Embed Size (px)
Citation preview
1
Climate change impacts on groundwater storage in the Central Valley, California 1
Sarfaraz Alam1*, Mekonnen Gebremichael1, Ruopu Li2, Jeff Dozier3, Dennis P. Lettenmaier4 2
Authors Affiliation: 3
1Department of Civil and Environmental Engineering, University of California, Los Angeles, 4
California, USA 5
2Department of Geography and Environmental Resources, Southern Illinois University-6
Carbondale, Illinois, USA 7
3Bren School of Environmental Science & Management, University of California, Santa Barbara, 8
California, USA 9
4Department of Geography, University of California, Los Angeles, California, USA 10
Emails: [email protected] (S. Alam), [email protected] (M. Gebremichael), 11
[email protected] (R. Li), [email protected] (J. Dozier), [email protected] (D. P. 12
Lettenmaier) 13
ORCID: S. Alam 0000-0002-9592-2782; M. Gebremichael 0000-0002-6216-9966; R. Li 14
0000-0003-3500-0273; J. Dozier 0000-0001-8542-431X; D.P. Lettenmaier 0000-0002-15
0914-0726 16
*Corresponding author: Sarfaraz Alam ([email protected]) 17
18
19
20
21
22
23
2
Abstract 24
Groundwater plays a critical supporting role in agricultural production in the California Central 25
Valley (CV). Recent prolonged droughts (notably 2007-2009 and 2012-2016) caused dramatic 26
depletion of groundwater, indicating the susceptibility of the CV’s water supply to climate 27
change. To assess the impact of climate change on groundwater storage in the CV, we combine 28
integrated surface water and groundwater models with climate projections from 20 global 29
climate models and thereby explore the vulnerability of CV groundwater for climate scenarios 30
RCP4.5 and RCP8.5. We find that groundwater has been declining over the past decades (3 31
km3/yr on average during 1950-2009). In the absence of future mitigating measures, this decline 32
will continue, but at higher rate due to climate change (31% and 39% increase in loss rate under 33
RCP4.5 and RCP8.5). The highest loss, more than 80% of the total loss, will occur in the semi-34
arid southern Tulare region. We perform computational experiments to quantify the relative 35
contribution of future crop-water use and headwater inflows to total storage change. Our results 36
show that, absent management changes, continuing declines in future groundwater storage are 37
mainly attributed to ongoing over-use of groundwater. However, changes in the seasonality of 38
streamflow into the CV, as well as (small) changes in annual inflows and increased crop water 39
use in a warmer climate, will add 40-70% to the current average annual groundwater losses, 40
bringing them to about 5 km3/yr. 41
Keywords: groundwater; Central Valley; integrated modeling; water management 42
43
1 Introduction 44
Global warming poses serious threats to water resources, particularly in the Western U.S (Bates 45
et al. 2008; Cohen et al. 2000; Gleick et al. 2000). Climate change can directly or indirectly 46
affect different components of the hydrological cycle; precipitation, snowmelt, and 47
3
evapotranspiration are linked to agricultural water supply and demand (Garrote et al. 2015; 48
García-Garizábal et al. 2014). Increases in water demand and decreases in surface water supply 49
caused by a warming climate can negatively affect groundwater storage through increased 50
groundwater exploitation, especially in regions like the Central Valley of California where 51
groundwater is already stressed. 52
The Central Valley (CV) of California is one of the most productive agricultural regions in the 53
world. On average, sources of irrigation water are about equally divided between surface water 54
and groundwater (Li et al. 2018), but during droughts surface water availability is reduced, and 55
the difference is mostly made up by groundwater, along with some reduction in irrigated area to 56
reduce demand. Throughout the CV, groundwater sources relative to surface water sources vary, 57
partially due to the natural north-south precipitation gradients, with groundwater relatively more 58
important to the south. Surface water comes mostly from streams that head in the Sierra Nevada 59
(hereafter referred to as headwater watersheds). Precipitation is highly seasonal, mostly 60
occurring from October through March. Pronounced spatial differences in the amount and timing 61
of surface water supplied from the headwater watersheds arise from generally higher total 62
precipitation to the north and greater winter snowpack storage to the south. To address these 63
differences and the overall semi-arid climate, the CV relies on a complex network of reservoirs 64
and manmade canals to store, import and transfer water, especially to its driest southern portions. 65
In recent decades, a gap has developed between water demand and surface water supply. This 66
gap has been caused by reductions in water available from the Colorado River Basin, a shift from 67
row crops to tree crops with higher water demands, and pervasive drought over much of the last 68
decade. This gap has been met during the 2012-2016 drought mostly by extraction of 69
groundwater, along with some reduction in irrigated area (Xiao et al. 2017). Overexploitation of 70
4
groundwater was especially pervasive. Posing a serious threat to the long-term viability of 71
agriculture in the CV, the dropping water tables motivated the California legislature to enact the 72
Sustainable Groundwater Management Act (SGMA), which requires that groundwater use in the 73
CV be brought into balance with recharge (California State Legislature 2014). 74
75
Historically, the area devoted to irrigated agriculture in the CV expanded over the 100-year 76
period from the 1870s to the 1970s, at which point the area under irrigation became roughly 77
stable (Olmstead and Rhode, 2017). However, aside from exceptionally high runoff years, the 78
water balance of the CV has been in deficit (Xiao et al. 2017; Famiglietti 2014; Scanlon et al. 79
2012b), effectively meaning that groundwater is being mined. This situation worsened during 80
two drought periods during the past decade (2007-09 and 2012-16). Using satellite data from the 81
Gravity Recovery and Climate Experiment (GRACE), Famiglietti et al. (2011) estimated that the 82
CV experienced net groundwater depletion during 2003-2010 of about 20 km3. Xiao et al. (2017) 83
estimated cumulative groundwater storage change (ΔGW) using multiple data sources including 84
GRACE data and in situ records and model estimates of the CV’s water balance and estimated a 85
depletion rate of 7.2 ± 1.0 km3/yr from April 2006 to March 2010 and 11.2 km3/yr for the 2012-86
2016 drought. Other relevant studies support similar findings that CV groundwater has been in 87
decline for at least the past 50 years (e.g., Famiglietti et al. 2011; Scanlon et al. 2012a; Chen et 88
al. 2016), with most finding that the imbalance has accelerated over the last decade or so. The 89
loss of groundwater in recent droughts is a key indicator of the susceptibility of the CV 90
groundwater system to climate variability and change, to an extent that has not previously been 91
estimated. 92
93
5
We address the question of how the CV groundwater system is likely to respond to climate 94
change over the next century, and how the effect of additional stress that is likely to be placed on 95
the system by warmer climate will compare with the ongoing effects of transitioning crop types, 96
primarily from row to tree crops. Essentially, a warming climate changes the timing and amount 97
of streamflow into the CV, and increased crop-water use results directly and indirectly from 98
warmer temperatures. 99
100
To address these issues, we utilize a suite of global climate model output archived for the Fifth 101
Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) to represent 102
uncertainty in patterns of future climate that will affect both runoff and crop water use. We test 103
projections both for modest global emissions increases (RCP4.5) and a more draconian scenario 104
(RCP 8.5). Compared to previous efforts (e.g. Massoud et al. 2018; Hanson et al. 2012), we use a 105
full suite of climate models and scenarios to better represent current understanding of future 106
uncertainties in the factors that will affect CV groundwater in the coming decades. Starting with 107
climate scenarios from Global Climate Models (GCMs), we downscale to 6 km resolution, and 108
then we use a hydrology model to represent the effects of headwater watersheds in producing 109
streamflow into downstream reservoirs and subsequent releases from reservoirs that produce 110
surface flows into the CV. Finally, we represent the implications of future surface runoff into the 111
CV on groundwater use. In contrast to many past studies, our domain is the entire Sacramento-112
San Joaquin-Tulare lowland and upland domain (SSJT), as described in the next section. 113
114
2 Study area 115
6
The study area consists of the entire CV from north of Redding south to the Tulare River basin 116
lowlands, and all the drainages that flow to the CV so defined (see Fig. 1). In total, the study area 117
(160,000 km2) constitutes about one third of the area of California. The CV portion of the 118
domain is nearly flat, contrasting with the high elevation headwater watersheds which are the 119
source of most of the surface water that flows to the CV. Strong elevational precipitation 120
gradients characterize the transition from the valley floor to the mountains, and a subtler 121
precipitation gradient exists from the humid north to the semi-arid south (Bales et al. 2006). 122
Averaged over the domain, more than 85% of the average annual precipitation occurs between 123
October and March. Notwithstanding that headwater basins to the south are more snow-124
dominated than those to the north, northern watersheds generate more than two-thirds of the total 125
annual inflow to the CV. This imbalance has motivated the construction of dams and conveyance 126
structures that transfer water from the north to the south. This complex network of surface 127
storage and conveyance structures (most of the former are in the uplands, and the latter in the 128
CV) control the distribution of surface water throughout domain and throughout the year. The 129
reservoirs store high winter and spring flows and release them during the high-demand summer 130
season, while providing flood protection and environmental services (mostly supplementing 131
summer low flows) as well. There are around 18 major reservoirs located in headwater 132
watersheds (shown in Fig. 1) that regulate flows to the CV. There are five major hydrologic 133
regions (HR) in CV (Brush and Dogrul 2013), specifically Sacramento (SC), San Joaquin (SJ), 134
Tulare (TL), Delta and East-Delta. In this study, we compare groundwater storage changes in 135
three major hydrologic regions, the Sacramento, San Joaquin and Tulare. 136
3 Approach and data 137
3.1 Modeling approach 138
7
We used a range of climate model projections and hydrologic and water resources management 139
models to evaluate groundwater storage changes in major hydrologic regions. In general, our 140
headwaters to groundwater modeling framework can be divided into three major components, 141
which include models representing the headwater watershed, headwater reservoir regulation, and 142
valley integrated surface-groundwater. We linked each of the components modeled 143
unidirectionally and forced with different climate projections and scenarios. Such integration of 144
systems of models can effectively estimate potential changes in groundwater due to climate 145
change over the CV (Hanson et al. 2012). The sequence of steps we followed are to (1) identify 146
representative climate models and extract downscaled climate variables; (2) simulate headwater 147
watershed runoff using the Variable Infiltration Capacity (VIC) model (Liang et al. 1994); (3) 148
model headwater reservoir storage and releases using the Central Valley model (CVmod: Van 149
Rheenen et al. 2004); (4) model the integrated surface-groundwater system using the California 150
Central Valley Groundwater-Surface Water Simulation Model (C2VSIM: Brush et al. 2013a); 151
and (5) assess future trends and variability of groundwater for multiple climate scenarios (Fig. 2). 152
153
3.2 Headwaters to groundwater modeling framework 154
The headwater watersheds in our modeling framework are represented by the VIC model - a 155
physically based semi-distributed hydrological model that solves the land surface water and 156
energy budgets and provide outputs at a grid cell level (Liang et al. 1994). VIC has a long history 157
of development and successful application in numerous studies, including previous applications 158
over part or all of the CV system by Van Rheenen et al. (2004), Xiao et al. (2017), and others. 159
For the current study, we use VIC simulations by Pierce et al. (2018) that are available at daily 160
timestep and 1/16th degree spatial resolution for both a historical period (driven by the Livneh et 161
8
al. (2015) gridded historical data set) and future period (driven by LOCA downscaled climate 162
projection for 32 climate models by Pierce et al. 2014). We obtained the VIC simulation data 163
from the Cal-Adapt online portal managed by the Geospatial Innovation Facility at University of 164
California, Berkeley. We aggregated VIC-simulated daily runoff to monthly to generate inflows 165
to reservoirs and other CV inflow points. To check that VIC simulated headwater watershed 166
flows with reasonable accuracy we compared annual flow from major watersheds with available 167
unimpaired flow estimates (CDWR 2016). The simulated flow from the model and unimpaired 168
flow are generally in agreement (see Table S1 and Fig. S1 in supplementary information). We 169
then used the headwater inflows as input to the headwater reservoir model (where reservoirs 170
exist) and surface-groundwater model (in headwater watersheds without reservoirs). 171
172
We simulated headwater reservoir operation using the reservoir simulation model CVmod (Van 173
Rheenen et al. 2004). The original version of CVmod simulated reservoir storage in the 174
Sacramento and San Joaquin basins. CVmod operates at a monthly time step. The model takes 175
natural inflows from surrounding headwater basins (observed or simulated by a hydrological 176
model such as VIC) as input and simulates reservoir storage and releases based on operations 177
related to flood control, hydropower production, navigation, instream flow requirement for fish, 178
and water supply demands. We updated the model’s parameters to better match historical storage 179
and release observations. We also extended the model to represent four additional reservoirs in 180
the Tulare basin (Pine Flat, Isabella, Success, and Kaweah Lake), calibrated with historical flows 181
and evaluated performance relative to observed reservoir storage and releases. We ran CVmod 182
using VIC simulated headwater inflows to obtain reservoir release and storage for both historical 183
9
and future climate simulations (Fig.1 shows reservoirs considered in this study). Details of model 184
performance are given in Text S2, Table S2 and Fig. S2 of supplementary information. 185
186
We used C2VSIM to simulate CV surface water and groundwater dynamics. C2VSIM was 187
originally developed by CDWR and has a long history of development and applications 188
including use with climate change studies (Brush et al. 2013a; Brush and Dogrul 2013; Dogrul et 189
al. 2016). The model simulates groundwater flow in three layers on finite element grids and 190
dynamically couples with a one-dimensional simulation of land surface hydrologic processes, 191
streamflow, vertical movement in the vadose zone, lake, and flow from adjacent ungauged 192
watersheds (Brush and Dogrul, 2013). The groundwater flow equations are solved at 1392 finite 193
element grids (51,000 km2) and land surface hydrologic processes are computed at 21 water 194
balance subregions - CDWR and the US Bureau of Reclamation have subdivided the CV for 195
purpose of estimating surface water delivery and groundwater use. We used a coarse grid version 196
of the model with element sizes ranging from 5.4 km2 to 87 km2 (average 37 km2). The model 197
estimates water demand given irrigated land use as determined by agricultural management 198
parameters, crop types, water supply from surface water diversions groundwater pumping, and 199
climatic forcing (Dogrul et al. 2016). Brush et.al. (2013b) calibrated C2VSIM for the period 200
1975-2003 to produce a good match between simulated and observed streamflow, groundwater 201
heads, and head differences between wells. C2VSIM as we implemented it runs for the period 202
October 1921 to September 2009 at a monthly time step (see Brush et al. 2013a for details). 203
204
The inputs to C2VSIM include precipitation, crop evapotranspiration (ETc), crop types, surface 205
water diversions, urban water demand and, inflow from surrounding watersheds. To demonstrate 206
10
the direct impact of climate change on agriculture and groundwater resources, we set urban and 207
agricultural land use patterns, and irrigation efficiency to 2003 values. However, given that urban 208
water demand is anticipated to increase in the future (Johnson 2009), we applied a 1.2% annual 209
rate of increase in urban water demand, similar to Hanson et al. (2012), from year 2000 through 210
2098. We computed ETc as the product of the crop coefficient (Kc) and reference crop 211
evapotranspiration following Allen et al. (1998). The mean values and seasonal variations in Kc 212
are obtained from Brush et al. (2004) and ITRC (2003). We estimated reference crop 213
evapotranspiration (ETo) using the Penman-Monteith equation applied following Allen et.al. 214
(1998) at 1/16-degree spatial resolution for the study region. This ETo calculation requires air 215
temperature, vapor pressure deficit, relative humidity, solar radiation, and wind speed (Eq 1). We 216
extracted all the required variables from VIC (driven by Livneh et al, 2015 data for the historical 217
period and LOCA climate projections from Pierce et al. 2014) for the future. We temporally 218
aggregated ETo to obtain monthly data from daily time series and extracted the subregion-219
averaged ETo. Twelve major crop types are used by C2VSIM including virtual crops that 220
aggregate multiple actual crops (e.g. grains, orchard, field crops, truck crops). To verify that 221
C2VSIM performed well during the historical simulation driven by the Livneh et al. (2015) data, 222
we compared our groundwater storage change results with previous studies (shown in section 4). 223
𝐸𝑇0 =0.408∆(Rn − G) + 𝛾
900T + 273 𝑒2(𝑒𝑠 − 𝑒𝑎)
∆ + 𝛾(1 + 0.34𝑢2) (1) 224
In equation 1, Rn is net radiation, G is ground heat flux (assumed zero), T is mean daily air 225
temperature at 2 m height (℃), 𝑢2 is wind speed at 2 m height, 𝑒𝑠 is saturation vapor pressure, 𝑒𝑎 226
is actual vapor pressure, Δ is the slope of the saturation vapor pressure curve, and 𝛾 is the 227
psychrometric constant. 228
11
Additionally, C2VSIM requires user-defined surface water diversion information along the 229
stream network (shown in Fig. 1). Previous studies of the CV and its contributing watersheds 230
used relatively simple methods to estimate surface water deliveries. For instance, Hanson et al. 231
(2012 and 2005) estimated future diversions based historical analogs (partitioning years into wet 232
and dry periods), Miller et al. (2009) used historic diversion statistics (mean, median). We 233
developed a simple linear programming (LP) model with surface water networks similar to 234
C2VSIM to estimate surface water diversions (Fig. 1 shows the network and diversion locations; 235
a more detailed description of the diversion locations is provided by Brush and Dogrul, 2013). 236
The objective of the LP model is to maximize the surface water delivery under constraints like 237
mass balance, and maximum/minimum instream flow (see Text S3 of supplementary 238
information). We used upstream headwater watershed flows as the upstream boundary to the LP 239
model. We first ran the LP model at a monthly time step for the projection period and then 240
applied the surface water diversion output to C2VSIM. In C2VSIM, surface water deliveries are 241
used first to meet demand; any deficit in surface water supply is then met by pumping 242
groundwater, which implies that maximizing diversions reduces groundwater pumping. 243
244
3.3 Climate models, and crop and hydrologic scenarios 245
To account for the uncertainty associated with future projections from different GCMs, we 246
performed a preliminary analysis to select suitable climate models for this study. We compared 247
the empirical cumulative distribution functions (ECDFs) of mean annual runoff (VIC-simulated 248
annual total runoff averaged over 2006-2099 for the Sacramento, San Joaquin, and Kings Rivers 249
for RCP 8.5) among different combinations of GCMs (for more information refer to Text S4 and 250
Fig. S3 of supplementary information). Based on our comparison we selected 20 GCMs that 251
12
capture a range of variability out of the 32 GCMs and archived in the Coupled Model 252
Intercomparison Project phase 5 (CMIP5; the selected GCMs are shown in Table S3 of 253
supplementary information). We considered both RCP 4.5 and RCP 8.5 global emissions 254
scenarios. We first ran integrated chain of models for the historical period (1950-2009), where 255
land use and cropping patterns vary annually (referred to as “Hist” hereafter). Next, we ran the 256
models with historical climate (1950-2009) and assumed fixed (to 2003 levels) agricultural 257
practices and cropping patterns (referred to as “Baseline”). To assess the impact of climate 258
change, we ran the models with future climate (2010-2098) forcing based on projections for 259
RCP4.5 and RCP8.5 (we refer to this simulation as ‘CCR’ hereafter), where we assume fixed (at 260
2003 levels) agricultural practices and cropping pattern. We also conducted scenario analyses to 261
better understand the underlying causes of future climate impact on groundwater storage. We 262
analyzed the simulation results for three future periods: beginning of century (BOC; years 2010-263
2039), mid-century (MOC; years 2040-2069) and end of century (EOC; years 2070-2098). The 264
formulation of scenarios is detailed in sections 3.3.1 and 3.3.2 as well as in Table 1. 265
266
3.3.1 Future crop-water use and headwater inflow change scenarios 267
The amount of groundwater being pumped to meet agricultural demands in CV is essentially a 268
balance between the crop water demand, precipitation and surface water supply from headwater 269
watersheds. Crop evaporative demand increases with temperature, or precisely, with 270
temperature-dependent variables such as the vapor pressure deficit, net radiation, and the slope 271
of the vapor pressure relationship with temperature). Historically, most of the total crop-water 272
demand in the northern part of the CV has been satisfied by precipitation (52.7%) and surface 273
water supply (25.3%), whereas lower precipitation (18.6%) and surface water (24.6%) 274
13
availability in the southern part of the domain has resulted in higher groundwater pumping (Li et 275
al. 2018). As the climate continues to change, both demand and supply are expected to change, 276
and the magnitudes of which will vary over regions within the CV depending on climate. To 277
better understand the effect of climate change on groundwater storage, we explored a range of 278
scenarios that quantify ΔGW associated with crop-water use (which is a balance between crop 279
evaporative demand and precipitation) in the CV and surface water supply from headwater 280
watersheds (headwater inflow hereafter). As a proxy for current crop-water use and headwater 281
inflow, we randomly sampled the baseline (1950-2009) historical analysis to generate a set of 282
synthetic time series of monthly data for 30 years. We repeated the random sampling two more 283
times to create 90 years of data (we used same 30 years data for each future period to be 284
consistent in comparing the outputs between the periods). We implemented those 90 years with 285
baseline agricultural management (at 2003 levels). To identify the ΔGW resulting from crop-286
water use versus headwater inflow, we performed two scenario analyses: (1) current crop-water 287
use and future inflows (FI), and (2) current inflows and future crop water use (FC). 288
289
3.3.2 Future headwater inflow volume and seasonality change scenarios 290
Both the volume (of annual inflow to the CV) and inflow timing (seasonality) of headwater 291
inflows are expected to change in the future. The effect of these projected changes on 292
groundwater storage is not well known. To better understand the ΔGW that are likely to be 293
associated with future headwater inflow volume vs seasonality, we conducted two scenario 294
analyses by adjusting the projected inflows. We multiplied the projected headwater inflows 295
(2010-2098) by perturbation ratios between projected inflows and historical inflows following 296
Wang et al. (2011) and Miller et al. (2003). The scenarios are described as follows: 297
14
Annual inflow volume adjustment 298
We determined annual perturbation ratios (𝐴𝑐) from the 30-year (1971-2000) historical mean 299
annual inflows (𝑅) and 30-year projected mean annual inflows (𝑅′𝑐) for three future periods 300
(2010-2039, 2040-2069 and 2070-2099). We then multiplied the projected inflows (𝑅′𝑐𝑖𝑗, 𝑐 = 301
2010-2039, 2040-2069 and 2070-2099, i = 1, . . . ,12 for each month, j = 1, . . ,90 years) by these 302
ratios. The future time series (𝑇𝑐𝑖𝑗) are modified to have the same annual means as the historical 303
inflows (1971-2000), but their seasonality is unchanged. 304
𝐴𝑐 =
𝑅
𝑅′𝑐
(2)
𝑇𝑐𝑖𝑗 = 𝐴𝑐 × 𝑅′𝑐𝑖𝑗 (3) 305
The modified inflows corresponding to each GCM were used as input to the integrated chain of 306
models. We refer to this scenario as future seasonality change only or I_SC hereafter. The 307
difference in groundwater estimation relative to the CCR simulation shows the effect of future 308
annual volume (changes) of future inflows. 309
310
Monthly inflow perturbation 311
We also generated headwater inflow time series that keep the projected annual inflow volumes 312
unchanged in the future, with monthly (mean) fractions of mean annual inflows the same as in 313
the historical (1971-2000) period. We determined the mean monthly fractions of annual flows 314
from the 30-year historical period (𝐵𝑖, 𝑖 = 1, … ,12 for each month). We then multiplied the 315
projected annual flows (𝑅′𝑗) by the fractions (𝐵𝑖) to obtain inflow time series (𝐺𝑖𝑗) with the 316
historical seasonality. Accordingly, the projected annual volumes were unchanged. 317
𝐺𝑖𝑗 = 𝐵𝑖 × 𝑅′𝑗 𝑖 = 1, . . ,12; 𝑗 = 1, . . ,90 (4) 318
15
The modified inflows were then used as input to the integrated chain of models. This scenario is 319
referred to as annual volume change only or I_AC hereafter. The difference in groundwater 320
estimation relative to the CCR simulation shows the effect of timing (changes) of future inflows. 321
322
3.4 Cropping pattern scenario 323
We quantify the ΔGW that would occur due to a shift from mostly row crops to tree crops 324
(together with future climate). C2VSIM, which is the core of this analysis, has 12 crop categories 325
(pasture, alfalfa, sugar beet, field crop, rice, truck crop, tomato, orchard, grains, vineyard, cotton, 326
citrus-olives). For the scenario analysis, we specified 40% and 60% row crops (classification of 327
row and tree crops obtained from Xiao et al. 2017) to high water consumptive tree crops 328
(orchard). One motivation behind choosing 40% change is from Xiao et al. (2017), who showed 329
that between 2007-2016 there was an average 40% shift from row crops to tree crops, suggesting 330
that much of the change we “project” has already occurred. In our case, we applied the shift 331
relative to the baseline cropping pattern (2003). We ran the models with 40% and 60% crop 332
shifts for all (20) future GCM climate scenarios (referred to hereafter as CS). The difference 333
between the CS and CCR scenarios (ΔGW) is due to a shift from row to tree crops. We use the 334
additional ΔGW estimates from crop-shift scenarios (40% and 60%) to identify percent shift 335
(row to tree crops) at which groundwater change due to crop-shift and climate change (RCP8.5) 336
are equal. 337
338
4 Results 339
4.1 Changes in hydrology 340
16
Fig. 3 shows mean annual changes (%) in precipitation (P), headwater inflows (I) and reference 341
evapotranspiration (ETo) for all 20 climate models and scenarios during three future periods 342
(BOC, MOC, and EOC representing beginning, middle, and end of century) compared to the 343
historic climate reference period (1971-2000). Multi-model precipitation estimates over CV have 344
a median annual decrease of -2.2% and -4.4% for RCP4.5 and RCP8.5 respectively near EOC, 345
however the variability of the multi-model estimates is quite high (standard deviations of 8.9% 346
and 15.2% for RCP4.5 and RCP8.5 respectively). On the other hand, temperature increases in the 347
future will increase crop evaporative demands (here we show ETo change as proxy). Climate 348
change will cause ETo over the CV to increase by 6.6 ± 1.5% and 10.5 ± 1.8% over the CV near 349
the end of the century (EOC) for RCP4.5 and RCP8.5 respectively. The rate of ETo increase is 350
higher in RCP8.5 due to continuing increase in temperature, whereas, the rate of increase 351
decreases slightly after mid-century (MOC) in RCP4.5. Additionally, projected changes in 352
temperature and precipitation over the headwater watersheds results in varying impacts on 353
headwater inflows (I). Mean annual inflows have negative trends but high variability especially 354
towards the end of the century, with changes of -4.20% ± 12% and -4.37% ± 19% at MOC; -355
3.59% ±16% and -8.19% ± 23% at EOC for RCP4.5 and RCP8.5 respectively (changes of 356
hydrologic variables shown in Table S4 of supplementary information). 357
358
Headwater inflow changes can be both positive and negative (in contrast to ETo which increases 359
in all models). The variability (standard deviation) of future inflow changes is higher than the 360
median in all cases; the headwater runoff projections range from negative to positive depending 361
on the GCM. Moreover, there is a contrasting trend in the cool season (October through March) 362
versus the warm season (April through September) flows where cool season flows are expected 363
17
(in the median) to increase, and warm season flows are expected to decrease. The variability in 364
multi-model estimates increases towards the end of century (percentage changes in seasonal 365
inflows are shown in Fig. 3). The median changes in annual inflows is negative, this is 366
dominated by warm season decreases vs. cool season increases (volumetric changes in inflow are 367
shown in Fig. S4 of supplementary information). To get an idea of multi-model agreement as to 368
the sign of the median change- we calculated the number of concordant minus discordant pairs, 369
divided by the numbers of pairs for annual precipitation, annual and seasonal inflows (shown in 370
Fig. S5 and Fig. S6 of supplementary information). We find that there is little agreement in the 371
annual estimates of inflows (concordance 0.1-0.3) and precipitation (concordance 0.1-0.6, lowest 372
during EOC) among models over the CV. There is, however, greater agreement at the seasonal 373
level of inflow changes (concordance in warm season 0.9-1 and cool season 0.4-0.9). On 374
average, concordance for decreases in inflow during the warm season is higher than for increases 375
in inflow during the cool season. 376
377
At the regional level, changes in precipitation are smaller (in absolute value) in the Sacramento 378
than in the San Joaquin and Tulare, where across-GCM variability is also larger (Fig. 3). 379
Similarly, mean annual runoff decreases in SJ and TL, whereas there is no clear trend in the 380
multi-model median for SC (percent changes in annual volume are higher in TL and SJ than in 381
SC). However, the multi-model median of headwater inflows in all the regions increases during 382
the cool season and decrease during the warm season, with a high degree of concordance in both 383
measures. 384
385
4.2 Historical changes in groundwater storage 386
18
Fig. 4 shows the historical (1950-2009) changes in precipitation, headwater inflows, agricultural 387
demand and cumulative simulated changes in groundwater storage (ΔGW hereafter). 388
groundwater storage decreased monotonically during the first half of our historical record (1950-389
1979) with a depletion rate of 4.1 km3/yr. During the latter half (1980-2009), the interannual 390
variability in ΔGW increased due to intermittent wet and dry years (as seen in the precipitation 391
and inflow time series shown in Fig. 3). The long term (1950-2009) historical groundwater 392
depletion rate is around 3.0 km3/yr. For the Baseline scenario (2003 cropping patterns), the 393
simulated cumulative ΔGW rate is -3.1 km3/yr (time series ΔGW shown in Fig. S7 of 394
supplementary information). The Baseline groundwater depletion rate is slightly higher than the 395
historical rate, mainly due to difference in the historical evolution of land-use and cropping 396
patterns; the historical run considers changes (i.e. agricultural area increased monotonically 397
between 1950 through 1975, then became relatively stable) while the Baseline run considers 398
land-use and cropping patterns to be fixed at 2003 levels. 399
400
We compared the simulated ΔGW from C2VSIM with other available estimates. The simulated 401
changes (mean trend) in ΔGW for April 2006 to September 2009 from C2VSIM is -8 km3/yr, 402
compared to -7.9 ± 1.3 km3/yr by Xiao et al. (2017) and -7.8 km3/yr by Scanlon et al. (2012a). 403
GRACE-based estimates include -7.2 ± 1 km3/yr by Xiao et al (2017) and -6.0 km3/yr by 404
Famiglietti et al. (2011) for the period April 2006 to March 2010 which is also close to our 405
simulated estimate of ΔGW. Fig. 4e compares monthly GRACE ΔGW from Xiao et al. (2017) 406
and C2VSIM; it is visually evident that C2VSIM reasonably captures the increasing/decreasing 407
trends. 408
409
19
4.3 Climate change impacts on groundwater storage 410
We quantify ΔGW for the projection period (2010-2098) where the forcing comes from 20 411
climate models and 2 RCP scenarios. Fig. 5a shows ΔGW for the entire CV and its major 412
hydrologic regions (ΔGW time-series for each climate model are shown in Fig. S8 of 413
supplementary information). It is visually evident that in both RCP scenarios, the CV is expected 414
to go through a continuous groundwater depletion with higher multi-model variability towards 415
the end of the century. Median ΔGW at the end of the century (2098) is expected to be -373 ± 88 416
km3 for RCP4.5 and -406 ± 116 km3 for RCP8.5. Furthermore, we fit linear regressions and 417
estimated trends in ΔGW over the future periods (Fig. 5b). The long-term (2010-2098) median 418
(over GCMs) trend in ΔGW is -4.1±1.1 km3/yr for RCP4.5 and -4.4±1.4 km3/yr for RCP8.5 419
respectively. These represent median trend increases of 31% and 39% compared to the baseline 420
trend (2003 cropping patterns) for RCP4.5 and RCP8.5 respectively. The rate of depletion and 421
uncertainty is higher during EOC, with a median rate ΔGW of -4.4±1.1 km3/yr (RCP4.5) and -422
4.9±1.4 (RCP8.5) km3/yr, which is equivalent to increases by 40% and 56% increase in ΔGW 423
depletion rate compared to the baseline trends (2003 cropping patterns), respectively. Overall, 424
there will be 40-70% increase (interquartile range of all percent changes) in GW depletion over 425
the entire CV. 426
427
The rate of groundwater decline under future climate change scenarios is not uniform across the 428
CV. The southernmost part (the TL region) will experience higher groundwater declines 429
compared to the central and northern parts (SJ and SC). On average, due to future climate 430
conditions, TL will experience 88% more groundwater decline than either SJ and SC. Separating 431
the results according to the major hydrologic regions of CV, median rates of change in ΔGW in 432
20
TL during the projection period are -2.5 ± 0.7 km3/yr (RCP4.5) and -2.5 ± 0.9 km3/yr (RCP8.5), 433
which is equivalent to 26% and 28% increases in ΔGW depletion rate compared to the baseline 434
trends (-2.0 km3/yr) for RCP4.5 and RCP8.5 respectively in TL. The second highest depletion is 435
expected in SJ where long-term rate of ΔGW change are -0.78±0.2 km3/yr (RCP4.5) and -436
0.86±0.3 km3/yr (RCP8.5), which is equivalent to 49% and 65.7% increases in ΔGW depletion 437
rate compared to Baseline (-0.52 km3/yr) for RCP4.5 and RCP8.5 respectively. ΔGW in SC is 438
even smaller, with rate of change in ΔGW is expected to be -0.38±0.11 km3/yr (RCP4.5) and -439
0.48 ± 0.14 km3/yr (RCP8.5), which is an increase in ΔGW depletion rate by 31.3% and 63.5% 440
compared to Baseline (-0.29 km3/yr) for RCP4.5 and RCP8.5 respectively. Groundwater losses 441
(in absolute value) are higher for all regions at the end of century (2070-2098) and higher for 442
RCP8.5 than for RCP4.5. In particular, the median (over GCMs) ΔGW rates will increase by 443
44.4%, 47.1% and 33.3% respectively under RCP4.5, and 80.9%, 53.9% and 47.5% for SC, SJ, 444
and TL, respectively under RCP8.5 at the end of century (regional time series of ΔGW are 445
shown in Fig. S9 of supplementary information). 446
447
4.4 Crop-shift scenario 448
We computed the ΔGW associated with a 40% and 60% shift towards tree crops (see section 3.4) 449
for all climate models (RCP8.5 scenario only). In general, a shift towards tree crops is expected 450
to increase water use requirements, but some of the increased water demand is due to climate 451
change, so we partitioned the climate- and crop change-related components. We found that a 452
40% shift towards tree crops (compared to 2003) causes an additional 0.91 km3/yr groundwater 453
loss, which is equivalent to a 29% increase ΔGW depletion rate compared to the baseline ΔGW 454
rate. Groundwater depletion from the combination of future climate (RCP8.5) and crop-shift 455
21
(40%) shows that the relative contributions of climate and crop-shift are about 58% and 42% 456
respectively. However, it is important to note that the 40% shift we considered is with respect to 457
the 2003 cropping pattern. As noted above much of this shift is already in place as of 2018 (see 458
Xiao et al. 2017). For this reason, we also tested a scenario with 60% crop-shift (row to tree), the 459
results of which are that groundwater declines at a rate of (increase relative to baseline) 1.39 460
km3/yr (equivalent to 44% increase in groundwater depletion rate relative to the baseline). We 461
find that a shift of about 56% (compared to year-2003 and without climate change) results in 462
about the same groundwater depletion as does the climate change (RCP 8.5 and baseline 463
cropping pattern-2003) over the entire 2010-2098 period. While the shift towards trees from row 464
crops causes increased groundwater loss in all the major regions, the percentage changes are 465
greatest in SJ (46% and 70% increase for two crop-shift scenarios), and the absolute changes are 466
(by far) the largest in TL (0.55 km3/yr and 0.83 km3/yr for 40% and 60% shifts respectively). 467
468
5 Interpretations and discussions 469
In section 4, we partitioned groundwater change into a climate change contribution and a crop 470
shift contribution. However, there are three main elements of the climate change contribution: 471
1) changes (increases) in crop water demand due to warmer temperatures, 2) the effect (mostly 472
negative) of changes in annual mean inflows to the groundwater, and 3) changes in the 473
seasonality (generally, more flow in winter and less in spring and summer) of the inflows. 474
475
5.1 Role of future crop-water use versus headwater inflow 476
Here, we analyze the effect of changes in the demand (crop-water use) versus supply (surface 477
water inflows) component of the CV water budget. Fig. 6 shows the annual average ΔGW 478
22
(additional) between FC (future crop-water use) and FI (future inflow) scenarios for three future 479
periods. We calculate the change associated with future crop-water use (FC) from the difference 480
in CCR (climate change run, which includes the effect of future inflow and crop-water use 481
changes) and FI, whereas the changes due to future inflow are obtained from the difference 482
between the CCR and FC scenarios. Here, we first calculate the difference in groundwater 483
anomalies (between CCR and FI or FC) and determine the median of the multi-model ensembles, 484
then we take the mean of the annual changes for each future period (example shown in Fig. S10 485
of supplementary information). 486
487
In all cases, climate change is expected to increase the groundwater depletion. We find that 488
future crop-water use (change) will have a greater (negative) impact on total groundwater storage 489
in the CV than inflow changes (FI) (predominantly during mid to end of century under RCP8.5); 490
the relative contribution attributable to FC and FI are around 60% and 40% respectively. At a 491
regional level, future crop-water use will have greater (negative) impact in SC (on average 60% 492
of ΔGW), whereas future inflow will have almost equal or greater (on average 50-60% of 493
change) impact (negative) in SJ (except RCP8.5 during EOC), indicating that the SJ is 494
susceptible to both inflow changes and increased crop-water use. However, the changes in 495
groundwater in TL due to future inflow and crop-water use are almost the same at BOC and 496
MOC, and changes (negative) due to crop-water use is higher at EOC (on average the relative 497
contribution is 50%). We find similarities in the relative effect of FI and FC in snow dominated 498
regions (SJ and TL) as opposed to the rain dominated region (SC). Our results also indicate that 499
the interannual variability (hence uncertainty) in all cases is higher for FI, which is due to higher 500
23
uncertainty in inflows (as shown in Fig. 3) and relatively less uncertain future increase in crop-501
water use (because temperature 502
rises in all cases). 503
504
5.2 Role of headwater inflow volume and seasonality shifts under future climate 505
Fig. 6 shows the effect of FI and FC, where FI is made up of a combination of changes in annual 506
inflow volumes, and changes in seasonality (Stewart et al. 2004; Regonda et al. 2005). Both 507
effects contribute to ΔGW. To determine the relative magnitude of these effects, we conducted 508
an experiment in which we partitioned the future inflows into seasonal and annual changes 509
(I_SC: seasonality changes in the future, and I_AC: annual volume changes in the future). We 510
then identified the relative differences of I_SC and I_AC from the CCR simulation. Fig. 7 shows 511
the changes in annual average groundwater storage (additional) due to future inflow seasonality 512
(I_SC) and annual volume (I_AC) (higher negative ΔGW means greater groundwater loss) 513
changes. The changes associated with future seasonality are calculated from the difference 514
between CCR and I_AC, whereas the changes due to future volume change are calculated from 515
the difference between CCR and I_SC scenarios. We followed the same process discussed in the 516
previous section (but between CCR and I_SC or I_AC) and compared the changes, shown in Fig. 517
7. We find that ΔGW change attributable to FI is about 80% from future headwater seasonality 518
change and 20% from future annual volume changes. The highest effect of seasonality changes 519
is in SC and SJ. On the other hand, inflow volume change has almost equal effect as seasonality 520
change in the TL region. We note that the effect of changes in natural inflow are mediated to 521
varying degrees by headwater regulation (regulated inflows under different scenarios are shown 522
24
in Fig. S11 of supplementary information); this mediation is much greater for seasonality as 523
contrasted with volume changes. 524
525
6 Summary and Conclusions 526
Groundwater plays a critical role in the Central Valley’s agricultural production. The 527
vulnerability of groundwater to climate in the Central Valley is evident from the drastic decline 528
of this resource over the past few decades (Xiao et al. 2017), during which average annual 529
groundwater loss has exceeded 3.0 km3/yr. Given this backdrop, our objective was to assess the 530
impact of future climate change on Central Valley (CV) groundwater storage. Our methodology 531
involves (1) simulation of surface and groundwater in the region using chain of integrated 532
models (C2VSIM, CVmod and VIC), (2) forcing the models with climate model projections 533
under different climate change scenarios (RCP4.5 and RCP8.5), and (3) simulation experiments 534
for cropping pattern change scenario. Based on our results, we conclude that: 535
• Groundwater storage in CV has been declining in recent decades (1950-2009) at an average 536
rate of about 3 km3/yr. This decline in groundwater has resulted from crop water use 537
exceeding natural inflows and groundwater recharge, which has been exacerbated by a 538
combination of a gradually warming climate, and from progressive shifts in crop types that 539
have led higher crop water use. 540
• In the absence of mitigation measures, over the future period (2010-2098), groundwater 541
loss is likely to continue, but at an increased rate. Assuming no (further) change in crop 542
types (relative to 2003), the long-term (2010-2098) rate of groundwater decline in CV 543
under future climate will increase from about 3.0 km3/yr to 4.1 ± 1.1 km3/yr (RCP4.5; 544
25
modest greenhouse emission scenario) and 4.4 ± 1.4 km3/yr for RCP8.5 (worst scenario), 545
or 31-39% higher than our base case with 2003 crop patterns. 546
• Groundwater declines are associated both with (1) more groundwater pumping due to 547
increased crop-water demands associated with rising temperatures (and to a lesser extent, 548
reduced precipitation), and (2) reduced surface water supply to CV from headwater 549
watersheds. We find that, absent mitigating measures, the dominant cause of future 550
groundwater declines will be increases in future crop-water (~60% of total change). We 551
note that the projections of headwater inflow changes are much more uncertain (hence 552
projections variable) as they are dominantly associated with precipitation changes, whereas 553
projections of crop water use, which are dominated by temperature changes, are less 554
uncertain. 555
• Groundwater declines, both in the past and projected for the future, are dominated by the 556
Tulare (TL) region, which has accounted for roughly 80 percent of the historical decline. 557
However, the sensitivities of future groundwater loss to climate-related crop water use and 558
inflow changes are not necessarily greatest in the TL basin. 559
California’s new Sustainable Groundwater Management Act (SGMA) mandates that 560
groundwater depletions be brought into balance. Increases in crop water use, and reduced 561
surface water availability associated with climate warming constitute an additional stress on 562
the system; depending on specifics, these additional stresses will increase the current 563
imbalance by one-third to one-half. 564
565
Acknowledgements 566
26
The University of California Research Initiatives Award LFR-18-548316 supported this work. 567
We are grateful to Dr. Tariq Kadir and Dr. Emin Dogrul from California’s Department of Water 568
Resources for technical assistance in the use of C2VSIM. We are also grateful to the kind help of 569
Dr. Alan Hamlet of University of Notre Dame for his assistance with CVmod. 570
571
References 572
Allen RG, Pereira LS, Raes D, Smith M (1998) FAO Irrigation and drainage paper No. 573
56. Rome: Food and Agriculture Organization of the United Nations, 56(97), e156. 574
Bales R.C., Molotch N.P., Painter T.H., Dettinger M.D., Rice R. and Dozier J. (2006) Mountain 575
hydrology of the western United States. Water Resources Research, 42(8). 576
Bates BC, Kundzewicz ZW, Wu S, Palutikof JP (2008) Climate change and water. Technical 577
Paper of the Intergovernmental Panel on Climate Change, IPCC Secretariat, Geneva. 578
Brush CF (2013b) Historical Rim Inflows, Surface Water Diversions and Bypass Flows for the 579
California Central Valley Groundwater-Surface Water Simulation Model (C2VSim), Version 580
3.02-CG. 581
Brush CF, Belitz K, Phillips SP (2004) Estimation of a Water Budget for 1972-2000 for the 582
Grasslands Area, Central Part of the Western San Joaquin Valley, California. US Department of 583
the Interior, US Geological Survey. 584
Brush CF, Dogrul EC (2013) User's Manual for the California Central Valley Groundwater-585
surface Water Simulation Model (C2VSim), Version 3.02-CG. Bay-Delta Office, California 586
Department of Water Resources. 587
27
Brush CF, Dogrul EC, Kadir TN (2013a) Development and calibration of the California Central 588
Valley groundwater-surface water simulation model (C2VSim), Version 3.02-CG. Bay-Delta 589
Office, California Department of Water Resources. 590
California State Legislature (2014) Sustainable Groundwater Management Act. Accessed April 591
12, 2018. https://www.opr.ca.gov/docs/2014_Sustainable_Groundwater_Management_ 592
Legislation_092914.pdf. 593
CDWR (2016) Estimates of Natural and Unimpaired Flows for the Central Valley of California: 594
Water Years 1922–2014. Report from California Department of Water Resources. 595
Chen J, Famigliett JS, Scanlon BR, Rodell M (2016) Groundwater storage changes: present 596
status from GRACE observations. InRemote Sensing and Water Resources (pp. 207-227). 597
Springer, Cham. 598
Cohen SJ, Miller KA, Hamlet AF, Avis W (2000) Climate change and resource management in 599
the Columbia River Basin. Water international. 25(2):253-72. 600
Dogrul EC, Kadir TN, Brush CF, Chung FI (2016) Linking groundwater simulation and reservoir 601
system analysis models: The case for California’s Central Valley. Environmental modelling & 602
software. 77:168-82. 603
Famiglietti JS (2014) The global groundwater crisis, Nat. Clim. Change, 4(11), 945–948, 604
doi:10.1038/nclimate2425. 605
Famiglietti JS, Lo M, Ho SL, Bethune J, Anderson KJ, Syed TH, Swenson SC, De Linage CR, 606
Rodell M (2011) Satellites measure recent rates of groundwater depletion in California's Central 607
Valley. Geophysical Research Letters, 38(3). 608
28
García-Garizábal I, Causapé J, Abrahao R, Merchan D (2014) Impact of climate change on 609
Mediterranean irrigation demand: historical dynamics of climate and future projections. Water 610
Resour Manag 28(5):1449– 1462 611
Garrote L, Iglesias A, Granados A, Mediero L, Martin-Carrasco F (2015) Quantitative 612
assessment of climate change vulnerability of irrigation demands in Mediterranean Europe. 613
Water resources management, 29(2):325-38. 614
Gleick PH (2000) Water: The Potential Consequences of Climate Variability and Change for the 615
Water Resources of the United States, National Water Assessment Group for the U.S. Global 616
Change Research Program. 617
Hanson RT, Dettinger MD (2005) Ground-water/surface-water responses to global climate 618
simulations, Santa Clara–Calleguas Basin, Ventura County, California, 1950–93, J. Am. Water 619
Resour. Assoc., 41(3), 517–536. 620
Hanson RT, Flint LE, Flint AL, Dettinger MD, Faunt CC, Cayan D, Schmid W (2012) A method 621
for physically based model analysis of conjunctive use in response to potential climate changes. 622
Water Resources Research, 48(6). 623
ITRC (2003) California crop and soil evapotranspiration for water balances and irrigation 624
scheduling/design. California Polytechnic State University, web: 625
http://www.itrc.org/reports/pdf/californiacrop.pdf 626
Johnson H (2009) California population: Planning for a better future,report, Public Policy Inst. of 627
Calif., San Francisco. 628
29
Li R, Ou G, Pun M, Larson L (2018) Evaluation of Groundwater Resources in Response to 629
Agricultural Management Scenarios in the Central Valley, California. Journal of Water 630
Resources Planning and Management, 144(12):04018078. 631
Liang X, Lettenmaier DP, Wood EF, Burges SJ (1994) A simple hydrologically based model of 632
land surface water and energy fluxes for general circulation models. Journal of Geophysical 633
Research: Atmospheres, 99(D7):14415-28. 634
Livneh B, Bohn TJ, Pierce DW, Munoz-Arriola F, Nijssen B, Vose R, Cayan DR, Brekke L 635
(2015) A spatially comprehensive, hydrometeorological data set for Mexico, the US, and 636
Southern Canada 1950–2013. Scientific data, 2:150042. 637
Massoud EC, Purdy AJ, Miro ME, Famiglietti JS (2018) Projecting groundwater storage changes 638
in California’s Central Valley. Scientific reports, 8(1):12917. 639
Miller NL, Bashford, KE, Strem E (2003) Potential impacts of climate change on California 640
hydrology. JAWRA Journal of the American Water Resources Association, 39(4), 771-784. 641
Miller NL, Dale LL, Brush CF, Vicuna SD, Kadir TN, Dogrul EC, Chung FI (2009) Drought 642
Resilience of the California Central Valley Surface‐Ground‐Water‐Conveyance System 1. 643
JAWRA Journal of the American Water Resources Association, 45(4):857-66. 644
Olmstead AL, Rhode PW (2017) A History of California Agriculture. Giannini Foundation of 645
Agricultural Economics, University of California 646
Pierce DW, Cayan DR, Thrasher BL (2014) Statistical downscaling using localized constructed 647
analogs (LOCA). Journal of Hydrometeorology, 15(6), 2558-2585.-2585. 648
30
Pierce DW, Kalansky JF, Cayan DR (2018) Climate, Drought, and Sea Level Rise Scenarios for 649
the Fourth California Climate Assessment. California’s Fourth Climate Change Assessment, 650
California Energy Commission. Publication Number: CNRA-CEC-2018-006. 651
Regonda SK, Rajagopalan B, Clark M, Pitlick J (2005) Seasonal cycle shifts in hydroclimatology 652
over the western United States. Journal of climate, 18(2):372-84. 653
Scanlon BR, Faunt CC, Longuevergne L, Reedy RC, Alley WM, McGuire VL, McMahon PB 654
(2012b) Groundwater depletion and sustainability of irrigation in the US High Plains and Central 655
Valley. Proceedings of the national academy of sciences, 109(24):9320-5. 656
Scanlon BR, Longuevergne L, Long D (2012a) Ground referencing GRACE satellite estimates 657
of groundwater storage changes in the California Central Valley, USA. Water Resources 658
Research. 48(4). 659
Stewart IT, Cayan DR, Dettinger MD (2004) Changes in snowmelt runoff timing in western 660
North America under abusiness as usual'climate change scenario. Climatic Change, 62(1-3):217-661
32. 662
VanRheenen NT, Wood AW, Palmer RN, Lettenmaier DP (2004) Potential implications of PCM 663
climate change scenarios for Sacramento–San Joaquin River Basin hydrology and water 664
resources. Climatic change, 62(1-3):257-81. 665
Wang J, Yin H, Chung F (2011) Isolated and integrated effects of sea level rise, seasonal runoff 666
shifts, and annual runoff volume on California’s largest water supply. Journal of hydrology, 667
405(1-2):83-92. 668
31
Xiao M, Koppa A, Mekonnen Z, Pagán BR, Zhan S, Cao Q, Aierken A, Lee H, Lettenmaier DP 669
(2017) How much groundwater did California's Central Valley lose during the 2012–2016 670
drought?. Geophysical Research Letters, 44(10):4872-9. 671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
32
Table 692
Table 1: Summary of the scenarios considered in this study 693
Scenario Description Period
Hist Historical climate and cropping pattern 1950-2009
Baseline Historical climate and cropping pattern for 2003 1950-2009
CCR Climate change run (future forcing + 2003 cropping pattern) for RCP 4.5 and RCP 8.5 2010-2098
FC Future crop-water use (CCR forcing over CV + headwater inflows as Hist + 2003
cropping pattern)
2010-2098
FI Future inflows (Hist forcing over CV + headwater inflows as CCR + 2003 cropping
pattern)
2010-2098
I_SC Headwater inflows seasonality change (seasonality as CCR + mean annual volume as
Hist) + 2003 cropping pattern
2010-2098
I_AC Headwater inflows annual volume change (mean annual volume as CCR + seasonality as
Hist) + 2003 cropping pattern
2010-2098
CS Cropping shift (future forcing + 40% and 60% shift from row to tree crops) 2010-2098
694
695
696
697
698
699
700
701
702
703
704
705
33
Figures 706
707
Fig. 1. Map showing Central Valley and its surrounding watersheds (headwater watersheds). The 708
five major hydrologic regions and rivers flowing through them are shown in the map to the right. 709
Headwater watersheds and major reservoirs we represent in the study region are also shown 710
(right). Blue shades indicate elevations of headwater watersheds. 711
712
713
714
34
715
Fig. 2. Connectivity and flow of data in our integrated modeling system. Items in the dashed box 716
are the key analysis and outputs. 717
718
719
720
721
722
723
724
725
726
727
728
729
35
730
Fig. 3. Changes in annual precipitation (Pann), ETo and headwater inflow (Iann) in CV and major 731
regions for three future periods relative to 1971-2000. Seasonal variation in inflows are shown in 732
bottom two rows. (sky-blue: RCP4.5, brown: RCP8.5). Horizontal axis represents three future 733
periods - beginning of century (BOC; 2010-2039); middle of century (MOC; 2040-2069); end of 734
century (EOC; 2070-98). 735
36
736
737
Fig. 4. Historical estimate of (a) annual precipitation, (b) annual inflows, (c) agricultural 738
demand, (d) annual cumulative groundwater storage changes (ΔGW). (e) ΔGW comparison 739
between C2VSIM simulation and GRACE estimate at monthly timestep. 740
741
742
743
744
745
746
747
37
748
Fig. 5. (a) ΔGW (relative to 2010) in CV and major regions for RCP4.5 (sky-blue) and RCP8.5 749
(brown). The boxplots represent ΔGW obtained by forcing the chain of models with forcing from 750
20 climate models (ΔGW till 2040, 2070 and 2098 reported in the plot). (b) ΔGW rate during 751
three future period. Whiskers represent min (max) value or 1.5 times interquartile range from 752
first (third) quartile, whichever is bigger (smaller). 753
754
755
756
757
758
759
760
38
761
Fig. 6. Additional changes in groundwater storage (annual) due to future crop-water use and 762
inflow under RCP4.5 (a) and RCP8.5 (b) compared to historical (averaged over three future 763
periods). The barplot represent average changes, and the whiskers show the interannual 764
variability (one standard deviation). 765
766
767
768
769
770
771
772
773
39
774
Fig. 7. Additional changes in groundwater storage (annual) due to future inflow volume and 775
seasonality changes for RCP4.5 (a) and RCP8.5 (b) compared to historical (averaged over three 776
future periods). The barplot represents average changes, and the whiskers show the interannual 777
variability (standard deviation). 778
