Purpose and Scope

This study was done in cooperation with Kern County, California. The objective of this study is to utilize all available data to calibrate the various components of the water balance. By utilizing all available data, this study aims to provide a more refined and accurate estimate of the average annual natural recharge and other hydrologic water balance variables in the IWV Basin. This study validates the many historical estimates and provides a more realistic view of the basin's hydrologic conditions. Additionally, the calibrated model was used to apply future climate projections, thus providing insight into potential future hydrologic conditions. The information obtained from this study may be used to guide sustainable resource management planning for the IWV Basin.

Study Area

The IWV is in the Mojave Desert east of the Sierra Nevada, about 200 kilometers (km) north of Los Angeles (fig. 1). The valley floor covers an area of 777 square kilometers, ranging in elevation from 655 to 732 meters (m) above sea level. The valley is bounded on the west by the Sierra Nevada, on the north by a low ridge of volcanic rocks and the Coso Range, on the east by the Argus Range, and on the south by the El Paso Mountains and Spangler Hills. In contrast to the steep topography surrounding the valley, the valley floor gently slopes toward China Lake, a large dry lake or playa, in the central northeastern valley. At an altitude of 655 m, it is the primary natural groundwater discharge point in the valley (fig. 1).

The IWV has an arid climate, with annual precipitation ranging from 10 to 15 centimeters (cm) falling on the valley floor (primarily in winter and spring, October through March). During the remainder of the year, rain falls infrequently during summer thundershowers. Mean air temperatures range from about 4 degrees Celsius (°C) in winter to about 28 °C in summer. Native vegetation covers most of the IWV study area, with desert scrub covering more than 56 percent of the valley floor. Barren lands cover about 4 percent of the study area, mostly over the volcanic rocks and in the China Lake playa. The rest of the basin is covered with other vegetation types, including desert scrub, alkali desert scrub, sagebrush, juniper, mixed chaparral, grasslands, and pines (fig. 2).

The IWV basin is bounded to the west along the Sierra Nevada by a north trending fault that is present in the center and eastern part of the valley (Berenbrock and Schroeder, 1994; Todd Engineers, 2014; fig. 3). Surficial geologic types in the IWV Basin include granites, which surround the valley, volcanics mostly in the north, conglomerates in the south, and small inclusions of metasediments. Desert fill covers the entire valley floor and western drainages. Unconsolidated deposits from lakebed, stream, and alluvial fans make up an upper aquifer and a lower aquifer. The primary deep aquifer in the IWV consists of alluvial deposits as much as 300 m thick, covering the main valley floor. The lower aquifer is generally unconfined except in the eastern part of the valley where the aquifer is confined by silt and clay lenses, lake deposits, and playa deposits. In the east-central part of the valley, beneath China Lake and other dry lakebeds (fig. 3), the lower aquifer is overlain by an upper aquifer that consists of lacustrine, alluvial, playa, and sandstone deposits (Berenbrock and Schroeder, 1994; Todd Engineers, 2014).

Basin Characterization Model, version 8 Local Calibration for Indian Wells Valley

Local model calibration was done in two steps to match AET and streamflow measurements. The first step was done by parameterizing the BCMv8 to match estimates of AET reported by Reitz and others (2017) by vegetation type for the IWV Basin to create monthly time series of AET for October 2000 through September 2013. In IWV, there are 19 vegetation types, with desert scrub covering more than 56 percent of the study area (fig. 2). Parameterization was done first by calculating monthly scaling factors (Kv) for each vegetation type using the AET values estimated by Reitz and others (2017) divided by PET calculated using the BCMv8 for each month and averaged during the measurement period. These Kv values represent the proportion of AET below potential for each month that a vegetation type transpires. Additional vegetation parameters included adjustments to the effective rooting depth to increase the amount of effective soil storage and increase AET, which is often required to match the AET values estimated by Reitz and others (2017), while reducing the water available for recharge and runoff. There also were growth parameters that allowed for the seasonality of AET for vegetation types that are sensitive to seasonal and annual variations in precipitation. More details on parameterization can be found in Flint and others (2021b).

The second step to local model calibration was to assemble streamflow measurements and accumulate the BCMv8 recharge plus runoff values for drainages upstream from each streamgage. Streamflow data measured at 10 gaging stations (8 U.S. Geological Survey stations and 2 Kern County stations, and data are available in Saleh and Flint, 2019) in relatively unimpaired streams were used for model calibration (fig. 1). Because of scarcity of streamflow data within the basin, it was necessary to use streamflow data from outside the IWV sub-watersheds to calibrate all local soils, vegetation, and geologic types (fig. 1). Table 2 provides the location, collecting agency, numerical identifier, and the period of record available from these gaging stations used for calibration (Saleh and Flint, 2019). Note that two streamgages measure drainage from the Sierra Nevada to the west, streamgages 7 and 8 measure drainage from lower mountains to the northeast, streamgages 3 and 6 measure drainage from intermittent lower elevation sites to the south, and streamgages 4, 5, 9, and 10 measure drainage from locations at the Sierra Nevada front where excess flow becomes runoff instead of mountain front recharge. Post-processing in a spreadsheet was done to accumulate monthly recharge and runoff values upstream from every streamgage. Post-processing used exponential decay values to calculate basin discharge to match peak flows, recession flows, and baseflows of the measured hydrographs (Flint and others, 2021b). Calibration statistics included the coefficient of determination (R²) and Nash-Sutcliffe Efficiency (NSE; Nash and Sutcliffe, 1970) and were used to evaluate calibration results and optimize the goodness-of-fit. Bedrock permeability was adjusted to change the ratio of recharge to runoff in the BCMv8 runs and to optimize the shape of the calculated hydrograph for each subarea. Adjustments to AET parameters were done iteratively with bedrock permeability to ensure the best fit of modeled to measured streamflow and to optimize the different components of the water balance. Additional calibrations to the model included matching observations of ponding on the China Lake playa. This calibration was done by reducing the bedrock permeability of the consolidated valley fill materials to allow for intermittent ponding during high precipitation events.

Future Climate Scenarios

To evaluate future hydrologic conditions in IWV, the calibrated IWV model was run using three future climate models (table 3) from the Global Climate Models (GCMs) selected for California’s Fourth Climate Change Assessment (4CCA) as best representing historical climatic patterns in California (California Department of Water Resources Climate Change Technical Advisory Group, 2015). The three selected models correspond to the Representative Concentration Pathway (RCP) 8.5, representing a future with a high global population, slow economic growth and technological change, and a high dependence on fossil fuels, without any consideration to reductions into greenhouse gas emissions (Riahi and others, 2011; Pierce and others, 2014). These three climate scenarios were statistically downscaled using the Localized Constructed Analogs (LOCA) method described in Pierce and others (2014) from the native model resolution (table 3) to a 6-km resolution. The statistical downscaling was done to improve estimates of extreme events and reduce the common downscaling problem of too many light-precipitation days (Pierce and others, 2014). The data were then further downscaled spatially to 270 m for BCMv8 application following the methods described in Flint and Flint (2012). The selected future scenarios represent a range of projected conditions, with the Second-generation Canadian Earth System Model (CanESM2) representing hot-wet conditions, the Community Climate System Model version 4 (CCSM4) representing warm-moderately wet conditions, and the Hadley Centre Global Environmental Model version 2 climate configurations (HadGEM2-CC) representing hot-dry conditions (table 3). The resulting datasets provided monthly minimum and maximum air temperature and precipitation for the 1981–2099 period. A historical baseline climate dataset from water years 1976 to 2005 (Livneh and others, 2013) was used to train the LOCA GCMs. Bias correction was applied to each LOCA scenario to match the modeled historical climate statistics with those of the historical observational data for that period (Pierce and others, 2014). As a result, to account for any slight differences between the historical GCM climate data and PRISM data (Daly and others, 2008), all calculations comparing BCM-derived results for historical and future 30-year mean periods were made using BCMv8 results based on the GCM-simulated historical data.

Figure 4 shows temperature and precipitation data available from the PRISM dataset used to run the historical BCM, along with simulated data from the model projections, which are run from 1950 to 2099 with carbon dioxide forcings beginning in 2006. Figure 4 shows that the bias-correction of the LOCA downscaled data to historical climate data (Livneh and others, 2013) do not exactly match the PRISM data. The difference can be partially explained by the fact that even though the Livneh and others (2013) dataset was based on PRISM, it used a consistent lapse rate throughout the domain (allowing for the dataset to extend outside of the continental United States), whereas PRISM used a variable lapse rate. More importantly, the bias correction was done to match the statistics (mean and variance) for the period 1976–2005, not 1981–2010, as shown in table 4.

The mean temperature rises for all models for both future periods, ranging from 4 to more than 5 ^°C by end of century. Precipitation is variable among models. The CanESM2 model increases the most for precipitation and has the most annual variability in the future run for precipitation and air temperature of all models, with the mean increasing from 221 millimeters per year (mm/yr) in the historical period to 324 mm/yr by end of century. The standard deviation also increases by nearly 30 percent by end of century, indicating that the future is projected to have more extreme precipitation years that also are hotter. The CCSM4 model also has increases in precipitation, with additional variability mid-century but not in the end-of-century period. HadGEM2-CC has variable changes in precipitation, rising mid-century and lowering by end of century, but the variability indicates more extreme rises by end of century.

Uncertainty

There are many sources of uncertainty in this study; however, improving the spatial distribution of physical properties during parameterization, hydraulic function tables, and local data used for model calibration, can reduce uncertainty and improve estimates of potential hydrologic outcomes. The BCMv8 addresses some of these uncertainties by incorporating vegetation-specific AET and using available spatial data of vegetation, topography, physical soil properties, and geology. Calibrations are done to optimize the fit to measured data for each hydrologic component of the water balance calculated by the model. Additional uncertainties in the BCM and extrapolation to ungaged basins are discussed at length in Flint and others (2013). Addressing uncertainties related to future climate-hydrology scenarios using GCMs is particularly difficult; however, we reduced this uncertainty by selecting a subset of the GCMs that were analyzed specifically to represent the range of historical climatic conditions in California and were selected for California’s 4CCA (California Department of Water Resources Climate Change Technical Advisory Group, 2015). In this study, we selected the CanESM2, CCSM4, and HadGEM2-CC to represent a range of projected hydrologic conditions in IWV.

Historical Climate Evaluation

Historical climatological and hydrological conditions for all subareas in IWV averaged during the 1981–2010 period are summarized in table 6. Total recharge for IWV from BCMv8 local calibration using historical PRISM data was estimated at about 10.7 mm³ (8,680 acre-feet) per year. This recharge value corresponds well with estimates from previous studies (table 2), developing a consensus among estimates during the last 50 years. Spatial distribution of natural recharge in IWV is shown on figure 7 and is a function of climate, geology, soil storage, and vegetation type. Most of the recharge in IWV is mountain front recharge from the Sierra Nevada (48 percent), volcanics subarea (21 percent), and the Argus Range (12 percent). Rose Valley subarea is not included in the sum of total recharge. Rose Valley is a subbasin northwest of Indian Wells Valley that—under wetter conditions more than 10,000 years ago—was part of a river system that flowed from Owens Valley through Rose Valley and Indian Wells Valley to Death Valley. Surface flow between the basins now occurs rarely, if ever. Subsurface flow is obstructed by geologically recent basalts, which force most of the groundwater to the surface as it crosses between basins (Todd Engineers, 2014).

Figure 8 shows the relation between precipitation and natural recharge estimated from the BCMv8 for each of the subareas in IWV during the 1981–2010 period. In all subareas, the relation had a low slope until enough precipitation occurred to exceed the requirements of soil-moisture storage, ponding, evaporation, and transpiration, after which recharge occurred. This threshold is approximately 150–250 mm/yr for all of the subareas, except for the volcanics, with a threshold of less than 100 mm/yr because of this subarea’s higher permeability (fig. 8). The distance that the recharge curve is offset from the 1:1 line is controlled by the potential for precipitation to be diverted to the other pre-recharge processes. For example, in the arid valley floor, where there is less precipitation and substantial soil storage, the estimated recharge was about 8 percent of the total BCMv8 estimated IWV recharge, and most of the precipitation was diverted to evapotranspiration (table 6); whereas, the Sierra Nevada subareas have more precipitation, much of it as snow.

The CWD is correlated with landscape stress, forest die-off, wildfire risk, and irrigation demand, and its variability is largely dominated by air temperature and available water. The spatial distribution of CWD over IWV is shown on figure 9, with the mean annual CWD of 1,200 mm/yr in the Sierra Nevada subareas. The melting of the snowpack in the Sierra Nevada subareas provides a longer seasonal duration of available water, thus maintaining a lower annual deficit. The highest CWD is present in the valley floor subarea of IWV, with more than 1,450 mm/yr, because of increased air temperature and PET and lower precipitation (table 6).

Future Climate Projections

Future projected hydrological conditions from three future climate scenarios for all subareas in IWV averaged during the 2040–69 mid-century and 2070–99 end-of-century periods were compared to simulated historical LOCA 1981–2010 conditions. Tables 7 and 8 show the percentage of change in projected climatic and hydrologic conditions when compared to 1981–2010 historical LOCA conditions. All three models show projected increases in average air temperature for IWV, and this increase is greater for end of century (table 7). The CanESM2 model projected the highest (as much as 29 percent, about 5 ^°C) increase in temperature in the end-of-century period compared to the other two models (table 7). There was large variation in the projected precipitation from the three GCMs for both periods. The CanESM2 model, a hot-wet projection, showed an average of 20-percent (about 4 cm) increase in precipitation in the mid-century period and an average of 31-percent (about 6 cm) increase in the end-of-century period when compared to historical LOCA data. The HadGEM2-CC model, a hot-dry projection, showed a decrease in precipitation in both periods, with a larger decrease for all subareas in the end-of-century period and reaching anywhere from a 23-percent decrease in the Coso Range to a 40-percent decrease in the Spangler Hills and El Paso Mountains (table 7). The CCSM4 warm-moderately wet model showed a small change in precipitation in both periods, with a slightly larger increase in the end-of-century period (table 7).

Future projected runoff for IWV was estimated using the three GCMs for the mid-century and end-of-century periods and compared to historical LOCA data (table 7). Results from the CanESM2 and CCSM4 models projected an increase in runoff in both future periods. CanESM2 projected a greater increase in runoff in the end-of-century period, whereas the more moderate CCSM4 model projected a similar increase in both periods. The hot-dry HadGEM2-CC model projected an increase in runoff in the mid-century period and decrease in the end-of-century period. This decrease is most significant (38 percent and 39 percent) in the El Paso Mountains and valley floor subarea, respectively. These results compare well with the HadGEM2-CC projected decrease in precipitation during the end-of-century period (table 7).

Generally, the trends shown in table 7, based on comparisons of 30-year annual means, indicate that there is a consensus among models that temperature is predicted to rise approximately 20–29 percent by end of century, which corresponds with the prediction from all models that CWD will also rise for all locations. However, there is not a consensus among models about whether it is predicted for more or less precipitation available in IWV, which determines how much water is available for AET, runoff, or recharge. Table 8 shows the percentage of change in the standard deviation of climatic and hydrologic conditions in comparison to historical conditions. In all subareas for all models, the annual variability in precipitation and air temperature increased in the mid-century and end of century by 1–42 percent for precipitation and 8–38 percent for temperature. Results from the CCSM4 model indicated an increase in annual variability in precipitation and air temperature only during the mid-century, followed by a decrease in precipitation, AET, CWD, and recharge in some subareas by the end of the century (table 8). However, most importantly for planning purposes, there is strong model consensus that by mid-century, variability in annual precipitation, runoff, and recharge—the primary water-supply variables—will increase, with recharge projected to rise by as much as 57 percent in the Sierra Nevada and 76 percent on the valley floor. This rise is caused by the increased variability in precipitation and the increase in the precipitation thresholds for recharge projected by all the models. These projections point to wet years where decisions to capture or store water are prudent. Also notable, results for all models indicated that all hydrologic variables increased on average in the Rose Valley subarea by end of century, except for HadGEM2-CC for precipitation, AET, runoff, and recharge (table 7).

Table 7 shows the change in AET and CWD from historical (1981–2010) to mid-century (2040–69) and to end-of-century (2070–99) periods for each of the three GCMs. The three scenarios varied in their projections of AET because HadGEM2-CC has lower projected precipitation and less available water for AET. The CanESM2 scenario showed the largest projected increase in AET in both periods in comparison to the historical period, with a larger increase in the end-of-century period, reflected in a more than 60-percent increase in AET in some subareas (Spangler Hills, El Paso Mountain, and valley floor) during the end-of-century period. This increase in AET corresponds with an increase in CWD among all sites in mid-century and end-of-century periods, with average CWD increases of 4 and 6 percent, respectively (table 7). Results from the HadGEM2-CC model indicate an overall increase in mid-century AET projections; however, there are large variations in AET projections in this period when evaluated on a subarea scale. AET increased as much as 35 percent in the Argus Range subbasin and decreased 17 percent in the Spangler Hills (table 7). The HadGEM2-CC model projected a decrease in AET for all the subareas in the end-of-century period except for a 13-percent increase in the Argus Range subarea. The HadGEM2-CC model showed an increase in CWD for both periods; however, HadGEM2-CC showed a bigger increase (almost double) in CWD projections for all subareas in the end-of-century period (table 7). Results from the CCSM4 model projected an increase in AET in all IWV subareas during both periods except for Spangler Hills and valley floor subareas, where CCSM4 projected a 26-percent and 5-percent decrease in AET by end of century, respectively. The CCSM4 results also indicated an increase in CWD in both periods: an average of 4 percent in mid-century and 7 percent in the end-of-century period compared to historical conditions. Spatial distribution of percentage of change in CWD estimations in the IWV projected by the GCMs for mid-century (2040–69) and end of century (2070–99) compared to the historical LOCA data are shown on figure 10. Although all the GCMs projected an increase in CWD in both periods, CanESM2 and CCSM4, the wet models, projected the largest spatial variation in all the subareas in both periods, with the greatest increases projected to be in the end-of-century period in the Sierra Nevada and valley floor subareas. HadGEM2-CC projected a lower spatial variation in CWD in both periods, with the larger projected increase in CWD in the end-of-century period.

Table 7 shows the percentage of change in projected recharge from the three GCMs when compared to 1981–2010 historical LOCA conditions. The three scenarios varied in their estimations; however, most recharge projections from the three models were greater than the model’s projection of precipitation for each period, except for the HadGME2-CC model in the end-of-century period, because most recharge occurs in years when precipitation exceeds the recharge threshold illustrated on figure 8. The change in the annual variability of precipitation (table 8) is much greater than historical precipitation, particularly for the two wet models (CanESM2 and CCSM4), ranging from 9 to 36 percent in the mid-century period, thus resulting in more opportunities for recharge in the future scenarios. All models showed an increase in recharge in the mid-century period for all IWV subareas, except for the 8-percent (0.09 mm³; 750 acre-feet) decrease projected by HadGEM2-CC for the Argus Range. Overall, the CanESM2 scenario projected the largest increase in recharge during the mid-century period, with a projected 51–61 percent increase in recharge in the Sierra Nevada subareas; however, CCSM4 projected a 77-percent increase (the largest estimate, Rose Valley, not included in this evaluation) in recharge in the valley floor subarea (table 7). There is a great variation in recharge projections from the three GCMs during the end-of-century period. CanESM2 projected an increase in recharge for all IWV subareas (Rose Valley not included in this evaluation) from an average of 70-percent (4.3 mm³; 3,456 acre-feet) increase in recharge in the South Sierra Nevada subarea to a 67-percent (0.6 mm³; 490 acre-feet) increase in recharge in the valley floor subarea. The CCSM4 model also projected an increase in recharge in all the subareas during the end-of-century period (except for a small decrease in the Argus Range). However, CCSM4 projections are smaller in magnitude than CanESM2 estimates in all subareas, except for the valley floor and Rose Valley where the CCSM4 estimate was larger, with a 73-percent (0.2 mm³; 154 acre-feet) increase in recharge (table 7). The HadGEM2-CC model showed a different pattern than the CanESM2 and CCSM4 models in the end-of-century period. HadGEM2-CC results indicated a large decrease in recharge for all subareas in IWV, from a 45-percent (2.4 mm³; 1,900 acre-feet, Rose Valley not included in this evaluation) decrease in the Sierra Nevada, to a 75-percent (0.6 mm³; 486 acre-feet) decrease in the valley floor (table 7).

Spatial distribution of percentage of change in recharge estimations in all IWV subareas projected by the GCMs for mid-century (2040–69) and end-of-century (2070–99) periods compared to the historical LOCA data are shown on figure 11. As previously stated, natural recharge in IWV is mostly generated from mountain front subareas (Todd Engineers, 2014), and results from CanESM2, a hot-wet scenario, and CCSM4, a warm-moderately wet scenario, show an increase in recharge in mid-century and end-of century periods in comparison to the historical period, particularly in the mountain front subareas. CanESM2 shows a larger increase in recharge in the end-of-century period compared to the mid-century period (recharge averaged among all sites, except Rose Valley, was 12-percent higher in the end-of-century period; table 7). This increase can be related to the increase in other CanESM2 projected climate conditions, such as the 11-percent increase in precipitation and 8-percent increase in average air temperature (averaged among all sites, except Rose Valley) in the end-of-century period compared to the mid-century period. CCSM4, unlike CanESM2, shows a greater increase (about 5 percent more, when all sites except Rose Valley were averaged) in recharge at mid-century when compared to recharge at the end of the century. Results from the HadGEM2-CC hot-dry scenario indicated a greater variation in projected recharge in the two periods. Results indicated an increase in recharge at mid-century in IWV (as much as 22 percent in the valley floor), except for an 8-percent decrease in the Argus Range (table 7). However, there was a large decrease in recharge (as much as 75 percent in the valley floor) in the end-of-century period, accompanied by a large decrease in the HadGEM2-CC projected precipitation of 32 percent (averaged among all sites, except Rose Valley; table 7).

The BCMv8 model’s interannual variability in recharge and runoff was evaluated using the standard deviation of seasonal (winter: October–March; summer: April–September) means simulated for the historical (1981–2010) and mid-century (2040–69) periods projected by the three GCMs (fig. 12). The standard deviation reveals only the magnitude of the variability and nothing about the seasonality of when water-year extremes generally occur. Results indicate that for all the IWV subareas, there is a large interannual variation in historical (1981–2010) recharge and runoff, particularly in the winter months, as well as a large difference among subareas (fig. 12A). This variability is more significant in subareas with more estimated recharge and runoff like the Sierra Nevada and the volcanic subareas, which increases the differences among subareas for recharge (fig. 12A). Projections from the GCMs indicated a similar variability in recharge and runoff in the mid-century period, with variability being more significant in the winter than in summer. However, the wet scenario from CanESM2 and the dry scenario from HadGEM2-CC showed greater seasonal variation. As figures 12B and 12C show, regardless of the larger magnitude of the CanESM2 wet scenario, both models showed almost double recharge and runoff in winter than in summer, specifically in subareas like the Sierra Nevada and El Paso Mountains, which provide most of the surface runoff and mountain front recharge to IWV. CCSM4, the moderate scenario, showed lower seasonal variations in runoff and recharge (fig. 12D). Given the large seasonal increases projected in the variability of runoff, these results illustrate that runoff also may be a valuable water resource in some years and, if managed strategically, could be used to supplement the groundwater supply.