1. Introduction
In recognition of the vulnerability of freshwater resources to changing climate, many studies have sought to examine the effects of climate change on components of the hydrologic budget. The most common approach has been to combine basin-scale hydrologic models with climate-change scenarios derived from general circulation model (GCM) output (Buytaert et al. 2009). GCMs are considered the most advanced tool currently available for simulating the effects of increasing greenhouse gases on the global climate system (Alley et al. 2007). GCM simulations of future climate through 2099 project a wide range of possible scenarios (Alley et al. 2007) but often overlook numerous climatological details necessary for hydrologic modeling at the basin scale because of their coarse resolution (Wigley et al. 1990; Carter et al. 1994; Xu 1999).
Statistical or dynamical methods can be used to downscale information from coarse-resolution GCMs to the basin scale for hydrologic modeling. Statistical downscaling (Wilks 1995; Wilby et al. 1999; Antolik 2000; McCarthy et al. 2001; Alley et al. 2007) uses empirical relations between features reliably simulated by a GCM at grid scales and surface predictands at subgrid scales. Dynamical downscaling uses regional climate model simulations with initial and lateral boundary conditions from GCM output for much more spatially detailed climate simulations over a region of interest (Hay and Clark 2003; Leung et al. 2003; Giorgi et al. 2001).
Although considerable research has gone into these downscaling procedures (e.g., Fowler et al. 2007), the most straightforward means of obtaining higher-spatial-resolution scenarios is to impose the change in coarse-scale GCM projections to an observed climate baseline of station points: the “change factor” method or the “delta change” method (e.g., Arnell 2003a; Arnell 2003b; Arnell and Reynard 1996; Hay et al. 2000; Diaz-Nieto and Wilby 2005; Eckhardt and Ulbrich 2003; Pilling and Jones 1999; Prudhomme et al. 2002; Hay and McCabe 2010). Fowler et al. (Fowler et al. 2007) reviewed the current downscaling literature specifically for hydrological impacts. They concluded that simple statistical downscaling methods, such as the change factor method, seem to perform as well as more complicated methods in reproducing mean climatic characteristics.
In this study, 14 basins, representing different hydroclimatic regions, were selected as modeling sites to assess the sensitivity and potential effects of long-term climate change on the freshwater resources of the United States using the Precipitation-Runoff Modeling System (PRMS) watershed model. Because the scope of this study was limited to climate change (as opposed to weather change), the change factor method was used to capture how changes in climate might evolve through the twenty-first century. For each basin, simulated precipitation and temperature from five GCMs, using one baseline (twentieth century) and three future (years 2001 through 2099) emission scenarios (Alley et al. 2007), were downscaled using the change factor method. The output variables simulated by PRMS indicate the simulated hydrologic response and sensitivity of any particular basin to GCM-projected changes in climate.
2. Study sites
The 14 basins selected as modeling sites to assess the sensitivity and potential effects of long-term climate change on the freshwater resources of the United States are shown in Figure 1 and listed by drainage area in Table 1. For each basin, Table 1 lists the basin abbreviation (used throughout this paper for identification); U.S. Geological Survey stream gauge used for model calibration (if applicable); drainage area; number of modeling units [hydrologic response units (HRUs)]; and elevation range of the modeling units. Each basin has a current or past PRMS model application developed for a variety of reasons. If published, the reference for each of these past PRMS model application studies is listed in Table 1.
Location map of the United States showing the sites of the 14 selected basins. PRMS models were developed for the basin areas shown. Red triangle denotes location of stream gauge for model calibration (note that basins are not to scale; see Table 1 for relative scales).
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
Selected basins listed by drainage area.
The 14 basins (Figure 1) can be grouped into six U.S. geographical regions:
The Naches and Sprague basins are located in the Cascade Mountains of the Pacific Northwest.
The Feather and Sagehen basins are located in the Sierra Nevada.
The Flathead, East, and Yampa basins are located in the Rocky Mountains.
The Starkweather, Clear, Black Earth, and Trout Lake basins are located in the Midwest.
The Cathance and Pomperaug basins are located in the northeastern United States.
The Flint basin is located in the southeastern United States.
3. Simulation model descriptions
Two types of simulation models were loosely coupled in this study: a general-purpose watershed hydrology model (PRMS) and GCMs. PRMS models were loosely coupled with GCMs by downscaling the coarse-resolution GCM precipitation and temperature output and applying it to the basin-scale PRMS models. These two types of simulation models and the downscaling procedure are described below.
3.1. PRMS
The watershed hydrology model PRMS (Leavesley et al. 1983; Markstrom et al. 2008) is a deterministic, distributed-parameter, process-based model used to simulate and evaluate the effects of various combinations of precipitation, climate, and land use on basin response. Response to normal and extreme rainfall and snowmelt can be simulated to evaluate changes in water-balance relations, streamflow regimes, soil-water relations, and groundwater recharge. Each hydrologic component used for generation of streamflow is represented within PRMS by a process algorithm that is based on a physical law or an empirical relation with measured or calculated characteristics.
Distributed-parameter capabilities of PRMS are provided by partitioning a basin into HRUs, using characteristics such as slope, aspect, elevation, vegetation type, soil type, and precipitation distribution. These HRUs are assumed to be homogeneous with respect to their hydrologic response. For each HRU, a water balance is computed each day and an energy balance is computed twice each day. PRMS is conceptualized as a series of reservoirs (impervious zone, soil zone, subsurface, and groundwater; see Figure 2) where the outputs by HRU combine to produce the daily basin response using one of the following three options (Viger et al. 2010): 1) a no-flow routing procedure in which the sum of the responses of all HRUs, weighted on a unit-area basis, produces the daily basin response; 2) a cascading-flow routing procedure; or 3) a Muskingum-flow routing procedure.
Schematic diagram of a watershed and its inputs simulated by PRMS (from Markstrom et al. 2008).
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
PRMS uses daily climate values of measured precipitation and maximum and minimum temperature distributed to each HRU. All of the PRMS models used in this study distribute precipitation and maximum and minimum temperature data from a set of points (station locations) to each HRU. The HRU-distributed climate values are used to compute solar radiation (SR), potential evapotranspiration (PET), actual evapotranspiration (AET), sublimation, snowmelt, streamflow, and infiltration in a PRMS simulation.
3.1.1. Evapotranspiration
In PRMS, PET is computed to determine the theoretical absolute upper limit of simulated AET: when available water is nonlimiting, AET equals PET. AET is computed in PRMS using many factors, including the simulated PET, vegetation type and cover density, land-use characteristics, soil type, simulated atmospheric conditions (available energy, precipitation, and cloud cover), and soil-water availability. AET is the computed rate of water loss due to plant activity and evaporation, which reflects the availability of water to satisfy PET.
For this study, all PRMS configurations used the modified Jensen–Haise (JH) method to estimate PET (Jensen and Haise 1963). The method uses mean near-surface air temperature and diurnal saturation vapor pressure differences (estimated from air temperature), elevation, and mean incoming solar radiation (solar radiation is estimated from air temperature and precipitation and is adjusted to account for aspect and slope of the HRU). The JH method, therefore, uses temperature and precipitation to estimate the daily variation in solar radiation at the surface and daily temperature of the evaporation surfaces. The JH method does not explicitly account for daily variation in relative humidity and assumes complete mixing of the atmosphere above the evaporation surface and hence does not explicitly account for variation in the near-surface vapor pressure gradient and mass transfer of vapor into the atmosphere.
3.1.2. PRMS calibration and evaluation
The PRMS models used in this study were initially developed by U.S. Geological Survey personnel for previous studies (see Table 1). This means the models were calibrated for different purposes and under different climate conditions. An overlapping period of record, or baseline condition, was determined by examining the PRMS input data in the selected models. The water years (WYs) 1988–99 were chosen for baseline conditions. This period ended in 1999 because GCM simulations for current conditions ended in 1999 for many of the archived models (a necessary condition for the GCM downscaling described later). Baseline conditions were not extended back prior to WY 1988 because of the length of the Natural Resources Conservation Service’s Snow Telemetry (SNOTEL) record used by many of the snowmelt-dominated PRMS models in this study. For this study, all PRMS models were evaluated for accuracy on a mean monthly basis because simulated mean monthly values for future conditions were evaluated in this study.
Initial evaluation of PRMS-simulated basin mean monthly values of solar radiation, potential evapotranspiration, and streamflow showed varying degrees of accuracy. To ensure consistency in model simulations, all models were calibrated using Luca (Hay and Umemoto 2006; Hay et al. 2006b; Hay et al. 2006c), a multiple-objective, stepwise, automated procedure that uses the shuffled complex evolution global search algorithm to calibrate PRMS. For this study, the automated procedure included the sequential calibration of a model’s simulation of SR, PET, and water balance. This process ensures that intermediate model fluxes as well as the water balance are simulated consistently with measured values (see Hay et al. 2006c).
Reliably reproducing the baseline period is important because any biases are likely to transfer to the future simulations of flow (Prudhomme and Davies 2009). Assessments of maximum and minimum temperature and precipitation baseline conditions (mean monthly baseline conditions; WYs 1989–99) for all basins are summarized in Figure 3. For evaluation purposes, the measured and PRMS-simulated mean monthly SR, PET, and streamflow values for baseline conditions are shown in Figure 4. The measured values (when available) are shown in red, and the PRMS-simulated values are shown in black. The Feather River basin does not have measured streamflow at the outlet of the basin: measured and simulated streamflow are shown for the Middle Fork basin, one of the interior gauges used for calibration within the Feather River basin (see Koczot et al. 2005). After calibration, all PRMS models were judged to produce accurate mean monthly results for SR, PET, and the water balance (when applicable) for the baseline period (Figure 4).
Basin mean monthly (1988–99) (a) maximum temperature (b) minimum temperature, and (c) precipitation for the 14 basins.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
Calibration results by basin: basin mean monthly measured (red) and PRMS-simulated (black) SR, PET, and streamflow.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
3.2. GCMs
General circulation models (or global climate models) are numerical models that represent all components of the Earth system, including the atmosphere, oceans, cryosphere, and land surface. The climate-change signal from individual GCMs (i.e., the difference between simulated future climate and simulated historical climate) can vary in direction and magnitude because of the uncertainties associated with each GCM parameterization, initial forcings, emission scenarios, and representation of the Earth system components (Fealy and Sweeney 2008). The same forcings can produce different GCM responses because of differences in the method for modeling climate system and feedbacks and the initial conditions used for the climate simulations. Given this parametric and structural uncertainty in climate modeling, it is desirable to use output from more than one GCM to obtain a range of potential future climatic conditions. Ensembles of GCM output provide an envelope of possible regional changes and their accompanying uncertainties (Murphy et al. 2004; Boorman and Sefton 1997; Alley et al. 2007).
Representing GCM uncertainties using multiple GCM model outputs is now feasible because of the large data archives available at the GCM modeling centers. The World Climate Research Programme’s Coupled Model Intercomparison Project phase 3 (CMIP3) multimodel dataset archive, which was used in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report Special Report on Emission Scenarios (SRES) (Alley et al. 2007) contains 28 archived GCMs with multiple emission scenarios. The emission scenarios describe how greenhouse gas emissions could evolve between 2000 and 2100, depending on various hypotheses on what the state of the future world may be. The range in results from the various GCMs can then be compared by emission scenario.
An analysis of the monthly precipitation and maximum and minimum temperature output from the CMIP3 dataset archive indicated that five of the GCM archives had output for baseline and three future emission scenarios that were suitable for the PRMS models. Table 2 lists the five GCMs chosen for the study, and Table 3 describes the baseline (1988–99) and three future (2001–99) emission scenarios chosen for analysis for each of the GCMs in the study. The three future emission scenarios lead to very different patterns in greenhouse gas emissions and concentrations, from relatively low (B1), to medium (A1B), to high concentrations (A2) during the twenty-first century (Alley et al. 2007).
GCM outputs used in this study from the World Climate Research CMIP3 multimodel dataset archive. GCM definitions not expanded in the text: Bjerknes Centre for Climate Research Bergen Climate Model (BCCR-BCM2.0), Commonwealth Scientific and Industrial Research Organisation Mark version 3.0 (CSIRO Mk3.0), Institute of Numerical Mathematics Coupled Model, version 3.0 (INM-CM3.0), and Model for Interdisciplinary Research on Climate 3.2, medium-resolution version [MIROC3.2(medres)].
In this study, GCM outputs were downscaled because the GCM spatial scales are not appropriate for hydrologic modeling at the scale of the 14 basins. Although there are existing downscaled GCM products available at resolutions that might be considered appropriate for basin-scale hydrologic model applications (e.g., http://gdo-dcp.ucllnl.org/downscaled_cmip3_projections/dcpInterface.html; http://www.gisclimatechange.org/), these products are gridded. All of the PRMS models used in this study distribute precipitation and maximum and minimum temperature data from a set of points (station locations) to HRUs defined for each basin. To apply these gridded products, each PRMS model would have to have been reconfigured to use gridded information, changing each model considerably from its original configuration. Therefore, a downscaling procedure that goes from coarse GCM grids to the climate stations used in each PRMS model was desired.
The strength of the GCMs is their ability to project changes in climate (Le Treut et al. 2007). The approach used in this analysis examined changes in mean monthly hydrologic conditions given simulations of future climatic conditions. Mean monthly climate-change factors (percentage changes in precipitation and degree changes in temperature) were computed for 12-yr moving window periods (2001–99) using the 20C3M (1988–99), A2, B1, and A1B emission scenarios. A 12-yr moving window, starting in 2001 and ending in 2099, results in 1320 future emission scenarios [(eighty-eight 12-yr climatologies, one per year starting with 2001–12 and ending with 2088–99) × (3 emission scenarios) × (5 GCMs)]. Climate-change input files for PRMS were generated by modifying the daily PRMS precipitation and maximum and minimum temperature inputs (1988–99) with the mean monthly climate-change factors derived from the GCMs, resulting in 1320 PRMS input files for each study area. The first year of each 12-yr simulation was used as PRMS initialization and is not included in results analysis. Figure 5 shows a schematic of the climate-change factor method as applied in this study.
Schematic of the climate-change factor method as applied in this study.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
4. Results
Climate-change scenarios were generated for PRMS by modifying the 12-yr baseline period of daily PRMS precipitation and temperature with the mean monthly climate-change factors derived from the GCMs. The climate-change factor input files developed for PRMS incorporate the climate station records as well as the change factor from the closest GCM grid node. The only difference in the PRMS baseline conditions and the PRMS climate-change input files is in the mean monthly values. The temporal sequencing remains unchanged, limiting this method to studies examining changes in mean climatic conditions. If changes in wet- or dry-spell lengths are important to the effects assessment, then another downscaling choice would be more appropriate (Wilby et al. 2004).
The climate-change factor files were used as input to PRMS to estimate potential effects of climate change on basin-scale hydrology. The scope of this study is limited to climate change; therefore, to capture how changes in climate may evolve through the twenty-first century, the PRMS output was analyzed on a basin mean monthly and annual basis. PRMS produces over 200 output variables, which characterize the hydrologic conditions for all HRUs on any day of the simulated time period (see Markstrom et al. 2008). For this study, basin mean annual and monthly values of precipitation and maximum and minimum temperature, evapotranspiration, streamflow, and soil moisture are considered.
The central tendencies, or mean, of the five GCMs for each of the three emission scenarios by basin were calculated for the selected PRMS output variables. For each of the eighty-eight 12-yr windows (from window 2001–12 through window 2088–99), the central tendency was calculated as a mean annual value using the last 11 years of each 12-yr window (the first year was reserved for PRMS model initialization). The results of a regression analysis (Table 4) show the projected change (slope) in the central tendency and associated adjusted R2 (adjR2) based on the central tendencies of the five GCMs for each of the three emission scenarios by basin. The slope indicates the change in the central tendency for the selected variable by year (center of the 11-yr window); red indicates a significant negative trend (p < 0.05); and blue indicates a significant positive trend (p < 0.05), accounting for effects of lag-1 autocorrelation on the degrees of freedom (Lettenmaier 1976; McCabe and Wolock 1997). The adjusted R2 value gives an indication of the variability in the central tendency trend. For example, the higher the adjusted R2 value, the straighter the central tendency line.
Projected change by year (slope) and adjR2 based on the central tendencies of the five GCMs for the three emission scenarios by basin: (a) precipitation, (b) maximum temperature, (c) minimum temperature, (d) evapotranspiration, (e) streamflow, (f) snow-covered area, (g) snowpack water equivalent, and (h) soil moisture. Blue (red) indicates a significant negative (positive) trend (p < 0.05) accounting for lag-1 autocorrelation.
4.1. Precipitation
The central tendencies are shown graphically in mean annual plots of precipitation in Figure 6. The three solid colored lines indicate the 11-yr moving mean daily values (x axis indicates center of 11-yr window) for the three future emission scenarios (central tendencies of the five GCMs for A2, A1B, and B1). The mean annual plots show the range in the 11-yr moving mean daily values by emission scenario. Red, blue, and yellow represent the A2, A1B, and B1 emission scenarios, respectively. The shaded areas shown for each emission scenario indicate an envelope or range of potential future climatic conditions simulated by the five GCMs for each emission scenario.
Range in 11-yr moving mean daily precipitation values by emission scenario for the 14 basins.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
Projected mean annual changes in precipitation are highly variable; many of the basins show both increases and decreases in future precipitation, depending on the GCM and the emission scenario. The large range in the precipitation projections indicates a large amount of uncertainty in the GCM projections used in this study. The change in precipitation over time varies by basin and most of the significant trends are positive, but emission scenarios do not always agree on the direction of the trend (Table 4a). In general, the associated adjusted R2 values in Table 4a are low, indicating high variability associated with the overall trend.
An overall increase in the central tendencies for precipitation were projected using the A1B and A2 emission scenarios for the Flathead, Naches, and Cathance basins; the A1B emission scenario for the Sprague, Yampa, East, and Sagehen basins; and the B1 emission scenario for the Trout Lake basin. The only basin with projected overall decreases in the central tendencies was the Flint (A1B and A2) basin. None of the basins showed significant overall increases or decreases in the central tendencies for all emission scenarios.
4.2. Temperature
All climate-change emission scenarios evaluated with PRMS project a steady increase in maximum and minimum temperatures progressing through the twenty-first century (Tables 4b,c). Figures 7 and 8 show summaries of the range in 11-yr moving mean daily maximum and minimum temperature values by emission scenario for the 14 basins. The largest increases and variability in temperatures are associated with the A2 scenario, and the smallest increases and variability are associated with the B1 scenario. The projected increases in maximum temperature are larger than those projected for minimum temperature.
Range in 11-yr moving mean daily maximum temperature values by emission scenario for the 14 basins.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
Range in 11-yr moving mean daily minimum temperature values by emission scenario for the 14 basins.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
4.3. Actual evapotranspiration
The projected increases in temperatures result in increases in actual evapotranspiration (Table 4d) on an annual basis and, when soil moisture is available, on a monthly basis (not shown). Figure 9 shows summaries of the range in 11-yr moving mean daily evapotranspiration values by emission scenario for the 14 basins. In general, an overall increase in the central tendencies for actual evapotranspiration was projected with low variability based on the relatively high adjusted R2 values (Table 4d). This rate of increase can vary by a couple of orders of magnitude between basins reflecting the dependency of actual evapotranspiration on basin characteristics (not just the change in temperature).
Range in 11-yr moving mean daily AET values by emission scenario for the 14 basins.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
4.4. Streamflow
Figure 10 shows summaries of the range in 11-yr moving mean daily streamflow values by emission scenario for the 14 basins. The combination of the variability in the precipitation projections with the steady projected increases in temperature results in inconsistent trends in simulated streamflow for projected future climatic conditions because both positive trends in precipitation and negative trends in future streamflow can occur, depending on the GCM and the emission scenario (Tables 4a,e). However, some patterns do emerge. For example, the central tendencies of the PRMS simulations using the A1B and A2 emission scenarios project decreases in mean annual streamflow in the majority of the basins (Flint, Yampa, East, Starkweather, Clear, Trout Lake, Black Earth, and Cathance) with the Yampa and East basins projecting decreases using the B1 scenario as well. None of the basins project significant increases in central tendencies of the PRMS simulations when using the B1 emission scenario (the lowest greenhouse gas emission pathway tested in this study). The only increases in the central tendencies for streamflow were projected using the A1B and A2 emission scenarios for the Naches and the A1B emission scenario for the Flathead. The variability in all these projections is quite large based on the range in the emission scenarios in Figure 10 and the adjusted R2 values in Table 4e.
Range in 11-yr moving mean daily streamflow values by emission scenario for the 14 basins.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
The projected warming can have a dramatic effect on the timing of peak monthly streamflow, especially in the snowmelt-dominated basins. Figure 11 compares projected mean monthly average temperature versus simulated mean daily streamflow over time. In Figure 11, mean monthly temperature and mean daily streamflow are calculated using the five GCMs and three emission scenarios for each of the 88 windows for the 14 selected basins. The steady increase in projected temperature can then be compared with the corresponding changes in streamflow on an approximately monthly basis. The months that appear relatively flat (such as August through November in Sagehen Basin; Figure 11n) indicate that increasing temperatures will have little effect on the flow volumes for those months. Months that show positive slopes (such as December through March for the Sprague Basin; Figure 11d) indicate that increasing temperatures will increase the flow volumes for those months. Months that show negative slopes (such as June in the Yampa Basin; Figure 11f) indicate that increasing temperatures will decrease the flow volumes for those months. In some cases, the flow volumes steadily increase and then steadily decrease with increasing temperature (such as March in the Cathance Basin; Figure 11m). The months that show high variability in streamflow (such as November through June in the Trout Lake Basin; Figure 11k) indicate the larger influence of the highly variable precipitation projections (rather than temperature increases) on streamflow.
Mean monthly average temperature vs mean daily streamflow using the five GCMs and three emission scenarios calculated for each of the 88 windows for the 14 selected basins.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
For the snowmelt-dominated basins (Feather, Flathead, Sprague, Naches, Yampa, East, and Sagehen), a distinct triangular pattern emerges in Figure 11. The increasing temperatures increase the flow in the colder months because of a change in precipitation form and/or an increase in snowmelt, transitioning into large decreases in flow volumes in the spring (less snowpack to melt) and transitioning into small decreases in flow volumes from summer into early winter. Precipitation projections are highly variable, but it is the projected increasing temperatures in these snowmelt-dominated basins that drive the changes in the streamflow. In basins that do not have complex terrain (Flint, Clear, Pomperaug, Trout Lake, Black Earth, and Cathance), another pattern emerges from Figure 11. The high variability in the precipitation projections and the effect these projections have on the variability of the projected streamflow is evident.
Changes in the timing of peak flow are evident in Figure 11 but are more easily visualized in Figure 12, which shows the projected mean monthly hydrographs for baseline conditions (red dashed line; 1989–99) and projected future conditions using the five GCMs and three emission scenarios for 2030 (green line; 2025–35), 2060 (tan line; 2055–65), and 2090 (2085–95) for the 14 selected basins. The large variability between emission scenarios or GCMs is not shown in Figure 12. Depending on the temperature, peak flow timing can remain unchanged in the coldest basins (April in the Starkweather basin) to dramatic shifts in peak timing by the end of the twenty-first century (from May to March in the Sagehen basin). In many of the basins, the broader-scale influences of climate change on the streamflow indicates an overall decrease in the importance of snowmelt and increase in the evapotranspiration out of the basin, though the uncertainty captured by the envelope of climate-change scenarios is large. If these scenarios of climate change are representative of future conditions, the projected decreases in the snow-covered area and snow-water equivalent (see Tables 4f,g) with an attendant reduction in the importance of the spring snowmelt to the stream and a smaller increase of snowmelt during the winter months may alter the characterization of some of these basins as being spring snowmelt dominated.
Mean monthly hydrographs for baseline conditions (red dashed line; 1989–99) and future conditions using five GCMs and three emission scenarios for WYs 2030 (green line; 2025–35), 2060 (tan line; 2055–65), and 2090 (2085–95) for the 14 selected basins.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
4.5. Soil moisture
Figure 13 compares projected mean monthly temperature versus soil moisture over time. In Figure 13, the mean monthly temperatures and soil moistures are calculated using the five GCMs and three emission scenarios for each of the 88 windows for the 14 selected basins. The steady increase in projected temperature can then be compared with the corresponding changes in soil moisture on a mean monthly basis. On an annual basis, all basins project decreases in soil moisture with significant trends occurring for the following: all of the A1B; most of the A2; and half of the B1 emission scenario projections (Table 4k). This holds true on a monthly basis for all basins that are not snowmelt dominated. The snowmelt-dominated basins tend to show slight increases in soil moisture during the coldest months because of increasing snowmelt and change in precipitation form. When the average temperature reaches approximately 5°C in the early spring, those basins begin to show large decreases in projected soil moisture that continues until the following year’s snowmelt begins.
Mean monthly temperature vs streamflow using the five GCMs and three emission scenarios calculated for each of the 88 windows for the 14 selected basins.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
5. Discussion
There are numerous sources of uncertainty at each step of simulation associated with this study: uncertainty in the GCMs, in the downscaling technique, and in the hydrologic model. Starting with the GCMs, large uncertainties are associated with the representation of the physical processes, model structure, and feedbacks within the climate system (Alley et al. 2007). The GCMs were run with a range of future emission scenarios based on trajectories on how economic, social, political, and technological development in the world might change in the future. The emission scenarios chosen for this study (Table 3) represent different paths for the future, none of which may be close to the true path. In this study, the quantification of GCM uncertainties was handled by using output from multiple GCMs (Table 2). Whereas some researchers promote weighting or not using specific GCMs based on their ability to reproduce current climate (Murphy et al. 2004; Tebaldi et al. 2005), Stainforth et al. (Stainforth et al. 2007) argue that all GCM ensembles can be used to produce a lower bound on the maximum range of uncertainty.
Future work to better quantify sources of GCM uncertainty may be achieved by combining Bayesian approaches (cf. Draper 1995; Tebaldi et al. 2005; Raftery et al. 2005) with Earth models of intermediate complexity or energy balance models (e.g., Urban and Keller 2009) to sample deeply into the tails of the distribution of climate sensitivity. The combined approach can be used to derive probabilistic projections or posterior model weights. This method has the potential to reduce the risk of overconfident projections that can result from truncated estimates of GCM model uncertainty within a finite ensemble of GCMs.
The spatial scale at which the GCMs are run is an additional source of uncertainty. The change factor method was used to downscale the coarse-resolution GCM output to the finer-resolution input required by the hydrologic model. In this method, the changes in atmospheric processes at the GCM subgrid scale are not resolved, such as changing interactions between air masses and topography, and these subgrid interactions may be substantial in some cases (e.g., Ikeda et al. 2010). The uncertainty from the downscaling procedure may, however, be overwhelmed by the choice of driving GCM (Fowler et al. 2007; Khan et al. 2006), which has been shown to be consistently greater than uncertainty from the hydrologic model or natural variability (Prudhomme and Davies 2009). Hydrologic model output was analyzed on a mean monthly and annual basis. The uncertainty associated with this simple approach to downscaling is not directly addressed, but simple statistical downscaling methods seem to perform as well as more sophisticated methods in reproducing mean characteristics (Fowler et al. 2007).
An important source of uncertainty is the dependence of the hydrologic model on the method used to calculate PET (Kingston et al. 2009). PET is the maximum possible evaporation rate: when available water is nonlimiting, AET equals PET. In PRMS, AET is central to the hydrologic model calculations and is controlled by the available soil moisture. Ideally the PET algorithm incorporates four key variables: air temperature, net radiation, vapor pressure, and wind speed; however, the reliability of the input climate data needs to play a large role in the choice of PET algorithm. Will more reliable estimates of PET be produced from physically based methods with large input data uncertainties or from empirical methods that use more reliable input data such as air temperature (Donohue et al. 2010)?
Calculation of PET as an empirical function of air temperature, as is done in PRMS, may artificially enhance the sensitivity of evaporation and soil moisture to temperature. Indeed, Milly and Dunne (Milly and Dunne 2011) speculate that the overestimation of PET by PRMS leads to larger decreases in streamflow than what would be expected with correctly projected PET estimates. The assumption is that, under warming conditions, the JH algorithm will overestimate the actual amount of energy available based solely on the projected increases in temperatures from the GCMs. The JH algorithm is assuming much more energy will be available just because temperature is increasing. Nevertheless, changes in the simulated AET rates, from the PRMS simulations of the twentieth century, are comparable to measured (or estimates of measured) changes presented in the literature (Walter et al. 2004, p. 204; Szilagyi 2001). This may give some degree of confidence for using the PRMS simulations in this study for hydrologic analogs of the climate- change emission scenarios.
The method of PET characterization is particularly important in regions where the precipitation is closely in balance with PET (Kingston et al. 2009). PET formulation is most crucial in areas in which AET is controlled primarily by PET (energy-limited basins) and basins that switch from energy- to water-limited states (the seasonal cycle of AET follows PET for parts of the year) (Donohue et al. 2010). In an attempt to determine the influence of PET on the PRMS simulation results in this study, the mean annual aridity index (precipitation divided by PET; United Nations Environment Program 1992) was calculated using the five GCMs and three emission scenarios for each of the 88 windows for the 14 selected basins. The aridity index was plotted against the correlation of monthly precipitation and PET in Figure 14. The gray rectangles indicate the critical zone in Figure 14. This critical zone identifies the energy-limited areas in which the seasonal cycle of precipitation and PET are similar: precipitation and PET have similar magnitudes, and high periods of PET correspond to high periods of precipitation.
Aridity index (precipitation divided by PET) vs correlation of monthly precipitation and PET for the five GCMs by emission scenario for the 14 basins. The gray rectangles indicate the energy-limited areas in which the seasonal cycle of precipitation and PET are similar.
Citation: Earth Interactions 15, 17; 10.1175/2010EI370.1
It becomes apparent that the warmer Midwestern basins (Trout Lake, Clear Creek, and Black Earth) fall into this critical zone where the method of PET calculation will produce the largest uncertainties when estimating streamflow. The broader-scale influences of climate change on the flow regimes in these Midwestern basins project an overall slight drying of the basin as a consequence of increased evapotranspiration with large uncertainty captured by the envelope of climate-change emission scenarios, a result that needs to be examined further because of the extra uncertainty concerning the PET calculations. Further work is needed to improve our understanding of the uncertainty associated with PET projections, especially in areas that fall into the critical zone defined in Figure 14.
The uncertainty going from the GCM to the hydrologic model can propagate in ways that may either be enhanced or compensated for, depending on the structure and parameterization of the hydrologic model (Buytaert et al. 2009). There are many areas for hydrologic model improvement. Research toward improved process representation and methods to represent uncertainty in the choice of model structure and model parameters, such that you can evaluate the climate-change “signals” in light of the uncertainty in hydrologic model formulations, are needed. The loosely coupled model approach (as used in this study) ignores the important feedback mechanisms between the hydrology and the climate. The calibration of the hydrologic model assumes that the parameters for baseline conditions will hold under future conditions, but calibration to past data may not be relevant to the future (Bloschl and Montanari 2010). The use of PRMS models formulated for other applications and brought together for an integrated assessment leads to a bigger concern of model consistency in terms of formulation, parameterization, and calibration. With continued concerns over climate-change, large-scale applications of this type can be expected to increase in the future. As noted by Bloschl and Montanari (Bloschl and Montanari 2010), when two groups examine the same watershed for impacts of climate change, the projections can be quite different. The same can be said for comparisons using the same model (PRMS in this case) formulated for different reasons by different sets of researchers. A national hydrologic model with consistent formulation (or at least formulations that are “correct” for given area), parameterization, and calibration would greatly enhance this type of climate-change impact work.
6. Summary and conclusions
The hydrologic response of 14 basins across the United States to three climate- change emission scenarios were evaluated by modifying a baseline period of daily PRMS precipitation and temperature with the mean monthly climate-change factors derived from five GCMs. All climate-change emission scenarios evaluated with PRMS show a steady increase in maximum and minimum temperatures progressing through the twenty-first century. This generally results in PRMS simulating an earlier spring snowmelt and increases in evapotranspiration if soil moisture is available. These increases in temperatures are generally considered to be a somewhat robust projection, especially when compared to the highly variable changes projected for precipitation, both between GCMs and between emission scenarios. Therefore, higher confidence may be placed in hydrologic changes mainly controlled by increases in temperature (e.g., snow accumulation and melt) rather than changes in precipitation.
If the projected increases in temperature hold true, then large changes in the timing of streamflow will occur in the snowmelt-dominated basins with winter temperatures closest to freezing. This has long been recognized with an increase in winter temperature initially resulting in less snow (change of form from snow to rain) and eventually earlier snowmelt (Dettinger and Cayan 1995). For this study, results from PRMS simulations using three emission scenarios of global climate change indicate that streamflows could increase in the winter and early spring and decrease in late spring and through the summer for many of the basins (Figures 11, 12). Uncertainty associated with the magnitude of these changes in streamflow is large. In the snowmelt-dominated basins, the large uncertainty in projections of future precipitation and its effect on the streamflow are small compared with the dramatic changes associated with increasing temperatures. Basins in which precipitation form (rain or snow) is sensitive to the air temperature are the most sensitive to changes in temperature and precipitation. This is consistent with the findings of Stewart (Stewart 2009) that warmer temperatures at mid-elevation mountainous regions show decreasing snowpack and earlier melt despite the increase in precipitation, but high-elevation regions that remain well below freezing during winter show little effect.
A change in timing and volume of snowmelt has large implications for water managers who rely on the melt from the spring snowpack to replenish reservoirs and provide early season irrigation; earlier snowmelt may lead to increased water scarcity. Even when the annual streamflow shows a positive trend into the twenty-first century, spring streamflow from snowmelt is projected to decrease (Figure 11); therefore, water managers will need to store more winter streamflow to meet the irrigation and instream flow demands. Increasing temperatures and population will invariably lead to increasing water demands, making water-management decisions increasingly difficult because of the uncertainty in the quantity and timing of the spring snowmelt.
The ecological implications of changes in the timing of and volume of snowmelt are not well understood. PRMS results project changes in the timing of peak flows because of snowmelt in all but the coldest basins (Figure 12), which may affect freshwater mixing in estuarine ecosystems and increase the length of the summer drought that characterizes much of western North America (Stewart et al. 2004). A shorter snow-covered season, with smaller snowpack volumes and a tendency for midwinter melts, has obvious implications for winter recreation but could potentially affect phenological responses, with corresponding changes in the ecosystem. Earlier snowmelt and increases in evapotranspiration rates may result in drier forests and more wildfires and threaten fish and amphibian habitat. The soil moisture results indicate a reduction of soil moisture, which could potentially change the overall vegetation in the system. This has obvious ecosystem implications and could potentially result in a changed and less diverse plant assemblage. Further, the system would likely be more prone to fires, which could dramatically alter the hydrologic response after a fire event. Changes in some of these hydrologic characteristics can have substantial effects on the ecology of the basin while having no effect on production of agricultural or municipal water supply.
These results did not consider potential future population growth and land-use changes. It is possible that changes in land cover can substantially affect climate at the regional and local scales. Within this context, methodologies for adequate simulation of changes in urban systems, agricultural systems, ecosystem disturbance regimes, and soil impacts are not yet well represented. The combined effects of climate change and urbanization may alter both the quantity and timing of streamflow and has the potential to change the conditions of water quality that support biological diversity in aquatic communities.
These results did not consider potential future land-cover dynamics such as forest fire or insect damage. They also do not answer the question of whether the potentially adverse effects due to climate change can be mitigated with careful land-use planning. Thus, the interrelated effects of land use and climate change may exacerbate future adverse changes in the basin beyond what is shown in this work. However, potentially adverse effects due to climate change and related changes to water use in the basin might be offset by potential land-use practices that change the flow of water in a beneficial way.
In many of the basins, projected decreases in soil moisture will increase agricultural water consumption, resulting in an increase in water demand. A reduction in soil moisture will cause drier conditions in the root zone. A drier root zone is likely to cause other human responses such as increased irrigation groundwater withdrawals to maintain current agricultural production levels. Such simulations are beyond the scope of the current work as simulating effects of increased irrigation requires a coupled groundwater–surface-water model rather than the surface-water model used here.
In many of the basins, population growth will result in increasing water demand. Streamflow in many of the basins is under increasing demand from water users and recreationalists within the basins. Potential changes in streamflow resulting from future changes in climate may add to the stress that these basins will experience as a result of projected increases in domestic and industrial water use due to population growth.
At this stage of analysis, the uncertainty associated with these simulations of future climatic conditions is overwhelming. The large variability in the GCMs, when evaluated with a highly nonlinear hydrologic model, makes a very complex story. The hydrologic impact is further complicated by the critical difference when choosing a higher or lower emission scenario.
In the past, modeling studies based on watershed simulation have been developed independently and on a project-by-project basis. These modeling studies proceeded without considerations for collaboration or reuse and generally have not been relevant outside of the context in which they were developed. Specifically, consistent application of a hydrologic modeling structure would help address the issues of 1) differing temporal and spatial scales and resolutions, 2) evolving data availability and needs, 3) differing calibration methods and purpose, and 4) internal and external institutional constraints. Future research, including the development of a national hydrologic modeling structure with consistent formulation, would greatly enhance this type of climate-change impact work.
References
Alley, R. B., and Coauthors, 2007: Summary for policymakers. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 1–18.
Antolik, M. S., 2000: An overview of the National Weather Service’s centralized statistical quantitative precipitation forecasts. J. Hydrol., 239, 306–337.
Arnell, N. W., 2003a: Effects of IPCC SRES emissions scenarios on river runoff: A global perspective. Hydrol. Earth Syst. Sci., 7, 619–641.
Arnell, N. W., 2003b: Relative effects of multi-decadal climatic variability and changes in the mean and variability of climate due to global warming: Future streamflows in Britain. J. Hydrol., 270, 195–213.
Arnell, N. W., and N. S. Reynard, 1996: The effects of climate change due to global warming on river flows in Great Britain. J. Hydrol., 183, 397–424.
Bjerklie, D. M., J. J. Starn, and C. Tamayo, 2010: Estimation of the effects of land use and groundwater withdrawals on streamflow for the Pomperaug River, Connecticut. U.S. Geological Survey Scientific Investigations Rep. 2010-5114, 93 pp.
Bloschl, G., and A. Montanari, 2010: Climate change impacts—Throwing the dice? Hydrol. Processes, 24, 374–381.
Boorman, D. B., and C. E. M. Sefton, 1997: Recognizing the uncertainty in quantification of the effects of climate change on hydrological response. Climatic Change, 35, 415–434.
Buytaert, W., R. Célleri, and L. Timbe, 2009: Predicting climate change impacts on water resources in the tropical Andes: The effects of GCM uncertainty. Geophys. Res. Lett., 36, L07406, doi:10.1029/2008GL037048.
Carter, T. R., M. L. Parry, H. Harasawa, and S. Nishioka, 1994: IPCC technical guidelines for assessing climate change impacts and adaptations. University College London Special Rep., 59 pp.
Dettinger, M. D., and D. R. Cayan, 1995: Large-scale atmospheric forcing of recent trends toward early snowmelt runoff in California. J. Climate, 8, 606–623.
Diaz-Nieto, J., and R. L. Wilby, 2005: A comparison of statistical downscaling and climate change factor methods: Impacts on low flows in the River Thames, United Kingdom. Climatic Change, 69, 245–268.
Donohue, R. J., T. R. McVicar, and M. L. Roderick, 2010: Assessing the ability of potential evaporation formulations to capture the dynamics in evaporative demand within a changing climate. J. Hydrol., 386, 186–197.
Draper, D., 1995: Assessment and propagation of model uncertainty. J. Roy. Stat. Soc., 57, 45–97.
Dudley, R. W., 2008: Simulation of the quantity, variability, and timing of streamflow in the Dennys River basin, Maine, by use of a precipitation-runoff watershed model. U.S. Geological Survey Scientific Investigations Rep. 2008-5100, 44 pp.
Eckhardt, K., and U. Ulbrich, 2003: Potential impacts of climate change on groundwater recharge and streamflow in a central European low mountain range. J. Hydrol., 284, 244–252.
Fealy, R., and J. Sweeney, 2008: Statistical downscaling of temperature, radiation and potential evapotranspiration to produce a multiple GCM ensemble mean for a selection of sites in Ireland. Ir. Geogr., 41, 1–27.
Fowler, H. J., S. Blenkinsopa, and C. Tebaldi, 2007: Review. Linking climate change modelling to impacts studies: Recent advances in downscaling techniques for hydrological modeling. Int. J. Climatol., 27, 1547–1578.
Giorgi, F., and Coauthors, 2001: Regional climate information—Evaluation and projections. Climate Change 2001: The Scientific Basis, J. T. Houghton et al., Eds., Cambridge University Press, 739–768.
Hay, L. E., and M. P. Clark, 2003: Use of statistically and dynamically downscaled atmospheric model output for hydrologic simulations in three mountainous basins in the western United States. J. Hydrol., 282, 56–75.
Hay, L. E., and M. Umemoto, 2006: Multiple-objective stepwise calibration using Luca. U.S. Geological Survey Open File Rep. 2006-1323, 25 pp.
Hay, L. E., and G. J. McCabe, 2010: Hydrologic effects of climate change in the Yukon River basin. Climatic Change, 100, 509–523.
Hay, L. E., R. L. Wilby, and G. H. Leavesley, 2000: A comparison of delta change and downscaled SRES emission scenarios for three mountainous basins in the United States. J. Amer. Water Resour., 36, 387–397.
Hay, L. E., M. P. Clark, M. Pagowski, G. H. Leavesley, and W. J. Gutowski Jr., 2006a: One-way coupling of an atmospheric and a hydrologic model in Colorado. J. Hydrometeor., 7, 569–589.
Hay, L. E., G. H. Leavesley, and M. P. Clark, 2006b: Use of remotely-sensed snow covered area in watershed model calibration for the Sprague River, Oregon. Proc. Joint Eighth Federal Interagency Sedimentation Conf. and Third Federal Interagency Hydrologic Modeling Conf., Reno, NV, Subcommittee on Hydrology, 8 pp.
Hay, L. E., G. H. Leavesley, M. P. Clark, S. L. Markstrom, R. J. Viger, and M. Umemoto, 2006c: Step-wise, multiple-objective calibration of a hydrologic model for a snowmelt-dominated basin. J. Amer. Water Resour., 42, 877–890.
Hay, L. E., G. J. McCabe, M. P. Clark, and J. C. Risley, 2009: Reducing streamflow forecast uncertainty: Application and qualitative assessment of the upper Klamath River basin, Oregon. J. Amer. Water Resour., 45, 580–596.
Ikeda, K., and Coauthors, 2010: Simulation of seasonal snowfall over Colorado. Atmos. Res., 97, 462–477.
Jensen, M. E., and H. R. Haise, 1963: Estimating evapotranspiration from solar radiation. J. Hydraul. Div. Amer. Soc. Civ. Eng., 89, 15–41.
Khan, M. S., P. Coulibaly, and Y. Dibike, 2006: Uncertainty analysis of statistical downscaling methods. J. Hydrol., 319, 357–382.
Kingston, D. G., M. C. Todd, R. G. Taylor, J. R. Thompson, and N. W. Arnell, 2009: Uncertainty in the estimation of potential evapotranspiration under climate change. Geophys. Res. Lett., 36, L20403, doi:10.1029/2009GL040267.
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 predictability, water years 1971-97. U.S. Geological Survey Scientific Investigations Rep. 2004-5202, 82 pp.
Leavesley, G. H., R. W. Lichty, B. M. Troutman, and L. G. Saindon, 1983: Precipitation-Runoff Modeling System: User’s manual. U.S. Geological Survey Water-Resources Investigations Rep. 83-4238, 207 pp.
Le Treut, H., R. Somerville, U. Cubasch, Y. Ding, C. Mauritzen, A. Mokssit, T. Peterson, and M. Prather, 2007: Historical overview of climate change science. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 93–127.
Lettenmaier, D. P., 1976: Detection of trends in water quality data from records with dependent observations. Water Resour. Res., 12, 1037–1046.
Leung, L. R., L. O. Mearns, F. Giorgi, and R. L. Wilby, 2003: Workshop on regional climate research: Needs and opportunities. Bull. Amer. Meteor. Soc., 84, 89–95.
Markstrom, S. L., R. G. Niswonger, R. S. Regan, D. E. Prudic, and P. M. Barlow, 2008: GSFLOW—Coupled Ground-Water and Surface-Water Flow model based on the integration of the Precipitation-Runoff Modeling System (PRMS) and the Modular Ground-Water Flow Model (MODFLOW-2005). U.S. Geological Survey Techniques and Methods 6-D1, 240 pp.
Mastin, M. C., and J. J. Vaccaro, 2002. Watershed models for decision support in the Yakima River basin, Washington. U.S. Geological Survey Open-File Rep. 2002-404, 46 pp.
McCabe, G. J., and L. E. Hay, 1995: Hydrological effects of hypothetical climate change in the East River basin, Colorado. Hydrol. Sci. J., 40, 1–16.
McCabe, G. J., and D. M. Wolock, 1997: Climate change and the detection of trends in annual runoff. Climate Res., 8, 129–134.
McCarthy, J. J., O. F. Canziani, N. A. Leary, D. J. Dokken, and K. S. White, Eds., 2001: Climate Change 2001: Impacts, Adaptation, and Vulnerability. Cambridge University Press, 1032 pp.
Milly, P. C. D., and K. A. Dunne, 2011: On the hydrologic adjustments of climate-model projections: The potential pitfall of potential evapotranspiration. Earth Interactions, 15. [Available online at http://EarthInteractions.org.]
Murphy, J. M., D. M. H. Sexton, D. N. Barnett, G. S. Jones, M. J. Webb, M. Collins, and D. A. Stainforth, 2004: Quantification of modelling uncertainties in a large ensemble of climate change simulations. Nature, 430, 768–772.
Pilling, C., and J. A. A. Jones, 1999: High resolution equilibrium and transient climate change scenario implications for British runoff. Hydrol. Processes, 13, 2877–2895.
Prudhomme, C., and H. Davies, 2009: Assessing uncertainties in climate change impact analyses on the river flow regimes in the UK. Part 1: Baseline climate. Climatic Change, 93, 177–195.
Prudhomme, C., N. Reynard, and S. Crooks, 2002: Downscaling of global climate models for flood frequency analysis: Where are we now? Hydrol. Processes, 16, 1137–1150.
Raftery, A. E., T. Gneiting, F. Balabdaoui, and M. Polakowski, 2005: Using Bayesian model averaging to calibrate forecast ensembles. Mon. Wea. Rev., 133, 1155–1174.
Stainforth, D. A., T. E. Downing, R. Washington, A. Lopez, and M. New, 2007: Issues in the interpretation of climate model ensembles to inform decisions. Philos. Trans. Roy. Soc., 365A, 2163–2177.
Stewart, I. T., 2009: Changes in snowpack and snowmelt runoff for key mountain regions. Hydrol. Processes, 23, 78–94.
Stewart, I. T., D. R. Cayan, and M. D. Dettinger, 2004: Changes in snowmelt runoff timing in western North America under a “business as usual” climate change scenario. Climatic Change, 62, 217–232.
Szilagyi, J., 2001: Modeled areal evapotranspiration trends over the conterminous United States. J. Irrig. Drain. Eng., 127, 196–200.
Tebaldi, C., R. L. Smith, D. Nychka, and L. O. Mearns, 2005: Quantifying uncertainty in projections of regional climate change: A Bayesian approach to the analysis of multimodel ensembles. J. Climate, 18, 1524–1540.
United Nations Environment Program, 1992: World Atlas of Desertification. Edward Arnold, 69 pp.
Urban, N. M., and K. Keller, 2009: Complementary observational constraints on climate sensitivity. Geophys. Res. Lett., 36, L04708, doi:10.1029/2008GL036457.
Viger, R. J., L. E. Hay, J. W. Jones, and G. R. Buell, 2010: Accounting for large numbers of small water bodies in the application of the Precipitation-Runoff Modeling System in the Flint River basin, Georgia. U.S. Geological Survey Open File Rep. 2010-5062, 36 pp.
Vining, K. C., 2002: Simulation of streamflow and wetland storage, Starkweather Coulee Subbasin, North Dakota, water years 1981-98. U.S. Geological Survey Water-Resources Investigations Rep. 02-4113, 28 pp.
Walter, M. T., D. S. Wilks, J. Parlange, and R. L. Schneider, 2004: Increasing evapotranspiration from the conterminous United States. J. Hydrometeor., 5, 405–408.
Wigley, T. W. L., P. D. Jones, K. R. Briffa, and G. Smith, 1990: Obtaining sub-grid scale information from coarse resolution general circulation model output. J. Geophys. Res., 95, 1943–1953.
Wilby, R. L., L. E. Hay, and G. H. Leavesley, 1999: A comparison of downscaled and raw GCM output: Implications for climate change scenarios in the San Juan River basin, Colorado. J. Hydrol., 225, 67–91.
Wilby, R. L., S. P. Charles, E. Zorita, B. Timbal, P. Whetton, and L. O. Mearns, 2004: Guidelines for use of climate scenarios developed from statistical downscaling methods. Supporting Material of the Intergovernmental Panel on Climate Change, 27 pp.
Wilks, D. S., 1995: Statistical Methods in the Atmospheric Sciences: An Introduction. Academic Press, 467 pp.
Xu, C., 1999: Climate change and hydrologic models: A review of existing gaps and recent research developments. Water Resour. Manage., 13, 369–382.