GRACE storage-streamflow hystereses reveal the dynamics of regional watersheds

This discussion paper has been under review for the journal Hydrology and Earth System Sciences (HESS). Please refer to the corresponding final paper in HESS. The published article is copyrighted by the author(s) and published by Copernicus Publications on behalf of the European Geosciences Union. The published article can be found at: http://hdl.handle.net/1957/57160


Introduction
At the most fundamental level, watershed processes can be described as the collection, storage, and release of water (Black, 1996;McDonnell et al., 2007).At a more complex level, watersheds typically function as non-linear, dynamic systems governed by their unique climate and geology (Kirchner, 2009).Gaining insights into hydrologic processes and behaviors helps to provide a process-based understanding of watersheds as dynamic environmental systems (Aspinall, 2010), and to identify connections that advance hydrologic science and hydrologic prediction (Wagener et al., 2007).At streamflow and water stored in a watershed as snow, soil moisture, groundwater and reservoirs (Appleby, 1970;Brutsaert, 2008;Kirchner, 2009;Sayama et al., 2011).These four primary storage components, along with landscape and topography, govern the fluxes of water through a catchment, and play an important role in the hysteretic nature of storage and runoff dynamics (McGlynn and McDonnell, 2003;McNamara et al., 2011).Knowledge of these processes is fundamental to developing an understanding of a watershed's hydrologic behavior.However, observations over larger regions can be technically challenging and costly, and in situ measurements from small basins do not necessarily represent the complexity inherent to watersheds at more broad scales.This scaling problem limits our capacity to understand and predict regional hydrologic processes, which is often the practical scale of watershed management (Blöschl, 2001;Western et al., 2002;Skøien et al., 2003;Peel and Blöschl, 2011;Thompson et al., 2011).
In the absence of large scale observations, past hydrological studies have typically relied on in situ measurements as a proxy for regional scale hydrological processes.
For example, in higher latitude or mountainous regions measurements of snow water storage have provided a simple metric that has been used in water resource planning for decades (Cayan, 1996;United States Army Corps of Engineers, 2001), and are often correlated to streamflow gauged downstream (Dozier, 2011).While informative, this approach can often provide hydrological forecasts that are misleading, because point-based measurements do not fully represent the broad-scale variability of rugged mountain terrain (Dozier, 2011;Nolin, 2012;Webster et al., 2014;Ayala et al., 2014).Similarly, measurements of soil moisture in the upper 2000 mm of the soil rely on point-based data that are often distributed at the regional scale, but do not effectively represent the true variability of soil moisture found at the regional scale (Western et al., 2002;Brocca et al., 2010).A complete understanding of groundwater stores and fluxes (deeper than 2000 mm) at regional scales also remains elusive, despite its increasing importance in water resources management (Wagener et al., 2007;Gleeson et al., 2012;Famiglietti and Rodell, 2013;Barthel, 2014).In addition to contributing Introduction

Conclusions References
Tables Figures

Back Close
Full to streamflow, groundwater serves as an important water resource for consumptive use (Gleeson et al., 2012).While local-scale methods have been applied with moderate success in the past, current trends in climate and in consumptive water demand suggest that long-term changes in hydrological fluxes will have a major impact at the regional scale (Milly et al., 2008).As a result, the supply and demand of water is also expected to shift, especially at the regional scale (Wagener et al., 2010;Gleick, 2014a).Hydrologic models can help address the questions of scale and bridge the gap between local scale observations and regional-scale processes by estimating the primary components of water storage (snow, soil moisture, reservoir, and groundwater) across a larger spatial grid.Regional-scale modeling approaches are integrated into water resource management operations for navigation, human consumptive use, irrigation, and hydropower (Payne et al., 2004;Rodell et al., 2004).Models can also be applied diagnostically to test scientific hypotheses and provide a better understanding of the physical processes that govern real world systems, such as the connections between snowmelt, streamflow, and groundwater (Beven, 2007(Beven, , 2010;;Moradkhani and Sorooshian, 2008;Kirchner, 2009;Clark et al., 2011;Capell et al., 2012).Despite their utility, developing and validating a model can be both time consuming and reliant on multiple data inputs, which even in the most well-instrumented basins provides sparse geographic coverage (Bales et al., 2006;Zang et al., 2012).The lack of an integrated measurement of water storage and streamflow has limited regional-scale hydrologic insights to model-based studies (Koster et al., 2010;Mahanama et al., 2011).
Since 2002, broad-scale measurements of changes in the amount of water stored across and through the earth have been available from NASA's Gravity Recovery and Climate Experiment (GRACE) satellites (Tapley et al., 2004).GRACE measures monthly changes in the Earth's gravitational field that are proportional to regional changes in total water storage (Wahr et al., 2006).GRACE satellites provide a monthly record of terrestrial water storage anomalies (TWSA), which represent the changes in the vertical sum of water at the Earth's surface stored in snow, surface, soil and groundwater.Water losses to runoff and evapotranspiration are implicit in the GRACE Figures

Back Close
Full storage signal, removing the added layer of complexity typically required to model the terrestrial water balance.GRACE data, coupled with modeled and measured variations of water stored in snow, surface reservoirs and soils, have successfully been decomposed to quantify regional groundwater changes (Rodell et al., 2009;Famiglietti et al., 2011;Voss et al., 2013;Castle et al., 2014) and have contributed to improving water balance calculations (Zaitchik et al., 2008;Li et al., 2012).More recent efforts have quantified the relationship between regional water storage and specific streamflow events (Reager et al., 2014), and have described regional storage-streamflow hysteresis for large basins (Riegger and Tourian, 2014).Although these previous studies have provided new insights into regional watershed hydrology, their analyses are more diagnostic in nature and did not explore the processes behind the observed behavior.
In this paper, we use GRACE observations of terrestrial water storage observations to expand upon a fundamental concept in watershed hydrology -that the temporal relationship between storage and runoff can be used to quantify complex watershed behavior at broad scales, including groundwater recharge amounts and timing, baseflow recession characteristics, and long lead-time streamflow prediction (Brutsaert, 2008;Sayama et al., 2011;Reager et al., 2014;Riegger and Tourian, 2014).The temporal relationship between coincident TWSA and discharge observations at three scales in the Columbia River Basin (CRB) of western North America is investigated using climate and geology as a framing principle to describe the shape of the storage-streamflow hysteresis.We associate regional and temporal differences in the hystereses with varying watershed dynamics.Finally, we compare the prognostic abilities of GRACE observations to individual modeled estimates of snow and soil moisture to predict streamflow at regional scales.Introduction

Conclusions References
Tables Figures

Back Close
Full

Study area
Our study area is the Columbia River Basin (CRB; 41-53 • N and 110-122 • W).This basin has wet winters, with up to 70 % of annual precipitation falling between November and March, 50-60 % of which occurs as snow (Serreze et al., 1999;Nolin et al., 2012).The spring months (April to June) are also wet, but warmer.Precipitation during the spring combines with snowmelt to swell rivers and potentially exacerbate flooding.Snowmelt also serves as a critical component of the hydrologic cycle recharging aquifers and filling streams later in the year.These contributions bridge the temporal disconnect between wet winters and dry summers when demand is at its peak as farmers, fish, hydropower and municipal users vie for over-allocated water resources (United States Army Corps of Engineers, 2001;Oregon Water Supply and Conservation Initiative, 2008).However, concerns with winter surplus and summer scarcity are not uniform across the CRB, since climate and geology vary greatly.Two of the study watersheds, the Upper Columbia (155 000 km 2 ) and the Snake River basin (182 000 km 2 ), represent distinctly different climatic and geologic provinces of the CRB (described and illustrated in Fig. 1).The Upper Columbia is wet and is characterized by steep topography of fractured rock and poor groundwater storage.In contrast, the arid Snake River basin is bowl-shaped with mountains on three sides.The interior of Snake River basin is a broad plain with well-developed soils and high aquifer storage.The Columbia River at The Dalles (614 000 km 2 ) encompasses the Upper Columbia and the Snake River sub-basins, and its climate and geology are an integration of the two (Fig. 1).A distinct climatic feature of the Columbia River at The Dalles is the western slope of the Cascade Mountains where over 3000 mm of mean annual precipitation at higher elevations sustains a considerable seasonal snowpack.The scale of this study was constrained to watersheds larger than 150 000 km 2 , the optimal minimum geographic limit of GRACE data (Yeh et al., 2006;Landerer and Swenson, 2012).Introduction

Conclusions References
Tables Figures

Back Close
Full

Methods and data
We used 108 months of GRACE and streamflow data over nine water years (WY; October-September;2004-2012).This data comprises positive, neutral, and negative phases of the El Niño-Southern Oscillation and negative and positive phases of the Pacific Decadal Oscillation (Feng et al., 2014;Iizumi et al., 2014).As a result, the data provides years of above-and below-average precipitation, snowpack, and streamflow for the region.The three watersheds were delineated upstream from United States Geological Survey (USGS) stream gages at 1 • resolution, which is the resolution of GRACE data (described below).In the CRB, these grid cells represent a dimension of approximately 80 by 120 km.The Upper Columbia consists of the area upstream of the Columbia River at the International Boundary gage (USGS 12399500), just downstream of the confluence of the Columbia and Pend-Oreille Rivers.The Pend-Oreille is a major watershed in the upper portions of the CRB.The Snake River gage at Weiser (USGS 13269000) provides gauged streamflow data above Hell's Canyon Reservoir, the largest impoundment in the Snake River basin.The USGS gage at The Dalles (USGS 14105700) provides the most downstream streamflow data for the CRB.
Monthly mean runoff (Q; mm) was calculated for each of the three gages using the USGS streamflow data.Measurements of TWSA were obtained from the GRACE RL-05 (Swenson and Wahr, 2006;Landerer and Swenson, 2012) data set from NASA's Tellus website (http://grace.jpl.nasa.gov).The errors present in the gridded GRACE data exist primarily as a result of truncation (i.e., a low number of harmonics) in the spherical harmonic solution, and smoothing and systematic noise removal (called "de-striping") that is applied after GRACE level-2 processing to remove spatially correlated noise (called "stripes") (Swenson and Wahr, 2006).This smoothing tends to smear adjacent signals To restore the GRACE signal lost during processing, the data were scaled using 1 • Land-Grid Scale Factors produced by putting a 1 • land surface model through identical processing (truncation and filtering) as the GRACE solutions, then measuring the decrease in the signal amplitude at each 1 • grid.These procedures are described on the Tellus website and detailed in Landerer and Swenson (2012).Monthly 1 • GRACE estimates of TWSA, and the associated 1 • leakage and measurement errors, were spatially averaged over each of the three study watersheds following the procedures described in the Tellus website.GRACE represents monthly storage anomalies relative to an arbitrary record-length mean value, analogous to the amount of water above or below the long-term mean storage of a bucket, and should balance with the equation: where all components are at monthly time steps; GW represents groundwater, SM represents soil moisture (from 0-2000 mm depth), SWE represents snow water equivalent (the equivalent depth of water held in snowpack), and RES represents reservoir storage.The ∆ used here represents the anomaly from the study-period mean, rather than a monthly change.To isolate monthly groundwater storage anomalies (∆GW or GWSA) in the above equation, ∆SM, ∆SWE and ∆RES estimates were subtracted from the monthly TWSA data using methods described in Famiglietti et al. (2011).Monthly SM values over the study basins were obtained from the mean of the North American and Global Land Data Assimilation Systems (NLDAS at 1/8 • resolution (Cosgrove et al., 2003) and GLDAS at 1/4 • resolution (Rodell et al., 2004), respectively), and were spatially averaged over the three study watersheds.were calculated for the five largest reservoirs in the CRB (see Appendix A).Other smaller reservoirs in the CRB were excluded when it was determined that fluctuations in their levels were below the detection limits of GRACE.Like all measurements, estimates of TWSA from GRACE contain error.For all of the study basins, the range of error is well below the TWSA signal strength, approximately an order of magnitude below the annual amplitude (200-300 mm) of the TWSA signal in the CRB.The basin-averaged TWSA errors (time invariant) for the three study basins are 37 mm (Upper Columbia), 22 mm (Snake), and 25 mm (The Dalles), and are plotted as bounds on the TWSA time series in Fig. 2a-c.Calculation of the error in individual terms followed standard methodologies (Famiglietti et al., 2011), where error in SM is the mean monthly standard deviation, and standard errors for SWE and RES are 15 % of mean absolute changes.Groundwater anomaly error is the calculated as the sum of basin-averaged errors (added as variance) in the individual terms in the calculation of ∆GW (Eq.1), including the error in TWSA (Swenson et al., 2006).The basin-averaged error variance for GWSA (time invariant) in the three study basins are 45 mm (Upper Columbia), 26 mm (Snake), and 33 mm (The Dalles), and are plotted as bounds on the GWSA time series in Fig. 2d-f.The individual components (SM, SWE, RES respectively) for each basin are Upper Columbia (24, 6, 0.01 mm), Snake (14, 3, 0.01 mm), and The Dalles (21, 4, 0.01 mm).Note that these error estimates are distributed across an entire regional watershed and do not represent the error at individual monitoring sites.
Based on an approach similar to Reager et al. ( 2014) and Riegger and Tourian (2014), we plotted the temporal relationship between TWSA and Q to examine hysteresis relationships in all three of the study watersheds for each individual water year and for the monthly mean across all water years.Expanding from the integrated terrestrial component of water storage, we also plotted the relationship between GWSA and Q.We examined the branches of these hysteresis plots to measure basin groundwater recharge amount and timing, storage peak amount and storage peak month.In order to verify groundwater hysteresis, we compared the GRACE-derived GWSA Introduction

Conclusions References
Tables Figures

Back Close
Full to groundwater depths from well measurements at 33 sites throughout the study region (Fig. 1 and Appendix A).These data were normalized by their standard deviation, and the mean of the 33 wells was calculated.The standard deviation of the GRACEderived GWSA for The Dalles was normalized to provide a direct comparison of GWSA and in situ measurements.
We further hypothesized that spring TWSA could predict Q later in the year for individual months and after peak SWE accumulation (April-September).To test this prognostic hypothesis we used a two-parameter power function (The MathWorks, 2013) to evaluate the ability of TWSA in March and April to predict cumulative Q from April-September (Q season ) and for August (Q August ), the low-flow month when demand is near its peak.Additionally, we tested and compared the modeled-values of SWE and soil moisture from NLDAS, GLDAS, and SNODAS to predict Q season and Q August using the same power-function analysis.

Storage-discharge hysteresis plots
The filling and emptying of the study basins at the regional-scale over the course of an individual WY results in a hysteretic relationship between storage and runoff (Fig. 3a).The hysteresis loops begin at the onset of the wet season in October, with TWSA increasing (Figs.3a and 4a-c) as precipitation is stored as snow and soil moisture.An increase in storage that is not offset by an increase in discharge indicates a predominance of snow inputs and the freezing of soil water.The lower branch of the hysteresis plot (storage increase unmatched by streamflow) can be used to estimate cumulative snow water equivalent and soil moisture in the basin.This is the water that later contributes to streamflow and groundwater recharge in the spring.
The hysteresis shifts direction from February-April (inflection 1, Fig. 3a) when saturated soils and snow melt cause Q to rapidly increase.Each hysteresis loop contains Introduction

Conclusions References
Tables Figures

Back Close
Full a vertical branch of the curve during which storage is relatively constant, but streamflow increases rapidly.This also represents the groundwater recharge branch of the loop.
As snow melts and the ground thaws, runoff is generated, recharge into soils occurs, and basins tend to be at peak storage during this branch.Storage losses and additional precipitation inputs during this period are re-organized internally.A second shift (inflection 2, Fig. 3a) occurs from April-June when peak TWSA begins to decrease, representing spring snowmelt and a switch from precipitation that falls primarily as snow to rain; these combine to contribute to peak Q.
Once peak Q values are reached, the loop shifts direction a third time (inflection 3, Fig. 3a), receding on both axes as contributions from snowmelt diminish while groundwater sustains streams and provides a source for irrigated agriculture.During this period, the relationship between TWSA and discharge is linear, corresponding to baseflow-driven runoff processes in which each monthly change in storage causes a proportional change in the generation of streamflow.
The hysteresis plots of TWSA-Q for an individual water year demonstrate that the timing and quantity of precipitation governs the size of a hysteresis loop for an individual WY (Figs. 3a, 4a-c, and 5).For instance wet years (e.g., 2008) have bigger loops, while dry years (e.g., 2005) are more compressed along both axes.However, the general shape of the loops is distinct for each basin.Plotting multiple WYs provides a family of curves for each basin that helps describe the how climate and geology governs the timing and magnitude of the relationship between TWSA and Q (Figs. 3a-c and 5).

Individual basin hysteresis plots
Of the three study basins, the Upper Columbia is the most hydrologically responsive, with a mean range of 210 mm for TWSA and 50 mm for Q.The Upper Columbia's steep topography and wet climate fills the limited aquifer quickly, reaching a storage threshold and generating runoff.The steep topography moves snowmelt and rain quickly through the terrestrial system and into the river channel until cresting in June (Figs.4 and 5), followed by declines in TWSA and Q from June-September.Introduction

Conclusions References
Tables Figures

Back Close
Full In contrast to the rapid response of the Upper Columbia, the Snake River receives ∼ 60 % less annual precipitation, but has an annual TWSA range that is only 26 % less (160 mm; Figs. 4 and 5).However, the hysteresis loops for the Snake River are collapsed (mean annual range in Q = 7 mm).The climate and geology of the Columbia River at The Dalles (Figs. 4 and 5) are an integration of the Upper Columbia and Snake River, seen in the shape of the hysteresis loops (mean ranges TWSA = 180 mm; Q = 25 mm).The period from February-June more closely resembles the Snake River basin, with gradual increases in TWSA and sharp increases in Q.The slope of the recession from June-September has the same general shape for The Dalles as the Upper Columbia (Fig. 4a and c), presumably from snowmelt-generated runoff.

Groundwater-discharge hysteresis plots
The hysteresis loops describing the temporal relationship between GWSA and Q are equally informative, with one dramatic difference -they temporally progress in opposite directions of the hysteresis loops of TWSA and Q (Fig. 3).For all three watersheds, GWSA decreased from October-February/March (Fig. 4d-f), and does not shift towards positive gains until early spring and the initial stages of melt before reaching its maximum in June.The 33 point-specific well data show considerable individual variability throughout a water year, and the mean of the mean of the standard deviations was close to zero for all months (Fig. 6).

Streamflow forecasting
We next present how TWSA was applied prognostically to predict streamflow.TWSA March can predict the total Q from April through September (Q season ) in all three basins with an R 2 range of 0.83-0.98 and a mean R 2 (R 2 ) of 0.91 (Fig. 7a, Table 1).Applying TWSA for April also provided similar results, but with a lower degree of skill in predicting Q (range = 0.75-0.92,R 2 = 0.86).Introduction

Conclusions References
Tables Figures

Back Close
Full TWSA March also served as a good indicator of August runoff (Q August , range = 0.68-0.88,R 2 = 0.77).While this overall mean is high, the range of agreement between basins was not uniform (Fig. 7b).The skill of TWSA March in the Snake River (R 2 = 0.68) was considerably lower than the Columbia River at The Dalles (R 2 = 0.88).
Snowpack and soil moisture play a considerable role in the hydrology of the CRB and are commonly used to help predict water demand and availability later in the year (Koster et al., 2010).We compared the predictive capabilities of the modeled snow (SWE) and soil moisture (SM) products to the GRACE TWSA data (Table 1).Compared to SWE, TWSA March provided a better indicator of seasonal and August runoff in the Upper Columbia and at the Dalles across all nine years (Fig. 6).In the Snake River, SM provided a slightly higher degree of skill than TWSA March in predicting Q season and Q August .However in the Upper Columbia and The Dalles, SM provided inferior predictive skill for Q season and Q August as compared to TWSA.

Discussion
Decades of data collection and monitoring at individual gage sites indicate that watersheds collect, store and release water.Using one integrated measurement from the GRACE satellites our results show these same process at the regional scale in the hysteresis loops of storage (TWSA) and runoff (Q).While hystereic processes have previously been identified in local-scale measurements (McDonnell, 2003;McGlynn and McDonnell, 2003), only recently has streamflow-storage hysteresis been measured at the regional scale (Riegger and Tourian, 2014).
Our work builds on Riegger and Tourian's (2014) results, and employs GRACE data to describe how regional watersheds function as integrated, non-linear systems governed by climate and geology.Climate controls the size of the hysteresis loops by providing a first-order control on hydrologic inputs and the storage of solid water, which in turn governs the ranges of TWSA and Q.However, runoff response to precipitation Introduction

Conclusions References
Tables Figures

Back Close
Full and snowmelt does not act independently from geology (Jefferson et al., 2008;Tague et al., 2008), which controls how liquid water is stored and routed through a watershed, even at the regional scale.This in turn helps govern the shape of a watershed's hysteresis curve.For example, basin steep slopes and fractured bedrock geology in the Upper Columbia cause snowmelt and precipitation to run off quickly, and a relatively small amount is retained in the soil.This phenomena is seen in the more open shape of the hysteresis curve in the Upper Columbia -once maximum TWSA is reached in March, it quickly transitions to runoff (Fig. 4a).These data suggest that this is a watershed where snow storage is the primary component of water storage, which is also reflected in the comparison of SWE and SM data (not shown).
In contrast, the arid and groundwater dominated Snake River basin results provide a very different family of hysteresis curves (Figs. 4 and 5) that are compressed vertically (Q) as compared to the Upper Columbia basin, despite showing a similar intra-annual range horizontally (TWSA).Another distinction is that the onset of spring melt runoff in February does not deplete TWSA at the Snake River.Instead, TWSA continues to increase until May, when peak runoff occurs.These data support the conceptual model of a watershed that retains comparatively more winter precipitation in soils and aquifers throughout the spring season, and that drains to sustain flow later in the year.
The greater Columbia River Basin upstream from The Dalles integrates the climatic and geologic characteristics of the Snake River and Upper Columbia as well as other areas within the CRB.The western slope of the Cascades, which is outside of the Upper Columbia, accumulates up to several meters of SWE each winter.The region due east of the Cascades is characterized by a plain underlain by basalt that provides excellent aquifer storage and helps dampen the snowmelt pulse in the spring.The hysteresis loops for the CRB reflect these combined characteristics, where the onset of melt in February produces a pronounced increase in runoff similar to the Upper Columbia.However, as with the Snake River, increases in measured TWSA are found in the CRB through April.Introduction

Conclusions References
Tables Figures

Back Close
Full We applied these climatic and geologic insights to develop and test the hypothesis that spring TWSA could predict Q later in the year, based on two observations: First, the shapes of the hysteresis curves for each basin are similar (Figs.4a-c and 5), but vary by magnitude of annual TWSA.Second, peak TWSA occurs before the peak runoff.We show that the integrated GRACE signal is a good baseline measurement in predicting seasonal streamflow across a range of water years with regards to precipitation and streamflow.In essence, our data suggest that the water stored across and through the Columbia River Basin in March describes the water available for the remainder of the water year.In the CRB and in the northwestern United States, snowpack is commonly used as a metric for predicting spring runoff.Despite its importance to the hydrologic cycle of the region, measurements of TWSA March from GRACE provide a better prediction of seasonal and August runoff than model-derived estimates of snowpack.GRACE TWSA March also provided a better prediction for runoff than soil moisture, except for the Snake River basin.There March soil moisture provided a better indicator of runoff for the rest of the year, although TWSA March had a similar accuracy (Table 1).Despite a relatively short data record, the years of our study represent a wide-range of conditions with regards to climate and streamflow, which is captured in our models and is shown in the box plots to the right of Fig. 7a and b.
Although these results are promising with regards to using GRACE as a predictive tool for water resources, Fig. 7a and b suggest that Q is insensitive to TWSA March values below 100 mm.We recognize that all three of these regional watersheds are managed through a complex series of dams and reservoirs that create an altered runoff signal.Water resources managers use point-specific and model-based estimates of water storage in the region to optimize their operations.Additionally, in the fertile plains of the Snake River and lower CRB, broad-scale agriculture relies on both ground-and surface water for irrigation.Water withdrawals would be implicit in the TWSA signal and reduce Q.Subsequent, more detailed analysis would help quantify these effects.However the compilation of a complete irrigation dataset lies outside the scope of this study.We also acknowledge the length of record for this study would ideally be longer.Introduction

Conclusions References
Tables Figures

Back Close
Full Regardless of the length of record or anthropogenic influence, climate and geology still provide the first-order controls on water storage that are found in the hysteresis loops.GRACE encapsulates these hydrologic processes through measurements of TWSA.The hysteresis loops expand and contract accordingly during wet and dry years, as the intra-annual relationship between TWSA and Q represents the fluxes of water into and out of the watershed.Despite intra-annual differences, each of the subregional watersheds can be described by a family of hysteresis curves.The predicative capability using TWSA, the vertical sum of water, as compared to snowpack and soil moisture further highlights the integrated nature of water storage in regional hydrology.These integrated measurements of TWSA provide simple, but informative predictions of seasonal and monthly runoff.
GRACE-derived calculations of GWSA also provide insights into the hydrological processes governing groundwater recharge and depletion, as evidenced in the GWSA hysteresis loops.The GWSA-Q curves show an out-of-phase relationship between precipitation and groundwater recharge from the start of the wet season in October until February or March.This suggests that groundwater helps sustain stream flow during the wet fall and winter and that pore space in soils and geologic materials must fill to a certain threshold before groundwater begins to recharge and runoff is generated.The relationship between the TWSA and GWSA curves from October-March identifies how the onset of snowmelt also marks the beginning of groundwater recharge, and suggests that snowmelt inputs to groundwater are considerable.In the CRB this is critical as current climate trends are projected to reduce snowpack accumulation and exacerbate melt in the region (Wu et al., 2012;Rupp et al., 2013;Sproles et al., 2013).Additionally, our analysis identifies summer as the time of peak groundwater storage in all three regional watersheds.This peak corresponds to the timing of groundwater pump tests that are commonly conducted in June, and are used to develop groundwater withdrawal regulations (Jarvis, 2011(Jarvis, , 2014)).Our data suggest that groundwater pump tests should not be limited to an individual month, and should also include periods of reduced storage particularly during the winter months.The inclusion of multiple Introduction

Conclusions References
Tables Figures

Back Close
Full pump tests throughout the year could be particularly relevant as the population and water demand is projected to increase in the region.
The point-specific well data (Fig. 6) is ambiguous and show considerable variability with no consistent pattern regarding the timing of recharge and peak groundwater levels (light grey lines in Fig. 6).Rather than excluding these results or selecting individual wells that match GRACE data we include the results from all 33 wells to help demonstrate the variability that exists from well to well, and how site characteristics (i.e., usage, depth, location) might not represent regional groundwater characteristics (Jarvis, 2011(Jarvis, , 2014)).An in-depth study that couples GWSA and well data would potentially help explain the ambiguity, but lie outside the scope of this study.
GRACE is somewhat of a blunt instrument with regards to temporal (monthly) and spatial (1 • ) resolution.However, this emerging technology provides a new dimension to regional watershed analysis by providing an integrated measurement of water stored across and through the Earth.These measurements continue to prove their value in retrospective analysis of regional hydrology (Rodell et al., 2009;Castle et al., 2014).However, the hysteresis loops presented here and in Riegger and Tourian (2014) demonstrate the ability of GRACE data to help develop a process-based understanding of how regional watersheds function as simple, dynamic systems.As the temporal record of GRACE continues to develop, its value as both a diagnostic and predictive tool will continue to grow.In the mean time, these data have value in augmenting existing management strategies.
Perhaps one of the most important facets of GRACE data is that it does not distinguish political boundaries, and it is not linked to a specific in situ monitoring agency with limited data access, and has the capacity to bridge sparse and inconsistent onthe-ground hydrologic monitoring networks that exist in many regions of the world.
Previous GRACE-based analysis has shown its value in highlighting negative trends in terrestrial water storage in trans-boundary watersheds (Voss et al., 2013;Castle et al., 2014), and resulting regional conflict exacerbated by water shortages (Gleick, 2014b).Introduction

Conclusions References
Tables Figures

Back Close
Full GRACE provides an objective measurement of a region's water resources that can provide valuable insights into potential shortages or surpluses of water resources.

Conclusions
We have shown how GRACE-based measurements of TWSA distill the complexity of regional hydrology into a simple, dynamic system.TWSA and derived estimates of GWSA reveal hysteretic behavior for regional watersheds, which is more commonly associated with hydrologic measurements at local scales.While the magnitude of the hysteresis curves varies across years, they retain the same general shape that is unique to each watershed.We demonstrated the utility of these hysteresis curves by showing how TWSA during March can be used to predict Q during the drier summer months.Because GRACE-TWSA can augment prediction, managers could start to interpret each year's hysteresis curve for the upcoming spring and summer, providing greater clarity and validation for model-based forecasts presently used by water resource managers.Although this study focused on the CRB, which has a rich data record, GRACE data are available at a global scale and could be readily applied in areas with a paucity of data to understand how watersheds function and to improve streamflow forecasting capabilities.GRACE does not discern political boundaries and provides an integrated approach to understanding international watersheds (Voss et al., 2013).This resource could serve as a valuable tool for managers in forecasting surplus and scarcity, and in developing strategies that include changes in supply and demand due to human consumptive needs and current climate trends (Wagener et al., 2010;Gleick, 2014a).Introduction

Conclusions References
Tables Figures

Back Close
Full  Full the local scale, in situ instrumentation can quantify the non-linear relationship between Figures Back Close Full Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | together (within the radius of the filtering function), resulting in smaller signals being lost, and larger signals having a coarser footprint and a loss of spatial information.Discussion Paper | Discussion Paper | Discussion Paper | Monthly 1 km resolution SWE values were obtained from the mean of NLDAS and Snow Data Assimilation System (SNODAS; National Operational Hydrologic Remote Sensing Center, 2004) and were spatially averaged over the three watersheds.SNODAS data were used in place of the GLDAS data product, which considerably underestimated SWE in mountainous areas when compared to point-based measurements.Changes in monthly reservoir storage Introduction Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Zang, C. F., Liu, J., van der Velde, M., and Kraxner, F.: Assessment of spatial and temporal patterns of green and blue water flows under natural conditions in inland river basins in Northwest China, Hydrol.Earth Syst.Sci., 16, 2859-2870, doi:10.5194/hess-16-2859DiscussionPaper | Discussion Paper | Discussion Paper |

Figure 1 .Figure 2 .Figure 3 .
Figure 1.Context map and descriptions of the three study watersheds.

Table A2 .
The groundwater wells used in the analysis that compares GRACE-derived groundwater with location-specifc wells.USGS is the United States Geological Survey and IDWR is the Idaho Department of Water Resources.