Abstract

Water managers across the United States face the need to make informed policy decisions regarding long-term impacts of climate change on water resources. To provide a scientifically informed basis for this, the evolution of important components of the basin-scale water balance through the end of the twenty-first century is estimated. Bias-corrected and spatially downscaled climate projections, from phase 3 of the Coupled Model Intercomparison Project (CMIP3) of the World Climate Research Programme, were used to drive a spatially distributed Variable Infiltration Capacity (VIC) model of hydrologic processes in the Salt–Verde basin in the southwestern United States. From the suite of CMIP3 models, the authors select a five-model subset, including three that best reproduce the historical climatology for the study region, plus two others to represent wetter and drier than model average conditions, so as to represent the range of GCM prediction uncertainty. For each GCM, data for three emission scenarios (A1B, A2, and B1) were used to drive the hydrologic model into the future. The projections of this model ensemble indicate a statistically significant 25% decrease in streamflow by the end of the twenty-first century. The primary cause for this change is due to projected decreases in winter precipitation accompanied by significant (temperature driven) reductions in storage of snow and increased winter evaporation. The results show that water management in central Arizona is highly likely to be impacted by changes in regional climate.

1. Introduction

There is ample evidence that the climate is changing in the southwestern United States, but the physical mechanisms of how this will affect streamflow, especially in the lower Colorado basin, are poorly understood. Numerous studies have assessed the potential impacts of climate change on the upper Colorado River basin (e.g., Nash and Gleick 1991; Christensen et al. 2004; Christensen and Lettenmaier 2007; McCabe and Wolock 2008; Rajagopalan et al. 2009; Hurkmans et al. 2009). There is overwhelming consensus that streamflow in the Colorado River basin, which comes largely from the upper basin, will likely decrease under future climate conditions with increased greenhouse gas forcing, although there is significant spread among the projections (Vano et al. 2014). The lower basin is relatively less well studied, due mainly to the fact that most rivers in the lower basin were dammed during the twentieth century and therefore do not contribute much surface water flow to the Colorado River basin (except in very wet years). Nonetheless, Colorado River tributaries in the lower basin (including two watersheds studied in this paper) are very important from a water supply perspective, providing a total of 54% of the water used in the state of Arizona. A significant proportion (~70% based on analysis of observed flows) of this water comes from the springtime melting of accumulated snowpack.

In central Arizona, the dynamics of snow accumulation and melt are controlled by ambient temperatures from November to March. Harpold et al. (2012) analyzed Snowpack Telemetry (SNOTEL; www.wcc.nrcs.usda.gov/snow/) data for the Intermountain West and reported a significant decreasing trend in snowpack accumulation, especially in the lower Colorado River basin. Proxy temperature reconstructions provide evidence for rising temperatures in the twentieth century, in comparison to the Holocene (Marcott et al. 2013), and general circulation model (GCM) [based on phase 3 of the Coupled Model Intercomparison Project (CMIP3)] outputs support this trend (e.g., Seager et al. 2007), the latter being frequently used in climate change attribution studies (e.g., Maurer et al. 2010; Christensen and Lettenmaier 2007). For example, GCM outputs from 12 models that participated in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) have been used to assess drought (Cayan et al. 2010) and the ability to sustain water supply to meet demands (MacDonald 2010) in the southwestern United States during the twenty-first century.

The U.S. Bureau of Reclamation conducted the West-Wide Climate Risk Assessment that use the Variable Infiltration Capacity (VIC) hydrologic model with projections from the CMIP3 dataset (Water and Environmental Resources Division 2011), but this study leaves out two watersheds in central Arizona (Water and Environmental Resources Division 2011, p. 58) that are critical to surface water supply to this region. A subsequent study using the climate data generated by Maurer et al. (2007) was used to run the VIC hydrology model for the entire Colorado basin (the results from this study are archived at http://gdo-dcp.ucllnl.org/downscaled_cmip_projections/dcpInterface.html#Projections:%20Complete%20Archives). We extracted the monthly total runoff from this archive and compared it to the observed streamflow in central Arizona; a positive bias in simulated streamflow was found, and hence, we did not use their simulations in this study. Gober et al. (2010) used a planning model for the city of Phoenix to simulate “what if” scenarios under varying policy decisions and future climates based on the IPCC AR4. An important result from this study is that the current levels of per capita water consumption cannot be supported in the future without unsustainable groundwater withdrawals. While the studies mentioned above indicate statistically significant declines in the water resources of the southwestern United States as a whole, a comprehensive study of climate change impacts on watersheds in the lower Colorado River basin will help water managers in the Salt River Project (SRP) and the City of Phoenix to better understand the impacts on water availability, reservoir operations, and hydropower generation. Here, we study two watersheds in central Arizona, namely, the Salt and Verde River basins (SVRB), which are part of the larger Colorado River basin (Fig. 1). Through the Colorado River Compact, signed in 1922, 9.247 billion m3 [7.5 million acre feet (maf)] was apportioned to the lower Colorado River basin, and of this, 3.452 billion m3 (2.8 maf) is allocated to Arizona. This water comes from upper-basin watersheds and is stored in Lake Powell and Lake Mead. Water from these reservoirs is then delivered to the cities of Phoenix and Tucson via the Central Arizona Project’s (CAP) 336-mi-long system of canals, pumping stations, and pipelines that divert water into central and southern Arizona. In addition to the supply from CAP, water from the SVRB is a major component of water supply for the region. The SRP is the largest provider of water and power to the greater Phoenix metropolitan area, and it manages six reservoirs with a storage capacity of about 2.467 billion m3 (2 maf) on SVRB. Unfortunately, if there is an official declaration of shortage in the Colorado River water supply, CAP will be subject to substantial reductions in its allocations because it holds junior priority water entitlement to the Colorado River water. This makes it critical for water managers to have information about water availability in the lower-basin watersheds, primarily the SVRB, to compensate for potential losses from the upper basin.

Fig. 1.

Map of the Colorado River basin, showing the Salt and Verde Rivers watershed study area in central Arizona.

Fig. 1.

Map of the Colorado River basin, showing the Salt and Verde Rivers watershed study area in central Arizona.

The goal of this study is to examine the potential impacts of climate change on the water balance of the SVRB. We do this by setting up a physically based hydrological model of the area and driving it with bias-corrected and spatially downscaled (BCSD) climate data for the historical period (1960–2000) and a future period (2010–98). For both periods, the climate data are taken from five carefully selected GCMs that participated in CMIP3, and for the future climate projections, we examine three greenhouse gas emission scenarios (A1B, A2, and B1). While one could debate the relative merits and demerits of statistical downscaling versus dynamical downscaling (e.g., Fowler et al. 2007), the major focus of this paper is to understand the physical mechanisms driving hydrologic change in the SVRB.

2. The study area

The Salt and Verde Rivers together drain an area of approximately 35 100 km2 (see Fig. 1). Because of the large variation in elevation (ranging from 280 to 3350 m MSL), vegetation varies from desert scrub in the valley bottoms to piñon–juniper at midelevations and ponderosa pine and fir–spruce in higher elevations. Streamflow in the Salt River is regulated downstream of Roosevelt Lake, which has a storage capacity of 2.04 billion m3 (1.6 maf). Four smaller downstream reservoirs are used to satisfy multiple demands for power generation, municipal and agricultural water supply, recharge to groundwater storage, and recreation. Streamflow in the Verde is regulated downstream of the Horseshoe Dam, which has a storage capacity of 112 million m3 (110 000 acre feet).

Based on observational records (Maurer et al. 2002), the watersheds together receive ~160 mm of precipitation during the winter (December–March) and ~172 mm during summer (July–September). Although the amount of precipitation has a bimodal distribution annually, its character varies. Winter precipitation is typically from synoptic-scale extratropical cyclones that produce precipitation over large areas for long durations (from hours to days), and some of this is stored in the form of a snowpack at the upper elevations. Summer precipitation is mainly a result of the monsoon, which results in highly localized thunderstorms (producing intense precipitation) that typically last on the order of only 20–30 min. Hydrologically, winter precipitation is the more important contributor to water storage and runoff, whereas monsoonal precipitation occurs so quickly and is so localized that it sometimes has very little impact on streamflow. Corresponding seasonal streamflow averages are 820 million m3 (December–May) and 148 million m3 (July–September). Clearly, the major contributor to total annual streamflow in the Salt–Verde system is precipitation that falls during the winter months, in fact; 39% of annual precipitation at SNOTEL stations within the basin is recorded to be snowfall (Serreze et al. 1999). Based on temperature and elevation zones that stay below 0°C in the winter, approximately 26% of the watershed is snow covered during the winter.

3. Methods

To understand how climate change could impact physical hydrological processes in the two basins, the VIC soil moisture accounting model (Liang et al. 1994; Wood et al. 1992) was used. Briefly, the model partitions the precipitation input into various components, such as soil moisture, evapotranspiration (ET), snow storage, quick runoff, and base flow. VIC also requires wind speed and climatological averages were used for this. A routing algorithm developed by Lohmann et al. (1998) combines the quick runoff and base flow to produce streamflow in the basin. Two sets of atmospheric forcing data are required to drive the hydrologic model—precipitation and temperature. We used observed data for historical simulations and the GCM-generated simulated values for evaluating future impacts.

  1. Historical simulations: The Maurer et al. (2002) dataset, available at ⅛° spatial resolution for the continental United States, was used to force the land surface hydrology model to generate simulations of historical streamflow. The atmospheric forcing included precipitation and temperature at a daily time step. This dataset spans the period 1949–2000.

  2. Future impact studies: To evaluate the impacts of potential future climate change on local hydrologic variables, a second set of atmospheric forcing data was generated based on simulations provided by five of the GCMs that participated in the IPCC AR4 (available at www.ipcc.ch/publications_and_data/publications_and_data_reports.shtml); we used three future (2010–98) greenhouse gas emissions scenarios (A2, A1B, and B1). These models also provide data for the historical period (1960–2000). The five GCMs are MPI, HadCM3, CCSM3, PCM, and MIROC. Table 1 provides a summary of the data used in this study and their primary references.

Table 1.

Datasets used in this study.

Datasets used in this study.
Datasets used in this study.

There are multiple approaches to selecting GCM model output (e.g., Brekke et al. 2008; Pierce et al. 2009; Dominguez et al. 2010) for conducting climate change studies; our choice of the three primary GCMs (HadCM3, MPI, and CCSM3) is based on the approach suggested by Dominguez et al. (2010). The authors select the three models, out of the 23 models that participated in the IPCC AR4, that best capture the historical precipitation and temperature climatology and that best simulate El Niño–Southern Oscillation (ENSO) for the southwestern U.S. region. Two additional GCM simulations, that is, MIROC and PCM for the A2 emission scenario, were added to represent the simulated driest and wettest extremes, respectively, based on their simulations of precipitation means for the twenty-first century.

The Program for Climate Model Diagnosis and Intercomparison (PCMDI) collects climate model output contributed by leading modeling centers around the world in response to proposed activities of the World Climate Research Programme’s (WCRP) Working Group on Coupled Modelling (WGCM). These climate simulations, which included past, present, and future climate, were archived in 2006 and are the primary data for CMIP3 (Covey et al. 2003). Bias correction and spatial downscaling of the WCRP CMIP3 dataset was performed at Santa Clara University and was archived on the Lawrence Livermore National Laboratory website (http://gdo-dcp.ucllnl.org/downscaled_cmip_projections/dcpInterface.html). This was the source for the downscaled GCM data used in this study. The methodology for bias correction and spatial downscaling follows Wood et al. (2002, 2004) and Maurer et al. (2007) . The approach uses a quantile mapping technique to remove distributional biases in the GCM simulations when their simulations of historical climate conditions tend to be too wet/dry/warm/cold relative to the observations.

Prior to evaluating the future impacts of climate change, we first tested the ability of the hydrologic model to simulate historical observed streamflow in the basin. Whereas several of the parameters of the VIC model can be derived from satellite data or geological surveys, some of the parameters are conceptual and do not correspond to physically measurable quantities. For the latter, it is common to use a calibration strategy (Sorooshian et al. 1993) that constrains the model in such a way that the difference between the model simulations and observations is minimized at a daily time step over some representative historical period. Parameter adjustments were performed using the shuffled complex evolution global search algorithm (Duan et al. 1992) that has been widely used for hydrologic model calibration (e.g., Pokhrel et al. 2008; Tang et al. 2006; Vrugt et al. 2005). To obtain suitable model performance in terms of flow volume and timing, the Nash–Sutcliffe efficiency (NSE; Nash and Sutcliffe 1970) was used as a performance metric.

Figure 2 shows the calibration and evaluation performance of the model for the Salt River for the periods 1981–90 (the calibration period) and 1991–2000 (an evaluation period). The solid line indicates observed monthly streamflow and the dashed line indicates the monthly discharge simulated by the model using the calibrated parameters. Performance for the Salt River at a monthly time step is satisfactory with NSE value of 0.78. Note the NSE metric evaluated at the monthly time scale shows significant improvement over the uncalibrated value of NSE of 0.64 achieved for a model run generated with a priori parameter values (before calibration) for the entire period of record. Since, in climate-change-type studies, one typically evaluates seasonal changes and Shi et al. (2008) showed that for seasonal streamflow forecasting monthly calibrations of streamflow is sufficient, we did not go beyond monthly evaluations for this study.

Fig. 2.

Model calibration and evaluation results for the Salt River basin. The graphs show aggregate monthly streamflows.

Fig. 2.

Model calibration and evaluation results for the Salt River basin. The graphs show aggregate monthly streamflows.

Apart from the ability of the model to simulate historical observed streamflow, we also evaluate whether the raw GCM data provide a satisfactory simulation of the historical observed climate. Figure 3 shows the comparison in the form of a climatology plot of observed precipitation [based on the Maurer et al. (2002) dataset], raw GCM output, and the necessity to BCSD the precipitation data (Fig. 4 is similar, but for temperature). Raw results from the GCMs show that winter precipitation is represented somewhat well but that summer precipitation is much worse because of the inability of the GCMs to adequately simulate the summer monsoon. In our case this is not a big issue, since we are considering water resource projections and since streamflow in the basin under study is dominated by winter precipitation. The observed and BCSD precipitation and temperature data were used to drive the VIC hydrological model for the study area to simulate historical streamflow in the basin. Figure 5 shows the simulated streamflow for the Salt River basin, using the observed climate and the BCSD GCM climate forcings. The climatology plots shown in Figs. 35 are based on averaging daily data between 1960 and 2000. These figures show that there is sufficient skill in terms of predicting the historical hydroclimate of the basins, and hence, we can use these models to investigate the physical process of change in the twenty-first century. The dampened variability in the simulated discharge when VIC is driven by the raw GCM (as opposed to BCSD) data is also clear from these plots. The following section discusses the results of merging the projected future climate multimodel ensemble data with the hydrologic model to estimate potential hydrological changes.

Fig. 3.

Comparison of the historic (1960–2000) BCSD GCM simulated and observed precipitation over the Salt River basin. The basin averages were calculated over the A1B, A2, and B1 scenarios for each model.

Fig. 3.

Comparison of the historic (1960–2000) BCSD GCM simulated and observed precipitation over the Salt River basin. The basin averages were calculated over the A1B, A2, and B1 scenarios for each model.

Fig. 4.

As in Fig. 3, but for temperature.

Fig. 4.

As in Fig. 3, but for temperature.

Fig. 5.

Comparison of the historic (1960–2000) simulated streamflow using observed climate and observed and BCSD GCM simulated streamflow over the Salt River basin. The averages were calculated over the A1B, A2, and B1 scenarios for each model.

Fig. 5.

Comparison of the historic (1960–2000) simulated streamflow using observed climate and observed and BCSD GCM simulated streamflow over the Salt River basin. The averages were calculated over the A1B, A2, and B1 scenarios for each model.

4. Results

a. Projected annual changes to hydrologic variables

Figure 6 shows the time series of both historical and projected changes in temperature, precipitation, and streamflow. For the historical period, the solid red line shows the observed values and the gray shaded region shows the range of values (from minimum to maximum) estimated by the GCMs used in this study. For the future period, individual time series from the various GCMs are grouped by A1B, A2, and B1 emission scenarios and indicated using dashed and dash–dotted lines. The two extreme scenarios are shown in dark gray. Also shown is the ensemble mean of the different scenarios for each of the variables under consideration, smoothed by using a 5-yr moving average window to bring out the trends.

Fig. 6.

Historical and future change to (a) temperature, (b) precipitation, and (c) streamflow in the Salt River basin. The asterisk indicates that the ensemble mean had a significant trend using a nonparametric test at the p value 0.05.

Fig. 6.

Historical and future change to (a) temperature, (b) precipitation, and (c) streamflow in the Salt River basin. The asterisk indicates that the ensemble mean had a significant trend using a nonparametric test at the p value 0.05.

Based on the ensemble mean, temperatures are projected to rise by about 3.5°C in comparison to the historical average, precipitation is projected to decrease by about 5%, and streamflow is projected to decline by approximately 25% by the end of the twenty-first century. A nonparametric trend test based on the Mann–Kendall method (Helsel et al. 2006) indicates that the projected increasing trend in temperature and the projected decreasing trend in streamflow are both statistically significant (p value < 0.05), whereas the projected decreasing trend in annual precipitation is not. The annual average change in runoff during the near future is negligible; however, in the far future, the decrease in runoff is −1.04 mm month−1, and this is a result primarily of annual average decrease in precipitation of −0.79 mm month−1 and increases in evapotranspiration of 0.272 mm month−1.

Similar results in terms of trends were observed (not reported here for conciseness) for the Verde River basin. While the streamflow hydrograph for the Verde is similar to that of the Salt, the magnitudes of flow are much lower, due mainly to the fact that the Verde basin receives, on average, less precipitation than the Salt. Note that the Verde basin also shows a statistically significant decline in streamflow into the twenty-first century.

In summary, both of these basins are projected to have an increasing trend in temperature that is statistically significant, small decreases in precipitation that are not significant, and significant decreases in streamflow. To understand the causes of the decrease in streamflow, we next perform a seasonal analysis by subdividing the annual time series.

b. Projected seasonal changes to water balance components

An understanding of the physical mechanisms that would lead to decreased streamflow requires an analysis of the seasonal changes, that is, not only in precipitation minus evapotranspiration but also changes in snow water equivalent (SWE). This is important since approximately 70% of the annual streamflow is generated in the winter (see section 2). Figure 7 shows how precipitation is transformed to streamflow for different (yearly, winter, summer) periods. The left column shows mean annual values of the water balance variables, except for precipitation (which are total annual), and the middle and right columns show the cool season (December–May) and warm season (June–November) values. The choice of these months was guided by the fact that we can then sample the entire year, rather than use the typical definition of winter and summer months, which are usually used to indicate only 3 months [e.g., December–February (DJF) or June–August] each.

Fig. 7.

Time series of various water balance components at annual, cool (December–May), and warm (June–November) time scales for the Salt River basin. In each graph, the solid black line indicates the ensemble mean, and the blue dashed and the green dash–dotted lines indicate the two extreme scenarios. The shaded region indicates one standard deviation above and below each of the ensemble members.

Fig. 7.

Time series of various water balance components at annual, cool (December–May), and warm (June–November) time scales for the Salt River basin. In each graph, the solid black line indicates the ensemble mean, and the blue dashed and the green dash–dotted lines indicate the two extreme scenarios. The shaded region indicates one standard deviation above and below each of the ensemble members.

Figure 7 shows that decreases in ensemble mean annual streamflow for the Salt basin are caused primarily by both decreases in winter precipitation (which are statistically significant) and by reduced snowmelt contributions to streamflow. There is an increasing trend in ET in the winter, but its absolute magnitude, compared to the ET in the summer, is small. These results demonstrate the significance of winter processes on water supply in this basin. Under current conditions, ~70% of annual streamflow is realized during the winter and spring seasons, with winter (DJF) average temperatures about 2°C below freezing for the 2100–2750-m elevation band. Consequently, if the projected 3.5°C increase in temperature by the end of the twenty-first century occurs, we can expect a significant amount of snow coverage to vanish as the snow line (the elevation above which precipitation falls as snow) shifts to a higher elevation.

Similar results in terms of direction of change were observed for the Verde basin. The projected decreases in streamflow are again attributable to reduced winter precipitation. Compared to the magnitude of snowmelt contribution to the Salt River, the Verde River does not have a big snow component even in the historical period, mostly because the average elevations in the basin are much lower than in the Salt. Even though the magnitude contribution of snowmelt is different, there is a similar reduction in snow contribution in the Verde River.

To evaluate the sensitivity to exclusion of the extreme scenarios, namely, the wet (PCM-A2) and dry (MIROC-A2) climate model runs, a similar analysis as in Fig. 7 was performed with these models removed. The results from this analysis clearly showed that the trends in the water balance components do not change because of the exclusion of the wet and dry models.

Figure 7 shows the projected impacts, on snowpack, of rising temperatures into the second half of the twenty-first century. To examine the impacts on spatial distribution of snow, we plot the relationship between spatial distribution of winter average temperatures and elevation. For simplicity, we assume that all grid points within the watershed for which daily average temperatures fall below 0°C will be snow covered during the winter. The VIC model also accounts for precipitation falling as rain or snow based on a temperature index.

Figure 8 shows two subplots of winter average temperatures versus elevation for all grid cells within the Salt and Verde basin, one for the historic period (1970–2000) and one for the future projections (2068–98). We see that, whereas about 26% of the watershed has historically been covered by snow during the winter (Fig. 8a), the model projections indicate that this area will reduce to only 5% by the end of the century (Fig. 8b). Simultaneously, the snow line is expected to shift from 2000 m MSL (historical) to about 2500 m MSL (future). The snow line in this study is defined as the lowest elevation that has an average temperature below 0°C in the winter. This positive trend in the snow line shifting to higher elevations has already been observed in the historical period as shown by Svoma (2011).

Fig. 8.

Plot of winter average temperature vs elevation to show the spatial change in snow. Note that the snow line in this study is defined as the lowest elevation that has an average temperature below 0°C in winter.

Fig. 8.

Plot of winter average temperature vs elevation to show the spatial change in snow. Note that the snow line in this study is defined as the lowest elevation that has an average temperature below 0°C in winter.

In addition to analyzing temperature as a surrogate for snow cover, one can analyze the VIC model–simulated snow; Fig. 9 shows the change in model-simulated SWE for two future periods differenced from the historical observed period. Notice two features in this graph: by the middle of the twenty-first century, the values of SWE begin to go down in comparison to the historical observed period, and the area covered by snow is significantly reduced, leaving snow only in the highest elevations of the watershed (around the White Mountains in the western side of the watershed). This can be significant from a water management standpoint, especially impacting water supply, reservoir operations, and hydropower generation for the Salt River Project.

Fig. 9.

Spatial change in SWE for two future periods differenced from the historical reference period for the Salt River basin. (Note that the magnitudes are negative, meaning the future values are smaller than the historic.)

Fig. 9.

Spatial change in SWE for two future periods differenced from the historical reference period for the Salt River basin. (Note that the magnitudes are negative, meaning the future values are smaller than the historic.)

Figure 10 compares the historical and projected future water balances for the Salt River basin (results averaged over all models and scenarios). The water balance can be written as

 
formula

where the change in total storage S is due to soil moisture and snow (the two primary state variables of the system), P is the precipitation input, E is evapotranspiration, and runoff R consists of two components—fast runoff (RO) and slow base flow (BF)—as shown in Fig. 10 for the GCM model output averaged across all scenarios. This figure helps verify the closure of the water balance and also provides insights into what is causing the eventual decrease in streamflow.

Fig. 10.

Monthly water balance change in the Salt River basin for (left) historic and (middle, right) percentage change in these variables for two future periods compared to the historical period for the (top) Salt and (bottom) Verde Rivers.

Fig. 10.

Monthly water balance change in the Salt River basin for (left) historic and (middle, right) percentage change in these variables for two future periods compared to the historical period for the (top) Salt and (bottom) Verde Rivers.

For the Salt River (Fig. 10, top), the historical period (1970–2010) is characterized by a significant winter snow contribution, whereas the spring streamflow is driven primarily by melt of winter snow accumulations. For the future period (2011–50), the winter snow contribution to spring melt is significantly reduced, and there is not a decrease in winter precipitation and increases in winter evaporation are small. Hence, this first causes a shift toward early runoff and as precipitation falls more as rain than snow. Toward the latter half (2051–90) of the twenty-first century, the reason for decrease in runoff is due to small (statistically insignificant) decreases in precipitation, decreases in SWE, and increases in winter evaporation. The results are similar in magnitude and direction for the Verde River basin, as shown in Fig. 10 (bottom).

Temperatures in this semiarid basin rise significantly as early as April, causing large increases in ET, with this demand being met by soil moisture stored during the cool season. During the next 3 months, when there is very little precipitation in the region, soil moisture continues to be depleted until the monsoon arrives in July. Notice in Fig. 10 that the soil moisture storage is further depleted in the future during the transition from winter to spring and summer. While the monsoon precipitation adds small amounts to the soil moisture storage, this is largely lost to ET caused by very high temperatures during this time of the year. Notice that in the summer the precipitation causes the soil moisture to increase, but there is no significant increase in runoff.

5. Discussion

We have examined, quantitatively, the potential changes in hydroclimatology of the Salt and Verde Rivers that may be expected in response to projected climate change. The quantification is done in a way that can help water managers in central Arizona make better informed decisions. The precipitation in the basin is bimodal (as explained in section 2) and thus is necessary to quantify the seasonal variations in streamflow. By merging GCM climate information with a physically realistic hydrologic model, we find that significant declines may be expected in streamflow for both the Salt and the Verde River basins (Fig. 6). While the expectation of streamflow decline in the southwestern United States in response to projected climate change has been previously reported (e.g., Seager et al. 2007) and our results are in agreement with these findings, the mechanisms causing such change in the lower Colorado has been further explored in the current research.

A fundamental argument in the Seager et al. (2007) paper is that the twenty-first-century drying of the Southwest will be caused by decreases in precipitation accompanied by increases in temperature, such that available water (defined as precipitation minus ET) is reduced. Our study of the Salt–Verde system suggests that this explanation, although correct, does not account for the trend in an important state variable, that is, SWE. Based on our current research, we propose a refinement of the causal mechanism for reduction in streamflow in central Arizona. To properly explain the eventual decline in system streamflow, it is true that ensemble mean temperature shows a statistically significant increasing trend and the ensemble mean streamflow shows a statistically significant decreasing trend. However, the corresponding declines in annual precipitation are not statistically significant. Figures 710 show that it is the combined effect of temperature increase and winter precipitation and snowmelt decline and increase in winter ET that is causing the projected streamflow reductions. The decline in snowmelt is, in turn, caused by reduced winter precipitation (although the change is not statistically significant) combined with reduced snow storage brought about by the increases in minimum daily temperatures (Fig. 7, see SWE) and increase in winter ET (Fig. 10, see right column). The current winter average temperatures are only ~1°–2°C below freezing. As temperatures are expected to increase by about ~3°C by the end of the twenty-first century, there will be significantly less snow at the higher elevations (Figs. 8, 9).

In contrast, projected changes in future summer precipitation bracket the historical average, with four simulations projecting slightly lower precipitation and four others projecting slightly higher precipitation. However, even in the latter case, the projected increases in precipitation are not enough to compensate for the shortfalls in streamflow volume brought about by declining accumulation of snowpack and its subsequent release as snowmelt. Further, summer ET is already high, is “water limited” in the basin, and only shows a small increase in future projections (Fig. 7). However, the result related to changes in summer precipitation should be viewed with some caution, as the GCMs do not simulate monsoon as a salient climatological feature. This is an area where dynamical downscaling may add some important additional insights.

Since winter accumulation of snow is important for water availability in the Salt–Verde basin, we also assess potential changes in spatial distribution of snow. Assuming that grid cells at temperatures less than 0°C temperature have snow cover, we find that the areal extent of snow cover can be expected to decrease from the historic value of 26% of the basin area to about 5% (Figs. 8, 9). As temperatures increase in a warmer climate, snowpack will be affected in three distinct ways: 1) by changing the fraction of total precipitation that falls as snow (Wi et al. 2012), 2) by affecting accumulation in the snowpack (Pierce et al. 2008), and 3) by changing the large-scale circulation that can in turn affect the storm tracks (Yin 2005). These impacts will be stronger in the Salt than in the Verde because the Salt River currently has larger snow storage than the Verde River.

6. Conclusions

In conclusion, these results can help water managers assess the impacts of projected climate change on the hydrological responses of the Salt and Verde watersheds in central Arizona. The metropolitan Phoenix area currently relies on water from both the upper Colorado River basin (through CAP) and the SVRB (through SRP). However, if water shortages were to occur in the upper basin, the CAP supply would be significantly reduced because it holds a junior priority water entitlement to Colorado River water. This makes it critical for water managers in the region to quantify potential changes in water availability that could affect water supply, reservoir operations, and hydropower generation in central Arizona. The projected increases in temperature of about 3.5°C, and slight decreases in precipitation, can be expected to result in a 25% reduction in streamflow by the end of the twenty-first century, caused mainly by small reductions in winter precipitation combined with reduced snowmelt contributions during the spring due to higher winter evaporative losses. Importantly, the snow line will climb from 2000 to about 2500 m by the end of the century, causing a reduction in snow-covered area from 26% to only 5%.

These findings are significant from the standpoint of water management in central Arizona. With continued growth in population and likely reductions in available surface water supply, tough policy and management decisions will be needed to help sustain water supply. This analysis provides important information on climate induced future streamflow projections useful for local water managers, namely, the Salt River Project and the City of Phoenix assessments of potential impacts of climate change on future operations.

Acknowledgments

The first author received partial support from the Water Sustainability Program, through the University of Arizona Technology and Research Initiative Fund 2010/2011 Student Fellowship Program; NASA (Grant CAN NN-H-04-Z-YO-010-C); and the Salt River Project, Phoenix, Arizona. The third author received 2011–12 sabbatical support from Jesus Carrera of the Institute of Environmental Assessment and Water Research (IDAEA), Spanish Council of Scientific Research (CSIC), and the Spanish Ministry of Science and Innovation (MEC); additional support was also provided by Andy Pitman and the Australian Research Council through the Centre of Excellence for Climate System Science (Grant CE110001028), and by the Cooperative Research Programme of the Organization of Economic Cooperation and Development. We are also thankful to several modeling groups, the Program for Climate Model Diagnosis and Intercomparison (PCMDI), and the WCRP’s Working Group on Coupled Modelling (WGCM), for their roles in making available the WCRP CMIP3 multimodel dataset. Support for that dataset is provided by the Office of Science, U.S. Department of Energy. In addition, computer support by James Broermann at the Department of Hydrology and Water Resources and VIC modeling assistance by Matej Durcik are greatly appreciated. The authors thank Mohammed Mahmoud from CAP for his perspectives on Colorado water management issues. The authors also thank the anonymous reviewers who have helped refine this manuscript.

REFERENCES

REFERENCES
Brekke
,
L. D.
,
M. D.
Dettinger
,
E. P.
Maurer
, and
M.
Anderson
,
2008
:
Significance of model credibility in estimating climate projection distributions for regional hydroclimatological risk assessments
.
Climatic Change
,
89
,
371
394
, doi:.
Cayan
,
D. R.
,
T.
Das
,
D. W.
Pierce
,
T. P.
Barnett
,
M.
Tyree
, and
A.
Gershunov
,
2010
:
Future dryness in the southwest US and the hydrology of the early 21st century drought
.
Proc. Natl. Acad. Sci. USA
,
107
,
21 271
21 276
, doi:.
Christensen
,
N. S.
, and
D. P.
Lettenmaier
,
2007
:
A multimodel ensemble approach to assessment of climate change impacts on the hydrology and water resources of the Colorado River basin
.
Hydrol. Earth Syst. Sci.
,
11
,
1417
1434
, doi:.
Christensen
,
N. S.
,
A. W.
Wood
,
N.
Voisin
,
D. P.
Lettenmaier
, and
R. N.
Palmer
,
2004
:
The effects of climate change on the hydrology and water resources of the Colorado River basin
.
Climatic Change
,
62
,
337
363
, doi:.
Collins
,
W. D.
, and Coauthors
,
2006
:
The Community Climate System Model version 3 (CCSM3)
.
J. Climate
,
19
,
2122
2143
, doi:.
Covey
,
C.
,
K. M.
AchutaRao
,
U.
Cubasch
,
P.
Jones
,
S. J.
Lambert
,
M. E.
Mann
,
T. J.
Phillips
, and
K. E.
Taylor
,
2003
:
An overview of results from the Coupled Model Intercomparison Project
.
Global Planet. Change
,
37
,
103
133
, doi:.
Dominguez
,
F.
,
J.
Canon
, and
J.
Valdes
,
2010
:
IPCC-AR4 climate simulations for the southwestern US: The importance of future ENSO projections
.
Climatic Change
,
99
,
499
514
, doi:.
Duan
,
Q. Y.
,
S.
Sorooshian
, and
V.
Gupta
,
1992
:
Effective and efficient global optimization for conceptual rainfall–runoff models
.
Water Resour. Res.
,
28
,
1015
1031
, doi:.
Fowler
,
H. J.
,
S.
Blenkinsop
, and
C.
Tebaldi
,
2007
:
Linking climate change modelling to impacts studies: Recent advances in downscaling techniques for hydrological modelling
.
Int. J. Climatol.
,
27
,
1547
1578
, doi:.
Gober
,
P.
,
C. W.
Kirkwood
,
R. C.
Balling
,
A. W.
Ellis
, and
S.
Deitrick
,
2010
:
Water planning under climatic uncertainty in Phoenix: Why we need a new paradigm
.
Ann. Assoc. Amer. Geogr.
,
100
,
356
372
, doi:.
Gordon
,
C.
,
C.
Cooper
,
C. A.
Senior
,
H.
Banks
,
J. M.
Gregory
,
T. C.
Johns
,
J. F. B.
Mitchell
, and
R. A.
Wood
,
2000
:
The simulation of SST, sea ice extents and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments
.
Climate Dyn.
,
16
,
147
168
, doi:.
Harpold
,
A.
,
P.
Brooks
,
S.
Rajagopal
,
I.
Heidbuchel
,
A.
Jardine
, and
C.
Stielstra
,
2012
:
Changes in snowpack accumulation and ablation in the intermountain west
.
Water Resour. Res.
,
48
, W11501, doi:.
Helsel
,
D. R.
,
D. K.
Mueller
, and
J. R.
Slack
,
2006
: Computer program for the Kendall family of trend tests. USGS Scientific Investigations Rep. 2005–5275, 4 pp. [Available online at http://pubs.usgs.gov/sir/2005/5275/pdf/sir2005-5275.pdf.]
Hurkmans
,
R.
,
P. A.
Troch
,
R.
Uijlenhoet
,
P.
Torfs
, and
M.
Durcik
,
2009
:
Effects of climate variability on water storage in the Colorado River basin
.
J. Hydrometeor.
,
10
,
1257
1270
, doi:.
Jungclaus
,
J. H.
, and Coauthors
,
2006
:
Ocean circulation and tropical variability in the coupled model ECHAM5/MPI-OM
.
J. Climate
,
19
,
3952
3972
, doi:.
K-1 Model Developers
,
2004
: K-1 Coupled Model (MIROC) Description. K-1 Tech. Rep. 1, H. Hasumi and S. Emori, Eds., Center for Climate System Research, University of Tokyo, Tokyo, Japan, 34 pp. [Available online at http://ccsr.aori.u-tokyo.ac.jp/~hasumi/miroc_description.pdf.]
Liang
,
X.
,
D. P.
Lettenmaier
,
E. F.
Wood
, and
S. J.
Burges
,
1994
:
A simple hydrologically based model of land-surface water and energy fluxes for general-circulation models
.
J. Geophys. Res.
,
99
,
14 415
14 428
, doi:.
Lohmann
,
D.
,
E.
Raschke
,
B.
Nijssen
, and
D. P.
Lettenmaier
,
1998
:
Regional scale hydrology: I. Formulation of the VIC-2L model coupled to a routing model
.
Hydrol. Sci. J.
,
43
,
131
141
, doi:.
MacDonald
,
G. M.
,
2010
:
Water, climate change, and sustainability in the southwest
.
Proc. Natl. Acad. Sci. USA
,
107
,
21 256
21 262
, doi:.
Marcott
,
S. A.
,
J. D.
Shakun
,
U. P.
Clark
, and
A. C.
Mix
,
2013
:
A reconstruction of regional and global temperature for the past 11,300 years
.
Science
,
339
,
1198
, doi:.
Maurer
,
E. P.
,
A. W.
Wood
,
J. C.
Adam
,
D. P.
Lettenmaier
, and
B.
Nijssen
,
2002
:
A long-term hydrologically based dataset of land surface fluxes and states for the conterminous United States
.
J. Climate
,
15
,
3237
3251
, doi:.
Maurer
,
E. P.
,
L.
Brekke
,
T.
Pruitt
, and
P. B.
Duffy
,
2007
:
Fine-resolution climate projections enhance regional climate change impact studies
.
Eos, Trans. Amer. Geophys. Union
,
88
,
504
, doi:.
Maurer
,
E. P.
,
H. G.
Hidalgo
,
T.
Das
,
M. D.
Dettinger
, and
D. R.
Cayan
,
2010
:
The utility of daily large-scale climate data in the assessment of climate change impacts on daily streamflow in California
.
Hydrol. Earth Syst. Sci.
,
14
,
1125
1138
, doi:.
McCabe
,
G. J.
, and
D. M.
Wolock
,
2008
:
Joint variability of global runoff and global sea surface temperatures
.
J. Hydrometeor.
,
9
,
816
824
, doi:.
Nash
,
J. E.
, and
J. V.
Sutcliffe
,
1970
:
River flow forecasting through conceptual models part I—A discussion of principles
.
J. Hydrol.
,
10
,
282
290
, doi:.
Nash
,
L. L.
, and
P. H.
Gleick
,
1991
:
Sensitivity of streamflow in the Colorado basin to climatic changes
.
J. Hydrol.
,
125
,
221
241
, doi:.
Pierce
,
D. W.
, and Coauthors
,
2008
:
Attribution of declining western U.S. snowpack to human effects
.
J. Climate
,
21
,
6425
6444
, doi:.
Pierce
,
D. W.
,
T. P.
Barnett
,
B. D.
Santer
, and
P. J.
Gleckler
,
2009
:
Selecting global climate models for regional climate change studies
.
Proc. Natl. Acad. Sci. USA
,
106
,
8441
8446
, doi:.
Pokhrel
,
P.
,
H. V.
Gupta
, and
T.
Wagener
,
2008
:
A spatial regularization approach to parameter estimation for a distributed watershed model
.
Water Resour. Res.
,
44
, W12419, doi:.
Rajagopalan
,
B.
,
K.
Nowak
,
J.
Prairie
,
M.
Hoerling
,
B.
Harding
,
J.
Barsugli
,
A.
Ray
, and
B.
Udall
,
2009
:
Water supply risk on the Colorado River: Can management mitigate?
Water Resour. Res.
,
45
, W08201, doi:.
Seager
,
R.
, and Coauthors
,
2007
:
Model projections of an imminent transition to a more arid climate in southwestern North America
.
Science
,
316
,
1181
1184
, doi:.
Serreze
,
M. C.
,
M. P.
Clark
,
R. L.
Armstrong
,
D. A.
McGinnis
, and
R. S.
Pulwarty
,
1999
:
Characteristics of the western United States snowpack from snowpack telemetry (SNOTEL) data
.
Water Resour. Res.
,
35
,
2145
2160
, doi:.
Shi
,
X. G.
,
A. W.
Wood
, and
D. P.
Lettenmaier
,
2008
:
How essential is hydrologic model calibration to seasonal streamflow forecasting?
J. Hydrometeor.
,
9
,
1350
1363
, doi:.
Sorooshian
,
S.
,
Q.
Duan
, and
V. K.
Gupta
,
1993
:
Calibration of rainfall–runoff models: Application of global optimization to the Sacramento Soil Moisture Accounting Model
.
Water Resour. Res.
,
29
,
1185
1194
, doi:.
Svoma
,
B. M.
,
2011
:
Trends in snow level elevation in the mountains of central Arizona
.
Int. J. Climatol.
,
31
,
87
94
, doi:.
Tang
,
Y.
,
P.
Reed
, and
T.
Wagener
,
2006
:
How effective and efficient are multiobjective evolutionary algorithms at hydrologic model calibration?
Hydrol. Earth Syst. Sci.
,
10
,
289
307
, doi:.
Vano
,
J. A.
, and Coauthors
,
2014
:
Understanding uncertainties in future Colorado River streamflow
.
Bull. Amer. Meteor. Soc.
,
95
,
59
78
, doi:.
Vrugt
,
J. A.
,
C. G. H.
Diks
,
H. V.
Gupta
,
W.
Bouten
, and
J. M.
Verstraten
,
2005
:
Improved treatment of uncertainty in hydrologic modeling: Combining the strengths of global optimization and data assimilation
.
Water Resour. Res.
,
41
, W01017, doi:.
Washington
,
W. M.
, and Coauthors
,
2000
:
Parallel climate model (PCM) control and transient simulations
.
Climate Dyn.
,
16
,
755
774
, doi:.
Water and Environmental Resources Division
,
2011
: West-wide climate risk assessments: Bias-corrected and spatially downscaled surface water projections. Tech. Memo. 86-68210-2011-01, Technical Service Center, Bureau of Reclamation, Denver, CO, 122 pp. [Available online at www.usbr.gov/WaterSMART/docs/west-wide-climate-risk-assessments.pdf.]
Wi
,
S.
,
F.
Dominguez
,
M.
Durcik
,
J.
Valdes
,
H. F.
Diaz
, and
C. L.
Castro
,
2012
:
Climate change projection of snowfall in the Colorado River basin using dynamical downscaling
.
Water Resour. Res.
,
48
, W05504, doi:.
Wood
,
A. W.
,
E. P.
Maurer
,
A.
Kumar
, and
D. P.
Lettenmaier
,
2002
:
Long-range experimental hydrologic forecasting for the eastern United States
.
J. Geophys. Res.
,
107
, 4429, doi:.
Wood
,
A. W.
,
L. R.
Leung
,
V.
Sridhar
, and
D. P.
Lettenmaier
,
2004
:
Hydrologic implications of dynamical and statistical approaches to downscaling climate model outputs
.
Climatic Change
,
62
,
189
216
, doi:.
Wood
,
E. F.
,
D. P.
Lettenmaier
, and
V. G.
Zartarian
,
1992
:
A land-surface hydrology parameterization with subgrid variability for general circulation models
.
J. Geophys. Res.
,
97
,
2717
2728
, doi:.
Yin
,
J. H.
,
2005
:
A consistent poleward shift of the storm tracks in simulations of 21st century climate
.
Geophys. Res. Lett.
,
32
, L18701, doi:.

Footnotes

*

Current affiliation: Division of Hydrologic Sciences, Desert Research Institute, Reno, Nevada.