Hydrologic and Climatic Responses to Global AnthropogenicGroundwater Extraction

YUJIN ZENG

State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics,

Institute of Atmospheric Physics, Chinese Academy of Sciences, and College of Earth Science,

University of Chinese Academy of Sciences, Beijing, China

ZHENGHUI XIE

State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics,

Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

JING ZOU

Institute of Oceanographic Instrumentation, Shandong Academy of Sciences, Qingdao, China

(Manuscript received 11 March 2016, in final form 28 July 2016)

ABSTRACT

In this study, a groundwater (GW) extraction scheme was incorporated into the Community Earth System

Model, version 1.2.0 (CESM1.2.0), to create a new version called CESM1.2_GW, which was used to in-

vestigate hydrologic and climatic responses to anthropogenic GW extraction on a global scale. An ensemble

of 41-yr simulationswith andwithoutGWextraction (estimated based on local water supply and demand)was

conducted and analyzed. The results revealed that GW extraction and water consumption caused drying in

deep soil layers but wetting in upper layers, along with a rapidly declining GW table in areas with the most

severe GW extraction, including the central United States, the north China plain, and northern India and

Pakistan. The atmosphere also responded to GW extraction, with cooling at the 850-hPa level over northern

India and Pakistan and a large area in northern China and central Russia. Increased precipitation occurred in

the north China plain due to increased evapotranspiration from irrigation. Decreased precipitation occurred

in northern India because the Indian monsoon and its transport of water vapor were weaker as a result of

cooling induced by GW use. Additionally, the background climate change may complicate the precipitation

responses to the GW use. Local terrestrial water storage was shown to be unsustainable at the current high

GWextraction rate. Thus, a balance between reducedGWwithdrawal and rapid economic developmentmust

be achieved in order to maintain a sustainable GW resource, especially in regions where GW is being

overexploited.

1. Introduction

With economic development and population growth,

water demands are rapidly increasing andwater resources

are becoming scarce inmany regions of the world (Rodell

and Famiglietti 2002; Rodell et al. 2009; Alvarez et al.

2012; Shi et al. 2013; Devic et al. 2014). Groundwater

(GW), because of its convenience for extraction and good

quality, is widely extracted to supplement human de-

mands of freshwater, especially in regions where surface

water (SW) supplies are limited (Liu et al. 2001). How-

ever, GW overextraction lowers GW tables, reduces total

terrestrial water storage, weakens hydraulic connections

between aquifers and rivers, and may even decrease lake

area (Coe and Foley 2001). Variations in a regional water

table can directly influence soil moisture content (Yuan

et al. 2008; Xie and Yuan 2010; Di et al. 2011; Xie et al.

2012). GW consumption, such as for irrigation, also has

been shown to increase local evapotranspiration and de-

crease the temperatures near the surface and in the lower

Corresponding author address: Dr. Zhenghui Xie, LASG, Institute

of Atmospheric Physics, Chinese Academy of Sciences, P.O. Box

9804, Beijing 100029, China.

E-mail: zxie@lasg.iap.ac.cn

Denotes Open Access content.

1 JANUARY 2017 ZENG ET AL . 71

DOI: 10.1175/JCLI-D-16-0209.1

� 2017 American Meteorological Society

troposphere by affecting soil moisture content (Yu et al.

2014; Zou et al. 2014). Increased water vapor resulting

from GW irrigation can induce local convection and fur-

ther alter atmospheric water balances (Haddeland et al.

2006; Lo and Famiglietti 2013). Irrigation using GW can

alter the carbon cycle and thewater cycle (Xie et al. 2014).

Therefore, continually overextracting a GW resource for

consumption not only reduces GW table levels but also

changes the regional, and even global, environment and

climate (Pokhrel et al. 2012b; Chen and Hu 2004).

Numerous studies have demonstrated the effects of

GW withdrawal and consumption on hydrological or

land surface processes. Most of these studies have used

hydrological models or land surface models to simulate

the effects of anthropogenic water management. For

example, Haddeland et al. (2006) incorporated an irri-

gation module into the Variable Infiltration Capacity

model and found that irrigation increases evapotranspi-

ration and decreases surface air temperature in the

Mekong and Colorado River basins. Hanasaki et al.

(2008) constructed a water resource model that consid-

ered GW withdrawal and anthropogenic reservoir oper-

ations. Ozdogan et al. (2010) incorporated crop

information from satellite remote sensing and an irriga-

tion scheme into a land surface model and found that

irrigation increases evapotranspiration by 12% and re-

duces surface sensible heat flux (SH) by a similar per-

centage in the growing season. Döll et al. (2012) applied awater use model [Water—A Global Assessment and

Prognosis (WaterGAP)] to study the impacts of GW

extraction on total terrestrial water storage; they found

that irrigation had a negative influence on local water

storage, especially in regions where the source of irriga-

tion water was dominated by GW. Pokhrel et al. (2012a)

incorporated an anthropogenic water management

scheme into the Minimal Advanced Treatments of Sur-

face Interaction and Runoff model and found that irri-

gation could cause a maximum increase of 50Wm22 in

latent heat flux (LH) in the summertime.Zou et al. (2015)

incorporated a GW extraction scheme into the Commu-

nity Land Model (CLM), version 3.5, to investigate the

effects of anthropogenic extraction of GW on land sur-

face processes in the Haihe River basin in northern

China; they found that GW extraction for human activi-

ties causes wetting and cooling at the land surface and

reduces GW storage. Pokhrel et al. (2015) presented an

integrated hydrologic model to explicitly simulate the

GWdynamics and pumping on a global scale and showed

that global GW withdrawal and depletion for year 2000

were 570 and 330km3yr21, respectively. However, these

studies only considered the hydrological effects of GW

use. They did not consider subsequent climatic effects of

GW pumping and consumption.

GW extraction and consumption can influence the at-

mosphere and climate through their effects on soil mois-

ture and surface heat fluxes (Chen and Hu 2004). In

regions with GW overextraction, soil moisture may de-

crease because of declining GW tables, even thoughmost

of the GW exploited is used for irrigation (Siebert et al.

2010), which should increase soil moisture. Until now,

only a few studies have investigated the climatic responses

to human GW management and all these studies have

been conducted on a regional scale. For example, Lo and

Famiglietti (2013) investigated the impacts of irrigation in

California’s Central Valley using a climate model driven

by a regional estimate of agricultural water use supplied

by surface and subsurface sources. Zou et al. (2014)

incorporated a GW extraction scheme into the Regional

ClimateModel, version 4 (RegCM4), for theHaihe River

basin in northern China; they revealed that GW extrac-

tion andwater consumption causedwetting and cooling in

the lower troposphere and increased precipitation.

In these regional investigations, nonlocal effects of

GW extraction were not fully considered. This study

aims to investigate both the hydrologic and climatic re-

sponses on a global scale to GWmanagement, including

GW pumping, GW irrigation, and other activities re-

lated to industrial and domestic GW use.

In section 2, we first describe the Earth system model

CESM1.2.0 and the GW extraction scheme we have

developed and incorporated into CESM1.2.0. The data

used to estimate the water supply and demand and the

model simulations are described in section 3. The results

from the simulations are discussed in section 4. Con-

clusions and discussion are presented in section 5.

2. Model development

a. The Community Earth SystemModel, version 1.2.0

The Community Earth System Model, version 1.2.0

(CESM1.2.0; Hurrell et al. 2013), consists of atmo-

sphere, land, ocean, and sea ice components linked by a

coupler that communicates among these components

and merges and regrinds fields as needed. The

CESM1.2.0 model was developed from the Community

Climate System Model, version 4 (CCSM4; Gent et al.

2011). The atmospheric component is the Community

Atmosphere Model, version 4 (CAM4; Neale et al.

2010), which applies the Lin–Rood finite-volume dy-

namical core (Lin 2004) with an atmospheric grid spac-

ing of 0.98 3 1.258 horizontally and 26 levels in the

vertical direction arranged in a hybrid pressure sigma

coordinate (Neale et al. 2013). The ocean component is

the Parallel Ocean Program (POP; Jones et al. 2005)

from the Los Alamos National Laboratory (LANL).

72 JOURNAL OF CL IMATE VOLUME 30

The sea ice component is an extension of the LANL sea

ice model (Hunke and Lipscomb 2008). The land com-

ponent is the Community Land Model, version 4.5

(CLM4.5; Oleson et al. 2013), which simulates bio-

geophysical exchange of radiation, LH, and SH; mo-

mentum between land and atmosphere as modified by

vegetation and soil; heat transfer in soil and snow; and

the hydrologic processes and snow dynamics (Lindsay

et al. 2014). To represent the heterogeneity within a

model grid, CLM4.5 applies a nested subgrid hierarchy

in which each grid cell is composed of multiple land

units, snow and soil columns, and plant function types

(PFTs). This allows different land covers, such as vari-

eties of vegetation and crops, as well as land unit types

(lake, urban, and glacier), to be treated differently

depending on their own biogeophysical and bio-

geochemical processes, even when they are coexist in

the same model grid box. More information about the

subgrid structure of CLM4.5 can be found in Oleson

et al. (2013).

b. Scheme for GW extraction and its implementationinto CESM1.2.0

The GW extraction and consumption scheme de-

veloped by Zou et al. (2014) was incorporated into

CESM1.2.0 by integrating it into the land component (i.e.,

the land surface model CLM4.5) of CESM1.2.0, as shown

in Fig. 1. The resultingmodel (called CESM1.2_GW)was

then used to investigate the effects of anthropogenic GW

withdrawal and use. As shown in Fig. 1, GW is pumped

FIG. 1. Framework of simulated human-induced water resource extraction and consumption

processes and their incorporation into the CLM4.5 and CESM1.2.0 models.

1 JANUARY 2017 ZENG ET AL . 73

from an aquifer and apportioned among agricultural, in-

dustrial and domestic uses. Water consumed by agriculture

is considered as ‘‘effective precipitation’’ reaching the soil

surface. The industrial and domestic consumptions have

two components: 1) the waste water produced by industry

and human daily life, which is treated as a discharge into

local rivers and 2) the net water consumption, which is

treated as evaporation to the atmosphere.

The aquifer recharge rate is updated by subtracting

the GW extraction rate from the original aquifer re-

charge rate, which can be expressed as

Qaqu_new

5Qaqu_ori

2Qg, (1)

where Qaqu_ori (mms21) is the original aquifer recharge

rate calculated by CLM4.5, Qg (mms21) is the GW ex-

traction rate (see section 2c), and Qaqu_new (mms21) is

the new aquifer recharge rate after considering anthro-

pogenic GW pumping, which is used by CLM4.5 for the

next groundwater table calculations.

The irrigation consumption supported by GW is

expressed in the model using

Qtopsoil_new

5Qtopsoil_ori

1 ragrQ

g, (2)

whereQtopsoil_ori (mms21) is the original net water input

into the soil surface simulated by CLM4.5, ragr is the

ratio of agricultural consumption rate to the total Qg

(see section 2c), andQtopsoil_new (mms21) is the new net

water input into the soil surface after consideration of

the GW irrigation effect, which is used by CLM4.5 in the

next surface runoff and infiltration calculation.

The waste water produced by industrial and domestic

water uses is described as

Qrunoff_new

5Qrunoff_ori

1a(rind

1 rdom

)Qg, (3)

where Qrunoff_ori (mms21) is the original surface runoff

calculated by CLM4.5, rind and rdom (see section 2c) are,

respectively, the ratios of industrial and domestic con-

sumption rates to the totalQg, a is the waste water ratio,

and Qrunoff_new (mms21) is the surface runoff revised to

include waste water.

Except for waste water, all other components of in-

dustrial and domestic GW consumption go into the at-

mosphere as described by

Qevp_new

5Qevp_ori

1 (12a)(rind

1 rdom

)Qg, (4)

where Qevp_ori (mms21) is the original total evapotrans-

piration flux, which is the sum of vegetation transpiration

and evaporation from soil, leaves, and stems, as simulated

by CLM4.5. The variable Qevp_new (mms21) is the new

total evapotranspiration flux increased by the industrial

and domestic net consumption;Qevp_new is transferred to

the coupler and then to CAM4, which calculates the ef-

fect of this variable on the atmospheric condition.

As shown in Fig. 1, based on the structure of coupling,

other variables in CLM4.5 are influenced by Qaqu_new

andQtopsoil_new, each of which is affected directly byGW

extraction and consumption. The atmospheric variables

in CAM4 are changed byQevp_new. In addition, changes

in both the land and atmosphere subsequently affect

each other through land–atmosphere coupling.

c. Estimation of GW withdrawal

Before starting a CESM1.2_GWmodel simulation, the

parameters Qg, ragr, rind, rdom, and a must be estimated

using historical GW withdrawal and consumption data.

Three sources of data were combined to estimate GW

withdrawal and application. The first data source was the

Food andAgricultureOrganization of theUnitedNations

(FAO) global water information system, developed by the

Land and Water Division of the FAO (http://www.fao.

org/nr/water/aquastat/main/index.stm). This dataset con-

tained regional water withdrawal data and expressed ag-

ricultural, industrial, and municipal water withdrawals as

percentages of the totalwithdrawal. TheFAOdataset also

gave the size of area equipped for irrigation in 1970, 1990,

and 2009 and it estimated the global water withdrawal by

agriculture (in the period around 2003) to be 2710km3yr21.

The second dataset was the Global Map of Irrigation

Areas, version 5.0 (GMIA5; Siebert et al. 2005, 2013),

which uses a raster format having a grid resolution of 50 toshow the size of area equipped for irrigation in each grid,

and the size of area that is irrigated usingGW in each grid.

The third dataset consisted of historical monthly soil

moisture and saturated soil moisture simulated by

CLM4.5 offline using the atmospheric forcing dataset

described in Qian et al. (2006) for years 1965–2000.

The methodology used to combine these data was as

follows. First, the total amount of global water with-

drawal by agriculture around 2003, provided by the

FAO dataset, was allocated to every model grid cell

(0.98 3 1.258) and weighted by the agricultural area of

each grid cell (as provided by GMIA5) as well as by the

soil water deficit (the difference between climatological

saturated soil moisture and historical soil moisture, both

of which were provided by the CLM4.5 offline simula-

tion), as described by

Qagr(i, j)5Q

agr_glob

Airr(i, j)W

def(i, j)

�i, jA

irr(i, j)W

def(i, j)

, (5)

where i and j are the indices for longitude and latitude,

respectively, of the model grid, Qagr(i, j) (m3) is the ag-

ricultural water withdrawal both from surface and

74 JOURNAL OF CL IMATE VOLUME 30

subsurface at the grid point (i, j) in 2003,Qagr_glob (m3) is

the global agricultural water withdrawal in 2003 provided

by FAO, Airr(i, j) (m2) is the size of area equipped for

irrigation on the grid point (i, j) provided byGMIA5, and

Wdef(i, j) (m3m23) is the soil water deficit calculated from

the CLM4.5 offline simulation. The soil water deficit here

indicates the irrigation intensity, which should be stron-

ger over drier soil and weaker over wetter soil. Then, the

agricultural water use supplied by the GW in each grid

point in 2003 [Qagr_gw(i, j) (m3)] was calculated using the

GW irrigation data provided by GMIA5 as

Qagr_gw

(i, j)5Qagr(i, j) 3 f

gw(i, j), (6)

where fgw(i, j) is the fraction of irrigated area fed by GW

at the grid point (i, j), which was estimated by GMIA5.

Next, using the regional dataset about municipal, in-

dustrial, and agricultural water consumption provided by

the FAO, the GW used by these three sectors, as well as

the totalGWwithdrawal (around 2003), were calculated as

Qtot_gw

(i, j)5Qagr_gw

(i, j)/fagr(i, j), (7)

Qind_gw

(i, j)5Qtot_gw

(i, j) 3 find(i, j), and (8)

Qlive_gw

(i, j)5Qtot_gw

(i, j) 3 flive

(i, j), (9)

where Qtot_gw(i, j) (m3), Qind_gw(i, j) (m3), and

Qlive_gw(i, j) (m3) are the total, industrial, and do-

mestic GW consumption at the grid point (i, j), re-

spectively, and fagr(i, j), find(i, j), and flive(i, j) are,

respectively, the fractions of the GW consumption by

agricultural, industrial, and domestic water use. These

fractions were calculated by combining the FAO-

provided fractions of the total water consumption (from

both GW and SW) for agricultural, industrial, and do-

mestic water use fagr0(i, j), find0(i, j), and flive

0(i, j), and the

global values evaluated by Döll et al. (2012) that GW

contributes 42%, 27%, and 36% of water used for irri-

gation, manufacturing, and households, respectively, as:

8>>>>>>>>>><>>>>>>>>>>:

fagr(i, j)5

0:42f 0agr(i, j)

0:42f 0agr(i, j)1 0:27f 0ind(i, j)1 0:36f 0live(i, j)

find(i, j)5

0:27f 0ind(i, j)0:42f 0agr(i, j)1 0:27f 0ind(i, j)1 0:36f 0live(i, j)

flive

(i, j)50:36f 0live(i, j)

0:42f 0agr(i, j)1 0:27f 0ind(i, j)1 0:36f 0live(i, j)

.

(10)

Figure 2a shows the resulting global spatial patterns of

Qtot_gw around year 2003. As shown in Fig. 2a, the three

largest contiguous areas with high GW extraction rates

are located in northern India and Pakistan, the north

China plain, and the central United States. Most areas in

these regions were suffering a GW withdrawal rate of

more than 90mmyr21 (around 2003). Besides, a large

number of scattered areas with severe GW extraction

(more than 90mmyr21) are seen in northern Italy, parts

of western Europe, parts of the Euphrates–Tigris basin

and southern Iran, northwestern China, eastern coastal

Australia, and the west coastal area of Florida in the

United States. In other areas across the world, such as

southern China, southern and eastern Brazil, eastern and

western Europe, South Africa, and across the entire

United States, GW extraction is significant, with a rate

of a few to tens of millimeters per year. These results are

corroborated by many recent studies (Khan et al. 2008;

Siebert et al. 2010; Döll et al. 2012; Leng et al. 2013;

Shi et al. 2013; Leng et al. 2014; Zou et al. 2014).

To estimate annual data, the amount of GW extrac-

tion was assumed to change linearly from 1970 to 1990

and from 1990 to 2009. The linear regression coefficients

were calculated for these two time periods using the

regional component of the FAO water use dataset,

which provides data only for 1970, 1990, and 2009. We

also extrapolated the groundwater withdrawal data

forward to 1965 using the same regression coefficient of

1970–90.With these coefficients and theGWwithdrawal

data for approximately 2003, the GW pumping rate in

every year from 1965 to 2005 was estimated using linear

interpolation. We recognized that the linear in-

terpolation may result in large uncertainties because the

water consumption and GW pumping rate in a certain

year depend on the water availability (such as pre-

cipitation) and the climate of that year. However, long-

term changes in annual GW pumping rates, which are

the focus of this study, aremore steadily increased by the

population growth and socioeconomic development

than by climatic variations, as demonstrated by many

previous studies (Llamas 1988; Liu et al. 2001; Omole

2013; Wu et al. 2014; Zou et al. 2015). Because annual

data for GW withdrawal were unavailable on a global

scale, the linear interpolation based on the scattered years

of available with GW withdrawal data provides a rea-

sonable and convenient way to represent the long-term

change inGWpumping rates from 1965 to 2005. Figure 2b

shows the time series of interpolated annualGWpumping

rate in the central United States, the north China plain,

and northern India and Pakistan, where GW extraction

was largest in the world (Fig. 2a). It is clear that the ex-

traction rate increases at a similar pace before and after

1990, further confirming the linearity of the change rate.

In reality, the seasonal allocation of annual water con-

sumption is very complex. Seasonal water consumption

depends on local climate, phenology of crops, economy,

1 JANUARY 2017 ZENG ET AL . 75

policy, and other natural and social conditions.Accurately

depicting seasonal water consumption on a global scale is

an extremely difficult task. Here, we will focus only on the

long-term annual-mean effects of GW extraction (rather

than the seasonal variations) to reduce the large un-

certainties from intra-annual allocation of the annual

water consumption. For this purpose, the GW consump-

tion for industrial and domestic water uses was simply

distributed evenly to every model time step, and the GW

consumption for agriculturewas evenly distributed only to

model time steps in the spring and summer (March–May

and June–August, respectively, for the Northern Hemi-

sphere, and September–November and December–

February, respectively, for the Southern Hemisphere)

when irrigation is dominated. Zou et al. (2014) showed

that on the north China plain the annual effects of GW

extraction were similar regardless of whether an even

allocation or a statistically based allocation was applied.

Thus, our simple allocation scheme should be acceptable

for studying the annual effects of GW withdrawal.

In this study, a (the waste water ratio) was set as 30%

based on Mao et al. (2000). The remaining 70% of in-

dustrial and domestic GW withdrawal includes not only

the amount of water actually consumed by factories and

residential communities, but also the amount of water

dissipated when waste water is discharged into streams.

3. Experimental design

To investigate the impacts of human GW extraction

on the land surface and climate, two sets of numerical

experiments were conducted using the original CESM1.2.0

(referred to as CTL) and using CESM1.2_GW coupled

with aGWwithdrawal and consumption scheme (referred

FIG. 2. (a) Global groundwater extraction rate around year 2003 and (b) the area-averaged ex-

traction rate in the central United States, the north China plain, and northern India and Pakistan

from1965 to 2005.As designated by the three boxes in (a), the centralUnited States is defined as the

region within 338–428N, 978–1058W; the north China plain is defined as the region within 348–408N,

1108–1208E; and northern India and Pakistan is defined as the region within 238–338N, 688–788E.

76 JOURNAL OF CL IMATE VOLUME 30

to as EXP). Thus, the CTL runs do not include the GW

extraction whereas the EXP runs do include it, and the

EXP-minus-CTL difference would provide a measure of

the impact of the GW extraction as simulated by the

CESM1.2_GW model. In these experiments, only the at-

mospheric (CAM4) and land (CLM4.5) components were

integrated, while the ocean surface conditions were pre-

scribed using monthly observations of sea surface tem-

peratures (SSTs) and sea ice concentration from the

HadISST dataset (Rayner et al. 2003). Each of the CTL

and EXP experiments contained six ensemble simulations

using different initial conditions. The ensemble averages

used here reduce the internal noise and enhance the forced

signal caused by the GW extraction. All simulations were

conducted with horizontal spacing of 0.98 3 1.258 and a

time step of 30min for both the CAM4 and CLM4.5. The

simulations were conducted for a 41-yr duration from 1965

to 2005 (using the period of 1965–69 for spinup), during

which time-varying data for GW extraction and other ex-

ternal drivers were used. We used the default external

forcing settings for CAM4 and CLM4.5 as described in

Neale et al. (2010) and Oleson et al. (2013), respectively.

These forcing data, such as greenhouse gas concentrations,

aerosol and ozone concentrations, solar irradiance, volca-

nic activity, and land use data were estimated from ob-

servations or remote sensing.

Because of internal variability, model results were

anticipated to be sensitive to initial conditions. One way

to reduce the sensitivity is to conduct ensemble simu-

lations using different initial conditions and then aver-

age the results over the ensemble of simulations (Koster

et al. 2002, 2006). In the current study, six individual

simulations, typically differing only in their initial at-

mospheric and land surface conditions, comprised each

ensemble (CTL or EXP). The method described by

Koster et al. (2006) was used to generate the six sets of

different ‘‘restart files,’’ which were obtained from a pre-

existing 70-yr 1850 control run (which was started from

another 500-yr 1850 control run by the CESM1.2.0)

with a separation time of 10 yr between the restart

conditions. Simulations with corresponding ensemble

members in EXP and CTL were started from the same

restart files; thus, EXP number 1 and CTL number 1

shared the same initial conditions. Ensemble averages of

the simulations for the years from 1970 to 2005 were

analyzed to study the effects of GW extraction.

4. Results

a. Validation of CESM1.2_GW

Key to validating the simulation ability of the

CESM1.2_GW model was verifying whether it could

reproduce observed long-term changes in the water

table induced by GW withdrawal. Unfortunately, the

worldwide data on GW table depth, especially time se-

ries data, were scarce. However, because there are only

three dominant contiguous areas in the world that are

experiencing severe GW extraction, we will focus on

validating the model’s ability to accurately represent the

GW table trends in these three areas (i.e., the north

China plain, northern India and Pakistan, and the cen-

tral United States; the outlined areas in Fig. 2a). The

north China plain has been studied by many researchers

because of its GW overextraction (Liu et al. 2010; Zou

et al. 2014). A dataset of GW table variations from 1991

to 2000 for 173 observation wells in the basin was pro-

vided by the Ministry of Water Resources and the In-

stitute forGeo-environmentalMonitoring of China. The

central United States has also suffered from over-

extraction of GW, as shown by recent studies (Döll et al.2012); water table data from this region were abundant,

and were provided by United States Geological Survey.

Data from 163 observation wells in the central United

States for 1965–2005 were used for validation purposes

here. Northern India and Pakistan experienced water

storage declines as shown by National Aeronautics and

Space Administration’s (NASA) Gravity Recovery and

Climate Experiment (GRACE) satellites (Tapley et al.

2004; Rodell et al. 2009; Tiwari et al. 2009). Neverthe-

less, GW table data in this region are unavailable. Thus,

we compared the model-simulated water storage with

the GRACE observations of total water storage varia-

tions for northern India and Pakistan. Figure 3 shows the

time series of simulated annual area-averagedGW table

depth against the observations in the central United

States and the north China plain, and the simulated

annual total water storage changes in northern India and

Pakistan against GRACE satellite observations.

Figure 3a indicates that although the long-term mean

values of the simulated GW table in the central United

States were significantly different from the observations,

the EXP ensemble of model simulations, which showed

downward trends of GW table with 0.32myr21, gener-

ally reproduced (although with a slight overestimation)

the observed downward trends (0.2myr21), while the

CTL ensemble did not shown any trends in the GW

table. Considering that the GW module in CESM is

relatively simple (it does not contain any expression of

GW lateral flow and bedrock depth distribution) and

that the aim of this study is to investigate the climate

change caused by GW extraction, it is not surprising to

see large mean biases in the simulated water table in

Fig. 3a. The overestimated GW table downward trends

of the EXP ensemble were also shown for the north

China plain (Fig. 3b), while the EXP ensemble showed

downward trends of 0.16myr21, which were significantly

1 JANUARY 2017 ZENG ET AL . 77

higher than the value of 0.09myr21 shown in observa-

tions. Figure 3c shows that the EXP simulations also

predicted the decreasing trend (20.62 cmyr21) in total

terrestrial water storage in northern India and Pakistan,

although the predictions were somewhat greater than

indicated by the GRACE satellite observations

(20.50 cmyr21). In the CTL ensemble, there was almost

no any trend for the terrestrial water storage

(only 20.02 cmyr21). The ability of the CESM1.2_GW

model in reproducing the observed declining trends of

the GW table and water storage in key geographic re-

gions provides confidence about the GW extraction

module for applying it to study the effects of GW ex-

traction on global climate.

b. Differences in spatial distribution of hydrologicvariables

Figure 4 shows the multiyear (1970–2005) mean spa-

tial distributions of the EXP-minus-CTL difference for

the GW table depth, top 10-cm soil moisture content,

runoff, total water storage, LH, and SH. The ensemble-

averaged results in Fig. 4a show that contiguous areas

with significant GW level declines were found in the

central United States, in parts of the north China plain,

and in northern India and Pakistan. Compared with

changes predicted by the CTL simulations, the GW ta-

bles in most areas declined more than 4m when aver-

aged from 1970 to 2005. These results correspond with

those of Zou et al. (2014) for the north China plain,

Sophocleous (2005) for the central United States, and

Kumar and Singh (2008) for northern India. Overall, the

water table depth pattern described by the difference in

EXP and CTL simulations was similar to the pattern of

GW extraction shown in Fig. 2a.

Figure 4b shows that the contiguous areas with sta-

tistically significant increases in surface soil moisture

were located in northern India along with Ganges River,

in theHuaiheRiver basin in the northChina plain, in the

Mississippi River basin in the central United States, and

in a large area of high-latitude regions in the Northern

Hemisphere. The first three areas correspond with the

distribution of GW extraction shown in Fig. 2a, implying

that the increased soil moisture at these three regions

was caused by GW irrigation, as shown in Fig. 1. How-

ever, the soil moisture increase in the central United

States was relatively weaker than in northern India and

China. This is because, according to the FAO’s statistics,

91% and 64% of GW extraction was consumed by ag-

riculture in India and China, respectively, whereas in the

United States only 38% of extracted GW was used by

agriculture. Areas that were extensively irrigated (north

of India and China) would be expected to have higher

soil moisture contents than areas with less irrigation

(central United States). There were large areas of high-

latitude regions in the Northern Hemisphere also

showing a high content of soil water even though GW

consumption by agriculture here could be negligible. This

might be caused by the mixed effects of atmospheric in-

ternal variability and land–atmosphere interactions (in-

ternal noise). Although the increases of 10-cm soil

moisture passed the significance tests over some area of

the north China plain, northern India and Pakistan, and

central United States, in fact the magnitude of changes

was relatively small (within 0.04m3m23 in China and

FIG. 3. Time series of simulated annual area-averaged groundwater table depth against observations in (a) the

central United States and (b) the north China plain, and (c) the simulated annual total water storage anomalies of

northern India and Pakistan against GRACE satellite observations.

78 JOURNAL OF CL IMATE VOLUME 30

FIG. 4. Spatial distribution of ensemble-averaged EXP-minus-CTL differences averaged over 1970–2005 for

annual-mean (a) GW table depth (m), (b) 10-cm soil moisture (m3m23), (c) runoff (mmyr21), (d) terrestrial water

storage (cm), (e) latent heat flux (Wm22), and (f) sensible heat flux (Wm22). The stippled areas indicate the

regions where the difference passed the 95% confidence level in the Student’s t test.

1 JANUARY 2017 ZENG ET AL . 79

FIG. 5. Spatial distribution of ensemble-averaged EXP-minus-CTL differences averaged over 1970–2005 for

annual-mean (a) 850-, (b) 500-, and (c) 200-hPa level temperature (8C) and (d) 850-, (e) 500-, and (f) 200-hPa level

geopotential height (m). The stippled areas indicate the regions where the difference passed the 95% confidence

level in the Student’s t test.

80 JOURNAL OF CL IMATE VOLUME 30

FIG. 6. Spatial distribution of ensemble-averaged EXP-minus-CTL differences averaged over 1970–2005 for

annual-mean JJA (a) precipitation (mmday21); (b) 850-, (c)500-, and (d) 200-hPa level wind field (m s21); and

(e) horizontal vapor flux (kgm21 s21). The gray shading indicates the regions where the magnitude of the vector

difference passed the 95% confidence level in the Student’s t test.

1 JANUARY 2017 ZENG ET AL . 81

India and within 0.02m3m23 in the United States). Fur-

thermore, at a deeper range of soil, such as to 50- or 100-cm

depth, the increases in soil moisture will be further

weakened because the wetting effects of irrigation de-

crease rapidly with increasing soil depth (a relationship

that will be highlighted in section 4d).

Figure 4c shows the multiyear (1970–2005) mean

spatial distributions of the EXP-minus-CTL difference

for the annual runoff. Areas of the central United States

and the north China plain that were shown to be suf-

fering from severe GW extraction also showed increases

in runoff; however, runoff decreased significantly in

northern India even though the GW extraction was also

rigorous. The increased runoff in the United States and

China can be explained by Eqs. (2) and (3), in which the

waste water produced by industrial and domestic water

use was directly converted to runoff in the model,

whereas the irrigation water also added the runoff by

increasing the amount of water reaching the topsoil. The

decreased runoff in India may mainly be a consequence

of decreased precipitation due to regional GW with-

drawal and consumption (as discussed later in section 4c).

In other regions across the world, there were also a large

number of areas showing visible runoff differences due to

precipitation change, but most of themwere the results of

internal variability and failed to pass the Student’s t test.

Figure 4d shows the multiyear (1970–2005) mean

spatial distributions of the EXP-minus-CTL difference

for the total terrestrial water storage. A comparison of

Fig. 4d with Fig. 4a shows that changes in terrestrial

water storage had a similar pattern as changes in the

water table depth. The mechanism for this water storage

reduction is simple: GW extraction was a process of

transferring water from underground reserves to the

land surface. The water on the land surface was then

consumed by evapotranspiration and sent to the

FIG. 7. Vertical profiles of ensemble-averaged EXP-minus-CTL differences for annual-mean (a) soil moisture

(m3m23), (b) specific humidity (1021 g kg21), and (c) temperature (8C) in the centralUnited States, the northChina

plain, and northern India and Pakistan.

82 JOURNAL OF CL IMATE VOLUME 30

FIG. 8. Spatial distribution of ensemble-averaged EXP2-minus-CTL2 differences averaged over last 10 simulation

years for annual-mean (a) latent heat flux (Wm22), (b) sensible heat flux (Wm22), and 850-hPa level (c) temperature

(8C) and (d) geopotential height (m). The stippled areas indicate the regions where the difference passed the 95%

confidence level in the Student’s t test.

1 JANUARY 2017 ZENG ET AL . 83

atmosphere. Through horizontal wind advection, part of

the lifted water vapor was transported to other places

and a local water deficit occurred. Figure 4d reveals that

the regions that were experiencing serious GW over-

extraction (especially the central United States, the

north China plain, and northern India) were indeed

suffering losses of water resources. These results are

supported by Rodell et al. (2009) and Döll et al. (2012).Figures 4e and 4f show the multiyear (1970–2005) mean

spatial distributions of the EXP-minus-CTL differences for

the LH and SH, respectively. The three contiguous over-

extraction areas (the central United States, the north China

plain, and northern India and Pakistan) were also identified

to be experiencing significant LH increases (5–20Wm22)

and SH decreases (2–10Wm22). According to Dirmeyer

et al. (2013), the evapotranspiration in these three places is

restricted more by water availability than by temperature;

thus, when a greater supply of water was available, more

evaporation and stronger LH would occur, and with more

energy taken away from land by LH, the SH would sub-

sequently decrease.

c. Differences in spatial distribution of atmosphericvariables

As noted earlier, the effects of GW withdrawal and

consumption on land can affect climate through flux

FIG. 9. Spatial distribution of ensemble-averaged EXP2-minus-CTL2 differences averaged over last 10 simulation

years for annual-mean JJA (a) precipitation (mmday21), (b) 850-hPa level wind field (m s21), and (c) horizontal vapor

flux (kgm21 s21). The gray shading indicates the regions where the magnitude of the vector difference passed the 95%

confidence level in the Student’s t test.

84 JOURNAL OF CL IMATE VOLUME 30

exchange between the land surface and the atmosphere.

Figures 5a–c show the multiyear (1970–2005) mean

spatial distributions of the EXP-minus-CTL differences

for the air temperature at 850-, 500-, and 200-hPa levels.

From Fig. 5a, at the 850-hPa level, two contiguous areas

experiencing a significant cooling effect of approxi-

mately 0.38C were obvious. One of these areas was lo-

cated in northern India and Pakistan; the other area was

northern China plus a large area of central Russia. A

comparison of Figs. 5a and 2a indicates that the cooling

effect was caused by the GW irrigation. However, in

northern China and central Russia, the size of the area

experiencing lower temperature was much larger than

the area experiencing severe GW extraction. This situ-

ation was probably induced by the advection of air be-

cause in the irrigation seasons the prevailing wind of this

region at the 850-hPa level is from south to north (Atlas

et al. 1996; Young 1999). Thus, the prevailing wind takes

the cooling air farther north and reduces the ambient

temperature of the area north of the main sites of irriga-

tion. Such a process is a clear example of how irrigation

could affect the climate outside the local placewhere it was

used, a scenario that could not be simulated by an offline

model (Döll et al. 2012). This type of interaction stresses

the necessity of using a global scale land–atmosphere

coupling model to study GW extraction effects. There

was also a contiguous area, the Tibetan Plateau in eastern

China bordering northern India, experiencing obviously

higher temperature (more than 0.48C) at the 850-hPa level.This temperature effect might be caused by the irrigation-

induced thermal circulation and will be discussed later. At

the 500- and 200-hPa levels (Figs. 5b and 5c, respectively),

the cooling effects were almost invisible, indicating that the

impacts of GWuse on temperature weremore apparent in

the lower atmosphere than in the upper atmosphere.

Figures 5d–f show the multiyear (1970–2005) mean

spatial distributions of the EXP-minus-CTL differences

for the geopotential height at 850, 500, and 200 hPa.

FromFig. 5d, besides the internal noise in theArctic and

Antarctic, the Tibetan Plateau in China experienced

lower geopotential height at the 850-hPa level, which

passed the Student’s t test. This might be caused by the

cooling effects of irrigation in northern India (Fig. 5a),

which increased the air density and lifted the local ge-

opotential height. The added air in northern India

originated from its east side, so the geopotential height

in the Tibetan Plateau decreased, forming thermal cir-

culation between northern India and the Tibetan Plateau.

The thermal circulation transported the heat from

northern India to the Tibetan Plateau and increased the

temperature in theTibetanPlateau (Fig. 5a).At the 200-hPa

level of the upper atmosphere, the geopotential height

in the Tibetan Plateau increased, demonstrating the

existence of the irrigation-induced thermal circulation

(Fig. 5c). In other regions across the world, there was

no difference in geopotential height that was statisti-

cally significant according to the Student’s t test.

Figure 6a shows the multiyear (1970–2005) mean

spatial distributions of the EXP-minus-CTL differences

for June–August (JJA) precipitation. The pattern is in-

teresting. For GW overextracted regions of the north

China plain and northern India, the former experienced

increased precipitation while the precipitation in the

latter decreased. The contrasting effects of GW extrac-

tion on precipitation might come from the different

precipitation generating mechanisms of these two

regions. According to recent studies, northern China

belongs to areas that have a strong summer land–

atmosphere coupling interaction (Dirmeyer et al. 2013;

Zeng and Xie 2015), meaning that the precipitation

changes in the region are significantly affected by the

local soil moisture changes. Thus, with a wetter surface

soil (Fig. 4b), the north China plain received more pre-

cipitation. To understand the precipitation decrease in

northern India, we plotted the EXP-minus-CTL differ-

ences for the JJA wind field at the 850-, 500-, and

200-hPa levels in Figs. 6b–d. Figure 6b indicates that, at

the 850-hPa level, a wind anomaly from the Indian

subcontinent to the Arabian Sea is induced by GW ir-

rigation, the magnitude of which passed the Student’s t

test. In the upper level at 200hPa, the wind anomaly

reversed in direction from the sea to the continent, in-

dicating that a thermal circulation was produced by the

GW irrigation between the Arabian Sea and the Indian

subcontinent. Because most of the water vapor was lo-

cated in the lower atmosphere, the irrigation-induced

thermal circulation might have obstructed the water

vapor flowing from the Arabian Sea to the northern

Indian subcontinent in monsoon season. To demon-

strate this, we mapped the EXP-minus-CTL differences

for JJA horizontal vapor flux integrated from the 850- to

200-hPa level; the result is shown in Fig. 6e. The horizontal

vapor flux was calculated as:

DF5D

ðz2z1

qvdz , (11)

where F (kgm21 s21) is the flux of vertically accumu-

lated horizontal vapor; z1 (m) and z2 (m) are the heights

of the 850- and 200-hPa pressure surfaces, respectively;

q (kgm23) is the specific humidity in each layer; v (m s21)

is the wind vector in each layer; dz (m) is the thickness

of each layer; and D represents the difference between

the EXP and CTL ensemble simulations. As shown

in Fig. 6e, corresponding to the wind anomaly at the

850-hPa level, a water vapor flow anomaly was induced

1 JANUARY 2017 ZENG ET AL . 85

from the Indian subcontinent to the Arabian Sea in

summer, indicating that the vapor flow from sea to land

in the monsoon season was obstructed, and the JJA

precipitation in northern India was subsequently re-

duced. The vapor flow anomaly might be induced by the

cooling effects over northern India because the mon-

soon was forced by the land–ocean temperature differ-

ences. Besides, in India the vegetation also played an

important role. Lee et al. (2009) showed that the

anomalies in the normalized difference vegetation index

(NDVI) have increased over the Indian subcontinent as

the irrigated area has increased. Consequently the

evapotranspiration is increased and takes more heat

away from ground due to the irrigation and increased

NDVI. As a result, the surface temperature in July is

cooled, leading to a reduction of land–sea thermal con-

trast, which is a key factor driving the Indian monsoon.

Therefore, the monsoon circulation is weakened and

rainfall is reduced as both irrigation and the area cov-

ered by vegetation increased. It should be noticed that

the magnitude of precipitation change by GW use was

small although it could be statistically detected. Xie and

Arkin (1997) found that the precipitation rate of the

north China plain and northern India can exceed 500 and

1000mmyr21, respectively, while the magnitude of

change in our research is restricted to only 50mmyr21.

Our findings correspond with several previous studies

about the precipitation feedback to irrigation on a re-

gional scale (Barnston and Schickedanz 1984; DeAngelis

et al. 2010; Lo and Famiglietti 2013; Zou et al. 2015).

d. Differences in the vertical profiles

In addition to the mean spatial distributions, the mean

vertical profiles of EXP-minus-CTL differences for soil

moisture in the central United States, the north China

plain, and northern India and Pakistan (where GW ex-

traction was the most severe in the world) are shown in

Fig. 7a. As indicated in Fig. 7a, the effects of GW use on

soil moisture were different for the top of the soil profile

and deeper soil layers. The topsoil became wetter and

the deep soil became drier in all three regions. The

wetting effects on topsoil were caused by GW irrigation.

In contrast to the topsoil, the deep soil became drier

because of the declining GW table that results from

severe extraction. The vertical profiles of EXP-minus-

CTL differences for air temperature and specific hu-

midity of the three regions are shown in Figs. 7b and 7c,

respectively. As shown in Fig. 7b, the effects of GW

extraction on increased moisture in the lower tropo-

sphere appeared in all three regions. The moisture

effects in surface air were mainly induced by the

wetter topsoil in these areas compared to nonirrigated

areas. However, even in India where the world’s highest

consumption of GW by irrigation occurred, the en-

hanced water vapor content at the ground surface was

no more than 0.12 g kg21; compared with the climatol-

ogy value of nearly 70 g kg21, the effects were nearly

negligible. As shown in Fig. 7c, the cooling effects were

transmitted to the lower troposphere. The cooling ef-

fects decreased as height above the ground surface in-

creased. In the north China plain, where the cooling

effects were most significant, the decreased temperature

was nearly 0.38C.

e. Climatic responses to GW extraction excludingbackground climate change

The former runs of CTL and EXP ensembles used the

observed SSTs, atmospheric CO2 concentration, land

cover, and other sets of annually varying data from 1965

to 2000 as the external forcing of the simulations. Their

results reflected the effects of actual GW extraction.

However, on the other hand, the year-to-year varied

SSTs (as well as other forcing variables) complicated the

climatic response to GW extraction, making the results

not just related to GW forcing. To isolate the effects of

GW extraction, we conducted another two ensemble

simulations called CTL2 and EXP2with and without the

anthropogenic GWextraction, respectively. For the new

simulations, the SSTs, atmospheric CO2 concentration,

land cover, and other forcing datasets were fixed to the

climatological monthly values for year 2000, and the

GW extraction forcing values were fixed at 2003 levels.

As for the former CTL andEXP simulations, each of the

new ensembles contained six runs with different initial

conditions. All the runs were conducted for 15 yr (the

first 5 yr for spinup), cyclically using the forcing data.We

applied the ensemble-mean EXP2-minus-CTL2 differ-

ences averaged over the last 10 yr to investigate the

climatic responses to GW extraction excluding the year-

to-year background climate change.

Figure 8 shows the annual-mean spatial distributions

of the EXP2-minus-CTL2 differences for the land–

atmosphere LH, SH, and the air temperature and geo-

potential height at 850 hPa, which were shown in the

former CTL and EXP simulations to be significantly

modified by GW use. The effects of GW extraction on

LH and SH shown in Figs. 8a and 8b were the same as in

Figs. 4e and 4f, demonstrating that the background cli-

mate change did not greatly affect the irrigation-induced

LH and SH changes. However, Figs. 8c and 8d reveal

that the magnitudes of EXP2-minus-CTL2 differences

for temperature and geopotential height were much

higher than that they were in the EXP-minus-CTL dif-

ferences (Figs. 5a and 5d), although their spatial patterns

were similar. This result indicates that the mechanism

controlling how the GW use impacted the temperature

86 JOURNAL OF CL IMATE VOLUME 30

and geopotential height was the same in both cases

(including background climate change or not), but the

background climate change might weaken the magni-

tude of the impacts induced by GW extraction.

Figure 9 shows the annual-mean spatial distributions

of the EXP2-minus-CTL2 differences for the pre-

cipitation, 850-hPa level wind field and horizontal vapor

flux (all in JJA). Comparison of Fig. 9a with Fig. 6a

shows that northern India still experienced a pre-

cipitation decrease due toGWuse when the background

climate changes were removed. However, in the north

China plain, the expected precipitation increase (which

was shown in Fig. 6a) disappeared. The explanations for

this are shown in Figs. 9b,c. In the new runs, GW use

induced a cyclone anomaly at the 850-hPa level around

the north China plain, and the cyclone anomaly weak-

ened the water vapor flowing from the Yellow Sea and

East China Sea to the China mainland. This weaker

vapor flow offset the irrigation-increased evapotranspi-

ration, cancelling significant change demonstrated in

earlier runs for the JJA precipitation in the north China

plain. Additionally, in the new runs, the mechanism for

the decrease in JJA precipitation in northern India was

slightly different from that in the former CTL and EXP

simulations. In the new simulations, the vapor flow

anomaly was from the Indian subcontinent to the Bay of

Bengal (Fig. 9c), whereas in the former simulations the

flow was conveyed to the Arabian Sea. The comparison

between Figs. 9 and 6 indicates that the precipitation

responses to theGWusemight be strongly related to the

background climate change with large uncertainties.

The consumedGW seems to be easily conveyed into the

sea, and the precipitation of northern India in the

monsoon season was likely to be reduced by GW use.

5. Conclusions and discussion

In this study, a GW extraction and consumption

scheme was incorporated into the CESM1.2.0. Two en-

sembles of simulations with and without GW extraction,

respectively, were conducted to detect changes in the

global hydrological cycle and climate due to GW

extraction.

The main conclusions of this study are as follows. 1)

There were three contiguous areas in the world that

were experiencing high-level GWextraction rates (more

than 90mmyr21 around 2003); these were located in

northern India and Pakistan, the north China plain, and

the central United States. 2) The combined processes of

GW extraction and consumption caused significant de-

clines in GW tables in the central United States, parts of

the north China plain, and northern India and Pakistan.

The water table below the ground surface declinedmore

than 4m from 1970 to 2005 in these areas. With these

GW table declines, the total terrestrial water storage

was also depleted. In regions with intenseGW irrigation,

the upper soil became wetter while deeper soil became

drier as a consequence of irrigation and GW table de-

clines. The wet surface soils in turn increased LH by

5–20Wm22 and decreased SH by 2–10Wm22 in most

key areas of GW extraction. 3) The processes of GWuse

significantly cooled the air at 850 hPa, decreasing tem-

perature by approximately 0.38C in northern India and

northern China plus a large area of central Russia. A

thermal circulation was induced by the cooling effects

and decreased the geopotential height at 850hPa in the

Tibetan Plateau. The north China plain experienced an

increase of JJA precipitation, whereas the precipitation

in northern India decreased because the Indian mon-

soon and its transport of water vapor were weaker as a

result of cooling induced by GW use. 4) Background

climate change may complicate the climatic effects of

GW use, especially for the precipitation and wind field

responses.

The research demonstrated possible global hydrologic

and climatic responses to GW extraction. However, the

assumptions and limitations in the study should be

noted. Besides the uncertainties within the original

version of CESM1.2.0 (Gent et al. 2011), the scheme

developed in the current study to describe GW with-

drawal and consumption is highly simplified. In fact, not

all water consumed in agriculture is used for irrigation.

For example, the water used to supply livestock and fish

ponds is not counted in this research, although it is much

smaller in magnitude than the water consumed by irri-

gation (Siebert et al. 2010). In addition, the amount of

water consumed in industrial and domestic water use is

uncertain due to varieties of complex consumptive uses,

such as in cooling, heating, electricity generation,

washing, drinking, and catering. In the current study,

water consumption was divided only into net expense

and waste water. Moreover, the water withdrawal and

consumption data used as input in the model contained

many uncertainties. The ratios of water use for agricul-

tural, industrial, and domestic water use are archived by

the FAOon a regional basis, which was inconsistent with

the scale of the simulation grid cells. However, since

there are only several centers of severe GW extraction

worldwide, the regional data were probably adequate

for the study. Furthermore, the assumption that the

amount of GW extracted changed linearly from 1965 to

1990 and from 1990 to 2009 was a simple representation

that surely caused some deviations between the GW

extractions used in the model and actual extraction.

There are also some uncertainties within the GMIA5

data, which can be seen in Siebert et al. (2005). In

1 JANUARY 2017 ZENG ET AL . 87

addition, setting the waste water percentage of in-

dustrial water use and daily life to 30% for the whole

world is not accurate enough, and may produce some

uncertainties of the simulation results for the United

States and Europe, while in other regions the results

would be relatively unaffected since agriculture is the

main consumer of water in these areas. Another as-

pect of the uncertainties in this study relates to the

topographic heterogeneity and the spatial resolution

of the model. In reality, the interaction processes be-

tween land and atmosphere are tightly related to the

surface elevation at a much finer spatial scales than

our model resolution, as many previous studies have

shown (Avissar and Liu 1996; Avissar and Schmidt

1998). Although CAM4 and CLM4.5 include some

subgrid schemes that evaluate the standard deviation

of elevations within each model grid based on the

high-resolution data of topographic height and use the

standard deviation to calculate the fluxes between

land and atmosphere, the runoff processes, SW stor-

age, and snow dynamics to show the spatial hetero-

geneity (Neale et al. 2010, Oleson et al. 2013), these

simple parameterizations are relatively inaccurate

and thus produced more uncertainties in our results.

However, at this time, because the aim of this paper is

to make a first step toward showing the global impacts

of GW extraction and withdrawal on climate, we give

our results with a caution about the elevation-related

uncertainties. In the future, using an updated ver-

sion of CESM with its improved subgrid schemes for

the finescale processes, we will rerun our model to

see how the results are affected by topographic

heterogeneity.

The sustainable use of water resources must be

stressed. Although water reallocation by GW with-

drawal and consumption can temporally increase upper

soil moisture and runoff, all the results from this re-

search (i.e., GW table declines, reduced soil moisture in

deep layers of the soil profile, and decreases in total

terrestrial water storage) emphasize that available water

resources are unsustainable in regions that have high

GW pumping rates. In the future, water demand will

dramatically increase as the global population continues

to grow and economic development expands. To address

these difficulties, measures such as large-scale water

diversion, improvements in irrigation efficiency, and

water recycling should be implemented, instead of in-

tensifying GW extraction. Among those areas experi-

encing severe water resource problems due to GW

extraction, northern India may require the most urgent

interventions because the monsoon circulation is

weakened by GW use here, and both precipitation and

runoff are being reduced concurrently.

Future research related to this work is necessary in at

least two aspects. First, the climate model should be

further developed, especially in regard to the charac-

terizations of processes tightly connected to the GW

dynamic. A GW flow model that considers both vertical

and lateral GWflow has been tested at basin scale (Zeng

et al. 2016a,b) and will be incorporated into CESM that

better applies on the global scale in the future. The

coupling of a crop model with the CLM has also been

finished and tested in China (Chen and Xie 2012). These

two models will be applied into CESM1.2_GW and will

help improve the calculation of GW table levels and

water balances. Second, the scheme and data related to

GW extraction should be more specific and realistic to

more accurately reflect the variability of climate effects

at intra-annual and interannual scales. These improve-

ments will enable simulations that are more accurate in

assessing water sustainability and the hydrologic and

climatic effects of human activities with regard to water

use. As a consequence, the improved models will sub-

stantially benefit the management of water resources

and the adaptation to climate change with society

development.

Acknowledgments. This study was supported by the

National Natural Science Foundation of China

(Grants 91125016 and 41575096), and by the Chinese

Academy of Sciences Strategic Priority Research

Program under Grant XDA05110102. We thank Xing

Yuan, Xiangjun Tian, Shuang Liu, and Yan Yu for

their assistance with this work and helpful discussion.

We also thank Minghua Zhang and two other anony-

mous reviewers for the helpful comments that im-

proved the manuscript.

REFERENCES

Alvarez, M. D., M. M. Trovatto, M. A. Hernández, and

N. González, 2012: Groundwater flow model, recharge esti-

mation and sustainability in an arid region of Patagonia,

Argentina. Environ. Earth Sci., 66, 2097–2108, doi:10.1007/

s12665-011-1435-8.

Atlas, R., R. N. Hoffman, S. C. Bloom, J. C. Jusem, and

J. Ardizzone, 1996: A multiyear global surface wind velocity

dataset using SSM/I wind observations. Bull. Amer. Meteor.

Soc., 77, 869–882, doi:10.1175/1520-0477(1996)077,0869:

AMGSWV.2.0.CO;2.

Avissar, R., and Y. Liu, 1996: Three-dimensional numerical study

of shallow convective clouds and precipitation induced by land

surface forcing. J. Geophys. Res., 101, 7499–7518, doi:10.1029/

95JD03031.

——, and T. Schmidt, 1998: An evaluation of the scale at which

ground-surface heat flux patchiness affects the convective

boundary layer using large-eddy simulations. J. Atmos.

Sci., 55, 2666–2689, doi:10.1175/1520-0469(1998)055,2666:

AEOTSA.2.0.CO;2.

88 JOURNAL OF CL IMATE VOLUME 30

Barnston, A. G., and P. T. Schickedanz, 1984: The effect of irri-

gation on warm season precipitation in the southern Great

Plains. J. Climate Appl. Meteor., 23, 865–888, doi:10.1175/

1520-0450(1984)023,0865:TEOIOW.2.0.CO;2.

Chen, F., and Z. H. Xie, 2012: Effects of crop growth and devel-

opment on regional climate: A case study over East Asian

monsoon area. Climate Dyn., 38, 2291–2305, doi:10.1007/

s00382-011-1125-y.

Chen, X., and Q. Hu, 2004: Groundwater influences on soil mois-

ture and surface evaporation. J. Hydrol., 297, 285–300,

doi:10.1016/j.jhydrol.2004.04.019.

Coe, M. T., and J. A. Foley, 2001: Human and natural impacts on

the water resources of the Lake Chad basin. J. Geophys. Res.,

106, 3349–3356, doi:10.1029/2000JD900587.

DeAngelis, A., F. Dominguez, Y. Fan, A. Robock, M. D. Kustu,

and D. Robinson, 2010: Evidence of enhanced precipitation

due to irrigation over the Great Plains of the United States.

J. Geophys. Res., 115, D15115, doi:10.1029/2010JD013892.

Devic, G., D. Djordjevic, and S. Sakan, 2014: Natural and

anthropogenic factors affecting the groundwater quality in

Serbia. Sci. Total Environ., 468–469, 933–942, doi:10.1016/

j.scitotenv.2013.09.011.

Di, Z., Z. Xie, X. Yuan, X.-J. Tian, Z. Luo, and Y. Chen, 2011:

Prediction of water table depths under soil water–

groundwater interaction and stream water conveyance. Sci.

China Earth Sci., 54, 420–430, doi:10.1007/s11430-010-4050-8.

Dirmeyer, P. A., Y. Jin, B. Singh, andX. Yan, 2013: Trends in land–

atmosphere interactions from CMIP5 simulations.

J. Hydrometeor., 14, 829–849, doi:10.1175/JHM-D-12-0107.1.

Döll, P., and Coauthors, 2012: Impact of water withdrawals

from groundwater and surface water on continental water

storage variations. J. Geodyn., 59–60, 143–156, doi:10.1016/

j.jog.2011.05.001.

Gent, P. R., and Coauthors, 2011: The Community Climate System

Model version 4. J. Climate, 24, 4973–4991, doi:10.1175/

2011JCLI4083.1.

Haddeland, I., D. P. Lettenmaier, and T. Skaugen, 2006: Effects of

irrigation on the water and energy balances of the Colorado

and Mekong River basins. J. Hydrol., 324, 210–223,

doi:10.1016/j.jhydrol.2005.09.028.

Hanasaki, N., S. Kanae, T. Oki, K. Masuda, K. Motoya,

N. Shirakawa, Y. Shen, and K. Tanaka, 2008: An integrated

model for the assessment of global water resources—Part 1:

Model description and input meteorological forcing. Hydrol.

Earth Syst. Sci., 12, 1007–1025, doi:10.5194/hess-12-1007-2008.

Hunke, E. C., and W. H. Lipscomb, 2008: CICE: The Los Alamos

sea ice model user’s manual, version 4. Los Alamos National

Laboratory Tech. Rep. LA-CC-06-012, 76 pp.

Hurrell, J. W., and Coauthors, 2013: The Community Earth

System Model: A framework for collaborative research.

Bull. Amer. Meteor. Soc., 94, 1339–1360, doi:10.1175/

BAMS-D-12-00121.1.

Jones, P. W., P. H. Worley, Y. Yoshida, J. White, and J. Levesque,

2005: Practical performance portability in the Parallel Ocean

Program (POP). Concurr. Comput. Pract. Exper., 17, 1317–

1327, doi:10.1002/cpe.894.

Khan, S., T. Rana, H. Gabriel, and M. Ullah, 2008: Hydrogeologic

assessment of escalating groundwater exploitation in the In-

dus Basin, Pakistan.Hydrogeol. J., 16, 1635–1654, doi:10.1007/

s10040-008-0336-8.

Koster, R. D., P. A. Dirmeyer, A. N. Hahmann, R. Ijpelaar,

L. Tyahla, P. Cox, and M. J. Suarez, 2002: Comparing the

degree of land–atmosphere interaction in four atmospheric

general circulation models. J. Hydrometeor., 3, 363–375,

doi:10.1175/1525-7541(2002)003,0363:CTDOLA.2.0.CO;2.

——, and Coauthors, 2006: GLACE: The Global Land–

Atmosphere Coupling Experiment. Part I: Overview.

J. Hydrometeor., 7, 590–610, doi:10.1175/JHM510.1.

Kumar,M. D., andO. P. Singh, 2008: How serious are groundwater

over-exploitation problems in India: A fresh investigation into

an old issue. International Water Management Institute Rep.,

20 pp. [Available online at http://publications.iwmi.org/PDF/

H041881.pdf.]

Lee, E., T. N. Chase, B. Rajagopalan, R. G. Barry, T.W. Biggs, and

P. J. Lawrence, 2009: Effects of irrigation and vegetation ac-

tivity on early Indian summer monsoon variability. Int.

J. Climatol., 29, 573–581, doi:10.1002/joc.1721.Leng, G., M. Huang, Q. Tang,W. J. Sacks, H. Lei, and L. R. Leung,

2013: Modeling the effects of irrigation on land surface fluxes

and states over the conterminous United States: Sensitivity to

input data and model parameters. J. Geophys. Res. Atmos.,

118, 9789–9803, doi:10.1002/jgrd.50792.

——, ——, ——, H. Gao, and L. R. Leung, 2014: Modeling the

effects of groundwater-fed irrigation on terrestrial hydrology

over the conterminous United States. J. Hydrometeor., 15,

957–972, doi:10.1175/JHM-D-13-049.1.

Lin, S.-J., 2004: A ‘‘vertically Lagrangian’’ finite-volume dynamical

core for global models. Mon. Wea. Rev., 132, 2293–2307,

doi:10.1175/1520-0493(2004)132,2293:AVLFDC.2.0.CO;2.

Lindsay, K., and Coauthors, 2014: Preindustrial-control and

twentieth-century carbon cycle experiments with the Earth

system model CESM1(BGC). J. Climate, 27, 8981–9005,

doi:10.1175/JCLI-D-12-00565.1.

Liu, C., J. Yu, and E. Kendy, 2001: Groundwater exploitation and

its impact on the environment in the North China Plain.Water

Int., 26, 265–272, doi:10.1080/02508060108686913.

Liu, J., D. Qin, H. Wang, M. Wang, and Z. Yang, 2010: Dualistic

water cycle pattern and its evolution in Haihe River basin.

Chin. Sci. Bull., 55, 1688–1697, doi:10.1007/s11434-010-3043-5.

Llamas, M. R., 1988: Conflicts between wetland conservation and

groundwater exploitation: Two case histories in Spain. Envi-

ron. Geol. Water Sci., 11, 241–251, doi:10.1007/BF02574814.

Lo, M. H., and J. S. Famiglietti, 2013: Irrigation in California’s

Central Valley strengthens the southwestern U.S. water cycle.

Geophys. Res. Lett., 40, 301–306, doi:10.1002/grl.50108.

Mao, X., J. Ni, and Y. Guo, 2000: A case study on characteristics of

wastewater effluents in accelerated economic growth areas (in

Chinese). Acta. Sci. Circumstantiae, 20, 219–224.

Neale, R. B., and Coauthors, 2010: Description of the NCAR Commu-

nity Atmosphere Model (CAM 4.0). NCAR Tech. Note NCAR/

TN-4851STR, 212 pp. [Available online at http://www.cesm.ucar.

edu/models/cesm1.0/cam/docs/description/cam4_desc.pdf.]

——, J. Richter, S. Park, P. H. Lauritzen, S. J. Vavrus, P. J. Rasch, and

M. H. Zhang, 2013: The mean climate of the Community Atmo-

sphere Model (CAM4) in forced SST and fully coupled experi-

ments. J. Climate, 26, 5150–5168, doi:10.1175/JCLI-D-12-00236.1.

Oleson, K. W., and Coauthors, 2013: Technical description of

version 4.5 of theCommunity LandModel (CLM).NCARTech.

Note NCAR/TN-5031STR, 420 pp. [Available online at http://

www.cesm.ucar.edu/models/cesm1.2/clm/models/lnd/clm/

doc/UsersGuide/book1.html.]

Omole, D. O., 2013: Sustainable groundwater exploitation in

Nigeria. J. Water Resour. Ocean Sci., 2, 9–14, doi:10.11648/

j.wros.20130202.11.

Ozdogan, M., M. Rodell, H. K. Beaudoing, and D. L. Toll, 2010:

Simulating the effects of irrigation over the United States in a

1 JANUARY 2017 ZENG ET AL . 89

land surface model based on satellite-derived agricul-

tural data. J. Hydrometeor., 11, 171–184, doi:10.1175/

2009JHM1116.1.

Pokhrel, Y. N., N. Hanasaki, S. Koirala, J. Cho, P. J.-F. Yeh,

H. Kim, S. Kanae, and T. Oki, 2012a: Incorporating anthro-

pogenic water regulation modules into a land surface model.

J. Hydrometeor., 13, 255–269, doi:10.1175/JHM-D-11-013.1.

——, ——, P. J. Yeh, T. J. Yamada, S. Kanae, and T. Oki, 2012b:

Model estimates of sea-level change due to anthropogenic

impacts on terrestrial water storage. Nat. Geosci., 5, 389–392,

doi:10.1038/ngeo1476.

——, S. Koirala, P. J. F. Yeh, N. Hanasaki, L. Longuevergne,

S. Kanae, and T. Oki, 2015: Incorporation of groundwater

pumping in a global land surface model with the representa-

tion of human impacts. Water Resour. Res., 51, 78–96,

doi:10.1002/2014WR015602.

Qian, T., A. Dai, K. E. Trenberth, and K. W. Oleson, 2006: Sim-

ulation of global land surface conditions from 1948 to 2004.

Part I: Forcing data and evaluations. J. Hydrometeor., 7, 953–975, doi:10.1175/JHM540.1.

Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V.

Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003:

Global analyses of sea surface temperature, sea ice, and night

marine air temperature since the late nineteenth century.

J. Geophys. Res., 108, 4407, doi:10.1029/2002JD002670.

Rodell, M., and J. S. Famiglietti, 2002: The potential for satellite-

based monitoring of groundwater storage changes using

GRACE: TheHigh Plains aquifer, CentralUS. J. Hydrol., 263,

245–256, doi:10.1016/S0022-1694(02)00060-4.

——, I. Velicogna, and J. S. Famiglietti, 2009: Satellite-based es-

timates of groundwater depletion in India. Nature, 460, 999–

1002, doi:10.1038/nature08238.

Shi, J., R. Ma, J. Liu, and Y. Zhang, 2013: Suitability assessment of

deep groundwater for drinking, irrigation and industrial pur-

poses in Jiaozuo City, Henan Province, north China.Chin. Sci.

Bull., 58, 3098–3110, doi:10.1007/s11434-013-5952-6.

Siebert, S., P. Döll, J. Hoogeveen, J. M. Faures, K. Frenken, and

S. Feick, 2005: Development and validation of the global map

of irrigation areas. Hydrol. Earth Syst. Sci., 9, 535–547,

doi:10.5194/hess-9-535-2005.

——, J. Burke, J. M. Faures, K. Frenken, J. Hoogeveen, P. Döll,and F. T. Portmann, 2010: Groundwater use for irrigation—A

global inventory. Hydrol. Earth Syst. Sci., 14, 1863–1880,

doi:10.5194/hess-14-1863-2010.

——, V. Henrich, K. Frenken, and J. Burke, 2013: Global Map of

Irrigation Areas, version 5. Rheinische Friedrich-Wilhelms-

University and FAO. [Available online at http://www.fao.org/

nr/water/aquastat/irrigationmap/index10.stm.]

Sophocleous, M., 2005: Groundwater recharge and sustainability in

the high plains aquifer in Kansas, USA.Hydrogeol. J., 13, 351–

365, doi:10.1007/s10040-004-0385-6.

Tapley, B. D., S. Bettadpur, M.Watkins, and C. Reigber, 2004: The

Gravity Recovery and Climate Experiment: Mission overview

and early results.Geophys. Res. Lett., 31, L09607, doi:10.1029/

2004GL019920.

Tiwari, V. M., J. Wahr, and S. Swenson, 2009: Dwindling ground-

water resources in northern India, from satellite gravity ob-

servations. Geophys. Res. Lett., 36, L18401, doi:10.1029/

2009GL039401.

Wu, B., and Coauthors, 2014: Systematic assessment of the un-

certainty in integrated surface water–groundwater modeling

based on the probabilistic collocation method. Water Resour.

Res., 50, 5848–5865, doi:10.1002/2014WR015366.

Xie, P., and P. A. Arkin, 1997: Global precipitation: A 17-year

monthly analysis based on gauge observations, satellite es-

timates, and numerical model outputs. Bull. Amer. Meteor.

Soc., 78, 2539–2558, doi:10.1175/1520-0477(1997)078,2539:

GPAYMA.2.0.CO;2.

Xie, Z., andX. Yuan, 2010: Prediction of water table under stream–

aquifer interactions over an arid region.Hydrol. Processes, 24,160–169, doi:10.1002/hyp.7434.

——, Z. Di, Z. Luo, andQ.Ma, 2012: A quasi-three-dimensional

variably saturated groundwater flow model for climate

modeling. J. Hydrometeor., 13, 27–46, doi:10.1175/

JHM-D-10-05019.1.

——, N. Zeng, H. Wang, Z. Lin, X. Tian, and B. Jia, 2014: Past,

present and future of the carbon cycle.Natl. Sci. Rev., 1, 18–21,

doi:10.1093/nsr/nwt021.

Young, I. R., 1999: Seasonal variability of the global ocean wind and

wave climate. Int. J. Climatol., 19, 931–950, doi:10.1002/

(SICI)1097-0088(199907)19:9,931::AID-JOC412.3.0.CO;2-O.

Yu, Y., Z. H. Xie, and X. B. Zeng, 2014: Impacts of modified

Richards equation on RegCM4 regional climate modeling

over East Asia. J. Geophys. Res. Atmos., 119, 12 642–12 659,

doi:10.1002/2014JD021872.

Yuan, X., Z. Xie, and M. Liang, 2008: Spatiotemporal prediction

of shallow water table depths in continental China. Water

Resour. Res., 44, W04414, doi:10.1029/2006WR005453.

Zeng, Y., and Z. Xie, 2015: Projection and evaluation of the land–

atmosphere coupling strength over China by CMIP5 models

(in Chinese).Climate Environ. Res., 20, 337–346, doi:10.3878/

j.issn.1006-9585.2015.14242.

——, ——, Y. Yu, S. Liu, L. Wang, B. Jia, P. Qin, and Y. Chen,

2016a: Ecohydrological effects of stream–aquifer water in-

teraction: A case study of theHeihe River basin, northwestern

China. Hydrol. Earth Syst. Sci., 20, 2333–2352, doi:10.5194/hess-20-2333-2016.

——, ——, ——, ——, ——, J. Zou, P. Qin, and B. Jia, 2016b:

Effects of anthropogenic water regulation and groundwater

lateral flow on land processes. J. Adv. Model. Earth Syst., 8,1106–1131, doi:10.1002/2016MS000646.

Zou, J., Z. Xie, Y. Yu, C. Zhan, and Q. Sun, 2014: Climatic re-

sponses to anthropogenic groundwater exploitation: A case

study of theHaihe River Basin, northern China.ClimateDyn.,

42, 2125–2145, doi:10.1007/s00382-013-1995-2.

——, ——, C. Zhan, P. Qin, Q. Sun, B. Jia, and J. Xia, 2015:

Effects of anthropogenic groundwater exploitation on

land surface processes: A case study of the Haihe River

Basin, northern China. J. Hydrol., 524, 625–641, doi:10.1016/

j.jhydrol.2015.03.026.

90 JOURNAL OF CL IMATE VOLUME 30