GRACE Groundwater Drought Index: Evaluation of ...

Remote Sensing of Environment 198 (2017) 384–392

Contents lists available at ScienceDirect

Remote Sensing of Environment journal homepage: www.elsevier.com/locate/rse

GRACE Groundwater Drought Index: Evaluation of California Central Valley groundwater drought Brian F. Thomas a,b,⁎, James S. Famiglietti a,c,d, Felix W. Landerer a, David N. Wiese a, Noah P. Molotch a,e, Donald F. Argus a a

Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, USA Department of Geology and Environmental Science, University of Pittsburgh, Pittsburgh, PA 15260, USA Department of Earth System Science, University of California, Irvine, CA 92697, USA d Department of Civil and Environmental Engineering, University of California, Irvine, CA 92697, USA e Department of Geography, Institute of Arctic and Alpine Research, University of Colorado at Boulder, Boulder, CO 80309, USA b c

a r t i c l e

i n f o

Article history: Received 30 October 2016 Received in revised form 30 May 2017 Accepted 25 June 2017 Available online 3 July 2017 Keywords: Groundwater drought Remote sensing Drought indices GRACE

a b s t r a c t Quantitative approaches to assess the complexity of groundwater drought are hindered by the lack of direct observations of groundwater over space and time. Here, we present an approach to evaluate groundwater drought occurrence based on observations from NASA's Gravity Recovery and Climate Experiment (GRACE) satellite mission. Normalized GRACE-derived groundwater storage deviations are shown to quantify groundwater storage deficits during the GRACE record, which we define as the GRACE Groundwater Drought Index (GGDI). As a case study, GGDI is applied over the Central Valley of California, a regional aquifer undergoing intensive human activities and subject to significant drought periods during the GRACE record. Relations between GGDI and other hydrological drought indices highlight our ability to capture drought delays unique to groundwater drought. Further, GGDI captures characteristics of groundwater drought that occur as a result of complex human activities and natural changes, thus presenting a framework to assess multi-driver groundwater drought characteristics. © 2017 Elsevier Inc. All rights reserved.

1. Introduction and background By the end of 2015, the prolonged and well-documented drought in California, recognized as starting in 2012 (Griffin and Anchukaitis, 2014), had reached historically unprecedented conditions. The Palmer Drought Stress Index (PDSI) (Palmer, 1965), a commonly used index to evaluate drought based on a simple water budget model (Alley, 1984), identified June and July of 2014 as having the two lowest indices in the Central Valley of California, where records date from 1900 to present, while the spring of 2015 exhibited PDSI values in the lowest 95% quantile (Fig. S1). Equally severe deficits were documented in the Standardized Precipitation Index (McKee et al., 1993) (Fig. S1). A comparison between drought indices and streamflow reconstruction further illustrates the historic nature of the drought, which potentially represents the worst California drought in the last 1200 years (Griffin and Anchukaitis, 2014). The prolonged drought, attributed to a persistent upper level weather dipole (Wang et al., 2014), resulted in a cascade of impacts affecting rangelands (Larsen et al., 2014), forests (Baguskas ⁎ Corresponding author at: Geology and Environmental Science, University of Pittsburgh, 4107 O'Hara Street, Room 200 SRCC Building, Pittsburgh, PA 15260, USA. E-mail address: [email protected] (B.F. Thomas).

http://dx.doi.org/10.1016/j.rse.2017.06.026 0034-4257/© 2017 Elsevier Inc. All rights reserved.

et al., 2014; Asner et al., 2016), agriculture and socioeconomics (Howitt et al., 2014) and groundwater (Faunt et al., 2015). The term drought, used ambiguously above, refers to prolonged dryness manifesting itself in various water deficits including meteorological (precipitation), hydrological (streamflow), agricultural (soil moisture), socioeconomic and groundwater (Dracup et al., 1980; Mishra and Singh, 2010). Drought is largely driven by a change in climatic forcing, for example decreases in precipitation, which develop slowly and can last months to years (Tallaksen and van Lanen, 2004; Tallaksen et al., 2009). The lack of a formal definition of drought (Wilhite et al., 2007) combined with the difficulty in investigating its precursors have resulted in compartmentalized drought indices (for example, hydrological drought: PDSI; meteorological drought: SPI; groundwater drought: SGI (Bloomfield and Marchant, 2013)). Recent studies have sought to evaluate integrated drought indices (Hao and AghaKouchak, 2014; Hao et al., 2014; Ma et al., 2014). Thomas et al. (2014) developed a framework to evaluate a holistic drought characterization using data from the Gravity Recovery and Climate Experiment (GRACE) satellites, focusing on the total water storage deficits to characterize drought occurrence. As the effect of drought cascades from meteorological to hydrological to agricultural drought, groundwater storage may be impacted

B.F. Thomas et al. / Remote Sensing of Environment 198 (2017) 384–392

(Eltahir and Yeh, 1999). Groundwater drought is a distinctive class of drought resulting from the decrease in groundwater recharge (Goodarzi et al., 2016) and the decrease in groundwater storage and discharge (Mishra and Singh, 2010; Bloomfield and Marchant, 2013; Bloomfield et al., 2015). The direct and indirect consequences of anthropogenic influences can be readily linked to exacerbating groundwater drought (Tallaksen and Van Lanen, 2004). During hydrological and agricultural drought periods, water demands are often satisfied by groundwater withdrawals since groundwater storage provides resiliency (Hughes et al., 2012). Excessive groundwater withdrawals can thus magnify perceived drought (Famiglietti et al., 2011; Famiglietti and Rodell, 2013; Castle et al., 2014). Understanding groundwater drought is important in regions where subsurface storage influences regional security (Famiglietti, 2014) given the persistence in groundwater drought (Peters et al., 2003; Hughes et al., 2012). Identifying groundwater drought is important, especially in arid regions where the interplay between groundwater recharge and abstraction results in variable groundwater stress conditions (Richey et al., 2015a). Our ability to identify groundwater drought, however, is hindered by our inability to directly observe changes in groundwater storage. In California's Central Valley (Fig. 1), it has long been recognized that an overreliance on groundwater to meet water demands has resulted in substantial decreases of groundwater storage (Famiglietti et al., 2011; Scanlon et al., 2012a, 2012b) and subsequent land subsidence (Poland et al., 1975; Faunt et al., 2015; Sneed and Brandt, 2015). In the recent drought, decreases in groundwater storage intensified (Wang et al., 2016) resulting in unprecedented land surface deformation (Farr et al., 2015). As the occurrence of more severe and longer duration droughts are predicted to increase as a result of climate change (Cayan et al., 2006; Cook et al., 2015), our ability to observe hydrologic information to investigate the prevailing characteristics of drought become vital, especially in regards to water resources management (Aghakouchak et al., 2014). In particular, characterizing groundwater

385

storage at the onset of groundwater drought conditions is difficult due to the lack of continuous and spatially representative groundwater observations and the potential temporal offset between various observable drought conditions (i.e. precipitation and soil moisture) and groundwater drought (Calow et al., 1997; Changnon, 1987). Satellite remote sensing has been established as a powerful tool to observe water storage dynamics at large scales (Rodell and Famiglietti, 1999; Wahr et al., 2006) with the launch of the GRACE satellites. Observations from GRACE gravity anomalies may be converted into changes of water equivalent height thus tracking changes in total water storage across the globe. Previous applications of GRACE have isolated groundwater storage changes (Rodell et al., 2009; Famiglietti et al., 2011; Scanlon et al., 2012a, 2012b; Castle et al., 2014; Scanlon et al., 2015) given auxiliary data to remove other components of the water budget to draw important understanding of regional groundwater storage changes during the GRACE record. Whereas previous studies evaluated GRACE-derived groundwater storage changes as a response to drought (Famiglietti et al., 2011; Scanlon et al., 2012a, 2012b), this study explicitly introduces and evaluates a groundwater drought index based on GRACE observations in an effort to understand and identify groundwater drought. A case study focused on the Central Valley of California is described given the documented episodes of drought in the region (Famiglietti et al., 2011; Cook et al., 2015; Wang et al., 2014). The development of groundwater drought indicators has employed water budget approaches (Mendicino et al., 2008), statistical applications using in situ groundwater observations (Bloomfield and Marchant, 2013) or hydrologic model simulations (Houborg et al., 2012; Li and Rodell, 2015). In this paper, the water budget approach to derive a groundwater drought index is explored. Given drought propagation through various components of the hydrologic budget (Changnon, 1987; Eltahir and Yeh, 1999; Peters et al., 2003; Peters et al., 2005), we hypothesize that the observable expression of groundwater drought would occur some time after drought is expressed in soil moisture and precipitation indices. Previous groundwater drought

Fig. 1. Site map illustrating groundwater observation wells, reservoirs and the GRACE region used for the study.

386


studies (Goodarzi et al., 2016) sought to identify indices to make an explicit distinction between anthropogenic impacts and natural groundwater drought. In our evaluation, we recognize that changes in groundwater storage are largely influenced by anthropogenic impacts related to groundwater withdrawals, particularly within the Central Valley, and natural impacts, which are readily observed by GRACE (i.e. a change in groundwater storage (Alley and Konikow, 2015)). Our groundwater drought index captures the combination of human activities and natural changes, thus providing an intertwined index of drought advocated by Van Loon et al. (2016). Key to our analysis is a comparison between GRACE-derived groundwater storage changes and in situ observation of changes, using the groundwater drought index introduced by Li and Rodell (2015), to detect and document groundwater drought. Further, we compare our remote sensing-based groundwater drought index to common drought indices to highlight drought propagation represented as temporal response lags. 2. Data Our remote sensing-based drought characterization uses a combination of Earth observation data and in situ data to evaluate groundwater drought in the Central Valley of California (Fig. 1) over the study period of 10/2003–03/2016. We used monthly GRACE (Tapley et al., 2004) gravity coefficients from the JPL-RL05.1M mascon solution (Watkins et al., 2015; Wiese, 2015), available for the period 04/2002 through 03/ 2016. The mascon approach avoids the use of empirical post-processing filters to reduce errors (e.g., Swenson and Wahr, 2006), and instead uses a set of carefully evaluated a priori signal constraints to derive optimal monthly gravity fields (see Watkins et al. (2015) for details). This approach also greatly reduces the need for corrective data scaling (Landerer and Swenson, 2012), which we estimate here to be only 1.2 (compared to 2.35 in previous GRACE solutions, e.g., Famiglietti et al. (2011)). We calculated GRACE error for the monthly basin-mean water storage (Wahr et al., 2006) to be 4.46 cubic kilometers, or 29 mm water equivalent height (weqh) considering both measurement and leakage errors (Wiese et al., 2016) over the combined SacramentoSan Joaquin-Tulare basin (Fig. 1). Average water storage changes for the Central Valley were computed as anomalies of total water storage in equivalent water height relative to the baseline time period of 01/ 2004–12/2009. Basin volume anomalies were converted to cubic kilometers of water. GRACE and water balance comparisons were completed to assess our ability to evaluate water storage changes. Water balance data included TRMM-3B43 precipitation (P) data, the research-grade precipitation product from the Tropical Rainfall Measuring Mission (TRMM) satellite (Huffman et al., 2007), satellite-based evapotranspiration (ET) data, which integrates physically-based ET estimates from remotely-sensed atmospheric and vegetation conditions (MOD16 (Mu et al., 2007)), and in situ streamflow observations (Q) obtained from the U.S. Geological Survey (USGS) (http://waterdata.usgs.gov/ nwis) (see Fig. S2). Groundwater storage changes are estimated using a water mass balance approach whereby auxiliary datasets permit isolation of a groundwater storage signal from total water storage changes (Rodell and Famiglietti, 2002). We assume that the total water storage anomaly (TWSA) is comprised of anomaly changes in soil moisture (SMA), snow water equivalent (SWEA), surface water/reservoir storage (SWA) and groundwater (GWA) as TWSAt ¼ SMAt þ SWEAt þ SWAt þ GWAt

ð1Þ

where the subscript t indicates that the hydrologic components are functions of time. Groundwater storage anomalies (Fig. 2a) were estimated by rearranging Eq. (1) while errors for the GRACE-GW anomaly time series (shaded regions in Fig. 2(a)) were estimated by propagating TWS, SM, SW and SWE error following Rodell and Famiglietti (2002). The influence of deep soil moisture (Houborg et al., 2012; Swenson

and Lawrence, 2015) is recognized to influence drought indicators and isolation of a representative groundwater time series from GRACE. Our approach in isolating a groundwater residual from GRACE TWSA data employs approaches commonly used in GRACE-derived groundwater storage studies (Rodell et al., 2009; Famiglietti et al., 2011; Scanlon et al., 2012a, 2012b; Castle et al., 2014) using representative soil moisture up to 2 m depth. A summary of auxiliary datasets used to isolate the GRACE-derived groundwater time series is included in the Supplemental material. In situ groundwater observations were collected from the California Department of Water Resources (CADWR) California Statewide Groundwater Elevation Monitoring (CASGEM) database (www.water. ca.gov/groundwater/casgem/) for N 8000 observation wells in the study area. A well selection algorithm was used to isolate suitable observation well data with a minimum of a seasonal reading (i.e. at least one observation per 3-month period) to capture seasonal groundwater behaviors. Gap filling using spatial and temporal methods was completed to produce a monthly dataset of observed groundwater head for each observation well (see Supplemental for details). In this analysis, at most 2 consecutive months of missing data was estimated using gap filling methods that account for spatial correlation (neighboring wells) and temporal correlation (at-site variability) (Zeng and Levy, 1995). To assess volumetric changes in groundwater storage, ΔGWt, it is necessary to account for changes in groundwater head (ΔHt) and the aquifer specific yield, Sy, where ΔGW t ¼ Sy ΔH t

ð2Þ

In the Central Valley, spatial estimates of Sy are available from a calibrated USGS groundwater flow model for the region (Fig. S3) (Faunt, 2009). Average monthly statistics of depth of groundwater storage above an arbitrary datum were calculated using Eq. (2) to yield a dataset of 406 point observations (Fig. 1). Various drought indices, including the Palmer Drought Stress Index (PDSI, Palmer (1965)) and the Standardized Precipitation Index (SPI, McKee et al. (1993)), were downloaded to compare with our GRACEderived groundwater drought index. For this study, PDSI and SPI time series were calculated as the average over the Sacramento-San Joaquin-Tulare watershed with data obtained from the WestWide Drought Tracker (wrcc.dri.edu/wwdt/, accessed Sept 5, 2016). Similar to previous studies (Bloomfield and Marchant, 2013; Hao et al., 2014; Mendicino et al., 2008), we compare correlation and lag times between drought indices (PDSI, SPI, GGDI) to investigate similarities to the lag behaviors suggested by Changnon (1987) between precipitation, soil moisture and groundwater storage. 3. Methods 3.1. GGDI Analyses of groundwater storage changes using GRACE in the Central Valley have assessed changes during drought (Famiglietti et al., 2011; Scanlon et al., 2012a, 2012b) but did not identify characteristics of groundwater drought. Most recently, Wang et al. (2016) linked observed changes in groundwater storage in the Central Valley to drought as defined by PDSI. Classic definitions of drought refer to water deficits (Dracup et al., 1980; Mishra and Singh, 2010). Here, we extend the GRACE deficit approach of Thomas et al. (2014) to quantify groundwater storage deficits and surpluses, termed groundwater storage deviation, using our GRACE-derived groundwater storage time series (Fig. 2(a)). A monthly climatology (Ci, the climatology for month i) is calculated where n

Ci ¼

∑1i GWSAi where i ¼ 1; …; 12 ni

ð3Þ


387

Fig. 2. Time series of the (a) GRACE-derived groundwater storage anomaly (line) and monthly error (shaded region), (b) the monthly climatology over the GRACE period (line) with a 1standard deviation (1σ) variation illustrated as the shaded region, (c) the difference of 2(a)–2(b), defined as the groundwater storage deviation, which represents the deficit (in red) and surplus (in blue) in groundwater storage following Thomas et al. (2014) and (d) normalized groundwater storage deviation, which we define as the GRACE Groundwater Drought Index (GGDI).

Monthly GWSA climatology is illustrated in Fig. 2(b). It is important to note that our definition of monthly climatology adheres to the approach of Thomas et al. (2014) and deviates from previous definitions of climatology (for example, see Lloyd-Hughes and Saunders (2002), Hao et al. (2014)). In our analysis, we use monthly climatology to remove the influence of seasonality in groundwater storage changes (Jasechko et al., 2014). We subtract the monthly climatology (Fig. 2(b)) from GWSA (Fig. 2(a)) to obtain a groundwater storage deviation (GSD, Fig. 2(c)), which represents the net deviation in the volume of groundwater storage based on seasonal variability. Finally, we normalize the GSD by removing the mean (xGSD ) and dividing by the standard deviation (sGSD) where GGDI ¼

GSDt −xGSD sGSD

ð4Þ

GGDI represents the normalized net deviation in groundwater storage volumes. GGDI for the Central Valley is depicted in Fig. 2(d). Our derivation of GGDI is based on groundwater storage changes which we hypothesize should exhibit similarity to groundwater drought indices including GWI. 3.2. Calculation of in situ groundwater drought time series To assess groundwater drought in the Central Valley, we first must obtain a time series of groundwater storage changes from in situ groundwater observation records. Given the spatial pattern of

groundwater observations (Fig. 1), it is necessary to identify a time series which can characterize groundwater storage changes across the entire region. We use a simple Thiessen polygon analysis to account for the relative spatial weight of individual monthly groundwater storage observations. For each month, an area-weighted storage change for the Central Valley region was calculated as n

GWSt ¼

∑i¼1 Ai GW t i n

∑i¼1 Ai

ð5Þ

where Ai represents the area of the Thiessen polygon for well i and GWt is calculated from Eq. (2). To calculate GWI, we follow the procedure of Li and Rodell (2015) by first removing seasonality from our in situ groundwater storage change whereby X t ¼ GWSt m −GWSt m ∈m ¼ 1; …; 12

ð6Þ

given Xt represents the deseasonalized groundwater storage change time series and m represents the month of interest. Secondly, we normalize Xt by removing the time series mean (X t ) and dividing by the standard deviation (sXt) where GWIt ¼

X t −X t sX t

ð7Þ

388


The time series, GWIt, represents the average seasonal deviation from the average condition for an area weighted average of groundwater storage change. 4. Results The monthly climatology depicted in Fig. 2(b) illustrates a decrease in groundwater storage over time, as the monthly climatology fails to rise above zero for any particular month during the GRACE record. In the interpretation of Fig. 2(b), it is important to note that anomalies of GRACE TWSA and GWSA are with respect to the baseline period of 01/ 2004–12/2009. As the recent drought beginning in 2012 falls outside the baseline period, drastic changes in groundwater storage that occurred during the recent drought would shift the monthly climatology time series. Further, the addition of newly collected data would shift the monthly climatology time series. GGDI clearly documents 2 periods of groundwater drought during the GRACE record that are coincident to previously documented drought conditions in the Central Valley (Famiglietti et al., 2011; University of California Center for Hydrologic Modeling, 2014; Wang et al., 2014), where groundwater drought is expressed by negative values of GGDI. Further, it is clear that the potential severity of groundwater drought in the Central Valley increased during the recent drought as compared to the 2007–2011 groundwater drought event. It is important, however, to recognize that GGDI represents a groundwater drought index with respect to the average condition of the groundwater system which takes into account the combination of anthropogenic and natural changes. A comparison of GWI, GWSA and GGDI is illustrated in Fig. 3. In Fig. 3, it is clear that each time series captures groundwater storage changes

attributed to the 2007–2011 drought and the most recent drought beginning in 2012. An evaluation between our estimate of GWI, which represents in situ based observations of groundwater drought and GRACE-derived groundwater storage anomalies (Fig. 3(a)) depict moderate correlation (ρ = 0.63, p b 0.0001) (Fig. 3(b)). The relation between GGDI and GWI (Fig. 3(c)) exhibited strong correlation (ρ = 0.80, p b 0.0001) (Fig. 3(d)). GRACE-derived groundwater storage changes and GGDI result in moderate correlation (ρ = 0.74, p b 0.001). If we assume GWI, based on in situ observations, best represents groundwater storage changes as a response to groundwater drought, our results support our hypothesis that GRACE-derived groundwater storage anomalies alone cannot capture groundwater drought as readily as GGDI. This result could potentially be influenced by deep soil moisture not represented in NLDAS output (Houborg et al., 2012) and the lack of human activities including irrigation in NLDAS. A temporal lag between GGDI and GRACE-derived groundwater storage changes was noted in correlations with GWI. In each case, storage-based metrics lagged GWI by 2 months. A correlation of GGDI and GWI increased to 0.83 (p b 0.001) at a lag of 2 months (Fig. 3(b)), while a correlation of GRACE-derived groundwater storage changes and GWI increased slightly to 0.65 (p b 0.001) (Fig. 3(d)) at a lag of 2 months. A statistical comparison of the standardized drought indices (SPI and PDSI) and GGDI was conducted to assess our ability to capture representative drought anomalies using remote sensing approaches. Correlations between GGDI and PDSI (ρ = 0.32, p = 0.021) were moderate (Fig. 4(a) and Fig. 4(b)), while correlations with SPI (ρ = −0.10, p = 0.038) were weak (Fig. 4(c) and Fig. 4(d)). A temporal lag of 5 months was identified for PDSI and SPI, which when accounted for increased correlations (PDSI: ρ = 0.43, p b 0.027; SPI: ρ = 0.13, p b 0.010). The

Fig. 3. Time series of (a) the groundwater drought index (GWI) based on Li and Rodell (2015) (black) and GWSA (blue) with lag correlations (b). A comparison of (c) GGDI (blue) and GWI (black) depicts similar drought response as suggested by increased correlations (d).


temporal offset can be identified in Fig. 4; the drought beginning in 2007 was noted in 11/2006 by PDSI, 03/2007 by SPI but not identified in GGDI until 07/2007 in GGDI. Further, it is apparent that deficits in precipitation (SPI) and soil moisture (PDSI) were fulfilled more rapidly than observed in GGDI, whereby the event of the drought in 2011 lingered until 08/2011 in GGDI, a full 6 months after PDSI and SPI.

5. Discussion The remote-sensing approach to detect, evaluate and quantify groundwater drought presented here provides a framework to identify regional groundwater drought anomalies. Understanding the cascading changes of drought as moisture deficits move through the hydrologic cycle to impact groundwater resources are hindered by the lack of in situ groundwater observations. Conversely, remote sensing techniques hold much promise to understand hydrologic changes (Famiglietti, 2014; Famiglietti et al., 2015) and permit an assessment of groundwater drought as evidenced in results presented here. Our results document strong correlation (Fig. 3) between a GRACE-based groundwater drought index (GGDI) as compared to in situ-based groundwater indices (GWI), suggesting use of our approach can effectively characterize groundwater drought. Our results document that GRACE-derived groundwater storage anomalies may permit an assessment of groundwater changes but, as compared to our groundwater storage deviation approach, does not capture groundwater drought as observed using an in situ-based groundwater drought index (GWI). GGDI was found to exhibit higher correlation to GWI, suggesting that further processing of GRACE-

389

derived groundwater storage anomalies is necessary to characterize groundwater drought. Our comparison of GGDI with representative drought indices (PDSI and SPI) identified a temporal offset of 5 months, suggesting that traditional drought indicators demonstrate drought conditions 5 months prior to evidence of groundwater drought. These results were clearly identified in the drought period of 2007 to 2011 and the offset of the recent drought across the Central Valley, which identified drought onset by 5 months for both PDSI and SPI. In California, such behavior is expected given the complex allocation of surface water for irrigation, which, during dry periods, is reduced requiring agricultural irrigation to shift use from surface water to groundwater. Furthermore, the well-recognized offset of storage changes from precipitation to soil moisture to groundwater (Changnon, 1987; Eltahir and Yeh, 1999) likely factors into the observed groundwater drought response. Piecewise correlations, split into time periods concurrent to droughts (2007–2011, 2012–2016) exhibited variable temporal lag relationships between GGDI and PDSI/SPI. During the 2007–2011 drought, correlations were negative (PDSI: ρ = −0.80, p b 0.001; SPI: ρ = −0.83, p b 0.001) and exhibited temporal lags of −6 and 0 months, respectively. These behaviors indicate that groundwater storage is largely driven by soil moisture deficits, which were fulfilled by a complex mix of surface water allocations and groundwater use to sustain agriculture in the Central Valley. Similar results were speculated given the relationship between groundwater storage changes and surface water allocations by Famiglietti et al. (2011). During the most recent drought, correlations remained strong (PDSI: ρ = 0.85, p b 0.001; SPI: ρ = 0.72, p b 0.009) with temporal offsets of 2 and 9 months, respectively. The sign change in correlation suggests that the consistent overuse of

Fig. 4. Comparison of GGDI (black) and smoothed PDSI (blue) (panel a) showing persistent nature of groundwater drought. A temporal lag of 5 months was found between PDSI and GGDI (panel b). A comparison of GGDI (black) and SPI (blue) in panel c highlights reduced correlations (panel d) and a temporal offset of 5 months between SPI and GGDI. Shaded regions represent drought periods as determined from negative PDSI and SPI values.

390


groundwater had continued from the previous drought, and consequently exacerbated perceived drought characteristics lagging agricultural drought (PDSI) and meteorological drought (SPI) by several months. 6. Conclusions The proper identification of drought is dependent on the choice of an appropriate drought index. Our application of a GRACE-deficit analysis to characterize groundwater drought in the Central Valley captures anthropogenic effects and natural drought responses. As climate variability and human alterations of natural systems increases, it is imperative that drought indices capture the hydrologic response of a system during drought. Future risks of groundwater drought will likely rise due to increasing reliance of groundwater to fulfill water demands (Gleick, 2000; Famiglietti and Rodell, 2013) and changes in climate and groundwater recharge (Meixner et al., 2016; Zhang et al., 2016; Tashie et al., 2016; Thomas et al., 2016). In regions such as the Central Valley, groundwater resources are relied upon to fulfill water demands in normal climatological years and withdrawals are increased during drought periods to offset decreases in surface water allocations (Famiglietti et al., 2011; Castle et al., 2014; Thomas and Famiglietti, 2015). Whereas surface water reservoirs and soil moisture storage may be replenished rapidly as a result of a storm event (Dettinger, 2013) or by a seasonal increase in snowpack, groundwater drought conditions tend to persist (Peters et al., 2005; Mishra and Singh, 2010). An example of the persistent nature of groundwater drought can be noted in the response times illustrated in Fig. 4, which indicates that groundwater drought typically occurs some time after drought is expressed in soil moisture (PDSI) and precipitation (SPI). The future ability of groundwater systems to rebound from groundwater drought may be impacted by groundwater depletion at regional (Rodell et al., 2009; Famiglietti et al., 2011; Scanlon et al., 2012a, 2012b; Thomas and Famiglietti, 2015) and global scales (Wada et al., 2010; Gleeson et al., 2012; Döll et al., 2014; Richey et al., 2015a, 2015b). GGDI, which captures both natural and anthropogenic influences, may thus provide a framework to evaluate groundwater drought and the hindrance of groundwater storage recovery after drought caused by unsustainable groundwater use. Given the historical trajectory of groundwater storage depletion in the Central Valley (Faunt, 2009) and the likelihood of increased drought (Cook et al., 2015), the potential to magnify the apparent severity and duration of groundwater drought is likely to escalate. Although groundwater depletion in the Central Valley has been well documented, the characterization of groundwater drought highlighted here illuminates the adverse impacts posed by drought in conjunction with groundwater overdrafts to water security in a region (Famiglietti, 2014; Richey et al., 2015a). The possibility of similar behaviors globally is of concern (Wada et al., 2010; Richey et al., 2015a, 2015b), raising fears of our ability to effectively manage groundwater resources during drought (Famiglietti, 2014). Our evaluation illustrates the capability of GGDI to characterize groundwater drought in regions without adequate in situ observations, which provide vital information to evaluate groundwater risk as climates and water use change (Villholth et al., 2013). A limitation of our approach resides in the fact that GRACE observes a 1-D change in groundwater storage (Alley and Konikow, 2015). Our drought indicator does not account for the horizontal movement of water within aquifer systems, which may impact the spatiotemporal groundwater dynamics and thus the groundwater drought characteristics for a given region. Furthermore, GRACE does not observe changes in groundwater discharge, for example baseflow, which can be important factors in the evaluation of groundwater drought (Mishra and Singh, 2010). GRACE provides monthly observations of water storage changes at a temporal resolution which may fail to capture changes in groundwater storage during drought (Chew and Small, 2014). As suggested in Thomas et al. (2014), we advise GGDI be used to identify

groundwater drought for periods that exhibit negative GGDI for periods of 3 or more months. In this evaluation, we adhered to previous GRACEderived groundwater storage methods to isolate a residual groundwater time series (Rodell et al., 2004; Rodell et al., 2009; Famiglietti et al., 2011; Scanlon et al., 2012a, 2012b; Castle et al., 2014). Simulated soil moisture with regard to drought is problematic as deep soil layer moisture reflects long-term drought (Sheffield et al., 2004). Swenson and Lawrence (2015) evaluated controls of soil depth in CLM, highlighting the hydrologic influence of uniform soil moisture depth simulations as used in this study. Variable soil moisture depth simulations (Brunke et al., 2016) have indicated great promise in constraining simulated soil moisture uncertainty given recognized model deficiencies (Clark et al., 2015). As shown in Fig. S5 in the Supplemental material, soil moisture anomalies at various soil depths did not result in significantly different GGDI estimates, thus highlighting the robust metric to assess groundwater drought. Despite these limitations, our groundwater drought index captured lag times and persistence in drought characteristics which, given a long history of understanding the delayed fashion of drought through the hydrologic cycle (Changnon, 1987), have yet to be apparent in groundwater drought indices. Acknowledgements The research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. GRACE land data are available at http://grace.jpl.nasa.gov, supported by the NASA MEaSUREs Program; other data used in the paper are cited within the text and described in the Supplemental. Support from the NASA GRACE Science Team is gratefully acknowledged. We wish to thank 3 anonymous reviewers whose comments contributed to substantial improvements to the manuscript. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.rse.2017.06.026. References Aghakouchak, A., Feldman, D., Stewardson, M.J., Saphores, J.D., Grant, S., Sanders, B., 2014. Australia's drought: lessons for California. Science 343 (6178), 1430–1431. Alley, W.M., 1984. The Palmer drought severity index: limitations and assumptions. J. Clim. Appl. Meteorol. 23 (7), 1100–1109. Alley, W.M., Konikow, L.F., 2015. Bringing GRACE down to earth. Ground Water 53 (6), 826–829. Asner, G.P., Brodrick, P.G., Anderson, C.B., Vaughn, N., Knapp, D.E., Martin, R.E., 2016. Progressive forest canopy water loss during the 2012–2015 California drought. Proc. Natl. Acad. Sci. 113 (2), E249–E255. Baguskas, S.A., Peterson, S.H., Bookhagen, B., Still, C.J., 2014. Evaluating spatial patterns of drought-induced tree mortality in a coastal California pine forest. For. Ecol. Manag. 315, 43–53. Bloomfield, J.P., Marchant, B.P., 2013. Analysis of groundwater drought building on the standardised precipitation index approach. Hydrol. Earth Syst. Sci. 17, 4769–4787. Bloomfield, J.P., Marchant, B.P., Bricker, S.H., Morgan, R.B., 2015. Regional analysis of groundwater droughts using hydrograph classification. Hydrol. Earth Syst. Sci. 19 (10), 4327–4344. Brunke, M.A., Broxton, P., Pelletier, J., Gochis, D., Hazenberg, P., Lawrence, D.M., Leung, L.R., Niu, G.Y., Troch, P.A., Zeng, X., 2016. Implementing and evaluating variable soil thickness in the community land model, version 4.5 (CLM4. 5). J. Clim. 29 (9), 3441–3461. Calow, R.C., Robins, N.S., MacDonald, A.M., Macdonald, D.M.J., Gibbs, B.R., Orpen, W.R.G., Mtembezeka, P., Andrews, A.J., Appiah, S.O., 1997. Groundwater management in drought-prone areas of Africa. Int. J. Water Resour. Dev. 13 (2), 241–262. Castle, S.L., Thomas, B.F., Reager, J.T., Rodell, M., Swenson, S.C., Famiglietti, J.S., 2014. Groundwater depletion during drought threatens future water security of the Colorado River Basin. Geophys. Res. Lett. 41 (16), 5904–5911. Cayan, Dan, Maurer, E., Dettinger, M., Tyree, M., Hayhoe, K., Bonfils, C., Duffy, P., Santer, B., 2006. Climate Scenarios for California. http://bairwmp.org/docs/climate-change/ Climate%20Action%20Team%20Reports/Cayan%20et%20al%202006%20Climate% 20Scenarios%20for%20California.pdf accessed April 4, 2017. Changnon, S.A., 1987. Detecting drought conditions in Illinois (Vol. 169). Illinois State Water Survey. Chew, C.C., Small, E.E., 2014. Terrestrial water storage response to the 2012 drought estimated from GPS vertical position anomalies. Geophys. Res. Lett. 41 (17), 6145–6151.

B.F. Thomas et al. / Remote Sensing of Environment 198 (2017) 384–392 Clark, M.P., Fan, Y., Lawrence, D.M., Adam, J.C., Bolster, D., Gochis, D.J., Hooper, R.P., Kumar, M., Leung, L.R., Mackay, D.S., Maxwell, R.M., 2015. Improving the representation of hydrologic processes in Earth System Models. Water Resour. Res. 51 (8), 5929–5956. Cook, B.I., Ault, T.R., Smerdon, J.E., 2015. Unprecedented 21st century drought risk in the American Southwest and Central Plains. Sci. Adv. 1 (1), e1400082. Dettinger, M.D., 2013. Atmospheric rivers as drought busters on the US West Coast. J. Hydrometeorol. 14 (6), 1721–1732. Döll, P., Mueller Schmied, H., Schuh, C., Portmann, F.T., Eicker, A., 2014. Global-scale assessment of groundwater depletion and related groundwater abstractions: combining hydrological modeling with information from well observations and GRACE satellites. Water Resour. Res. 50 (7), 5698–5720. Dracup, J.A., Lee, K.S., Paulson, E.G., 1980. On the definition of droughts. Water Resour. Res. 16 (2), 297–302. Eltahir, E.A., Yeh, P.J.F., 1999. On the asymmetric response of aquifer water level to floods and droughts in Illinois. Water Resour. Res. 35 (4), 1199–1217. Famiglietti, J.S., 2014. The global groundwater crisis. Nat. Clim. Chang. 4 (11), 945–948. Famiglietti, J.S., Rodell, M., 2013. Water in the balance. Science 340 (6138), 1300–1301. Famiglietti, J.S., Lo, M., Ho, S.L., Bethune, J., Anderson, K.J., Syed, T.H., Swenson, S.C., de Linage, C.R., Rodell, M., 2011. Satellites measure recent rates of groundwater depletion in California's Central Valley. Geophys. Res. Lett. 38 (3). Famiglietti, J.S., Cazenave, A., Eicker, A., Reager, J.T., Rodell, M., Velicogna, I., 2015. Satellites provide the big picture. Science 349 (6249). Farr, T.G., Jones, C., Liu, Z., 2015. Progress Report: Subsidence in the Central Valley, California. http://water.ca.gov/groundwater/docs/NASA_REPORT.pdf. Faunt, C.C. (Ed.), 2009. Groundwater availability of the central valley aquifer, California. US Geological Survey, Reston, VA, p. 225. Faunt, C.C., Sneed, M., Traum, J., Brandt, J.T., 2015. Water availability and land subsidence in the Central Valley, California, USA. Hydrogeol. J. 1–10. Gleeson, T., Wada, Y., Bierkens, M.F., van Beek, L.P., 2012. Water balance of global aquifers revealed by groundwater footprint. Nature 488 (7410), 197–200. Gleick, P.H., 2000. A look at twenty-first century water resources development. Water Int. 25 (1), 127–138. Goodarzi, M., Abedi-Koupai, J., Heidarpour, M., Safavi, H.R., 2016. Development of a new drought index for groundwater and its application in sustainable groundwater extraction. J. Water Resour. Plan. Manag. 142 (9), 04016032. Griffin, D., Anchukaitis, K.J., 2014. How unusual is the 2012–2014 California drought? Geophys. Res. Lett. 41 (24), 9017–9023. Hao, Z., AghaKouchak, A., 2014. A nonparametric multivariate multi-index drought monitoring framework. J. Hydrometeorol. 15 (1), 89–101. Hao, Z., AghaKouchak, A., Nakhjiri, N., Farahmand, A., 2014. Global integrated drought monitoring and prediction system. Sci. Data 1. Houborg, R., Rodell, M., Li, B., Reichle, R., Zaitchik, B.F., 2012. Drought indicators based on model-assimilated Gravity Recovery and Climate Experiment (GRACE) terrestrial water storage observations. Water Resour. Res. 48 (7). Howitt, R., Medellín-Azuara, J., MacEwan, D., Lund, J., Sumner, D., 2014. Economic analysis of the 2014 drought for California agriculture. Davis, CA: UC–Davis Center for Watershed Sciences Online at. https://watershed.ucdavis.edu/files/biblio/DroughtReport_ 23July2014_0.pdf. Huffman, G.J., Bolvin, D.T., Nelkin, E.J., Wolff, D.B., Adler, R.F., Gu, G., Hong, Y., Bowman, K., P., & Stocker, E. F., 2007. The TRMM multisatellite precipitation analysis (TMPA): quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeorol. 8 (1), 38–55. Hughes, J.D., Petrone, K.C., Silberstein, R.P., 2012. Drought, groundwater storage and stream flow decline in southwestern Australia. Geophys. Res. Lett. 39 (3). Jasechko, S., Birks, S.J., Gleeson, T., Wada, Y., Fawcett, P.J., Sharp, Z.D., McDonnell, J.J., Welker, J.M., 2014. The pronounced seasonality of global groundwater recharge. Water Resour. Res. 50 (11), 8845–8867. Landerer, F.W., Swenson, S.C., 2012. Accuracy of scaled GRACE terrestrial water storage estimates. Water Resour. Res. 48 (4). Larsen, R.E., Horney, M.R., Macon, D., 2014. Update of the 2014 drought on California rangelands. Rangel. Ecol. Manag. 36 (5), 52–58. Li, B., Rodell, M., 2015. Evaluation of a model-based groundwater drought indicator in the conterminous US. J. Hydrol. 526, 78–88. Lloyd-Hughes, B., Saunders, M.A., 2002. A drought climatology for Europe. Int. J. Climatol. 22 (13), 1571–1592. Ma, M., Ren, L., Singh, V.P., Yang, X., Yuan, F., Jiang, S., 2014. New variants of the Palmer drought scheme capable of integrated utility. J. Hydrol. 519, 1108–1119. McKee, T.B., Doesken, N.J., Kleist, J., 1993. The relationship of drought frequency and duration to time scales. Proceedings of the 8th Conference on Applied Climatology. Vol. 17, No. 22. American Meteorological Society, Boston, MA, pp. 179–183. Meixner, T., Manning, A.H., Stonestrom, D.A., Allen, D.M., Ajami, H., Blasch, K.W., Brookfield, A.E., Castro, C.L., Clark, J.F., Gochis, D.J., Flint, A.L., 2016. Implications of projected climate change for groundwater recharge in the western United States. J. Hydrol. 534, 124–138. Mendicino, G., Senatore, A., Versace, P., 2008. A Groundwater Resource Index (GRI) for drought monitoring and forecasting in a Mediterranean climate. J. Hydrol. 357 (3), 282–302. Mishra, A.K., Singh, V.P., 2010. A review of drought concepts. J. Hydrol. 391 (1), 202–216. Mu, Q., Heinsch, F.A., Zhao, M., Running, S.W., 2007. Development of a global evapotranspiration algorithm based on MODIS and global meteorology data. Remote Sens. Environ. 111 (4), 519–536. Palmer, W.C., 1965. Meteorological Drought. Vol. 30. US Department of Commerce, Weather Bureau, Washington, DC, USA. Peters, E., Torfs, P.J.J.F., Van Lanen, H.A.J., Bier, G., 2003. Propagation of drought through groundwater—a new approach using linear reservoir theory. Hydrol. Process. 17 (15), 3023–3040.

391

Peters, E., Van Lanen, H.A.J., Torfs, P.J.J.F., Bier, G., 2005. Drought in groundwater—drought distribution and performance indicators. J. Hydrol. 306 (1), 302–317. Poland, J.F., Lofgren, B.E., Ireland, R.L., Pugh, R.G., 1975. Land Subsidence in the San Joaquin Valley, California, as of 1972. Richey, A.S., Thomas, B.F., Lo, M.H., Reager, J.T., Famiglietti, J.S., Voss, K., Swenson, S., Rodell, M., 2015a. Quantifying renewable groundwater stress with GRACE. Water Resour. Res. 51 (7), 5217–5238. Richey, A.S., Thomas, B.F., Lo, M.H., Famiglietti, J.S., Swenson, S., Rodell, M., 2015b. Uncertainty in global groundwater storage estimates in a Total Groundwater Stress framework. Water Resour. Res. 51 (7), 5198–5216. Rodell, M., Famiglietti, J.S., 1999. Detectability of variations in continental water storage from satellite observations of the time dependent gravity field. Water Resour. Res. 35 (9). Rodell, M., Famiglietti, J.S., 2002. The potential for satellite-based monitoring of groundwater storage changes using GRACE: the High Plains aquifer, Central US. J. Hydrol. 263 (1), 245–256. Rodell, M., Famiglietti, J.S., Chen, J., Seneviratne, S.I., Viterbo, P., Holl, S., Wilson, C.R., 2004. Basin scale estimates of evapotranspiration using GRACE and other observations. Geophys. Res. Lett. 31 (20). Rodell, M., Velicogna, I., Famiglietti, J.S., 2009. Satellite-based estimates of groundwater depletion in India. Nature 460 (7258), 999–1002. Scanlon, B.R., Longuevergne, L., Long, D., 2012a. Ground referencing GRACE satellite estimates of groundwater storage changes in the California Central Valley, USA. Water Resour. Res. 48 (4). Scanlon, B.R., Faunt, C.C., Longuevergne, L., Reedy, R.C., Alley, W.M., McGuire, V.L., McMahon, P.B., 2012b. Groundwater depletion and sustainability of irrigation in the US High Plains and Central Valley. Proc. Natl. Acad. Sci. 109 (24), 9320–9325. Scanlon, B.R., Zhang, Z., Reedy, R.C., Pool, D.R., Save, H., Long, D., Chen, J., Wolock, D.M., Conway, B.D., Winester, D., 2015. Hydrologic implications of GRACE satellite data in the Colorado River Basin. Water Resour. Res. 51 (12), 9891–9903. Sheffield, J., Goteti, G., Wen, F., Wood, E.F., 2004. A simulated soil moisture based drought analysis for the United States. J. Geophys. Res.-Atmos. 109 (D24). Simon Wang, S-Y., Lin, Yen-Heng, Gillies, Robert R., Hakala, Kirsti, 2016. "Indications for Protracted Groundwater Depletion after Drought over the Central Valley of California*,+.". J. Hydrometeorol 17 (3), 947–955. Sneed, M., Brandt, J.T., 2015. Land subsidence in the San Joaquin Valley, California, USA, 2007–2014. Proceedings of the International Association of Hydrological Sciences 372, pp. 23–27. Swenson, S.C., Lawrence, D.M., 2015. A GRACE-based assessment of interannual groundwater dynamics in the Community Land Model. Water Resour. Res. 51 (11), 8817–8833. Swenson, S., Wahr, J., 2006. Post-processing removal of correlated errors in GRACE data. Geophys. Res. Lett. 33 (8). Tallaksen, L.M., Van Lanen, H.A., 2004. Hydrological Drought: Processes and Estimation Methods for Streamflow and Groundwater. Vol. 48. Elsevier. Tallaksen, L.M., Hisdal, H., Van Lanen, H.A., 2009. Space–time modelling of catchment scale drought characteristics. J. Hydrol. 375 (3), 363–372. Tapley, B.D., Bettadpur, S., Ries, J.C., Thompson, P.F., Watkins, M.M., 2004. GRACE measurements of mass variability in the Earth system. Science 305 (5683), 503–505. Tashie, A.M., Mirus, B.B., Pavelsky, T.M., 2016. Identifying long-term empirical relationships between storm characteristics and episodic groundwater recharge. Water Resour. Res. 52 (1), 21–35. Thomas, B.F., Famiglietti, J.S., 2015. Sustainable groundwater management in the arid Southwestern US: Coachella Valley, California. Water Resour. Manag. 29 (12), 4411–4426. Thomas, A.C., Reager, J.T., Famiglietti, J.S., Rodell, M., 2014. A GRACE-based water storage deficit approach for hydrological drought characterization. Geophys. Res. Lett. 41 (5), 1537–1545. Thomas, B.F., Behrangi, A., Famiglietti, J.S., 2016. Precipitation intensity effects on groundwater recharge in the southwestern United States. Water 8 (3), 90. University of California Center for Hydrologic Modeling, 2014. UCCHM Water Advisory #1, Water Storage Changes in California's Sacramento and San Joaquin River Basins from GRACE: Preliminary Updated Results for 2003–2013, Univ. of Calif., Irvine. http://dx.doi.org/10.13140/RG.2.1.2272.2084. Van Loon, A.F., Stahl, K., Di Baldassarre, G., Clark, J., Rangecroft, S., Wanders, N., ... Van Lanen, H.A.J., 2016. Drought in a human-modified world: reframing drought definitions, understanding, and analysis approaches. Hydrol. Earth Syst. Sci. 20 (9): 3631–3650. http://dx.doi.org/10.5194/hess-20-3631-2016. Villholth, K.G., Tøttrup, C., Stendel, M., Maherry, A., 2013. Integrated mapping of groundwater drought risk in the Southern African Development Community (SADC) region. Hydrogeol. J. 21 (4), 863–885. Wada, Y., van Beek, L.P., van Kempen, C.M., Reckman, J.W., Vasak, S., Bierkens, M.F., 2010. Global depletion of groundwater resources. Geophys. Res. Lett. 37 (20). Wahr, J., Swenson, S., Velicogna, I., 2006. Accuracy of GRACE mass estimates. Geophys. Res. Lett. 33 (6). Wang, S.Y., Hipps, L., Gillies, R.R., Yoon, J.H., 2014. Probable causes of the abnormal ridge accompanying the 2013–2014 California drought: ENSO precursor and anthropogenic warming footprint. Geophys. Res. Lett. 41 (9), 3220–3226. Watkins, M.M., Wiese, D.N., Yuan, D.N., Boening, C., Landerer, F.W., 2015. Improved methods for observing earth's time variable mass distribution with GRACE using spherical cap mascons. J. Geophys. Res. Solid Earth 120 (4), 2648–2671. Wiese, D.N., 2015. GRACE monthly global water mass grids NETCDF RELEASE 5.0. Ver. 5.0. PO.DAAC, CA, USA. Dataset accessed [2017-03-15] at. http://dx.doi.org/10.5067/ TEMSC-OCL05.

392


Wiese, D.N., Landerer, F.W., Watkins, M.M., 2016. Quantifying and reducing leakage errors in the JPL RL05M GRACE mascon solution. Water Resour. Res. 52. http://dx.doi.org/10. 1002/2016WR019344. Wilhite, D.A., Svoboda, M.D., Hayes, M.J., 2007. Understanding the complex impacts of drought: a key to enhancing drought mitigation and preparedness. Water Resour. Manag. 21 (5), 763–774.

Zeng, L., Levy, G., 1995. Space and time aliasing structure in monthly mean polar-orbiting satellite data. J. Geophys. Res. Atmos. 100 (D3), 5133–5142. Zhang, J., Felzer, B.S., Troy, T.J., 2016. Extreme precipitation drives groundwater recharge: the Northern High Plains Aquifer, central United States, 1950–2010. Hydrol. Process.