Snowpack and runoff response to climate change in Owens Valley and Mono Lake watersheds

Abstract

Precipitation from the Eastern Sierra Nevada watersheds of Owens Lake and Mono Lake is one of the main water sources for Los Angeles’ over 4 million people, and plays a major role in the ecology of Mono Lake and of these watersheds. We use the Variable Infiltration Capacity (VIC) hydrologic model at daily time scale, forced by climate projections from 16 global climate models under greenhouse gas emissions scenarios B1 and A2, to evaluate likely hydrologic responses in these watersheds for 1950–2099. Comparing climate in the latter half of the 20th Century to projections for 2070–2099, we find that all projections indicate continued temperature increases, by 2–5 °C, but differ on precipitation changes, ranging from −24 % to +56 %. As a result, the fraction of precipitation falling as rain is projected to increase, from a historical 0.19 to a range of 0.26–0.52 (depending on the GCM and emission scenario), leading to earlier timing of the annual hydrograph’s center, by a range of 9–37 days. Snowpack accumulation depends on temperature and even more strongly on precipitation due to the high elevation of these watersheds (reaching 4,000 m), and projected changes for April 1 snow water equivalent range from −67 % to +9 %. We characterize the watershed’s hydrologic response using variables integrated in space over the entire simulated area and aggregated in time over 30-year periods. We show that from the complex dynamics acting at fine time scales (seasonal and sub-seasonal) simple dynamics emerge at this multi-year time scale. Of particular interest are the dynamic effects of temperature. Warming anticipates hydrograph timing, by raising the fraction of precipitation falling as rain, reducing the volume of snowmelt, and initiating snowmelt earlier. This timing shift results in the depletion of soil moisture in summer, when potential evapotranspiration is highest. Summer evapotranspiration losses are limited by soil moisture availability, and as a result the watershed’s water balance at the annual and longer scales is insensitive to warming. Mean annual runoff changes at base-of-mountain stations are thus strongly determined by precipitation changes.

1 Introduction

Precipitation from the Eastern Sierra Nevada watersheds of Owens Lake and Mono Lake is a main water source, and the one of highest quality, for Los Angeles’ more than 4 million people. Winter precipitation is stored in the snowpack delivering roughly 0.25–0.6 km³ (0.2–0.5 million acre-ft) annually to the city in the dry season via the 340-mile long Los Angeles Aqueduct system, operated by the Los Angeles Department of Water and Power (LADWP). Seven reservoirs built along the aqueducts have a combined storage capacity of about 0.37 km³. Future availability of this water supply source is of critical importance to the city’s growing population and large economy.

Traditionally water-limited, Los Angeles now faces renewed water pressures. These include a Federal Court ruling limiting exports from the Sacramento-San Joaquin Delta; the city’s commitment in the last two decades to environmental mitigation and flow enhancements in Owens Valley and Mono Lake Basin (which have reduced the aqueduct’s average contribution to the city from a historical 2/3 to about 1/3 of total); the contamination of San Fernando valley groundwater; and sustained drought in the Colorado River (City of Los Angeles Water Supply Action Plan 2008). Given these pressures, an understanding of the foreseeable impacts of climate change on the city’s water supplies is vital for long-term planning.

Acknowledging the patterns of climatic and hydrologic changes identified in the Western US (see section 1 in suppl. mat. for a brief summary of literature), the goal of the present study is to quantify the foreseeable range of hydrologic impacts in the Owens and Mono Lake watersheds, given the best current projections of 21st Century climatic conditions. The focus is on the changes in mean hydrologic variables that may result from changes in mean climatic variables. The variables of interest are quantities aggregated spatially over the runoff-producing western slopes of the Mono and Owens watersheds and integrated in time over 30-year time horizons.

The cause-effect relationships between variables defined at coarse spatial and temporal scales can often be described by emergent dynamical relationships that are considerably simpler than at finer scales (e.g., Silberstein 2002; Auger et al. 2008; Wilson 2010; Sivapalan et al. 2011). We identify simple relationships between watershed-wide period mean hydrologic variables based on watershed-wide period mean temperature and precipitation.

2 Approach

The climate scenarios used are obtained from statistically-downscaled (to 1/8° of spatial resolution, approximately 12 km) simulations by 16 of the GCMs (listed in section 3 of suppl. mat.) that formed the basis of the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC 2007; Maurer et al. 2007a). For each GCM, two contrasting greenhouse gas emission scenarios are studied: A2 (higher emissions) and B1 (lower emissions), defined in Nakicenovic et al. (2000) (see also Costa-Cabral et al. this issue). This results in 32 projected time series of daily temperature, precipitation and wind speed, that we used as forcing to the hydrologic model.

The hydrologic model used is the Variable Infiltration Capacity model (VIC; Liang et al. 1994, see section 5 of suppl. mat. for additional VIC literature references). VIC is a distributed grid-based macroscale hydrologic model which represents explicitly the effects of vegetation, topography, and soils on the exchange of moisture and energy between land and atmosphere. VIC includes a full energy balance formulation and a comprehensive, physically-based representation of snow dynamics (Cherkauer et al. 2003; Cherkauer and Lettenmaier 2003). VIC has been previously applied for many climate change assessments (see section 5 of suppl. mat.).

The VIC model was calibrated and evaluated (at 1/8° resolution) using monthly unimpaired (or naturalized) flow data for 1950–2005 provided by LADWP from a network of stream gages at the base of the mountains on the slopes of the Eastern Sierra in the Owens-Mono watersheds (see Figure 5-1 in suppl. mat.). The VIC model was then run at daily time step over the 150-year period from 1950 to 2099, using as input the 32 aforementioned climatic time series, producing a range of plausible future hydrologic outcomes.

While there is a need to provide quantitative information for water resources planning, as done through this work, the underlying projections of climate change are subject to large and unquantifiable uncertainty. The main sources of uncertainty are unknown future emissions of greenhouse gases, uncertain response of the global climate system to increases in greenhouse gas concentrations, and incomplete understanding of regional manifestations that will result from global changes (e.g., Hawkins and Sutton 2010). The hydrologic projections developed in this work should therefore be considered to be plausible representations of the future, given the best current scientific information, and do not represent specific predictions.

3 Projected changes in climate and hydrology over the 21st Century

We express hydrologic volumes as water heights (in mm) averaged over the 2,140 km² simulation area. Values in mm over this area can be converted to million acre-feet (MAF) by multiplying by 1.736 × 10⁻³ MAF/mm.

Figure 1 shows the 32 time series of climatic and selected hydrologic variables, after smoothing by using 30-year running means. Plotting positions are the center of the 30-year period used in averaging. All GCMs, under either scenario B1 or A2, project temperature increases, reaching 2–5 °C by the end of the century. GCMs differ on precipitation, some projecting increases, the majority projecting mild decreases, but all exhibiting marked variability. All GCM runs project rises in the fraction of precipitation falling as rain from the recent historical value of about 0.2, to up to 0.5 (i.e., one-half of all precipitation falling as rain) by end of century. Most GCMs project a marked decline in April 1 snow water equivalent (SWE _A1), a variable used to forecast summer runoff in the Western United States. All GCMs project a shift (by up to 1 month) to earlier dates of arrival of half of the hydrograph’s volume (H ₅₀ ). The higher emission A2 scenario results in warmer temperatures for the end-of-century period, and a wider range for 21st Century precipitation and runoff projections.

Fig. 1

Thirty-year running means of simulated variables over water years 1951–2099, averaged over the Mono-Owens simulation area. Color codes for the GCM, as indicated in the legend of Fig. 2. Averaging over 30-year periods implies that no plotting points exist for the series’ first or last 15 years (1950–1964 and 2085–2099). While the variability of precipitation visibly influences that of runoff, it is the temperature’s long-term rising trend that determines the rise in the fraction of precipitation falling as rain and the earlier arrival of the hydrograph’s center

Figure 2 summarizes the projected changes in climate and hydrology for 30-year horizons. Timing and magnitude are represented in the x and y axis, respectively. For precipitation and runoff, the plotted timing corresponds to the date when half the water-year’s volume has been reached. For temperature, timing is the date when half of the water-year’s sum of deviations from the annual mean has been reached. For A2 but not for B1, several GCMs project a significantly earlier timing of precipitation (by a maximum of 2 weeks), while the average tendency of all GCMs is toward little change in either timing or magnitude. The projected shift in the timing of peak SWE is considerably larger, and is seen for both B1 and A2 and for all time horizons. The SWE peak shift—which for model CNRM for the late 21st Century reaches 1 month earlier under B1 and surpasses 5 weeks under A2—reflects the higher ratio of precipitation falling as rain and the earlier snow melt associated with higher temperatures in the late wet season. Warming also largely explains the projected shift to earlier runoff timing (measured by H ₅₀ ) which, for the late 21st century, reaches about 3 weeks earlier under B1 and over 4 weeks under A2, for several GCMs.

Fig. 2

Cabral-Zucker diagrams of projected changes in timing (x-axis) and magnitude (y-axis) of climatic and hydrologic variables under emissions scenarios B1 and A2, from the historical period 1951–1999 (black dots) to: (top panels) the early 21st Century period (2010–2039); (central panels) mid-21st Century period (2040–2069); and (bottom panels) late 21st-Century period 2070–2099

Daily simulated hydrologic storages and fluxes provide insight into processes not apparent from the 30-year average plots. Figure 3 displays the mean value for each day of the water year of hydrologic fluxes and storages, in two periods that differ greatly in temperature: 1951–1979 and 2070–2099. The three GCMs depicted project similar mean seasonal temperature rises for 2070–2099, but differ in precipitation: no change in precipitation for CCCMA, a 24 % decline for MIROC3 (the largest among GCMs), and a 56 % increase for IPSL (the largest among GCMs). The effects of temperature are seen by comparing the CCCMA results for 2070–2099 against the historical simulation. Lower values are simulated for both storages, SWE and soil moisture. Snowmelt starts earlier, the hydrograph peaks earlier and summer runoff is lower.

Fig. 3

Mean daily values of GCM climatological projections and VIC-simulated hydrologic fluxes and storage quantities for each day of the water year, for: a Historical period 1951–1979 as simulated by CCCMA (noting that all GCMs after BCSD downscaling and bias correction yield comparable simulations for this period); b 2070–2099 according to CCCMA (A2), which projects no significant change in mean annual precipitation compared to 1951–1979; c 2070–2099 according to MIROC3 (A2), which projects the largest decline in mean annual precipitation (−24 %); and d 2070–2099 according to IPSL (A2), which projects the largest increase in mean annual precipitation (+56 %). Projections for temperature rise by 2070–2099 are comparable for all GCMs in (b), (c) and (d). In each figure panel, the vertical axis on the left represents temperature (°C) as well as hydrologic fluxes (mm/day), which include precipitation, sublimation-evapotranspiration (SET) and total runoff (the sum of surface and sub-surface runoff). The vertical axis on the right is used for the hydrologic storages (in mm) soil moisture and snow water equivalent (SWE)

Apparent in Figure 1 is the resemblance in variability of sublimation-evapotranspiration (SET), runoff, and precipitation. Comparing, in Fig. 3a and b, the historical simulation against the CCCMA projections for 2070–2099 (which differs little in mean annual precipitation), SET rises only in spring (May and June). In summer, SET does not rise despite a temperature increase approaching 5 °C. This is explained by the depletion of summer soil moisture seen in Fig. 3b. For MIROC3 (Fig. 3c), the precipitation decline (−24 %) aggravates soil moisture depletion and the mean annual SET dropped by 8 %. Runoff dropped by 38 %. For IPSL (Fig. 3d), the precipitation rise (+ 56 %) feeds the snowpack and elevates soil moisture, allowing SET to rise by 22 %. Runoff rises the most, by 82 %, and several high winter runoff peaks occur, due to the higher fraction of precipitation falling as rain.

Figure 4 shows the annual cycle of daily SET and soil moisture for the same cases depicted in Fig. 3. In all cases, SET increases during spring due to rising soil moisture (from snowmelt) accompanied by higher temperatures, and peaks after the peak in snowmelt runoff. In summer, the depleted soil moisture leads to water-limited SET. By late summer, SET is nearly as low as in winter, despite the much higher temperature. The effect of warming is seen by comparing the CCCMA curve for 2070–2099 to the historical curve. These effects include lower annual minima and maxima of soil moisture, and a soil moisture rise that starts in early fall (instead of early spring), due to the higher rain/precipitation ratio. The effect of differences in precipitation on soil moisture and SET is seen by comparing the three GCM curves on the right panel, which have a similar degree of warming.

Fig. 4

Simulated a historical (1951–1979) and b projected (2070–2099) dependence of average daily SET on soil moisture and season. Each plotted line connects points representing each day of the water year. For example, the point representing August 1 in the left panel represents the average soil moisture and SET for all August 1 in 1951–1979. In (a) we see the dependence of SET on soil moisture, and also its dependence on temperature (which explains the difference between spring and summer SET for fixed soil moisture). All GCMs in the right panel project similar temperature rises by 2070–2099, but differ on projected precipitation changes: −24 % for MIROC3, approximately no change for CCCMA, and +56 % for IPSL

In Fig. 5, we return to the coarse (30-year) temporal scale and show the dependence of six hydrologic variables on climatic variables. All variables are 30-year averages representing the runoff-producing areas of the Owens-Mono watersheds, and the over-bar indicates their aggregate nature. Bivariate regression is used to derive relationships between the independent climatic variables $\overline T$ and $\overline P$ and the dependent hydrologic variables, yielding equations that apply to the ($\overline T$, $\overline P$) space sampled by the 32 GCM runs (see Figure 4-1 of suppl. mat.). In the equations below, $\overline T$ is expressed in °C, $\overline {{R_p}}$ is non-dimensional, ${\overline H_{{50}}}$ is in days, and all other variables are in mm/year. $\overline {SM}$ denotes mean soil moisture content, given in mm.

Fig. 5

Dynamical relationships between climatic and hydrologic variables that emerge at large scales. All variables are averaged over 30-year periods and over the simulated area of the watersheds. The relationships apply to the range of $\overline T$ and $\overline P$ values sampled in the 32 projected scenarios studied. In panel (f), arrow thickness is proportional to the value of the partial correlation coefficient, annotated next to it. For example, for $\overline {SW{E_{{A1}}}}$, the linear regression based on $\overline T$ alone has R ² = 0.410, and adding $\overline P$ to the model contributes with an additional $r_{{\overline {SW{E_{{A1}}}}, \left. {\overline P } \right|\overline T }}^2 = 0.543$, yielding R ² = 0.953 for the bivariate regression reported in panel (c). Conversely, if we have a linear regression model of $\overline {SW{E_{{A1}}}}$ based on $\overline P$ (R ² = 0.689), then adding $\overline T$ contributes with $r_{{\overline {SW{E_{{A1}}}}, \left. {\overline T } \right|\overline P }}^2 = 0.264$, yielding R ² = 0.953 for the bivariate regression. The influence of $\overline T$ decreases from Eqs. (1) through (5) (panels (a) through (e)), while the influence of $\overline P$ increases

$$ \overline {{R_p}} = 0.0071 {\overline T^2} + 0.179 $$

(1)

with R ² = 0.907. The 95 % confidence interval for the regression slope is [0.0067, 0.0075].

$$ \overline {{H_{{50}}}} = 0.032 \overline P - 6.739 \overline T + 249 $$

(2)

with R ² = 0.953. The 95 % confidence intervals are [0.027, 0.036] for the $\overline P$ regression slope and [−7.012, −6.466] for the $\overline T$ regression slope. Temperature is the main determinant, and the linear regression of $\overline {{H_{{50}}}}$ on $\overline T$ alone has coefficient of determination R ² = 0.898. The addition of $\overline P$ as an explanatory variable has a low partial correlation coefficient, $r_{{\overline {{H_{{50}}}}, \overline P |\overline T }}^2 = 0.060$, but improves $\overline {{H_{{50}}}}$ prediction for low $\overline P$ values.

$$ \overline {SW{E_{{A1}}}} = 0.691 \overline P - 28.146 \overline T + 11 $$

(3)

with R ² = 0.955. The 95 % confidence intervals are [0.656, 0.726] for the $\overline P$ regression slope and [−30.215, −26.082] for the $\overline T$ regression slope. Linear regression of $\overline {SW{E_{{A1}}}}$ on $\overline P$ alone, its main determinant, has R ² = 0.691, and the addition of $\overline T$ has a partial correlation coefficient $r_{{\overline {SW{E_{{A1}}}}, \overline T |\overline P }}^2 = 0.264$.

$$ \overline {SM} = 0.300 \overline P - 3.775 \overline T + 100 $$

(4)

with R ² = 0.982. The 95 % confidence intervals are [0.293, 0.308] for the $\overline P$ regression slope and [−4.220, −3.329] for the $\overline T$ regression slope.

$$ \overline Q = 0.871 \overline P - 179 $$

(5)

with R ² = 0.992. The 95 % confidence interval for the regression slope is [0.857, 0.885] and the 99 % confidence interval is [0.852, 0.889].

$$ \overline {SET} = 0.129 \overline P + 179 $$

(6)

where Eq. (6) is derived from Eq. (5) given the approximate equality for multi-year periods, $\overline Q + \overline {SET} = \overline P$. Direct linear regression of $\overline {SET}$ on $\overline P$ also yields Eq. (6), with the same slope and intercept values, with R ² = 0.721.

The dependencies expressed by these six equations represent dynamic relationships that emerge at the coarse temporal and spatial scale of these variables. Such relationships differ from those that rule finer-scale dynamics. In particular, the dependence of SET on temperature, seen at the seasonal scale (which, in Fig. 4, explains the difference in SET between spring and summer for a given soil moisture storage value), is not present at this multi-annual scale—or at the annual scale, as shown below. The explanation is that warming leads to a decline in $\overline {SW{E_{{A1}}}}$, the source of summer soil moisture, and summertime evapotranspiration becomes water limited. Thus, warming has two opposing effects on evapotranspiration, one through the rise of potential evapotranspiration, especially significant in summer; the other through the decline of summer soil moisture. The coefficient of determination for $\overline {SET}$ based on $\overline T$ is less than 0.1 for any GCM run; and based on $\overline P$ is 0.62–0.84 depending on the GCM and scenario.

Figure 6 shows water-year runoff plotted against precipitation for every individual water year that was simulated under emissions scenario A2 for the 149-year period 1951–2099 (small colored points). Observation-based annual data are also shown (in red). The observed ranges were 302–1,064 mm/y for $\overline P$, 169–682 mm/y (or 0.290–1.171 million acre-ft/y) for $\overline Q$, and 0.29–3.01 °C for $\overline T$. Equation (7) (see plotted line in Fig. 6) was fitted to the ensemble of data that includes not only these runs but also the corresponding runs for emissions scenario B1. Equation (7) captures the behavior of the ensemble of runs and is not intended to provide a predictive model for any individual GCM run. The precipitation and runoff in water year i are respectively denoted p _i and q _i .

$$ {q_i} = 0.707{p_i} - 79 $$

(7)

with R ² = 0.928.

Fig. 6

Main plot: The small colored points indicate runoff and precipitation for individual years, where runoff is simulated by VIC forced by a GCM (indicated by color) under emissions scenario A2 for 1951–2099. The observations-based dataset (water years 1951–1999) is in red, and the VIC-simulated runoff for that same period is in yellow. Also shown (in black) are the period means, which follow the linear relation in Eq. (5). Upper left inset plot: The period means, plotted in black in the main figure, are shown in color to indicate the emissions scenario and time period (similar to Fig. 5e). Lower right inset plot: The runoff anomaly, i.e. the deviation from Eq. (7), calculated from each colored point in the main graph, is plotted against the preceding year’s annual precipitation, revealing a log-linear relationship

Figure 6 shows (as black dots) the simulated 30-year means, and the observed 1951–1999 mean. These same data are plotted again in the top-left inset plot (similar to Fig. 5e), with color indicating the scenario and time horizon. The slope in Eq. (5) is steeper that in Eq. (7) due to the tendency for a multi-year period with high total precipitation to exhibit a high runoff ratio and fall above the Eq. (7) line, and the reverse tendency for low precipitation, owing to the effect of soil moisture antecedent conditions. The influence of the preceding year’s precipitation total on the current year’s runoff deviation from Eq. (7) is displayed in the lower-right inset plot in Fig. 6.

4 Discussion and conclusions

We have quantified the range of hydrologic impacts in the Owens and Mono Lake watersheds that may result from 32 plausible scenarios of 21st Century climatic conditions projected by 16 GCMs under greenhouse gas emissions scenarios B1 and A2. The VIC hydrologic model was used to obtain 32 daily hydrologic projections through year 2099. While VIC simulates complex relationships among daily variables, once their values are aggregated in time (30-year periods), simple relationships emerge relating each hydrologic variable to climatic variables. The emergent model (summarized in Fig. 5f) offers conceptual intelligibility of the watershed’s response to climatic changes.

For the variables averaged in space over the runoff-producing areas of the Owens-Mono watersheds, and averaged in time over 30-year periods, our findings for the late 21st Century (2070–2099) are summarized below. These findings apply only to the region of ($\overline T$, $\overline P$) space sampled by the 32 GCM runs, and neglect any effects that might arise from future changes in land cover, including changes in response to climate.

• All GCMs project increases in temperature ($\overline T$), by 2–5 °C, by the end of the century. Projected changes in precipitation ($\overline P$) vary from −24 % to +56 %, with most GCMs projecting small changes.

• The mean fraction of precipitation falling as rain ($\overline {{R_p}}$) increases approximately with the square of mean annual temperature (${\overline{T}^2}$) according to Eq. (1). By the end of the century, $\overline {{R_p}}$ is projected to rise from the recent historical value, estimated as 0.19, to values in the range 0.26–0.52.

• The average date when half of the water-year’s hydrograph volume has passed ($\overline {{H_{{50}}}}$) moves to earlier dates with $\overline T$ and later dates with $\overline P$, according to Eq. (2). All GCMs project an earlier arrival of the hydrograph, with a timing shift varying from 9 to 37 days.

• $\overline {SW{E_{{A1}}}}$ increases with $\overline P$ and decreases with $\overline T$, according to Eq. (3). Response to $\overline T$ corresponds to an estimated watershed-averaged decline by 28 mm (corresponding to 0.048 MAF) per 1 °C warming, absent $\overline P$ changes. Precipitation is the dominant determinant of $\overline {SW{E_{{A1}}}}$ variability. GCM projections range from little change in $\overline {SW{E_{{A1}}}}$ (such as for IPSL (A2) which projects the largest precipitation increase, by 56 % at the end of the century) to a decline by about 70 % (for MIROC3 (A2), which projects the largest precipitation decline, by 24 % by the end of the century)..

• The decline of $\overline {SW{E_{{A1}}}}$ and $\overline {{H_{{50}}}}$ contributes to summer soil moisture depletion, limiting $\overline {SET}$ and offsetting the increase in potential evapotranspiration associated with warming. The result is a lack of response of $\overline {SET}$ to $\overline T$. Instead, $\overline {SET}$ is shown to vary linearly with $\overline P$, and as a result so too does runoff ($\overline Q$) (Eqs. (5) and (6)).

• The response of $\overline Q$ to $\overline P$, described by Eq. (5) differs from the response of annual runoff (q _i ) to that year’s precipitation (p _i ), described by Eq. (7), owing to the influence of $\overline P$ on mean soil moisture conditions. According to Eq. (5), a future rise or fall in $\overline P$ by 100 mm/y (i.e., by 17 % of the historical 578 mm/y) originates a rise or fall in $\overline Q$ by about 87 mm/y (i.e., by 26 % of the historical 333 mm/y), regardless of $\overline T$. Compared to the change in $\overline P$, the change in $\overline Q$ is smaller in absolute value but larger as a percentage—a consequence of the negative intercept in Eq. (5).

Our estimate for these watersheds for a decline in $\overline {SW{E_{{A1}}}}$ by a watershed-averaged 28 mm per 1 °C rise in $\overline T$ absent any change in $\overline P$ (from Eq. (3)) is in agreement with the observations-based estimate for these watersheds by Howat and Tulaczyk (2005) of <50 mm/°C. Our modeled watersheds have high mean elevation (close to 3,000 m), and thus $\overline {SW{E_{{A1}}}}$ is expected to be relatively insensitive to all but the highest temperature increases (Maurer et al. 2007b), since much of the basin area will remain cold enough to support Spring snowpack. Howat and Tulaczyk (2005) also found that $\overline P$ rather than $\overline T$ has the dominant influence on $\overline {SW{E_{{A1}}}}$ in this region, and our results agree, with partial correlation coefficients $r_{{\overline {SW{E_{{A1}}}}, \left. {\overline P } \right|\overline T }}^2 = 0.543$ and $r_{{\overline {SW{E_{{A1}}}}, \left. {\overline T } \right|\overline P }}^2 = 0.264$. Our results further indicate that none of the 32 future climate scenarios considered here lead to a temperature-dominated $\overline {SW{E_{{A1}}}}$ dependence by the end of the century, because Eq. (3) holds for the entire range of $\overline T$ and $\overline P$ seen in all 32 scenarios.

Our results for $\overline {SET}$ and $\overline Q$ agree qualitatively with results previously obtained for eastern Sierra Nevada locations, e.g. by Jeton et al. (1996) and Lundquist and Loheide (2011). Using the PRMS model and prescribed scenarios of climate change for the East Fork Carson River (Eastern Sierra), Jeton et al. (1996) found that “[a]nnual total SET responds only weakly to the imposed changes in mean temperature and PET [potential sublimation-evapotranspiration]…” and that “[a]s a consequence, the annual total streamflow is relatively insensitive also.” Subsequent modeling-based studies concurred (see references in Lundquist and Loheide 2011), and the observational-based study of Lundquist and Loheide (2011) concluded that in the Sierra Nevada “…annual ET is limited more by moisture availability than energy availability.” While this relative insensitivity of mean runoff to energy (as expressed by temperature) exists in other water-limited and snow-dominated areas (e.g., Barnett et al. 2005), a considerably greater sensitivity to temperature is observed in many regions (Vogel et al. 1999), especially more humid areas (Milly and Dunne 2002).

The timing of hydrologic fluxes within the water year has well-recognized implications for water resources management, in particular for the sizing of artificial storage for summer supply. Less widely acknowledged is the role of timing on the watershed’s annual water balance, particularly in presence of marked climate seasonality and limited soil water storage capacity (Milly 1994). In the Owens-Mono watersheds, warming shifts hydrograph timing to earlier dates by increasing the fraction of precipitation falling as rain, reducing the volume of snowmelt, and initiating snowmelt earlier. These time shifts are responsible for the depletion of summer soil moisture, which limits summer evapotranspiration losses, and renders mean annual runoff insensitive to warming. Thus, complex dynamics operating at fine temporal scales give rise to simple emergent dynamics over longer time periods.