modeling the hydrology of climate change in … · hydrology, and soils data are used to model the...

MODELING THE HYDROLOGY OF CLIMATE CHANGE IN CALIFORNIA’SSIERRA NEVADA FOR SUBWATERSHED SCALE ADAPTATION1

Charles A. Young, Marisa I. Escobar-Arias, Martha Fernandes, Brian Joyce, Michael Kiparsky,Jeffrey F. Mount, Vishal K. Mehta, David Purkey, Joshua H. Viers, and David Yates2

ABSTRACT: The rainfall-runoff model presented in this study represents the hydrology of 15 major watershedsof the Sierra Nevada in California as the backbone of a planning tool for water resources analysis including cli-mate change studies. Our model implementation documents potential changes in hydrologic metrics such assnowpack and the initiation of snowmelt at a finer resolution than previous studies, in accordance with theneeds of watershed-level planning decisions. Calibration was performed with a sequence of steps focusingsequentially on parameters of land cover, snow accumulation and melt, and water capacity and hydraulic con-ductivity of soil horizons. An assessment of the calibrated streamflows using goodness of fit statistics indicatethat the model robustly represents major features of weekly average flows of the historical 1980-2001 time ser-ies. Runs of the model for climate warming scenarios with fixed increases of 2!C, 4!C, and 6!C for the spatialdomain were used to analyze changes in snow accumulation and runoff timing. The results indicated a reductionin snowmelt volume that was largest in the 1,750-2,750 m elevation range. In addition, the runoff center of massshifted to earlier dates and this shift was non-uniformly distributed throughout the Sierra Nevada. Because thehydrologic model presented here is nested within a water resources planning system, future research can focuson the management and adaptation of the water resources system in the context of climate change.

(KEY TERMS: climate variability ⁄ change; watershed management; runoff; planning.)

Young, Charles A., Marisa I. Escobar-Arias, Martha Fernandes, Brian Joyce, Michael Kiparsky, Jeffrey F.Mount, Vishal K. Mehta, David Purkey, Joshua H. Viers, and David Yates, 2009. Modeling the Hydrology ofClimate Change in California’s Sierra Nevada for Subwatershed Scale Adaptation. Journal of the AmericanWater Resources Association (JAWRA) 45(6):1409-1423. DOI: 10.1111 ⁄ j.1752-1688.2009.00375.x

INTRODUCTION

One of the most robustly described impacts of pro-jected climate warming will be on the water resources

of snowmelt-driven hydrologic systems, which providewater for one-sixth of the world’s population (Barnettet al., 2005). The Sierra Nevada of California is a par-ticularly well-studied example, in which climatechange will impact the natural and human systems

1Paper No. JAWRA-08-0199-P of the Journal of the American Water Resources Association (JAWRA). Received October 13, 2008; acceptedAugust 17, 2009. ª 2009 American Water Resources Association. Discussions are open until six months from print publication.

2Respectively, Senior Scientist (Young, Purkey), Staff Scientist (Escobar-Arias, Mehta), Stockholm Environment Institute, 133 D Street,Suite F, Davis, California; Project Engineer (Fernandes), MWH-Global, Boston, Massachusetts; Senior Scientist (Joyce), Stockholm Environ-ment Institute, Sommerville, Massachusetts; Ph.D. Candidate (Kiparsky), Energy and Resources Group, University of California, Berkeley,California; Professor (Mount), Department of Geology and Center for Watershed Sciences, University of California, Davis; Assistant ResearchScientist (Viers), Department of Environmental Science & Policy and Center for Watershed Sciences, University of California, Davis, Califor-nia; and Scientist (Yates), National Center for Atmospheric Research, Boulder, Colorado (E-Mail ⁄Young: [email protected]).

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION 1409 JAWRA

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

Vol. 45, No. 6 AMERICAN WATER RESOURCES ASSOCIATION December 2009

that rely on the hydrologic character of the mountainrange. Runoff of winter precipitation is delayed untilcloser to times of peak demand, as mountain snow-pack acts in essence as the state’s largest reservoir.

While the general implications of climate warmingfor California’s hydrology have caught policy maker’sattention, specific adaptive responses have yet tomaterialize. To explore possible strategies for adapta-tion and management, an analytical platform isneeded that will allow for a better understanding ofthe climate change impacts within individual riverbasins and, in turn, to assess how water and landmanagers may adapt to them (Poff, 2002; Palmeret al., 2008) as it is at this scale where many manage-ment decisions will need to be made (Hulme, 2005;Palmer et al., 2008). At the same time it is useful tohave the scope of such a tool be the entire SierraNevada as climate change impacts are expected to bespatially heterogeneous in terms of elevation and lati-tude. By being able to see more of the ‘‘big picture,’’planners can begin prioritizing conservation andmanagement actions.

Although several recent studies report on thepotential effects of rising temperatures on snowpackaccumulation and the earlier initiation of snowmelt(Cayan et al., 2001; Kim, 2001; Miller et al., 2003;Dettinger et al., 2004; Knowles and Cayan, 2004;Stewart et al., 2004; Knowles et al., 2006; Maureret al., 2007), to date, less effort has been spent tryingto calculate the local impacts of these changes withinindividual watersheds. Studies in other regions indi-cate that changes in hydrologic patterns will modifythe flow regime, sediment inputs, and water tempera-ture affecting physical habitat (Gibson et al., 2005),and similar impacts are also expected in the SierraNevada. For instance reduction in summer flows, cou-pled with increasing atmospheric temperatures, willincrease water temperatures and potentially the suit-ability of stream reaches as habitat for temperaturesensitive aquatic species (Myrick and Cech, 2004). Onland, changes in atmospheric temperature and CO2

concentrations may affect the distribution of plantcommunities as they respond to environmentalchange (Diffenbaugh et al., 2003; Kueppers et al.,2005; van Mantgem and Stephenson, 2007) which, inturn, will affect the hydrologic characteristics of thewatershed. Within the operated water resources sys-tem, changes in the timing and volume of snowmeltwill affect water supply, hydropower generation, floodcontrol, and recreational flows (Zhu et al., 2005;Vicuna et al., 2007).

Here, we present the first phase of development ofa water resources planning system, which in the endwill span the continuum from climate change tohydrologic response to management adaptation toimpact assessment. The underpinning of this system

is the Water Evaluation and Planning System(WEAP21), a spatially based model capable of calcu-lating changes in the hydrologic cycle, incorporatingchanging climate conditions, and representing thehuman managed system with dams, diversions, andhydropower projects. Importantly for our purposes,this platform allows users to incorporate downscaledglobal climate models and calculate impacts and pos-sible adaptation strategies within one platform (Yateset al., 2009).

Specific objectives of this paper are (1) to presentthe implementation of the WEAP21 model for 15 westslope Sierra Nevada watersheds from the Feather toKern Rivers (Figure 1), (2) describe the calibration ofthe models for unimpaired flows, and (3) to presentresults of scenarios of temperature increase on snowaccumulation and the timing of streamflow.

METHODS

In the WEAP21 model, climate, topography,land cover, surface water hydrology, groundwater

FIGURE 1. Project Watersheds and Topographic Map.

YOUNG, ESCOBAR-ARIAS, FERNANDES, JOYCE, KIPARSKY, MOUNT, MEHTA, PURKEY, VIERS, AND YATES

JAWRA 1410 JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

hydrology, and soils data are used to model the natu-ral or unimpaired hydrology of the western slope ofthe Sierra Nevada. Once the natural hydrology ismodeled, infrastructure elements can be added torepresent human hydrologic manipulations (Yateset al., 2005). The watersheds modeled, from north tosouth are the Feather, Yuba, Bear, American, Cosum-nes, Mokelumne, Calaveras, Stanislaus, Tuolumne,Merced, San Joaquin, Kings, Kaweah, Tule, and KernRivers. The extent of the model covers each contigu-ous watershed from the Sierran crest to thewatershed outlet near the Central Valley floor, anarea of roughly 650 · 125 km (Table 1). In this sec-tion, we present a discussion of the algorithms usedto model the hydrologic cycle and a description of theprocess to delineate ‘‘catchments’’ defined as the mod-eling units where the terrestrial components of thehydrologic cycle are simulated in WEAP21.

WEAP21 Algorithms

The WEAP21 model is a comprehensive, fully inte-grated water basin analysis system. It is a simulationmodel that includes a robust and flexible representa-tion of water demands from all sectors and flexible,programmable operating rules for infrastructure ele-ments such as reservoirs, canals, and hydropowerprojects. It has watershed rainfall-runoff modelingcapabilities that allow for calculation of hydrographcomponents including runoff, interflow, and baseflow. Soil moisture storage and snow accumulationand melt are also calculated. Water infrastructure

and demand elements can be dynamically nestedwithin the underlying hydrological processes. Ineffect, this allows the modeler to analyze how specificconfigurations of infrastructure, operating rules, andpriorities will affect water uses as diverse as in-stream flows, agricultural irrigation, and municipalwater supply under the umbrella of input weatherdata and physical watershed conditions. This makesit ideally suited to studies of dynamic change withinwatersheds, such as climatic shifts.

The physical hydrology model in WEAP21 repre-sents the terrestrial water cycle with a series ofsimultaneously solved equations. Rainfall is parti-tioned into snow, runoff, or infiltration depending ontemperature, land cover, and soil moisture status.Moisture in the root zone is partitioned into evapo-transpiration (ET), interflow, deep percolation, orstorage as a function of soil water capacity, hydraulicconductivity, potential ET, and vegetation-specific ETcoefficients. Deep percolation enters a second soilcompartment and is partitioned into base flow ordeep soil moisture storage. This partitioning is afunction of the soil moisture storage status, the waterholding capacity of the deep compartment, and thehydraulic conductivity of the deep sediments. For adetailed discussion of the model algorithms, see Yateset al., (2005). Due to the importance of snow pro-cesses in Sierra Nevada hydrology and modificationsto the WEAP21 algorithm used in this effort and notpresented in Yates et al. (2005), the snow accumula-tion and melt module is described here.

WEAP21 includes a simple temperature-index snow-melt model which computes an effective precipitation

TABLE 1. Watershed Name and Model Information.

Watershed Name and Code Watershed OutletNumber of

SubwatershedsNumber ofCatchments

American R. – AMR FOLSOM RESERVOIR 46 185Bear R. – BAR CAMP FAR WEST RESERVOIR 6 19Calaveras R. – CAL N HOGAN LAKE 3 14Cosumnes R. – COS COSUMNES R AT MICHIGAN BAR 11335000 5 27Feather R. – FEA LAKE OROVILLE 38 164Kaweah R. – KAW LAKE KAWEAH 8 61King R. – KNG PINE FLAT RESERVOIR 15 103Kern R. – KRN RIO BRAVO PP NR BAKERSFIELD CA, 11193010 10 68Merced R. – MER LAKE MCCLURE 6 42Mokelumne R. – MOK PARDEE RESERVOIR 13 56San Joaquin R. – SJN MILLERTON LAKE 38 185Stanislaus R. – STN NEW MELONES RESERVOIR 25 109Tule R. – TUL LAKE SUCCESS 9 56Tuolumne R. – TUO NEW DON PEDRO RESERVOIR 19 97Yuba R. – YUB DEER CREEK NR SMARTVILLE 11418500

AND HARRY L ENGLEBRIGHT LAKE20 82

Total number ofsubwatersheds and catchments

261 1,268

Average area (±SD) 182.5 km2 37.6 km2

MODELING THE HYDROLOGY OF CLIMATE CHANGE IN CALIFORNIA’S SIERRA NEVADA FOR SUBWATERSHED SCALE ADAPTATION


Pe. The model estimates snow water equivalent(SWE) and snowmelt from an accumulated snowpackin the catchment, where mc is the melt coefficientgiven as

mc !01

Ti"Ts

Tl"Ts

8<

: ifTi<Ts

Ti>Tl

Ts # Ti # Tl

$1%

with Ti the observed temperature for period i, and Tl

and Ts are melting and freezing temperature thresh-olds, with the melt rate is given as

mi ! min$Acimc;Em% $2%

Snow accumulation, Aci is a function of mc, mi, andthe observed total precipitation, Pi

Aci ! Aci"1 & $1"mc%Pi "mi $3%

Em is the available melt energy converted to anequivalent water depth ⁄ time. The available meltenergy is computed as

Em ! Rnet & Eother $4%

Rnet is the net radiation and Eother represents addi-tional forms of energy that contribute to snowmeltbeyond the incoming solar radiation. This lumpedparameter includes sensible, latent, advective, andground energies. It is calculated as the net radiation,Rnet, multiplied by an additional radiation factor, Rf,which was adjusted during calibration. The calcula-tion for net radiation considers the albedo which ismodeled using a simple algorithm that decreasesalbedo through time to represent the ‘‘ripening’’ of thesnow surface (USACE 1998, figure 5-5). The modeluser specifies a ‘‘new’’ snow albedo value, AN, and thelowest or ‘‘old’’ snow albedo, AO, value as a fraction.Albedo is set at the ‘‘new’’ value following snowfall, itis then decreased by 0.1 for each simulation week witha minimum of the ‘‘old’’ snow albedo value.

WEAP21 Catchment Delineation

Catchment nodes in WEAP21 represent a specificgeographic area in which the distribution of landcover and soil properties are specified. Input weatherdata are assumed to be uniform over the entire catch-ment area. To parameterize the catchments, publiclyavailable Geographic Information System (GIS)data including soils, vegetation, and digital elevationmodels (DEMs) were gathered. Catchment definition

was achieved using the following steps: (1) delinea-tion of watersheds, subwatersheds, and elevationbands using the DEM; (2) intersection of elevationbands with subwatersheds to create WEAP21 catch-ments; (3) classification of vegetation and soils; and(4) intersection of soils, vegetation, and catchments tocalculate fractional areas for each vegetation-soilcombination in each WEAP21 catchment.

To perform the delineation of watersheds, sub-watersheds, and elevation bands, 10 m DEMs avail-able from the U.S. Geological Survey (USGS) wereused (http://www.seamless.usgs.gov/). Watershed pourpoints were placed at the watershed outlets listed inTable 1. Subwatershed pour points were chosen basedon the location of important infrastructure includingdams, diversions, return flow structures, and USGSstreamflow gages (Figure 2A). USGS streamflowgages were used if they possessed approximately fiveyears or more of continuous record during the 1982-2001 calibration period. This delineation resulted in261 subwatersheds throughout the 15 watersheds(Table 1) with an average upstream discrete area of182.5 km2. Elevation bands were delineated for thefirst 500 meters above sea level and then for every250 meters up to the crest of the mountain range(500-4,000 m). This level of elevation discretizationwas chosen to provide resolution in the critical snowaccumulation elevation range. The creation of catch-ments required the intersection of the elevation bandswith each subwatershed. The resulting 1,268 geo-graphic regions with an average area of 37.6 km2

served as the basic modeling unit where hydrologicprocesses were modeled in WEAP21. At this base unitof accounting, where a catchment represents theintersection of subwatersheds and elevation bands,climatic data such as temperature, precipitation,humidity, and wind speed were assumed uniform andused as model inputs (Figures 2B and 2C).

Land cover and soils data were obtained from pub-licly available sources. Land cover information wasobtained from the National Land Cover Dataset(NLCD) (Homer et al., 2004). The NLCD vegetationdata included 29 land cover classes, which wereaggregated into seven land cover types: barren land,trees, agriculture, grasslands, shrubs, urban, andwater. The aggregations were based on the similari-ties in hydrological properties such as transpirationrates and runoff characteristics.

Soil depth classifications of greater than or lessthan 50 cm were assigned based on Natural ResourceConservation Service soil survey mapping unit mini-mum component depth to bedrock and the presenceof rock outcrop. The Soil Survey Geographic(SSURGO) database was used for 75% of the studyarea. The remaining 25% of the area was charac-terized using the coarser resolution State Soil



Geographic (STATSGO) data. The STATSGO datawas used for the entire Kern and Calaveras water-sheds and portions of the Kings, San Joaquin, Mok-elumne, Stanislaus, and Tuolumne watersheds.

The next step was the intersection of land coverand soil depth information with the catchments tocreate a unique land cover ⁄ soil depth combinationwithin each WEAP21 catchment (i.e., trees on deepsoils, trees on shallow soils, shrubs on deep soils,shrubs on shallow soils, etc). At this level of spatialintersection, model parameterization for plant andsoil characteristics was assumed uniform.

Weather Data

Due to the large variation in temperature andprecipitation over the Sierra Nevada caused by

orographic effects (Meyers and Cotton, 1992; Galew-sky and Sobel, 2005) and because observed weatherdata are relatively scarce, it was necessary to useinterpolated weather data as input to the model.The DAYMET dataset (Thornton et al., 1997) waschosen for this application due to its spatial resolu-tion (1 km grid) which is fine enough to provideresolved temperature and precipitation values forcatchments that may be in proximity but at very dif-ferent elevations. The DAYMET dataset includeddaily average temperature, vapor pressure deficit,and precipitation. DAYMET temperature and precip-itation time series were obtained for a single loca-tion within each catchment. The locations weredetermined by placing a point close to the polygoncentroid on the mid-elevation contour for the eleva-tion band, i.e., along the 1,125 m contour for the1,000-1,250 m band. These points were then used to

FIGURE 2. Subwatershed Delineation, Elevation Banding, and Catchment Definition.



obtain and incorporate the weather time series fromthe DAYMET website.

MODEL CALIBRATION

Model calibration was performed using manualcalibration techniques targeting observed SWE andcomputed unimpaired streamflow data obtained fromthe California Department of Water Resources(CDWR) and California Data Exchange Center(CDEC) webpage. SWE data were obtained for 15locations. Computed monthly unimpaired streamflowswere obtained for the watershed outlets of 13 of the15 modeled watersheds. Monthly data were usedbecause weekly data were not available during theentire calibration period for several watersheds. Nomonthly unimpaired flow data were available for theCalaveras or Bear River watersheds.

An initial set of parameters was developed thatcould be applied across all 15 watersheds and cap-tured the seasonal and inter-annual variability ofsnow and flow measurements across the Sierra

Nevada (Table 2). The most sensitive model parame-ters were then adjusted on a watershed by watershedbasis, including soil water capacity, hydraulic conduc-tivity, melt and freeze temperatures, additional radia-tion factor, and preferred flow direction (Table 3) toaccount for fine-scale differences in watershed charac-teristics not captured by the available data.

The accuracy of the model at predicting snow accu-mulation and melt is quantified by the correlationcoefficients between observed and modeled values ofSWE, shown in Figure 3. Pearson’s correlation coeffi-cients were calculated for weeks in which the observa-tional record was non-zero and averaged 0.87, andranged from 0.71 to 0.94. The largest shift from theuniform set of parameters in the snow model were forthe Merced, San Joaquin, and Kings watersheds(Table 3). Potential reasons for these differencesinclude the fact that SWE observations are point mea-surements and do not necessarily represent the spa-tial averages computed in the model. Additionally, wesuspect underpredictions of precipitation in the inputweather data, as will be discussed later in this section,had an effect on the calibrated parameter set.

For streamflow, calibration was based on compari-sons between the CDWR unimpaired streamflows andcomputed flows using the bias, root mean squareerror (RMSE), and 10th, 25th, 50th, and 75th flow ex-ceedence percentiles. In addition to the 10th flow ex-ceedence percentile, the RMSE of the log transformedflow values (RMSElog) was calculated to enable closerscrutiny of the ecologically important low summerand fall flows. Values of the BIAS ranged from )6%to 2% with an average of )1% (Table 4). The RMSEvaried from 38% to 65% with an average of 51%.Flow exceedence percentiles were within ±17% of

TABLE 2. Uniformly Applied Calibration Parameters.

Model Parameter Value

Crop coefficeint, kc 1.1Runoff resistance factor, RRF Ag = 8, Bare = 4, Grass = 12,

Shrubs = 14, Trees = 20,Urban = 4, Wet = 4

Albedo, new snow, AN 0.7Albedo, old snow, AO 0.3

TABLE 3. Non-Uniformly Applied Calibration Parameters.

Watershed Code FEA YUB BAR AMR COS MOK CAL STN TUO MER SJN KNG KAW TUL KRNAverageValue

Preferred flow direction 0.5 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.78Melting threshold (!C) 4 5 5 5 5 5 5 5 5 6 12 11 5 5 7 6.00Freezing threshold (!C) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 )5 )0.36Radiation factor 6 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4 4 3 4.5 4.5 4.5 4.43Shallow soil – root zonewater capacity (mm)

189 212 212 224 236 236 236 212 236 236 118 82 378 472 1,181 298

Deep soil – root zonewater capacity (mm)

480 540 540 570 600 600 600 540 600 600 300 210 960 1,200 3,000 756

Shallow soil – root zonehydraulic conductivity(mm ⁄week)

92 77 77 62 70 62 77 77 231 77 231 154 31 30 30 92

Deep soil – root zonehydraulic conductivity(mm ⁄week)

12 10 10 8 9 8 10 10 30 10 30 20 4 4 4 12

Deep water capacity (mm) 300 300 300 200 200 200 200 200 200 200 200 200 200 200 200 220Deep hydraulicconductivity (mm ⁄week)

20 30 30 35 55 85 20 35 45 30 55 35 55 40 35 40



observed values except for the 10th percentile valuesfor the Cosumnes and Tule watersheds and the 75thpercentile value for the Cosumnes watershed (Fig-ure 4). Values of RMSElog varied from 1% to 29%with an average of 9% (Table 4). Overall, the modelcaptured the major features of the flow hydrographsat the watershed outlets (Figure 5). However, duringthe model construction and calibration processsources of model, parameter specification, and inputdata error were identified.

Model error was identified during the analysis ofthe low flows using the RMSElog and the 10th flowexceedence percentile. The highest RMSElog valueswere in watersheds that had observed monthly flowsof zero during dry summers. This occurred in theAmerican, Cosumnes, Mokelumne, Stanislaus, Mer-ced, and Tule watersheds. The reason for the rela-tively poor fit in these watersheds is that thecontinuous mathematical functions of the hydrologi-cal algorithms in WEAP21 do not easily representextremely low flow events. For the watersheds inwhich there were no observed monthly flows of zero,the average RMSElog was 3%.

A potential source of parameter specification errorwas identified during the model construction process.As discussed earlier in this section, data from theSSURGO or STATSGO soil surveys were used to char-acterize soils as either deep or shallow. This classifica-tion was based on soil components for which the exactlocation within the mapping unit is not known. In thecase of the higher resolution SSURGO data, the aver-age mapping unit polygon area was 0.7 km2. As this ismore than an order of magnitude smaller than theaverage catchment size, the spatial uncertainty of com-ponent locations was considered unimportant. In con-trast, the STATSGO database had an averagemapping unit polygon size of 126 km2, which is overthree times the average catchment size. As the entiremapping unit was classified as shallow if one compo-nent had a measured bedrock depth less than 50 cm or

0.E+00

1.E+02

3.E+02A. Feather R, FOR (1570 m) r=0.75

0.E+00

1.E+02

3.E+02B. Feather R, HMB (1981 m) r=0.89

0.E+00

1.E+02

3.E+02C. Feather R, KTL (2225 m) r=0.89

0.E+00

1.E+02

3.E+02D. Mokelumne R, BLK (2438 m) r=0.84

0.E+00

1.E+02

3.E+02E. Mokelumne R, BLS (1981 m) r=0.83

0.E+00

1.E+02

3.E+02 F. Mokelumne R, MDL (2408 m) r=0.89

0.E+00

1.E+02

3.E+02

G. San Joaquin R, GRM (2408 m) r=0.91

0.E+00

1.E+02

3.E+02H. San Joaquin R, GRV (2103 m) r=0.94

0.E+00

1.E+02

3.E+02I. San Joaquin R, VLC (3063 m) r=0.80cm

0.E+00

1.E+02

3.E+02J. Kings R, CRL (3170 m) r=0.94

0.E+00

1.E+02

3.E+02K. Kings R, MTM (3018 m) r=0.92

0.E+00

1.E+02

3.E+02L. Kings R, BIM (2316 m) r=0.71

0.E+00

1.E+02

3.E+02

Oct-90

Oct-91

Oct-92

Oct-93

Oct-94

Oct-95

Oct-96

Oct-97

M. Kern R, UTY (3475 m) r=0.93

0.E+00

1.E+02

3.E+02

Oct-90

Oct-91

Oct-92

Oct-93

Oct-94

Oct-95

Oct-96

Oct-97

Time

N. Kern R, BCH (2332 m) r=0.87

0.E+00

1.E+02

3.E+02

Oct-90

Oct-91

Oct-92

Oct-93

Oct-94

Oct-95

Oct-96

Oct-97

O. Kern R, WTM (2728 m) r=0.93

FIGURE 3. Observed and Simulated (dashed lines) Snow Water Content Values for 15 Sites in the SierraNevada. Graph titles provide watershed name, CDEC station code, elevation, and correlation coefficient.

TABLE 4. Goodness of Fit Statistics for Predicted Monthly FullNatural Flows at Watershed Exits (WY 1982-2000, n = 228).

Watershed BIAS (%) RMSE (%) RMSElog (%)

FEA 0 53 1YUB 0 47 3AMR 0 39 14COS )3 38 29MOK 0 60 18STN 2 54 14TUO )5 55 4MER 1 50 10SJN )4 65 3KNG )6 49 3KAW 0 42 2TUL )2 58 14KRN )2 54 2

Note: RMSE, root mean square error. BIAS ! 100'$ !Qs " !Qo%= !Qo(,

and RMSE ! 100!Qo

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!Pn

i!1$Qs;i"Qo;i%2

n ;

rwhere Qs,i and Qo,i are simulated

and observed flow rates for each time step (i).



0

4001. Feather R

0

6002. Yuba R

0

6003. American R

0

4004. Cosumnes R

0

4005. Mokelumne R

0

4006. Stanislaus R

0

4007. Tuolumne R

0

4008. Merced R

0

4009. San Joaquin R

0

40010. Kings R

0

40011. Kaweah R

0

400

0 10 20 30 40 50 60 70 80 90 100

Probability of exceedence (%)

12. Tule R

0

400

0 10 20 30 40 50 60 70 80 90 100

Probability of exceedence (%)

13. Kern R

mm

FIGURE 4. Simulated (dashed line) and Observed Flow Exceedence ProbabilitiesatWatershed Outlets (1982-2000). Note different y-axis scales.

0

2001. Feather R

0

2002. Yuba R

0

2003. American R

0

2004. Cosumnes R

0

2005. Mokelumne R

0

2006. Stanislaus R

0

2007. Tuolumne R

0

2008. Merced R

0

2009. San Joaquin R

0

20010. Kings R

0

20011. Kaweah R

0

200

Month

12. Tule R

0

200 10 11 12 1 2 3 4 5 6 7 8 9

10 11 12 1 2 3 4 5 6 7 8 9

Month

13. Kern R

mm

FIGURE 5. Simulated (dashed line) and CDWR Estimated Average MonthlyFull Natural Flows at Watershed Outlets 1982-2000.



was rock outcrop, an overestimation of the shallow soilextent was likely. This was most prevalent in the Kernand Calaveras watersheds which only had STATSGOdata and is evident in the calibrated parameters forthe Kern watershed, which has the highest calibratedroot zone water capacity of all watersheds.

The relatively high root zone water capacity in theKern watershed is also likely caused by bias in theinput precipitation data. Analysis of the seven CDECstations in the Kern watershed reveal that inputDAYMET precipitation data for the study period hasan average positive bias of 19% at those locations.Presumably this precipitation bias extends to areaswhere there are no CDEC gages as the calibrationrequired a relatively large root zone water capacityand low hydraulic conductivity to increase transpira-tion and thereby reduce bias in predicted annualstreamflow at the watershed outlet. Additional exam-ples of bias in the input precipitation data were foundin the San Joaquin watershed where a negative bias(underprediction) was found at CDEC gages GreenMountain (GRM) and Volcanic Knob (VLC) during1993 and 1995. The volume of annual precipitation inthe input data is less than the observed SWE at theselocations. Additionally, during 1981-1984, 1986-1987,and 1991 the total input rainfall volume for thewatershed above the streamflow gage on the San Joa-quin River at Miller Crossing (#11226500) is lessthan the reported USGS streamflow. This negativeprecipitation bias resulted in different meltingthreshold and radiation factors in the snow accumu-lation and melt model for the Merced, San Joaquin,and Kings watersheds.

In general, interpolation of point observations ofprecipitation into spatially distributed fields usefulfor watershed modeling is difficult. In the case of theSierra Nevada, it is especially difficult in the higherelevation catchments due to the lack of observationaldata and pronounced orographic effects (Daly, 2006).Previous efforts at modeling rainfall-runoff in thisregion have encountered similar problems (Knowles,2000; Koczot et al., 2005) and have adjusted the pre-cipitation inputs to enable a better match betweenobserved and simulated streamflow values.

To assess model accuracy within the boundaries ofthe watersheds, weekly flows from USGS streamflowgages that measure runoff from undevelopedsubwatersheds were compared to model predictions(Figure 6 and Table 5), resulting in an average biasvalue of )1.3% for the 19 unimpaired flow gages (range)41% to 60%). Model bias was within ±10% for fivegages and ±20% for 14 gages. RMSE values rangedfrom 70% in the Kings watershed to 203% in the Mok-elumne watershed. Values of RMSElog ranged from 3%to 19% with an average value of 8%. While these statis-tics indicate a reduced goodness of fit when compared

with the values at the watershed exits (Table 4), it isimportant to note these gage records were not used inthe calibration process and instead were used to inde-pendently assess intra-basin model performance.

Future users of this model should consider the cali-bration results outlined above in determining how toapply the model. Depending on the question to beaddressed by the model there are varying levels ofconfidence in its accuracy. Overall, the calibrationresults indicate that the model is most accurate atpredicting streamflows at the watershed exits andsnow accumulation and melt which is to be expectedas this was the focus of the calibration effort. Valuesof streamflow RMSE at the watershed exits, whichweights high flows more than low flows, indicate thatthe worst fit to the observed data was in the San Joa-quin, Mokelumne, and Tule watersheds. Values ofRMSElog, which weights low flows more than highflows, show that the Cosumnes, Mokelumne, Ameri-can, Stanislaus, and Tule watersheds had the lowestperformance. An indication of model performance atthe subwatershed scale is provided by the comparisonwith 19 USGS flow gages. Values of RMSE indicatesubwatersheds in the Mokelumne, San Joaquin, andFeather watersheds had the lowest performance. Val-ues of RMSElog indicate that subwatersheds in theFeather, Tule, and Mokelumne watersheds had theworst low flow performance. With respect to predictedstreamflow center of mass timing, a majority of thesubwatersheds are within one week of the correcttime (Table 5), however, subwatersheds in theFeather, Stanislaus, and Tuolumne are less accurate.This information is useful in assessing the submonth-ly performance of the model and is directly relevantto the results presented below.

TEMPERATURE INCREASE EFFECTS ON SNOWACCUMULATION AND STREAMFLOW TIMING

Changes in the magnitude of snow accumulationand timing of runoff in the Sierra Nevada have beenstudied from both historical and predictive perspec-tives. Analyses of the historical streamflow recordhave shown that the timing of the runoff center ofmass has trended to earlier in the year for many SierraNevada watersheds over the past 50 years (McCabeand Clark, 2005; Regonda et al., 2005; Stewart et al.,2005). This change has been correlated to increases inspring and summer atmospheric pressure and temper-ature (McCabe and Clark, 2005; Regonda et al., 2005).The inter-annual variability in spring runoff timinghas been partially accounted for by the cycles ofthe Pacific decadal oscillation. Of the remaining



variability, a large fraction appears to be related toobserved increases in global temperatures (Stewartet al., 2005). Looking into the future, researchers have

found a continued shift in runoff center of mass to ear-lier in the year is to be expected as projected increasesin temperature affect snow accumulation and melt.

TABLE 5. Goodness of Fit Statistics for Weekly Flow Data at 19 Locations with USGS Streamflow Gages on Unimpaired Water Bodies.

Watershed USGS GageBIAS(%)

RMSE(%)

RMSElog

(%)Period ofRecord

Centerof Mass

Error (weeks)

FEA INDIAN C NR CRESCENT MILLS CA 11401500 )17 149 19 1982-1993 )2.5YUB N YUBA BL GOODYEARS BAR CA 11413000 1 101 4 1982-2000 0.7AMR NF AMERICAN A NORTH FORK DAM 11427000 5 78 4 1982-2000 1.7MOK COLE C NR SALT SPRINGS DAM 11315000 )10 137 14 1982-2000 )1.4MOK FOREST C NR WILSEYVILLE CA 11316800 60 203 7 1982-2000 1.5STN CLARK F STANISLAUS R NR DARDANELLE A 11292500 )13 127 8 1982-1993 2.3TUO M TUOLUMNE R A OAKLAND RECREAT CAMP CA 11282000 25 117 13 1982-1996

1998-20002.2

TUO SF TUOLUMNE R NR OAKLAND RECREATI CAMP 11281000 31 115 5 1982-19961998-2000

0.8

MER MERCED R A HAPPY ISLES BRIDGE NR YOSEMI CA 11264500 )18 101 6 1982-2000 0.1MER MERCED R A POHONO BRIDGE NR YOSEMITE CA 11266500 )20 99 5 1982-2000 )0.1SJN BEAR CREEK NR TA EDISON 11230500 )22 97 8 1982-1991 1.0SJN PITMAN CREEK BL TAMARACK CREEK 11237500 7 153 13 1982-1991 )0.2SJN SAN JOAQUIN A MILLER CROSSING 11226500 )41 92 7 1982-1991 1.2KNG KINGS R BL NORTH FORK NR TRIMMER CA 11218500 )12 70 5 1982-1993 0.5KAW MF KAWEAH R NR POTWISHA 11206501 )11 71 4 1982-2000 )0.7KAW E FK KAWEAH 11208731 )3 90 4 1994-2000 )0.8KAW MARBLE F KAWEAH 11208001 )12 84 4 1982-2000 )1.4TUL SF TULE R NR SUCCESS 11204500 2 103 19 1982-1990 )1.0KRN COMB FLOW KERN R AND KERN NO 3 11186001 )13 79 3 1982-2000 )0.4

Notes: RMSE, root mean square error; USGS, U.S. Geological Survey.

0

1002. Yuba R, N YUBA BL GOODYEARS BAR 11413000

0

25

0

1003. American R, NFAMERICAN 11427000

0

100

0

1005. Mokelumne R, COLE C NR SALT SPRINGS DAM 11315000

0

1006. Stanislaus R, CLARK FORK STANISLAUS R NR DARDANELLE CA 11292500

1. Feather R, INDIAN CR NR CRESCENT M 11401500

4. Mokelumne R, FOREST C NR WILSEYVILLE 11316800

0

1007. Tuolumne R, M TUOLUMNE R A OAKLAND RECREATION CAMP CA 11282000

0

1008. Tuolumne R, SF TUOLUMNE R NR OAKLAND RECREATION CAMP 11281000

0

1009. Merced R, A HAPPY ISLES BRIDGE NR YOSEMITE CA 11264500

0

10010. Merced R, A POHONO BRIDGE NR YOSEMITE CA 11266500

0

10011. San Joaquin R, BEAR CREEK NR TA EDISON 11230500

0

10012. San Joaquin R, PITMAN CREEK BL TAMARACK CREEK 11237500

0

10013. San Joaquin, A MILLER CROSSING 11226500

0

10015. Kaweah R, MF KAWEAH R NR POTWISHA 11206501

0

10016. Kaweah R, E FK KAWEAH 11208731

0

1001-

Oct

1-N

ov

1-D

ec

1-Ja

n

1-F

eb

1-M

ar

1-A

pr

1-M

ay

1-Ju

n

1-Ju

l

1-A

ug

1-S

ep

Week

17. Kaweah R, MARBLE F KAWEAH 11208001

0

25

1-O

ct

1-N

ov

1-D

ec

1-Ja

n

1-F

eb

1-M

ar

1-A

pr

1-M

ay

1-Ju

n

1-Ju

l

1-A

ug

1-S

ep

Week

18. Tule R, SF TULE R NR SUCCESS 11204500

mm

0

10014. Kings R, KINGS Bl N Fk 11218500

0

50

1-O

ct

1-N

ov

1-D

ec

1-Ja

n

1-F

eb

1-M

ar

1-A

pr

1-M

ay

1-Ju

n

1-Ju

l

1-A

ug

1-S

ep

Week

19. Kern R, COMB FLOW KERN R AND KERN NO 3 11186001

FIGURE 6. Observed and Simulated (dashed lines) Average Weekly Flowsfor 17 Sites. Length of record is shown in Table 5. Note different y-axis scales.



Shifts of one to two months in the timing of the meanmonthly peak flow are expected under a future sce-nario in which the temperature increases by 5!C(Miller et al., 2003). Changes in runoff timing arerelated to the loss of snowpack which is expected tooccur mostly in the 1,300-2,700 m elevation range(Knowles and Cayan, 2004). Depending on the CO2

emissions scenario used, others have found changes inrunoff center of mass are expected to vary from one toseven weeks by 2100 (Maurer, 2007).

For this analysis, we chose to focus on simple tem-perature increase scenarios similar to the methodemployed in Miller et al. (2003). The model is capableof easily incorporating downscaled weekly climatedata however in this first step we wanted to explorethe sensitivity of the model outputs to simple temper-ature increases. This was based on the conclusion

that the most common warming projections in globalcirculation models for California show an increase ofabout 5!C during the 21st Century. The expectedchange in rainfall is less certain but is expected to bemodest (Dettinger, 2005). To explore how changeswrought by increasing temperatures may evolve overtime and space and affect hydrologic responses weadded fixed increases of 2!C, 4!C, and 6!C to the his-torical 1982-2001 input temperature time series.

Snow Accumulation

The study area average change in snowmelt volumeas a function of elevation for the three temperatureincrease scenarios is presented in Figure 7A. Similarto other studies (Miller et al., 2003; Knowles and Ca-

0

0.1

0.2

0.3

0.4

0.5

0.6A

B

0-500 500-750

750-1000

1000-1250

1250-1500

1500-1750

1750-2000

2000-2250

2250-2500

2500-2750

2750-3000

3000-3250

3250-3500

3500-3750

3750-4000

Elevation (m)

Dec

reas

e in

Ann

ual S

now

Mel

t (m

of w

ater

) 2 deg4 deg

6 deg

0

10

20

30

40

50

60

70

80

90

100

0-500 500-750

750-1000

1000-1250

1250-1500

1500-1750

1750-2000

2000-2250

2250-2500

2500-2750

2750-3000

3000-3250

3250-3500

3500-3750

3750-4000

Elevation (m)

Dec

reas

e in

Ann

ual S

now

Mel

t (%

)

2 deg4 deg

6 deg

FIGURE 7. Decrease in Average Annual Snowmelt (1982-2001) Expressed as Depth of Water(A) and Percent Reduction (B) Under 2!C, 4!C, and 6!C Temperature Scenarios.



yan, 2004; Maurer, 2007) the 1,750-2,750 m elevationrange experienced the largest reduction in snowmelt.The peak reduction in snowmelt occurred in thesemid-elevations because lower elevations (<1,750 m)have less snow accumulation initially and higher(>2,750) elevations are cold enough that the modeledtemperature increases have less of an overall effect.

In another view of the same information, the per-cent reduction in snowmelt (Figure 7B) varied from70% at the lowest elevations to less than 10% at thehighest elevations under the 2!C warming scenario.With an increase of 4!C, over 75% of the low elevation(0-1,750 m) snowpack was lost and at the highestelevations around 20% was lost. In the 6!C warmingscenario over 90% of the snowpack is lost in theelevation range of 0 to 2,000 m and at the highestelevations (>3,750 m) there was a loss of 25%.

Shift in Runoff Center of Mass Timing

Similar to previous studies (Miller et al., 2003;Knowles and Cayan, 2004; Maurer, 2007), the reduc-tion in snowpack results in a shift in the runoff cen-ter of mass to earlier in the year (Figure 8). This is aresult of more precipitation falling as rain anddirectly becoming runoff instead of remaining in stor-age within the snowpack and due to earlier initiationof snowmelt. The higher spatial resolution of thismodel enables a distributed analysis of the change inrunoff center of mass. Impacts of temperature warm-ing are geographically non-uniform (Figure 9). Underthe 2!C scenario, the upper elevations of the Feather,Yuba, and American River watersheds will beaffected most as they fall predominantly in the 1,750-2,750 m range. Shifts in runoff center of mass timing

will exceed four weeks at some elevations in thesewatersheds. In contrast, the higher elevation moun-tainous regions in the southern portion of the studyarea will be impacted less in the 2!C scenario due tolower baseline temperatures. This indicates that asthe climate warms the highest elevations in thenorthern portion of the study area will be impactedfirst. If temperatures continue to increase the higherelevation mountains of the southern portion of thestudy area will also be impacted. In the 4!C scenario,shifts in runoff timing of more than six weeks willoccur in the highest mountains of the Feather Riverthrough Tuolumne River watersheds. Further south,the middle elevation portions of the Stanislaus Riverthrough Kings River watersheds will experienceshifts greater than six weeks. In the 6!C scenario,nearly all the highest mountains (>1,750 m) in thewatersheds from the Feather to the Kings River havea shift in runoff timing greater than six weeks. TheKern River watershed shows less sensitivity to thechanges in snowmelt. This is an artifact of the cali-brated soil water storage capacity, discussed above,which was nearly an order of magnitude larger thanmost other watersheds (Table 3).

Management Implications

The effects of increasing air temperature result innon-uniform hydrological responses over the region ofstudy. Analysis of the results presented in Figure 9and Table 6 reveal the importance of spatial scale inclimate change analysis for the Sierra Nevada. Formanagers of the large terminal reservoirs located atthe outlet of most watersheds, the model shows thata 2!C warming will result in an average shift in the

FIGURE 8. Shift in Runoff Center of Mass (1982-2001) Under 2!C, 4!C, and 6!C Temperature Scenarios.



center of mass of the runoff hydrograph of 1.8 weeks.However, relative shifts internal to the watershedsreveal that some regions will experience a shift of upto five weeks under a 2!C warming. Knowledge of themagnitude and location of such changes will beimportant to resource managers operating at scalessmaller than the watershed. For instance, manySierra Nevada hydropower projects rely on runofforiginating in the elevation bands most sensitive to

temperature changes and aquatic species that rely onrunoff from mid-elevation bands may be the firstimpacted by climate change. These observations illus-trate the utility of our modeling approach by provid-ing scenario-based simulations critical to decisionmaking within a localized context.

In future studies this model can be used to esti-mate the change in overall ET and its effect on runoffvolumes. After infrastructure elements such as damsand diversions are added, modification of operationsrules can be studied as a response to the shift in run-off timing. Model predictions of streamflow, soil mois-ture conditions, and ET will be useful in studyingchanges in physical habitat for terrestrial and aquaticenvironments. Potential changes in terrestrial habitatdue to climate change are related to changes in thelocal hydrologic balance and the resultant soil watercontent available to sustain vegetation cover neces-sary for species survival (van Mantgem and Stephen-son, 2007). Potential changes in aquatic habitat dueto climate change are related to changes in stream-flow temperature, volume, and timing which affect in-stream habitat for aquatic organisms (Poff, 2002).

CONCLUSIONS

A rainfall-runoff model that predicts natural flowshas been developed for the western slope of the

FEA

KRN

SJN

AMR

TUO

KNG

YUB

MER

STN

MOKCOS

KAW

TUL

CAL

BAR

FEA

KRN

SJN

AMR

TUO

KNG

YUB

MER

STN

MOKCOS

KAW

TUL

CAL

BAR

FEA

KRN

SJN

AMR

TUO

KNG

YUB

MER

STN

MOKCOS

KAW

TUL

CAL

BAR

Weeks0 - 1.51.5 - 33 - 4.54.5 - 6> 6

FIGURE 9. Shift in Runoff Center of Mass for 2!C, 4!C, and 6!C Increases inTemperature (from left to right). Values are averages for 1982-2001.

TABLE 6. Runoff Center of Mass at Watershed Outlets in WeeksFrom Oct. 1 and Shift in Timing of Center of Mass in Weeks Under2!C, 4!C, and 6!C Temperature Increase Scenarios (1982-2001).

Watershed

0!CCenter

ofMass

2!CCenter

ofMass

4!CCenter

ofMass

6!CCenter

ofMass D2!C D4!C D6!C

FEA 25.0 22.7 21.5 21.0 2.2 3.5 3.9YUB 23.2 20.9 19.6 19.1 2.3 3.6 4.1BAR 19.8 19.5 19.4 19.3 0.3 0.4 0.5AMR 23.6 21.3 19.8 19.1 2.3 3.7 4.5COS 21.6 20.5 20.1 19.9 1.0 1.5 1.7MOK 25.7 23.1 20.8 19.6 2.7 4.9 6.1STN 26.4 23.7 21.4 19.9 2.8 5.1 6.6TUO 25.8 23.6 21.8 20.4 2.2 4.0 5.4MER 26.9 24.8 22.7 21.1 2.2 4.2 5.8SJN 29.6 27.5 25.4 23.5 2.1 4.2 6.1KNG 30.9 28.8 26.7 24.8 2.0 4.1 6.1KAW 28.9 26.9 25.1 23.5 2.1 3.9 5.4TUL 26.5 24.8 23.6 22.8 1.6 2.8 3.6KRN 30.7 29.6 28.5 27.5 1.1 2.2 3.3CAL 22.0 21.8 21.7 21.7 0.2 0.3 0.4



Sierra Nevada including watersheds from theFeather River in the north to the Kern River in thesouth. The model provides the ability to analyze cli-mate change induced alterations in the hydrologiccycle within individual watersheds. The analysisshown here indicates that the impacts of warmingtemperatures will be experienced in a time and spacevarying manner that is largely a function of the dis-tribution of elevation in the Sierra Nevada. Themodel has been constructed in the WEAP21 frame-work which lends itself to the inclusion of the humanoperated system and therefore will be a powerful toolin assessing climate change impacts and adaptationsfor resource managers at the subwatershed scale.

ACKNOWLEDGMENTS

We would like to acknowledge the support of the Resources Leg-acy Fund Foundation and the Center for Watershed Sciences at theUniversity of California, Davis. Michael Kiparsky’s work was sup-ported by an NSF Graduate Fellowship, an NSF Doctoral Disserta-tion Research Improvement Grant, and by the California EnergyCommission.

LITERATURE CITED

Barnett, T.P., J.C. Adam, and D.P. Lettenmaier, 2005. PotentialImpacts of a Warming Climate on Water Availability in Snow-Dominated Regions. Nature 438:303-309, doi: 310.1038/nature04141.

Cayan, D.R., S.A. Kammerdiener, M.D. Dettinger, J.M. Caprio, andD.H. Peterson, 2001. Changes in the Onset of Spring in theWestern United States. Bulletin of the American MeteorologicalSociety 82:399-415.

Daly, C., 2006. Guidelines for Assessing the Suitability of SpatialClimate Data Sets. International Journal of Climatology 26:707-721.

Dettinger, M.D., 2005. From Climate-Change Spaghetti to ClimateChange Distributions for 21st Century California. San FranciscoEstuary and Watershed Science 3(1) (March 2005), Article 4.

Dettinger, M.D., D.R. Cayan, M. Meyer, and A.E. Jeton, 2004. Sim-ulated Hydrologic Responses to Climate Variations and Changein the Merced, Carson, and American River Basins, SierraNevada, California, 1900-2099. Climatic Change 62:283-317.

Diffenbaugh, N.S., L.C. Sloan, M.A. Snyder, J.L. Bell, J. Kaplan,S.L. Shafer, and P.J. Bartlein, 2003. Vegetation Sensitivity toGlobal Anthropogenic Carbon Dioxide Emissions in a Topo-graphically Complex Region. Global Biogeochemical Cycles17(2):1067, doi: 10.1029/2002GB001974.

Galewsky, J. and A. Sobel, 2005. Moist Dynamics and OrographicPrecipitation in Northern and Central California During theNew Year’s Flood of 1997. Monthly Weather Review 133:1594-1612.

Gibson, C.A., J.L. Meyer, N.L. Poff, L.E. Hay, and A. Georgakakos,2005. Flow Regime Alterations Under Changing Climate in TwoRiver Basins: Implications for Freshwater Ecosystems. RiverResearch and Applications 21:849-864, doi: 810.1002/rra.1855.

Homer, C., C. Huang, L. Yang, B. Wylie, and M. Coan, 2004. Devel-opment of a 2001 National Landcover Database for the United

States. Photogrammetric Engineering and Remote Sensing70:829-840.

Hulme, P.E., 2005. Adapting to Climate Change: Is There Scope forEcological Management in the Face of a Global Threat? Journalof Applied Ecology 42:784-794. doi: 710.1111/j.1365-2664.2005.01082.x.

Kim, J., 2001. A Nested Modeling Study of Elevation-DependentClimate Change Signals in California Induced by IncreasedAtmospheric CO2. Geophysical Research Letters 28:2951-2954.

Knowles, N., 2000. Hydroclimate of the San Francisco Bay-DeltaEstuary and Watershed. University of California, San Diego, LaJolla, CA, p. 292.

Knowles, N. and D.R. Cayan, 2004. Elevational Dependence of Pro-jected Hydrologic Changes in the San Francisco Estuary andWatershed. Climatic Change 62:319-336.

Knowles, N., M.D. Dettinger, and D.R. Cayan, 2006. Trends inSnowfall Versus Rainfall in the Western United States. Journalof Climate 19:4545-4559.

Koczot, K.M., A.E. Jeton, B.J. McGurk, and M.D. Dettinger, 2005.Precipitation-Runoff Processes in the Feather River Basin,Northeastern California, With Prospects for Streamflow Predict-ability, Water Years 1971-97. U.S. Geological Survey ScientificInvestigations Report 2004-5202, 82 p.

Kueppers, L.M., M.A. Snyder, L.C. Sloan, E.S. Zavaleta, andB. Fulfrost, 2005. Modeled Regional Climate Change and Califor-nia Enaemic Oak Ranges. Proceedings of the National Academyof Sciences of the United States of America 102(45):16281-16286.

van Mantgem, P.J. and N.L. Stephenson, 2007. Apparent Climati-cally Induced Increase of Tree Mortality Rates in a TemperateForest. Ecology Letters 10:909-916, doi: 910.1111/j.1461-0248.2007.01080.x.

Maurer, E.P., 2007. Uncertainty in Hydrologic Impacts of ClimateChange in the Sierra Nevada, California, Under Two EmissionsScenarios. Climatic Change 82:309-325, doi: 310.1007/s10584-10006-19180-10589.

Maurer, E.P., I.T. Stewart, C. Bonfils, P.B. Duffy, and D. Cayan,2007. Detection, Attribution, and Sensitivity of Trends TowardEarlier Streamflow in the Sierra Nevada. Journal of Geophysi-cal Research 112, D11118, doi: 10.1029/2006JD008088.

McCabe, G.J. and M.P. Clark, 2005. Trends and Variability inSnowmelt Runoff in the Western United States. Journal ofHydrometeorology 6:476-482.

Meyers, M.P. and W.R. Cotton, 1992. Evaluation of the Potentialfor Wintertime Quantitative Precipitation Forecasting OverMountainous Terrain With an Explicit Cloud Model 1.2-Dimen-sional Sensitivity Experiments. Journal of Applied Meteorology31:26-50.

Miller, N.L., K.E. Bashford, and E. Strem, 2003. Potential Impactsof Climate Change on California Hydrology. Journal of theAmerican Water Resources Association 39:771-784.

Myrick, C.A. and J.J. Cech, 2004. Temperature Effects on JuvenileAnadromous Salmonids in California’s Central Valley: WhatDon’t We Know? Reviews in Fish Biology and Fisheries 14:113-123.

Palmer, M.A., C.A. Reidy Liermann, C. Nilsson, M. Florke,J. Alcamo, P. Lake, and N. Bond, 2008. Climate Change andthe World’s River Basins: Anticipating Management Options.Frontiers in Ecology and the Environment 6(2):81-89. doi:10.1890/060148.

Poff, N.L., 2002. Ecological Response to and Management ofIncreased Flooding Caused by Climate Change. PhilosophicalTransactions of the Royal Society of London Series A – Mathe-matical Physical and Engineering Sciences 360:1497-1510.

Regonda, S.K., B. Rajagopalan, M. Clark, and J. Pitlick, 2005.Seasonal Cycle Shifts in Hydroclimatology Over the WesternUnited States. Journal of Climate 18:372-384.



Stewart, I.T., D.R. Cayan, and M.D. Dettinger, 2004. Changes inSnowmelt Runoff Timing in Western North America Under a‘Business as Usual’ Climate Change Scenario. Climatic Change62:217-232.

Stewart, I.T., D.R. Cayan, and M.D. Dettinger, 2005. ChangesToward Earlier Streamflow Timing Across Western NorthAmerica. Journal of Climate 18:1136-1155.

Thornton, P.E., S.W. Running, and M.A. White, 1997. GeneratingSurfaces of Daily Meteorological Variables Over Large Regionsof Complex Terrain. Journal of Hydrology 190:214-251.

USACE, 1998. Engineering and Design: Runoff From Snowmelt.Engineer Manual 1110-2-1406. United States Army Corps ofEngineers, Washington, DC.

Vicuna, S., E.P. Maurer, B. Joyce, J.A. Dracup, and D. Purkey,2007. The Sensitivity of California Water Resources to ClimateChange Scenarios. Journal of the American Water ResourcesAssociation 43:482-498, doi: 410.1111/j.1752-1688.2007.00038.x.

Yates, D., D. Purkey, J. Sieber, A. Huber-Lee, H. Galbraith, J.West, S. Herrod-Julius, C. Young, B. Joyce, and M. Rayej, 2009.Climate Driven Water Resources Model of the SacramentoBasin, California. Journal of Water Resources Planning andManagement 135(5):303-313.

Yates, D., J. Sieber, D. Purkey, and A. Huber-Lee, 2005. WEAP21 –A Demand-, Priority-, and Preference-Driven Water PlanningModel Part 1: Model Characteristics. Water International 30:487-500.

Zhu, T.J., M.W. Jenkins, and J.R. Lund, 2005. Estimated Impactsof Climate Warming on California Water Availability UnderTwelve Future Climate Scenarios. Journal of the AmericanWater Resources Association 41:1027-1038.



modeling the hydrology of climate change in … · hydrology, and soils data are used to model the...

Documents