WATER RESOURCES RESEARCH, VOL. 33, NO. 7, PAGES 1707-1719, JULY 1997

Hydrograph separations in an Arctic watershed using mixing model and graphical techniques

JamesP. McNamara, Douglas L. Kane, and Larry D. Hinzman Water and EnvironmentalResearch Center, University of Alaska Fairbanks

Abstract.Storm hydrographs in the Upper Kuparuk basin (142 km 2) in northern Alaska were separatedinto sourcecomponents using a mixingmodel and by recession analysis.In non-Arcticregions, storm flow is commonlydominated by old water, that is, water that existedin the basinbefore the storm.We suspectedthat this may not be true in Arctic regionswhere permafrostdiminishes subsurface storage capacity. during the snowmeltperiod was nearly all new water. However, all summerstorms were dominatedby old water. Stormsin a neighboringbasin were dominatedby new water but much lessthan was the snowmeltevent. Thus a large increasein old water contributions occurredfollowing the snowmeltperiod. This increasecontinued moderately through the summerin 1994 but not in 1995. We credit the seasonalchanges in old water contributionsto increasedsubsurface storage capacity due to thawing of the active layer.

Introduction suspectedthat storm flow may not be dominatedby old water, as is commonlyobserved. Permafrostis a ubiquitouspresence in the Arctic that influ- An analog for permafrostbasins may be watershedson the encesnearly all physicaland biologicalecosystem processes. southernCanadian Shield, where impermeablebedrock under- Severalstudies have shownthat permafrosthas significanthy- lies shallow soils. However, several workers have shown that drologicalconsequences which resultprimarily from the min- storm flow is indeed composedprimarily of old water in Ca- imal subsurfacestorage capacity due to frozen ground [Hin- nadian Shield watersheds,even with their diminished old water zman et al., 1993; Dingman, 1970; McNamara et al., 1997; reservoirs[Peters et al., 1995;Bottomley et al., 1986; Welset al., Rouletand Woo, 1988; Woo and Steer,1983]. This is of partic- 1991; Hinton et al., 1994]. An important distinctionbetween ular concernto the NSF Land-Air-Ice-Interaction(LAII) Arc- tic Flux Studyoperating in the Kuparuk River basinin north- Canadian Shield watershedsand the Kuparuk River basin is ern Alaska.The goal of the Arctic Flux Studyis to estimatethe that the subsurfacereservoir and consequentbasin storage regionalfluxes of massand energyin the Kuparuk River basin capacityin the Kuparuk River basin increasesas the ground between the land, the atmosphere,and the Arctic Ocean thawsduring the summermonths from essentiallyzero depth [Welleret al., 1995].This requiresa comprehensiveunderstand- in the to depths approachingthose in the Canadian ing of the mechanismsand pathwaysby which water travels Shield watersheds late in the summer. Other studies have throughthe system.Hence we investigatedthe compositionof shown that certain hydrologicprocesses undergo coincident stormflow in the Kuparuk River basinby askingthe following changeswith the thawingactive layer. Hinzman et al. [1991] questions:Is storm flow primarily composedof precipitation, showedthat the portion of the soil profile that contributesto callednew water, or subsurfacewater previouslyexisting in the hillslope runoff increasesthrough the summer. Further, Mc- basin, called old water, and what influence does permafrost Namara et al. [1997] suggestedthat runoff/precipitationratios have on storm flow composition?An understandingof both may decreaseas the active layer thicknessincreases. Thus we the partitioningof hydrographsand the mechanismsresponsi- suspectedthat the systematicincrease in activelayer thickness ble for the partitioningis a prerequisiteto understandingthe would produceconsequent changes in stormhydrograph com- relationshipsthat existbetween terrestrial and aquaticsystems. positionsthrough the summer. Severalcase studies in variousnonpermafrost regions have The specificobjectives of this paper are (1) determinethe shownthat old water typicallydominates storm hydrographs, proportionsof old and new water in stormflow in the Kuparuk includingsnowmelt events [Buttle and Sami, 1992;Bottomley et River basinduring 1994 and 1995,(2) investigatethe potential al., 1986;Dincer et al., 1970;Eshleman et al., 1993;Kennedy et influences,particularly of permafrost,on storm flow composi- al., 1986;Kobayashi et al., 1993;McDonnell et al., 1991;Rodhe, tion. Storm flow compositionswere determined from hydro- 1981;Peters et al., 1995]. This may have significantinfluences grhph separationsusing mixing model and graphical tech- on the transportof nutrientsfrom the terrestrialto the aquatic niques. The influences on storm flow composition were system,a primary area of researchin the Kuparuk River study. investigatedby constructingcorrelation matrices with variables The old water reservoirin a basinwith permafrostis severely restricteddue to the frozen ground.Essentially all subsurface includingold water composition,precipitation characteristics, flow occursin a shallowzone called the active layer that un- total flow, and storm date as a surrogatefor activelayer thick- dergoesannual freezing and thawingcycles. Consequently, we ness.Runoff generatingmechanisms were qualitativelyevalu- ated usinga techniquedeveloped by Eshlemanet al. [1993] to Copyright1997 by the American GeophysicalUnion. computecontributing areas based on hydrographseparations. Paper number 97WR01033. We focusedon summer stormsin the Upper Kuparuk River 0043-1397/97/97WR-01033509.00 basin(Figure 1), with a limited analysisof snowmeltprocesses.

1707 1708 MCNAMARA ET AL.: HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED

Alaska Location Map Inmavait Creek Watershed Area = 2.2 k•n2 •••'•Water track 7 = .026 k•

/ \ / Fairbanks

Watershed N Boundary / 0 500 km / •=•o.

Kuparuk Watershed '• U•perKuparuk Watershe•/' Area •Area = '142 kmz ./ . •. Arctic / west DOCk•• ••ean //

Watershed .

w•s•. g •1 Sa•w•n

/

K••kU99er •I=avait Creek

A A A A A A - A -- • . A .... • Watershed I I • • • Boundary Figure 1. Map showinglocations of the studysites in the KuparukRiver basin, northern Alaska. This study focusedon the Upper Kuparuk River basin (142 km2), with a limitedanalysis of ImnavaitCreek (2.2 km2).

Additional analyseswere performed in the much smaller This studyfocused on the Upper KuparukRiver, a headwa- neighboringImnavait Creek (2.2 km2). ter basin which drains 142 km 2 in the northern foothills of the BrooksRange. The slopesin the basin are coveredwith till from two glacialadvances, Sagavanirktok and Itkillik, from the Site Description middle and late Pleistocene[Hamilton, 1986]. At the intersec- The KuparukRiver flowsfrom the glaciatedfoothills of the tion with the Dalton highwaythe Upper Kuparuk River is a BrooksRange throughthe tundra flats of the coastalplain to fourth order on a U.S. Geological Survey (USGS) the Arctic Ocean near PrudhoeBay (Figure 1). The entire 1:63360 map. However, the hillslopesand valleys regionlacks trees, is underlainby continuouspermafrost, and containa complexnetwork of smallstreams that do not appear is coveredwith snowfor 7-9 monthseach year. The snowmelt on mapsat that scale.Two dominantstreams join togetherat event is generallythe dominanthydrologic event eachyear, the base of steep hills in the upper basin forming the main which typicallyoccurs over a 7-10 day period betweenearly channelwhich occupies a north-northwesttrending . The May andearly June [Kane et al., 1991].Approximately 30-40% main basinlength is 16 km, with a channellength of 25 km. of the annual precipitation falls as snow from September Vegetation in the basin consistsof alpine communitiesat throughMay. The averagesummer rainfall is around18 cm in higher elevationsand moist tundra communities,predomi- the foothills of the Brooks Range. The maximum snowfallis nantly tussocksedge tundra, at lower elevations.Patches of typically10-14 cm of water equivalent.Summer temperatures dwarfwillows and birchesup to 1 m in heightoccupy portions are typicallybetween 6 ø and 18øC,and winter temperatures are commonlyaround -15 øto -25øC. Permafrostthickness ranges of the banks.The averageelevation of the basin is 967 m. from around 300 m near the foothills to over 600 m near the ImnavaitCreek (2.2 km 2) is a smallbeaded stream occupy- coast[Osterkamp and Payne,1981]. Hence the regionis effec- ing a north-northwesttrending glacial valley which was formed tivelyisolated from deepgroundwater. Subsurface flow occurs during the Sagavanirktokglaciation (middle Pleistocene) in a shallowzone abovethe permafrostcalled the activelayer [Hamilton, 1986]. The dominantvegetation in the Imnavait whichundergoes annual freezing and thawing.Soils typically basinis tussocksedge tundra covering the hillslopes[Walker et thaw to maximumdepths of 25-40 cm but can thaw to 100 cm al., 1989].An organiclayer typically near 10 cm thick,but up to dependingon severalenvironmental factors including soil type, 50 cm thick in the valleybottom, overlies glacial till, where the slope,aspect, and soilmoisture [Hinzman et al., 1991]. soil rarely thawsdeeper than the extent of the organicpeat MCNAMARA ET AL.: HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED 1709 layer. The creek is essentiallya chain of ponds,called beads, flow, the new componentof the flow, and the old component that formed where the stream has eroded and melted massive of the flow, respectively.The streamflowattributed to old wa- ground-icedeposits. The streambottom rarely cutsthrough to ter at any time t is mineral soil but maintainsitself in the organiclayer. Imnavait Creek flowsanother 12 km beyondour stationbefore it joins Qo(t) = Q,(t)(C•(t) - Cn(t))/(Co(t) -- Cn(t)) (3) the Kuparuk River. and the new water flow is The hillslopesin the Kuparuk River basin are drained by a network of water tracks. A water track is essentiallya linear Qn(t) = Q•(t) - Qo(t) (4) zone of enhancedsoil moisturethat flowsdirectly down a slope Instantaneousproportions of total streamflow arising from and is best detectedby a changein vegetationfrom the sur- either new or old water are Qn(t)/Qs(t ) and Qo(t)/Qs(t), roundinghillslope [Hastingset al., 1989; Walkeret al., 1989]. respectively.The values of Qn, Qo, Qo/Qs, and Q•/Qs are Water tracksare generallyspaced tens of meters apart on the obtainedby summing(3) and (4) over the duration of the hillslopes,although their densityvaries [Walkeret al., 1989]. storm. Only intermittently do incisedchannels exist in water tracks, The natural tracer techniquerequires that the tracer signa- but they are significantcomponents of the hillslopehydrologic tures be conservative;that is, they do not changethrough a cycle.They often end in peat coveredvalley bottomsthrough storm, or the changescan be correctedfor. The signaturesof which water travels to the as diffuse subsurface flow old andnew water must also be distinct.Specific conductivity is through the active layer, or overland flow during extreme not entirely conservativebecause water dissolvessolutes events. as it passesthrough the soil. However, we believe our tech- nique, describedbelow, correctedfor changesin specificcon- Hydrograph Separation ductivitydue to soil contact. Early techniquesto separatestorm hydrographsinto source componentsinvolved graphical separation, or recessionanal- Definition of End Members ysis, to determine the portion of storm flow that originates Given the potential for spatial heterogeneityof soil water from groundwateror . The shapeof the hydrograph chemistry,it is difficultto obtain an accuratechemical signa- recessioncurve is usedto decipherthe timing and magnitude ture of the old water component.For this reason a common of surface and subsurfacerunoff. Newer, more physically based,techniques involve separating hydrographs into source method for estimatingthe old water componentis to assume that the stream water during the low flow periods between componentsusing naturally occurring tracers. Simple two- stormsrepresents an integratedsample of the old water in the componentmixing models are usedto partition stormflow into old and new water assumingflow sourceshave distinctchem- basin.A continuousrecord of specificconductivity during low flow periodsthen providesa simpleestimation of the old water ical or isotopicsignatures, where old water is water that existed component.However, Figure 2 showsthat the specificconduc- in the basinprior to a storm (i.e., soil moistureand ground- water) and newwater is the rain or snowmeltcontributed by a tivity of the Upper Kuparuk River during low flow periods increasedthrough the seasonso that poststormvalues were storm or snowmeltevent. Commonlyused tracersinclude ox- ygen isotopes,chloride, and specificconductivity. Advances in typicallyhigher than prestormvalues. We usedlinear interpo- hydrographseparation techniqueshave expanded the two- lation betweenthe prestormspecific conductivity and the post- storm specificconductivity to estimate the instantaneousold component mixing model to include soil water and deep water signaturesduring a storm[Hooper and Shoemaker,1986]. groundwateras separatecomponents in a three-component model [DeWalle et al., 1988]. The two-componentmixing In caseswhere the poststormconductivity was lower than the prestorm conductivitydue to dilution from the next storm we modelis acceptablein this studydue to the absenceof a deep groundwatersystem. used a constantconductivity through the storm equal to the We used specificconductivity as the primary tracer and prestormconductivity for the oldwater signature. For the •80 old water signature on storm 7 we used a constant value comparedthose resultsto recessionanalysis of the same hy- drographs.We used•80 as the tracerfor one of the hydro- throughthe stormequal to the stream•80 contentimmedi- ately prior to the storm. graphsas a checkon the specificconductivity approach and to The new water specificconductivity was more difficult to evaluate flow sourcesduring the snowmeltperiod. We used recessionanalysis when possible,but severalstorms contained estimate.Typically, the signatureof precipitationfor the event is used as the new water end-member.However, Pilgrim et al. multiple peaks, which made recessionanalysis impossible. Conductivitysignals of new water and old water in Imnavait [1979] showedthat the specificconductivity of dilute water changeswith soil contact time and is not a reliable end- Creek were often too closeto allowhydrograph separation by the mixing model. Consequently,we were only able to use member. They developed laboratory relationshipsfor the changesin specificconductivity and foundthat the conductivity recessionanalysis in Imnavait Creek. changeddramatically during the early stages,then approached Mixing Model equilibriumand increasedmore slowly.They then used these The mixingmodel is basedon the steadystate form of the relationshipsto correctfor the changingnew water signature, massbalance equationsfor water and a conservativetracer, successfullyseparating storm hydrographs using corrected spe- cific conductivity. Qs(t) = Qo(t) + Qn(t) (1) We estimatedthe new water signatureby measuringthe Qs(t)c•(t) = Qo(t)Co(t) + Qn(t)Cn(t) (2) specificconductivity of runoff in a small hillslopewater track whichdrains an areaof 0.026km 2 on thewest facing slope of where Q is the flow rate, C is the tracer concentration, t is the Imnavait Creek basin.We assumedthat the conductivityin time, and the subscriptss, n, and o refer to the total stream- the water track during the falling limb of a storm hydrograph 1710 MCNAMARA ET AL.: HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED

6O 50 .; •

E 203o u

3O 0 u•

Snowmelt • 20 5 E 3 4 1 6 7

o 1-May 21-May 10-Jun 30-Jun 20-Jul 9-Aug 29-Aug 18-Sep 1994

60 .•, Upper Kuparuk River Conductivity SøE

30•E 20 •-

....II Condu,ctivity 10

0 30 Snowmelt 8• E 20 7

•-•0 10 I 231 4 6 I 10

o 1-Ma• 21-Ma• 10•Jun 30-.Jun 20•Jul 9-Aug 29-Aug 18-$ep

1995 Figure2. Hydrographsand specific conductivity records in theUpper Kuparuk River basin for (a) 1994and (b) 1995.The stormsanalyzed in thisstudy are numbered chronologically.

was not significantlyinfluenced by old water and that the in- duration of the storm for the •80 content of new water on creasein specificconductivity of the water track duringthe storm 7. falling limb of a hydrographrepresented the slowincrease in specificconductivity as the rain waterapproaches equilibrium RecessionAnalysis with the soil.Hence, for eachstorm in the Upper Kuparuk Graphicalhydrograph separation has received considerable Riverbasin, the conductivityof the correspondingstorm on the criticismas there is no physicalbasis for its assumptions water track at peak flow was used as the new water end- [Freeze,1972]. Dingman [1994, p. 384] calledthe technique member. A potential error in this method is that the water "convenientfiction." However, recession analysis can provide a track runoffmay indeed have been influenced by old water.If qualitative way to evaluate runoff mechanismswhen tracer so, our estimatesof newwater conductivitymay be too high, techniquescannot be used.The techniquebecomes difficult to whichwould resultin erroneouslylow computationsof old useon complexhydrographs such as thosefrom overlapping water contributions.However, the specificconductivity in the stormsor from a rain on snowevent. Hydrograph recessions water track rangedbetween 6 and 9/•mho/cm, whichis closeto typicallyfollow an exponentialfunction. If plottedon a semi- that of local precipitation.Therefore underestimationof old logarithmicgraph with dischargeon the logarithmicscale, the water contributionswill not be significant.The specificcon- recessionshould be a straightline [Linsleyet al., 1982].How- ductivityof the water track was distinctlylower than the Ku- ever,the actualrecession typically occurs in separatelog linear paruk River for all summerstorms, which allowedus to use the segmentswith differentslopes for the differentsources of run- two-membermixing model for all storms.We used the •80 off. In a two-membersystem, surface flow and subsurfaceflow, contentof a bulk precipitationsample collected through the the breakin slopeis assumedto be the pointwhere surface MCNAMARA ET AL.: HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED 1711 runoff ceases.The remaining recessionis due to subsurface Upper Kuparuk River basin and identifiesthe stormsused in base flow. The old water recessioncurve is obtained by pro- this study.Storms 1, 2, 3, 4, 6, and 7 from 1994 and storms2, jecting the line representingbase flow recessionbackward in 3, 4, 6, 7, 8, and 10 from 1995 were separatedinto sourceflow time to the correspondingpeak of the hydrograph.A linear fit componentsusing the mixingmodel with specificconductivity from the initial stormresponse to the old flow peak completes as the tracer. Storms 2, 3, 4, 6, and 7 from 1994 storms and the old flow hydrograph.The proportionof old water contrib- storms3, 4, 6, and 7 from 1995 were separatedusing recession uting to the storm is calculatedby dividingthe area under the analysis.Storm 7 from1994 was also separated using •80 asthe old water hydrographby the area under the total hydrograph. tracer in the mixing model. Figures 3a-3c show the results for storm 7 of 1994 as an exampleof eachtechnique. The old water contributionfor this Field Methods stormcalculated by the conductivitymixing model, the •80 Streamflow was monitored at the Upper Kuparuk River mixing model, and recessionanalysis was 79, 81, and 81%, basin and Imnavait Creek outlets using stilling wells with respectively.These favorable comparisons confirm that hydro- Stevens F-1 water level recorders mounted with variable resis- graph separationusing specificconductivity as a tracer is ac- tance potentiometersto obtain digital data. Campbell Scien- ceptable. The old water and new water conductivityvalues tific CR10 data loggersrecorded stream stage every minute usedin the calculationsare shownon the plots. The calculated and averagedover hourly increments.An H-type was old water contributionsfor each storm using chemical and used at Imnavait Creek to aid dischargemeasurements. Dis- recessionseparation are shown in Table 1. The old water chargemeasurements were made followingUSGS standardsat contributionsfrom the mixingmodel rangedfrom 65 to 81% in severaldifferent stagesto produce rating curvesfrom which 1994 and 53 to 83% in 1995 with averagesof 72 and 68%, continuousrecords of dischargewere calculated. Two com- respectively.These resultsindicate that old water dominated plete meteorological stations recorded precipitation, wind storm hydrographsin the Upper Kuparuk River basin.Reces- speedand direction,air temperature,relative humidity, and sion analysisyielded similar results.Points would fall on the variousradiation terms. Hydrographswere producedfor both diagonalline on Figure 4 if the two techniquesproduced iden- basinsfrom the initiation of snowmeltin early May to just prior tical results.Although there is some scatter and clustering, to freeze-up in mid-September.To monitor activelayer thick- there is fairly good agreementbetween the two techniques, ness,thaw depthson a ridge top and on the west facing slope whichlends credence to the widelyused but physicallyunjus- in the Imnavait Creek basinwere estimatedby tracking the 0ø titled graphicalmethod of hydrographseparation. Kobayashi et isothermusing thermistors at variousdepths between 0 hnd al. [1993] alsofound that tracer techniquesand recessionanal- 150 cm. ysisproduced similar results. Specificconductivity was loggedhourly at the Upper Kupa- Eshlemanet al. [1993] found a negativecorrelation between ruk River, Imnavait Creek, and the water track usingCampbell precipitation intensity and old water contribution to storm Scientificconductivity probes. Water samplesfor oxygeniso- flow. Table 2 shows that no such correlation existed in the tope analysiswere collected every 3 hours during several Kuparuk River basinat the 5% significancelevel (a = 0.05). stormsusing an Isco automatic sampler and by hand during Table 1 containsthe supportingprecipitation data. Other po- snowmeltand betweenstorms in the summer.Snowpack melt- tential influences on old water contribution include total run- water was collected by digging a snow pit and inserting a off, total rainfall, rainfall duration,and activelayer depth. We high-densitypolyethylene tray at the base of the snowpack used storm date as a surrogatefor depth of thaw to test for before the initiation of melt. The pit was then covered with seasonaltrends as a result of active layer thawing. Table 2 foam board to reduce melting of the pit wall. All of the melt- showsthat the only significantcorrelation to old water contri- water in the tray at the end of each day was collectedusing a bution was storm date in 1994. There were no significantcor- plasticsyringe. Water samplesfor oxygenisotope analysis were relations to old water contribution in 1995. collectedwith no head space in glassscintillation vials and Additional resultsbased on the tracer separationsfollowing storedin a cool, dark room until they were analyzed.Isotopic the format of Eshlemanet al. [1993] are includedin Table 1. analysisof rain was completedonly on samplescollected for The newwater contributingarea (NWCA) is an estimateof the stormsin August 1994. Only bulk precipitationsamples were area of the basin that producesdirect runoff during a storm collectedthrough the storms.Consequently, we are unable to and is computedby dividingthe newwater flowvolume by the addressthe isotopicvariability of rain. Oxygen isotope mea- correspondingtotal rainfall for the event. The new water con- surementwas performed at the Universityof Alaska Fairbanks tributing portion (NWCP) is the NWCA dividedby the total Water and EnvironmentalResearch Center by extractingCO2 watershedarea and is an estimate of the percentageof the from the water samplesusing a vacuum extractionline, then watershedthat producesdirect runoff during a storm.In 1994, analyzingthe gas for •80 contenton a VG Isogasseries 2 mass NWCP in the Upper Kuparuk River basinranged from a high spectrometer. of 34% early in the seasonwith a decreasingtrend to a low of 5% later in the season.In 1995, NWCP rangedbetween 73% earlyin the seasonto 18% at summersend, althoughthere was Results no seasonal trend between the two extremes. The specific conductivitiesfor Imnavait Creek were too Summer Storms close to those for the water track to use the mixing model Both 1994 and 1995 were unusuallywet summerswith fre- technique for separatingImnavait Creek hydrographs.This quent storms.In 1994, 275 mm of rain fell at our gaugein the could mean that storm flow in Imnavait Creek is primarily Upper Kuparuk River basin, and 274 mm fell in 1995. The composedof new water, or that old water in the basin has 9-year averagerecorded at a nearby gaugein Imnavait Creek undergonevery few chemicaltransformations during its resi- was 183 mm. Figure 2 showsthe resultinghydrographs for the dencein the basin.Thus we usedrecession analysis to calculate 1712 MCNAMARA ET AL.' HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED

a) 5 50 Old Water Conductivity

40 • Stream Conductivity E

Total Hydrograph 30 E • 3 v

.>

20 = Old Water -- 79% o

•Ne lO •

• I wWater I NewWaterI Conductivity I o 6-Aug 7-Aug 8-Aug 9-Aug 10-Aug 11-Aug 12-Aug

b) 5 -16

New Water de1480 - -17 4 Total Hydrog

- -18

O

-2O

1 18 • Streamdel O -21 ,/ Old Water del •80

I' I I I I -22 6-Aug 7-Aug 8-Aug 9-Aug 10-Aug 11-Aug 12-Aug

c) 10

Total Hydrograph ••• SurfaceRunoff Recession .

0 6-Aug 7-Aug 8-Aug 9-Aug 1O-Aug 11-Aug 12-Aug 1994 Figure 3. Storm7, 1994hydrograph separations. (a) Specificconductivity as a tracerresulted in a calculated oldwater contribution of 79%. (b) Oxygenisotopes as a tracerresulted in a calculatedold water contribution of 81%. (c) Recessionanalysis resulted in a calculatedold water contributionof 81%. MCNAMARA ET AL.' HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED 1713

Table 1. HydrographSeparation Results for the Upper Kuparuk River

Total Total Rainfall New Water New Water Old Water Old Water , Rainfall,* Intensity, Discharge, Mixing Model, Mixing Model, Recession, NWCA,? NWCP,$ Storm Date m3 mm mm/hour m3 % % % km 2 %

1994 1 June 17 2049238 19.8 0.25 837964 35 65 NA 42 29.7 2 June 25 2007369 17.2 0.43 841169 33 67 66 49 34.4 3 July 4 2505238 33.9 1.02 1000386 30 70 65 29 20.8 4 July 12 1543135 19.6 1.97 450912 31 69 63 23 16.2 6 July 29 1115632 28.2 4.97 216286 19 81 75 8 5.4 7 Aug. 7 873229 12.3 0.56 192928 22 79 81 16 11.1 average 28 72 70 28 20 1995 2 June 4 2035175 6.7 0.13 691960 34 66 NA 104 73.2 3 June 11 1150804 5.2 0.26 195637 17 83 88 38 26.5 4 June 15 2380086 17.8 1.77 809229 34 66 67 45 31.9 6 July 1 2106950 18.1 1.12 505668 24 76 77 28 19.7 7 July 9 2627313 20.0 0.06 1234837 47 53 47 62 43.5 8 July 17 7499286 39.8 0.45 3299686 44 56 NA 83 58.4 10 Aug. 12 2917276 28.7 0.24 758492 26 74 NA 26 18.6 average 32 68 70 47 33

NA, not applicable. *Measured near Kuparuk headwatersstream gauge. ?New Water ContributingArea equal to new water flow volume/precipitationvolume. $New Water ContributingPortion equal to NWCA/basin area.

stormflow compositionsin Imnavait Creek. Most of the storm Snowmelt hydrographsin Imnavait Creek were complicatedby multiple peaks, which made recessionanalysis unreliable. Hence we We were unable to perform accuratehydrograph separa- were only able to perform recessionanalysis on storms3 and 6 tionsduring snowmelt due to the difficultiesin obtainingrep- in 1994 (Figure 5). Both stormsin 1994 had old water contri- resentativeend-member samples. However, the trendsin •80 butionsof 41%, indicatingthat newwater dominatedthe storm content and conductivityin the streamsduring snowmelten- hydrographsin Imnavait Creek. The lack of usable storms ablereasonable approximations. The •80 contentin theUpper prohibited determination of whether or not a seasonaltrend Kuparuk River, Imnavait Creek, and the water track increased existed. However, storm 6 is late in the season and still has a dramaticallythroughout snowmelt in remarkablysimilar pat- relativelylow old water contributioncompared to the 75% for terns (Figure 6). Cooperet al. [1993] reportedsimilar data in the recessionanalysis of the Upper Kuparuk River on the same Imnavait Creek. This initially appearsto representmixing of date. NWCPs for storms 3 and 6 in Imnavait Creek were 14 waters with distinct isotopic signatures.However, meltwater and 13%, respectively. collected immediately under the snowpack that had not

9o

o= 85

80

'o 75

• •-70

._o• 65

*• 60 o

0 50

45 45 50 55 60 65 70 75 80 85 90 Old water contributioncalculated by mixing model (%) Figure 4. Plot showingthe relationshipbetween results obtained from the mixing model analysisand recessionanalysis. Points would fall on the diagonalline in a perfect relationship. 1714 MCNAMARA ET AL.: HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED

Table 2. Correlation Matrices of Potential Influences on Old Water Contributions to Storm Flow

Storm Total Total Rainfall Rainfall Old Date Flow Rainfall Duration Intensity Water

1994' Storm Date 1 Total Flow -0.85 1 Total Rainfall -0.07 0.37 1 Rainfall Duration -0.69 0.43 -0.47 1 Rainfall Intensity 0.48 -0.36 0.71 -0.79 1 Old Water 0.93 -0.77 0.17 -0.64 0.62

1995P Storm Date 1 Total Flow 0.46 1 Total Rainfall 0.52 0.90 1 Rainfall Duration 0.43 0.93 0.75 1 Rainfall Intensity -0.11 -0.22 0.16 -0.50 1 Old Water -0.14 -0.59 -0.46 -0.42 0.01

*Correlationsbetween +0.62 and -0.62 are not significantat the 5% significancelevel. ?Correlationsbetween +0.61 and -0.61 are not significantat the 5% significancelevel.

reachedthe soil had a similar trend. This suggeststhat enrich- ing of fractionatingmeltwater sources without interactionwith mentof •80 in streamwaterduring snowmelt is dueto isotopic subsurface waters. fractionation,as opposedto mixing of differentsource waters. Without continuousdata we could not partition the hydro- Cooperet al. [1993]suggested this explanationto explainheavy graphsfor the snowmeltperiod. However, we calculatedthe isotopeenrichment in their data but did not have the meltwa- old water contributions at one time instant, the latest time for ter data to confirm their explanation. which we have meltwater isotopicdata. At that time, the melt- Thawed soil moisturethat originatedfrom the previousfall waterhad an •80 contentof -26.4 ppt,and the Kuparuk River precipitationis the potential old water sourceduring snow- had an •80 contentof -26.0 ppt. Usingthese values in the melt. Cooperet al. [1991,1993] reported soil moisture•80 two-componentmixing model (equation(3)) resultedin an old contentsaround -20 parts per thousand(ppt) in 2 different water contribution of 7%. If the valley bottom meltwater data years in the Imnavait Creek basin and reported that the vari- was indeed lighter than the basin average,then the old water ability around the basin was minimal. Further, the isotopic contributionwould be even lower. It is possiblethat -20.5 ppt contentof the water track throughmost of the summerof the is a heavy estimateof the soil moisture and that fractionation Cooperet al. [1993] studywas closeto their estimationof soil occursas soil moisturethaws, as well. If we subtract2 ppt from moisture isotopiccontent. In 1994 the isotopecontent of the the estimateof soilmoisture •80 content,which was the ap- water track rose to a plateau around -20.5 ppt. These favor- proximaterange of fractionationin the snowmeltwater data, able comparisonssuggest that this value can be used to ap- then the old water contribution to snowmelt increases to 10%. proximate the potential old water sourceduring snowmelt. This was on the risinglimb of the hydrograph,where old water That the •80 contentsin the KuparukRiver and Imnavait contributionsare typically highest. Cooper et al. [1991] re- Creek remain distinctlylower than the estimated soil water ported an old water contributionto stormflow of 14% at peak value further suggeststhat these waters are almost entirely flow in Imnavait Creek. derived from melting snow. The patternsin conductivityduring the snowmeltperiod for Cooperet al. [1993] reported that the lightestsnow (lowest the water track and the Upper Kuparuk River (Figure 2) con- •80 content)in theImnavait Creek basin occurred in thevalley firm the above results.The high conductivityat the onset of bottom.This may explain why the •80 valuesof thewater track snowmeltrunoff resultsfrom soluteexclusion in the snowpack. were distinctlyhigher than in Imnavait Creek and the Upper Ions migrate to the points of snowflakecrystals during the Kuparuk River but followed a similar pattern through the freezingprocess. Initial meltwatersflush these ions and result snowmeltperiod. The meltwatersampling location was in the in meltwater concentrationsmuch higher than the bulk snow- valley bottom, while the on the water track where the pack.The remainingsnowpack is depletedof ions,and further sampleswere collectedintegrated the areas higher on the meltwater therefore has low conductivity.After this process slope. If the runoff in the water track during the snowmelt occurredearly in the snowmeltperiod, the water track and the period was entirely from meltwaterheavier than meltwaterin Upper Kuparuk River had nearly equal conductivities,which the valley bottom, then Imnavait Creek shouldbe a mixture were similar to the conductivityof snowpackmeltwater col- betweenthe light meltwatercollected in the valleybottom and lected before it had contactwith the soil. This further suggests the heavier meltwater from the water track. Figure 6 shows that streamflow during the snowmeltperiod was almost en- that the •80 content of Imnavait Creek was indeed between tirely due to melting snowwith little contributionsfrom sub- the water track and the valley bottom meltwater. The Upper surface waters. KuparukRiver had an almostidentical pattern in •80 content Hydrologicbudget studiesin Imnavait Creek confirm that to Imnavait Creek. These patterns suggestthat the isotopic there can be essentiallyno mixingof meltwaterwith underlying contentsof streamwatercan be explainedentirely by the mix- soils[Kane et al., 1989, 1991;Hinzman et al., 1991]. Approxi- MCNAMARA ET AL.: HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED 1715

Storm 3

0.1

,-;-=

O.Ol

0.001 I ; I I I i 4-Jul 5-Jul 6-Jul 7-Jul 8-Jul 9-Jul 10-Jul 11-Jul

1994

b) I Storm 6

]

o o

0.01

0.001 29-JuI 30-JuI 31-,Jul 1-Aug 2-Aug 3-Aug 4-Aug 5-Aug 6-Aug 1994 Figure 5. Individual storm hydrographsused for recessionanalysis in Imnavait Creek. The only suitable stormswere (a) storm 3, and (b) storm 6.

mately 1.5 cm of meltwater infiltrates the desiccatedsurface rate of increase in active layer thicknessoccurs early in the soilsbefore runoff ensues [Kane et al., 1989],but thismeltwater summer. Thus, soon after snowmelt, rainfall is able to infiltrate refreezeswhen it contactsthe colder soil and essentiallyelim- the mineral soils,which was not possibleduring snowmelt,and inatesinfiltration [Hinzman et al., 1991;Kane and Stein,1983]. displaceold water into the streams.The basinstorage capacity and the potential old water reservoir continue to increaseas the active layer increasesthrough the summer.In the summer Discussion of 1994 a high correlation between storm date and old water Influences on Storm Flow Composition confirmed that old water contributionsincreased through the Summer storm flow in the Upper Kuparuk River basin is summer (Table 2), although at a moderate rate. Figure 7a dominated by old water, as is commonly observed in other illustrates the seasonaltrends in thaw depth and old water regions,despite the presenceof permafrost.However, whereas contributions in 1994. The lack of continued increase in old streamflowduring the snowmelt period in other regions is water contributionsin the summerof 1995 suggeststhat other dominatedby old water, streamflowduring the snowmeltpe- factorsinfluenced storm flow compositions. riod in the Upper Kuparuk River is almost entirely composed Eshlemanet al. [1993] demonstratedthat old water contri- of new water. Hence there is a dramatic shift in storm flow bution to stormflow decreasesas precipitation intensity in- compositionfrom the snowmeltperiod to the earliest summer creases.However, Hinzman et al. [1993] showedthat vertical storms.This changecan be credited to active layer thickness. hydraulic conductivitiesin the Imnavait Creek near surface Immediately followingsnowmelt, storage capacity of the soil is organic soils are so great that precipitation intensity rarely restrictedto a thin layer in the surfaceorganic soils. The fastest exceedsinfiltration capacityand that runoff occursmore com- 1716 MCNAMARA ET AL.: HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED

-22

--*-- Water Track --=- Imnavait Creek -23 -*- UpperKuparuk River -x- Snowpack Meltwater -24 -- UpperKuparuk River Discharge

=0-25 -

'• -26 • • 2520 -27 15 •E

-28 x 5

-29 I 0 4-May 8-May 12-May 16-May 20-May 24-May 1994 Figure6. The •80 contentsfor the UpperKuparuk River, Imnavait Creek, the hillslopewater track, and snowpackmeltwater during the 1994 snowmeltperiod. The snowmelthydrograph for the Upper Kuparuk River is shown for reference.

monly as a result of saturationof the active layer. Roulet and interaction with the subsurfacein nonpermafrost environ- Woo [1988] arrived at a similar conclusionin the Canadian ments. Low Arctic. They stated that in wetland soils,minimal runoff occurs until the soil saturates and the water table rises above Flow Sources and Hillslope Response the surface,effectively initiating a simultaneousresponse over Walkeret al. [1996] constructeda hierarchicgeographic in- the whole wetland area. This is a thresholdresponse of runoff formation system(GIS) of the Upper Kuparuk River basin initiation, and it suggeststhat total precipitation combined which showedthat open water, includingstreams, lakes, and with basin storagecapacity should be more significantthan pondscomprise 0.5% of the basin.The computedNWCPs are precipitationintensity in determininghillslope response to pre- significantlyhigher indicatingthat the new water storm flow cipitation events. responsecannot be accountedfor just by precipitationonto the Basin storagecapacity is dictated by the active layer thick- network and that there is indeed interaction with the ness and soil moisture conditions.Thus the depth of active surroundinghillslopes. However, by adding the riparian wet- layer thaw, in conjunctionwith precipitation patterns, influ- land areas (1.4%), the well-developedhillslope water tracks ences old water contributions to stormflow. Hinzman et al. (10.9%), and the poorly developedwater tracks(21.9%), the [1991] showedthat the soilmoisture in the surfaceorganic soil total area of the extended channel network is 34.7% of the layer in Imnavait Creek is highly sensitiveto recent precipita- total basin area. This is remarkably close to the 34% early tion patterns,while the soilmoisture in the underlyingmineral seasonNWCP in 1994 indicatingthat all of the early season soil remainsfairly constant.Therefore the influenceof active new water storm flow can be accountedfor by precipitation layer depth on basin storagecapacity is diminishedonce the onto the extended channel network. The hillslope contribu- thaw depth reachesthe mineral soil. Thus, in the period im- tionsto stormflow are likely comingfrom only the water tracks mediately following snowmelt,increases in active layer thick- with very little interaction with the unchanneledhillslopes. ness have dramatic influenceson storm flow compositions. Thus the water track networkmay act as the maximumpoten- However, later in the summer,any influenceof the activelayer tial saturatedarea during storms.The storagecapacity in the can be easilymasked by soil moistureconditions as dictatedby water tracks increasesas the seasonprogresses which could precipitationpatterns. account for the decrease in NWCP and the increase in old New water contributingportion (NWCP) results showed water contributionsthrough the summerof 1994.By the end of similarpatterns to old water contributionwith a seasonaltrend the seasonthe new water contributingarea is probably re- in 1994(Figure 7b) but not in 1995.NWCP maybe influenced strictedto narrow marginsaround the streams.The early sea- by changesin the activelayer thicknessin the samemanner as sonNWCP in 1995was 73%, whichis considerablyhigher than is old water contribution.However, a more significantresult of the area of the extended channel network. However, this storm the NWCP calculationsis the differencebetween permafrost was fairly closeto the end of the snowmeltperiod when large and nonpermafrostbasins. Eshleman et al. [1993] reported snowdriftspersisted on the eastfacing slopes and at the higher NWCP values between 0. i and 3% for the Reedy Creek wa- elevations.Consequently, the runoff during this storm was tershed in the Virginia coastalplain which has a humid sub- likely a mixture of precipitationand meltwater, which would tropicalclimate. Although these values range by a factor of 30, alter the contributing area of runoff generation. The mean they are considerablylower than those we report from the 1995 NWCP was 33%, indicatingthat throughoutthe season Upper Kuparuk River (Table 1). This suggestsmuch more new water storm flow can be accountedfor by precipitation MCNAMARA ET AL.: HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED 1717

a) 60 85

5O

75 •

• 30 •- 2O

65 lo

Analysis

0 I I I I I I 60 618 6118 6•28 718 7118 7•28 8/7

1994

b) 40

35

o 3o ._

o

c• 25

ß= m 20

i_ v

zg lO

5 -

o I I I I I I 618 6118 6128 718 7118 7128 817

1994

Figure 7. The 1994seasonal trends in stormresponse in the KuparukRiver basin.(a) The contributionof old waterto stormflow in the Upper KuparukRiver ascalculated by the mixingmodel and recession analysis is plottedagainst the right axis.Depth of thawis plottedon the left axisto illustratethe correlationto old water contributions.(b) The new water contributingportion (NWCP).

onto the extendedchannel network. These resultssuggest that moistureexisting in the water tracksthat are replenisheddur- the mechanismsthat are occurringon the unchanneledhill- ing storms.We havesuggested that the unchanneledhillslopes slopesare overwhelmedby the water tracksand are not sig- are not significantcomponents of new water runoff, and it is nificant in the basin responseexcept in extreme events. likely that they are not significantas old water sourcesas well. A goodagreement between NWCP and the extendedchan- We mustclarify our aboveexplanations relating hillslope thaw nel networkwas alsofound by Eshlemanet al. [1993] in Reedy depthto stormflow responseby statingthat thoseprocesses Creek, where runoff generationis dominatedby saturation describedare occurringwithin the hillslopewater tracks as overlandflow. Perhapsthen a commonmechanism of runoff well. It is impørtant to note that water tracks are distinct generationexists among watersheds in thesedifferent regions, geomorphølogicfeatures of drainagebasins in the Arctic,not despiteradically different hydroclimatological conditions. simplyzones of preferentialsaturation during storms.Essen- The commonlyaccepted ridging hypothesis tially, water tracks function as both hillslopesand channels. proposedby Sklashand Farvoiden [1979] states that oldwater Early in the season,the thaw depthsin both the unchanneled originatesin narrow marginsaround the streams.In the Ku- hillslopesand the water tracksare near zero. In the following parukRiver basin the water tracks are ephemeralstreams that summermonths the water tracksthaw deeper than the adja- existas zonesof enhancedsoil moistureduring dry periods.It centtundra regions. Hence the oldwater reservoir per unit of is likely then that old water contributionscome directlyfrom surfacearea is greaterin the water tracks,thereby enhancing 1718 MCNAMARA ET AL.: HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED the relationshipsbetween thaw depth and runoff described the water tracks connect directly with the streams without above, and lending further supportto our statementthat old passingthrough a wetland. water storm contributionscome primarily from water existing in the water tracks and saturatedvalley bottoms. The distinctdifferences in specificconductivity between Im- Summary navait Creek and the Upper Kuparuk River suggeststhat there We suspectedthat permafrostmay alter the compositionof is a sourceof solutespresent in the Kuparuk basinthat is not storm flow in the Kuparuk River basin from the common present in Imnavait Creek. The specificconductivity of Im- observationin other regions that old water dominatesstorm navait Creek is rarely greaterthan local precipitation,suggest- hydrographs.Further, we suspectedthat the gradualincrease ing that there is no significantinteraction with underlyingmin- in subsurfacestorage capacity due to thawing of the active eral soil and that the old water sourceexists in the organicsoil layer would imposeseasonal trends on storm flow character- horizon. A springlocated in the Kuparuk River headwaters8 istics.Using both a chemicalmixing model and graphicalre- kilometersupstream of our gaugingstation was sampledperi- cessionanalysis, we found that storm flow in the Upper Ku- odically through the summer, and the specificconductivities paruk River basinin 1994 and 1995was indeed dominatedby closelyfollowed the low flow streamconcentrations. Kreit et al. old water contributions. However, Imnavait Creek storm flow [1992] suggestedthat this springwater originatesfrom precip- was dominated by new water contributions.The difference itation within the basin that percolatesthrough coarseglacial betweenthe two basinsmay be a resultof the differencesin the .Also, many of the smallstreams and water tracksin potentialsaturated portions of eachbasin. Favorable compar- the headwatersof the Upper Kuparuk River have specific isonsbetween the specific conductivity mixing model, •80 mix- conductivitiessimilar to those of the spring. The water dis- ing model, and recessionhydrograph separation techniques chargingfrom the spring and in the upper streamsis likely supportour use of specificconductivity as a tracer and lend from the same source as the old water contributions to storm credenceto the physicallyunjustified recession analysis tech- flow. nique. Small streamsand water tracksmay have differingchemical In 1994, old water contributionsto storm flow in the Upper signaturesdepending on the bed material. Water tracksthat Kuparuk River increasedmoderately through the summer. existonly in the upper organicsoil layer havevery low specific Those seasonaltrends were not apparent in 1995, and no conductivitiesand will retain signaturesclose to new water seasonaltrends were observedin the storm flow dynamicsof throughout a storm, as in Imnavait Creek. Water tracks and Imnavait Creek in either year. However, there were large dif- streamsthat cut through to mineral soil pick up more solutes ferencesin the compositionsof stormflow betweenthe snow- and developan old water signaturesimilar to what we see in melt periodand the first summerstorms each year. Thus, in the the Upper Kuparuk River. The consequenceis that we may period immediately following snowmelt, increasesin active overestimatethe new water contributionsin peaty channels. layer thicknesshave dramatic influenceson storm flow com- The Upper KuparukRiver basincontains both peatyand stony positions.Later in the summerthe influenceof the depth of channels. However, that our recessionanalysis and mixing the active layer can be masked by soil moisture conditionsas model results are similar for the Upper Kuparuk River sug- dictatedby precipitationpatterns. geststhat our storm flow separationsare not significantlyin- New water contributingportion (NWCP) was muchgreater fluenced by this potential error. Further, this supportsour in the Upper Kuparuk River basin than in the basinwithout explanationthat the sourceof old water is from soil moisture permafroststudied by Eshlemanet al. [1993], which suggests within the water tracks and ephemeralstreams and that the that more interaction occurs between the surface and subsur- differing old water signaturesbetween basinsdo not require face in basinswithout permafrost.Further, NWCP decreased different explanationsof storm response. throughthe summerof 1994,which agreeswith the increasein The above discussionraises the question why Imnavait old water contributions.As the storagecapacity of the basin Creek storm flow is not dominatedby old water. This may be increasesthrough the thawing season,more new water enters due simply to the densityof hillslope water tracks and the the soil, as opposedto goingdirectly to runoff, and mixeswith presenceof a large wetland in the valley bottom of Imnavait old water to producerunoff. Creek. GIS mappingof the Imnavait Creek basinindicates that We credit both new water and old water sourcesto hillslope water tracks and suggestthat very little interaction occurs 56% of the basinis either part of the channelnetwork, riparian between unchanneled hillslopes and streams. The small wetland, or water track providinga very large potential satu- amount of interaction between the two zones is diminished rated area for quickflow of new water comparedto 35% in the even more with the increasingstorage capacity in the unchan- Upper KuparukRiver basin[Walker et al., 1996].The NWCPs neled hillslopesas the •a•on p•ogresses.That •to•m flow for ImnavaitCreek (14 and 13% for storms3 and 6 in 1994)are compositionhas even moderate dependenceon active layer considerably less than the potential saturated area, even thicknesshas implicationsthat a warmingclimate may impose though new water dominatesthe storm runoff. This suggests significantchanges in the hydrolog3,of watershedsin the Arc- that a relativelysmall portion of the basin contributesa ma- tic, which may then influencethe timing and magnitudeof the jority of the runoff during storms.That contributingportion delivery of nutrients to the aquatic system. may be the broad wetland in the valley bottom which occupies 12% of the basin and remains saturated most of the summer. A likely scenariois that the water tracksprovide a mixture of Acknowledgments. This work was supportedby the National Sci- old and new water to the wetland valley bottom, which pro- ence Foundation grants OPP-9214927 and OPP-9318535. We thank Robert Gieck for assumingresponsibility for much of the field work duces a saturated surface in the valley bottom from which and the many graduate students,technicians, and researchassociates continuedprecipitation runs off. A lower portion of the Upper at the Water and Environmental Research Center, UAF for their Kuparukbasin is valleybottom wetland (1.4%), thusmore of continualassistance. We thank the anonymousreviewers for the many MCNAMARA ET AL.: HYDROGRAPH SEPARATIONS IN AN ARCTIC WATERSHED 1719 constructivecomments that led to significantrevisions of the original Kennedy,V. C., C. Kendall, G. W. Zellweger,T. A. Wyerman, and R. manuscript. J. Avanzino,Determination of the componentsof stormflow using water chemistryand environmentalisotopes, Mattole River basin, California, J. Hydrol., 84, 107-140, 1986. References Kobayashi,D., Y. Kodama, M. Nomura, Y. Ishii, and K. Suzuki, Bottomley, J., D. Craig, and L. M. Johnson, Oxygen-18 studies of Comparisonof snowmelthydrograph separation by recessionanal- snowmeltrunoff in a small Precambrianshield watershed: Implica- ysisand by stream temperatureand conductance,IAHS Publ., 215, 49-56, 1993. tions for streamwater acidificationin acid-sensitiveterrain, J. Hy- drol., 88, 213-234, 1986. Kreit, K., B. J. Peterson,and T. L. Corliss,Water and sedimentexport Buttle, J. M., and K. Sami,Testing the groundwaterridging hypothesis of the upper Kuparuk River drainageof the North Slopeof Alaska, of streamflowgeneration during snowmelt,J. Hydrol., 135, 53-72, Hydrobiologica,240, 71-81, 1992. 1992. Linsley, R. K., M. A. Kohler, and J. L. H. Paulhus,Hydrology for Engineers,McGraw-Hill, New York, 1982. Cooper, L. W., C. R. Olsen, D. K. Solomon,I. L. Larsen, R. B. Cook, McDonnell, J. J., M. K. Stewart, and I. F. Owens, Effect of catchment- and J. M. Grebmeier, Stable isotopesof oxygenand natural and fallout radionuclidesused for tracingrunoff during snowmeltin an scalesubsurface mixing on streamisotopic response, Water Resour. Res., 27, 3065-3073, 1991. Arctic watershed, Water Resour. Res., 27, 2171-2179, 1991. Cooper,L. W., C. S. Solis,D. L. Kane, and L. D. Hinzman,Application McNamara, J.P., D. L. Kane, and L. D. Hinzman, An analysisof of oxygen-18tracer techniquesto arctichydrological processes, Arct. streamflowhydrology in the Kuparuk River basin:A nestedwater- Alp. Res.,25, ,247-255, 1993. shed approach,J. Hydrol., in press,1997. DeWalle, D. R., B. R. Swistock,and W. E. Sharpe,Three component Osterkamp,T. E., and M. W. Payne,Estimates of permafrostthickness from well logsin northernAlaska, Cold Reg.Sci. Technol.,5, 13-27, tracer model for storm flow on a small Appalachian forest catch- 1981. ment, J. Hydrol., 104, 301-310, 1988. Dincer, T., B. R. Payne,T. Florkowski,J. Martinec, and E. Tongiorgi, Peters, D. L., J. M. Buttle, C. H. Taylor, and B. D. LaZerte, Runoff Snowmeltrunoff from measurementsof tritium and oxygen18, Wa- productionin a forested,shallow soil, Canadian Shield basin, Water Resour. Res., 31, 1291-1304, 1995. ter Resour. Res., 6, 110-124, 1970. Dingman, S. L., Hydrology of the Glenn Creek watershed, Tanana Pilgrim, D. H., D. D. Huff, and T. D. Steele, Use of specificconduc- River basin,central Alaska, Res. Rep. 297, U.S. Army Cold Reg. Res. tance and contacttime relationsfor separatingflow componentsin storm runoff, Water Resour. Res., 15, 329-339, 1979. and Eng. Lab., Hanover, N.H., 1970. Dingman, S. L., PhysicalHydrology, Macmillan, Indianapolis,Ind., Rodhe, A., Springflood--Melt water or groundwater?,Nord. Hydrol., 1994. 12, 21-30, 1981. Eshleman, K. N., J. S. Pollard, and A. K. O'Brien, Determination of Roulet, N. T., and M. K. Woo, Runoff generation in a low Arctic contributingareas for saturationoverland flow from chemicalhy- drainagebasin, J. Hydrol., 101,213-226, 1988. drographseparations, Water Resour. Res., 29, 3577-3587, 1993. Sklash,M. G., and R. N. Farvolden,The role of groundwaterin storm Freeze, R. A., Role of subsurfaceflow in generatingsurface runoff, 1, runoff, J. Hydrol., 43, 45-65, 1979. Walker, D. A., N. A. Auerbach, L. R. Lestak, S. V. Mueller, and M.D. Baseflow contributionsto channel flow, Water Resour.Res., 8, 609- 623, 1972. Walker, A hierarchicGIS for studiesof process,pattern, and scale Hamilton, T. D., Late Cenezoic glaciation of the Central Brooks in an arcticecosystem: The Arctic SystemScience Flux Study,Ku- Range,in Glaciationin Alaska: Thegeologic record, vol. 99, edited by paruk Basin,Alaska. paper presented at the 2nd CircumpolarArctic Vegetation Mapping Workshop, Arendal, Norway, May 20-23, T. D. Hamilton, K. M. Reed, and R. M. Thorson,pp. 9-49, Alaska 1996. Geol. Soc.,Anchorage, 1986. Walker, M.D., D. A. Walker, and K. R. Everett, Wetland soils and Hastings, S. J., S. A. Luchessa,W. C. Oechel, and J. D. Tenhunen, Standingbiomass and productionin water drainagesof the foothills vegetation,Arctic foothills,Alaska, biological report, U.S. Dep. of of the Phillip SmithMountains, Holarct. Ecol., 12(3), 304-311, 1989. the Inter., Washington,D.C., 1989. Hinton, M. J., S. L. Schiff,and M. C. English,Examining the contri- Weller, G., F. S. Chapin, K. R. Everett, J. E. Hobbie, D. Kane, W. C. butions of glacial till water to storm runoff using two- and three- Oechel, C. L. Ping, W. S. Reeburgh,D. Walker, and J. Walsh, The componenthydrograph separations, Water Resour. Res., 30, 983-993, Arctic Flux Study:A regionalview of trace gasrelease, J. Biogeogr., 1994. 22, 365-374, 1995. Hinzman,L. D., D. L. Kane, R. E. Gieck, and K. R. Everett, Hydro- Wels, C., C. H. Taylor, and R. J. Cornet, Streamflowgeneration in a logic and thermal properties of the active layer in the Alaskan headwater basin on the PrecambrianShield, Hydrol. Processes,5, 185-199, 1991. Arctic, Cold Reg. Sci. Technol.,19, 95-110, 1991. Hinzman, L. D., D. L. Kane, and K. R. Everett, Hillslope hydrologyin Woo, M. K., and P. Steer,Slope hydrology as influencedby thawingof an Arctic setting, paper presentedat the Sixth International Con- the activelayer, Resolute,N. W. T., Can. J. Earth Sci., 20(6), 978- 986, 1983. ference on Permafrost,Chin. Acad. of Sci., Beijing, July 5-9, 1993. Hooper, R. P., and C. A. Shoemaker,A comparisonof chemicaland isotopichydrograph separation, Water Resour. Res., 22, 1444-1454, L. D. Hinzman, D. L. Kane, and J.P. McNamara, Water and En- 1986. vironmental Research Center, University of Alaska Fairbanks,Fair- Kane, D. L., and J. Stein, Water movementinto seasonallyfrozen soils, banks, AK 99775. (e-mail: [email protected];ffdlk@aurora. Water Resour. Res., 19, 1547-1557, 1983. alaska.edu;[email protected]) Kane, D. L., L. D. Hinzman, C. S. Benson,and K. R. Everett, Hydrol- ogy of Imnavait Creek, and arctic watershed,Holarct. Ecol., 12, 262-269, 1989. Kane, D. L., L. D. Hinzman, C. S. Benson, and G. E. Liston, Snow hydrologyof a headwaterArctic basin, 1, Physicalmeasurements (ReceivedNovember 18, 1996;revised February 10, 1997; and processstudies, Water Resour.Res., 27, 1099-1109, 1991. acceptedApril 4, 1997.)