer at Watana. Relative humidity, in particular, currently poses gr-eat prablems in assurance of reliable data. if not required, its omission would ease the data-reduction effort conriderabiy . As noted, the t-adiation measurements are also difficult to oh%tain reliably, especial1 y in the irinter, and are not very applicable to streamflow forecasting models. Wind direction sensors have occasional winter problems, too, but the data would be useful for analyzing storm movements if it is not too difficult to continue the measurements. It is possible that measurement of temperature may be satisfactorily performed by re- cording only the daily maximum and minimum temperatures. Continued measurement of all the parameters could be desirable dur- ing the summer at Watana Camp to expand the data base for reservoir temperature modelling. Sensors could be more easily maintained on a daily basis once a camp is operated continuously. 8. Where installation of new equipment is planned for one or more sites, consideration should be given to installing mechanical stations (rather than the electronic digital instruments now in use). Advantages that could be offered include the following: " bower initial cost. Since the stations record fewer paranders and are less technically sophisticated, they are less expensive. Less expensive repairs. Sophisticated electronics in the existing stations are difficult ts repair in the field. ' Greater reliability. ' No need for a special enclosure since artificial heat is not required. Existing stations do require a shelter and heat source, 63 An immediate visual record of the data provided on strip charts. This aids equipment troubleshooting at the site and can make data-reduction less costly, since the complete range of meteoroiogic analysis would not be required. 9. In consideration of the desire to improve reliability of the data reccrd and reduce the amount of missing data, one of two approaches could be taken. The first would be to instail a back-try system of sensors and recorder at each station, which would provide a partially redundant record with the primary system. The data from the secondary would not need to be reduced unless a problem caused some of the primary data to be lost. Then only the data to replace the missing recards would need tc be obtained. Back-up equipment would consist of' a thermograph to record contini.jous temperature and a precipitation can - to permit measurem~*nt of accumulated psecipi$atisn. The sther alterna"live WOIJ~~ be to instal1 data-collection platforms at the station and tefemeter the data for daily monitoring and review at the office. In this configuration, the redundant recorders would be useful but not necessary, but a commitment would have to be made to immediately maintain the field stations when data-collection or trans - mission problems are indicated. 90. The report recor3mendation to install data-collection platforms (DCP's) at data sites as soon as possible is very agreeable in concept. This would permit opportunity to get the system up and running before it is critically needed, would make real-time data available to benefit ongoing field operations, and wouid increase the reliability of the da- ta-collection system. However, the ability to limit instrument down-time whell problems are revealed via the telemetry depends on the ability to visit the site immediately. This is naturally subject "to weather, daylight, and helicopter or other logistical considerations. Along the same line as comment number 10, we do not agree with the report suggestion to remove on-site recorders after DCP's have been installed. Even with the telemetry, back-up recording systems are needed, especially if reliability of data-collection is a concern. Ex pe- sience with data-collection in the Susitna and other remote Alaskan basins has shown that data can easily be lost from problems besides just sensor malfunctions. Losses can also occur due to malfunctions of' the transmitter, receiver, or communication link or due to deizys in access to the site caused by weather, daylight, or helicopter availability. Back-up data does not necessarily need to be reduced unless data gaps occur in the telemetry system. Measurement of pan evaporation at \l'atana has been recommended in the report. These measurements have been collected at Watana since 1951. A pan anemometer, which aids in applying the data, has not been part of the instaliation but will be installed this season. Daily observations are recorded by hydrology staff or camp logistics per- son nel . There are several statements in the report that snow surveys at existing weather stations are of no value. The reason the surveys were initiated, even though some of the areas are extreme!y windblown, was because game biologists were very interested in snow conditions in sibu, 1.e. , the snow depths that moose and caribou actually had to contend with in the open areas. Since regular visits are made ta the sites, the data are very easily obtained. The state- ment is correct, however, in describing the windblo\*rn snow courses as of little value to streamflow forecasting, so perhaps snow surveys couf:2 be performed in nearby protected areas as well. 14. As was mentioned above, retention of observers to record daily observations at selected sites would increase system reliability and provide data at times when the instrumentation goes down. Such information could even be transmitted by radio or telephone to a processing center if necessary. Observers would probably be avail - able at the foilowing sites: 12) Denali (probably less than 100% of the year) (5) VJatana (as long as the camp is occupied) (6) Devil Canyon (if moved to High Lake Lodge) $90) Lake Louise An additional alternative would te relocation of the Devil Canyon station to Gold Creek instead of to High Lake. Goid Creek is beiow both damsites but would represent Middie Susitna areas (in place of the existing Sherman statics) and would offer reliable railroad per- sonnet as observers. 15. Rough cost estimates have been developed for various instrumentation alternatives considered. These are listed below: a. Approximate cost sf continuing existing stations (including recommended improvements to each): 84,400 per station $38,860 for 7 stations Costs for reduction, review, handling, editing, and reporting of the electronic station data are estimted to be $1,230 per month per station, including labor and computer costs. Reported data would include many of the same precipitation, temperature, and wind parameters currently reported (since the software al~eady exists) : 0 Precipitation (hour*ly and daily totals) ~ €3 Temperature (daily min/max/average) 0 Wind (daily resultant speed and direction, daiiy average speed, daily peak gust speed and direction, and daily prevailing direction. Wind roses could also continue to be prepared if desired. Wind sigma theta should continue to be measured at Watana, and reported when needed.) The back-up data system would permit reporting cf daily min/max/average temperatures and total accumulated precipitatjon since the: last inspection. The above estimate includes constructiibn of new shelters, relocation where necessary, construction of \Vyoming wind gages, ;j cb i+z,;3 - 21~~~ er ,-~2 , Susitna Joint Venture h il rv.,i;::*cF8 8, "Ega5 Page 7 a and installaticrn of a bac1.a:-up data recorder at each site. Elimination sf the back-up recorder would redldce the per-station cast approximately $1,200 (58,400 for 7 stations). The estimate does not include telejnetry costs or helicopter costs Crpr sling-loading tc new sites. These stations could be improved in turn, over several years if des i red, b. Approximate cost of acquiring mechanical weather stations: 86,600 per station $26,480 fw 4 sPa$lsns Costs for reduction, review, handling, editing, and reporting of the mechanical station data are estimated to be $1,250 per month per station. Reported data would include: Q% Precipitation (dai,ly total) o Temperature (daily min/max/average) 8 Wind (daily total wind run or daily average speed) The back-up system would permit reporting of daily minimax/average temperatures and total accumulated precipitation since the fast inspection, he estimate includes purchase of mechanical instruments (precipitation, temperature, wind), purchase of a backup system to provide redundant measurements, instailation of both systems, and construction of a Wyoming wind gage at each site. As above, elimination of the back-up system would reduce the per-station cost approximately $1,200 (54.800 for 4 stations). The estimate does not include telemetry costs or helicopter logistics for slil~g-loading to site. c. Approximate cost of hiring local observer for daily observation: $1,500 + $300/month at each station Costs for review, handling, and reporting of the observer data reports are estimated to be an additional 8250 per month per station. Reported data. would include the same values listed above for the mechanical shtians : EB Precipitation (daily tetal) o lemperatu re (daily min/max/average) $B Wind (daily total wind run or daily average speed) The estimate includes purchase 04 instruments (min/max thermometer, totalizing anemometer, and accumldlating precipitation can), instrument shelter, Alter wind screen. and installation. Observers to be paid $10 per day. 76. Another consideration for measurement of precipitation at higher elevations, where greater amounts of precipitation fall, is installation of large-volume storage gages and recorders. The advantage of the larger site is that the danger of the gage overflowing and losing data during large rainstorms and snowstorms is reduced. The SCS currently operates one of these at the Monahan Flat site. The cost of installing additional storage gages is estimated to be approximately 55,000 per site, not including helicopter logistics. Additional installations would be beneficial at sites 1, 4, 8, and 11 in the list under comment 4, above, and at the Upper Tsusena site if established. Operation of the 11-station network would increase field labor time and helicopter time during maintenance trips. The six existing stations can normally be maintained on a 1-2 day trip. A total of two to three days would probably be required to maintain the full 11 staticlns. Helicopter usage would be approximately four flight hours per day. The stations should continue to be inspected and main- tained once per month to verify their proper operation. The logistics costs for maintenance trips would be slightly higher when Watana Camp is closed, since helicopter flights would originate and end in Taikeetwa instead sf at the camp. -8. An alternative jogistics plan which should be considered in the net- work planning process is use of fixed-wing aircraft for access to the meteorologic stations. All stations recommended above for future station locations should be accessible by fixed-wing airplane most of the time, with the exception of the Upper Tsusena site suggested. For estimation purposes, using a helicopter cost of S320 per hour (Jet Ranger) and a fixed-wing cost of S2OO per hour (Cessna 2061, the foilowing are approximate monthly costs for each alternative: 1) Helicopter (2 days @ 6 hours per day 12 hours per month) 2) Fixed-wing (2 days @ 6 hours per day = 12 hours per month, plus 1 hour from Watana Camp by helicopter) 19.- As is emphasized in the report, data for streamflow forecasting will have to be available in a timely manner. This will best be accom- plished by telemetry of the data from the stations, either by the telephone-repeater network or the GOES satellite system, and then incorporation into the data storage and n~odeliing system, as Dr. Peck hss discussed. The alternative selected depends primarily on other communication requirements of the project. In summary, our recommendations at the present are: 7) To make sure field data needs of other environmental and engineering studies are considered and coordinated in design of the data-collection natwark. This wpuld include terrestrial studies, air quality management, river ice monitoring, and fisheries monitoring, as well as the water s~pply forecasting. 2) Operate a total of eleven meteorological stations i~ the basin above the Devil Canyon darnsite. Select sites in the vicinity of those proposed in the report in Figure 2 and Table 2, modified as discussed above in ccsmments number 3 and 4. 3) Continue precipitation, temperature, wind speed, and wind direction measu sements , 4) if not required for other purposes, measurement of relative humidity, solar radiatiop. and longwave radiation should be discontinued (with the possible excQption of Watana). 5) The existing electronic digital recording mete~rolagical stations should be maintained, There are seven rec~rders and associated sensors available for field use. The eighth unit also currently available should Be retained as a spare. The field installations and she1"rer.s should be upgraded as described above. 6) Mechanical recording meteorological ;tations should be acquired and installed at 'Four new sites to measure precipitation, "temperature, and wind speed. 7) Hire local observers where possible to improve data reliability. This could be done at Denali, Watana, Dwii Canyo11 and Lake Louise, and Gsid Creek. Bb install telemetry at one or more meteorologic stations to start develop- ment of the real-time data-collection system. Recorders should be retained at the stations even after installation sf data-coIIect'ron platforms. 91 Install back-up sensor and recorder systems to provide redundant measurements at stations ivfaich do not have either a locat observer- or a data-collection platform. !,i;4n-?*s-,lbaseo Susitnzi Joint Venture a [: 8, 198s 1"". r;ge 10 10) Continue snow surveys at existing weather sta4:lons but also make measurements in nearby protected areas to imprave the data quality for forecasting and for development of isohyetal maps. If) Consider conversion of the data-collection program to at least a partial fixed-wing aircraft network by selecting compatible station Iscations , It is expected that these recommendations will be considered along with those of Dr. Peck in his report, and the data-coilection programs for FY86 and beyond kill be refined over the next several months. We look forward to working with you in the system planning effort. Thank you for "ihe opportunity to comment on the report. If you have questions or comments on any of the above material, please do not hesitate to contact Jeff Coffin or myself. Very truly yours, Rdf4 CONSULTANTS, INC. Stephen Bredthauer, P.E. Susif na Hydt-ology Coordinator SB:JHC; bje cc: 8.H.Wang G. Gemperline