1. Introduction
Land use alteration of soil moisture could considerably affect the hydrologic cycle (Pielke et al. 2007; Kustu et al. 2010) and local and regional climate (Boucher et al. 2004; Bonfils and Lobell 2007), particularly in the wet–dry transition zones where land–atmosphere interactions play an important role in driving seasonal-to-interannual climate variability at the local and regional scales (Koster et al. 2004). In the context of long-term predictions, an improved understanding of hydrologic responses to anthropogenic perturbations related to both climate and land use changes are required. Recent studies have demonstrated important impacts of irrigation on land surface water budget and energy fluxes based on observations and model simulations. By adding water into the soil column in the growing season, irrigation practice increases soil moisture contents available for plants, which increases water vapor released into the atmosphere through evapotranspiration (ET; Boucher et al. 2004; Gordon et al. 2005) and greatly alters the hydrologic cycle (Haddeland et al. 2006; Kustu et al. 2010) in agricultural regions. Changes in soil moisture in turn could affect the surface energy balance, increasing latent heat fluxes, decreasing sensible heat fluxes, and consequently cooling the land surface (Haddeland et al. 2006; Shibuo et al. 2007; Tang et al. 2007; Sacks et al. 2009). Changes in the partitioning of the surface energy balance subsequently affect atmospheric boundary layer development (Kawase et al. 2008; Qian et al. 2013), including mass and energy exchanges between the land surface and the atmosphere in addition to precipitation-generating processes (e.g., convection; Lee et al. 2009; Douglas et al. 2009; DeAngelis et al. 2010; Lo and Famiglietti 2013; Qian et al. 2013; Harding and Snyder 2012). The changes in boundary layer development and shallow convection could potentially feed back to weather and climate systems (Sacks et al. 2009; Kueppers et al. 2007; Lobell et al. 2009; Puma and Cook 2010), with significant influence on local and regional hydrologic variability (Diffenbaugh 2009).
Although many studies have evaluated the impacts of irrigation on terrestrial hydrology using modeling approaches, uncertainties in input data such as irrigation fraction and model parameters (Leng et al. 2013), as well as irrigation schemes (Sorooshian et al. 2012), can lead to large uncertainties in the estimated impacts. A common approach in modeling irrigation that sets the soil moisture to field capacity or soil saturation (e.g., Adegoke et al. 2003; Lobell et al. 2008; Kueppers and Snyder 2011; Harding and Snyder 2012) or prescribes a fixed elevated amount of ET (e.g., Segal et al. 1998; Boucher et al. 2004) in irrigated areas can lead to overestimation or underestimation of irrigation effects. Moreover, few modeling studies considered the sources and availability of irrigation water and assumed unlimited irrigation water supply from local runoff or other unidentified sources (e.g., Sacks et al. 2009; Ozdogan et al. 2010; Sorooshian et al. 2012). This can have important implications to estimating irrigation effects, especially for regions dominated by groundwater pumping for irrigation.
In the United States, about 60% of irrigation relies on groundwater. Irrigation in California’s Central Valley and the High Plains accounts for about 50% of groundwater use nationwide (Scanlon et al. 2012). Reductions in groundwater storage caused by pumping for irrigation have become an important water resource management issue because of their long-lasting impacts on the aquifers. Furthermore, changes in groundwater storage and deep layer soil moisture can influence surface processes through surface–subsurface interactions (Liang et al. 2003; Leung et al. 2011). Thus, it is important to consider the effects of groundwater pumping for irrigation on surface and subsurface land surface hydrology in an integrative manner. Therefore, in addition to modeling realistic irrigation amounts, accurate representation of the water source for irrigation should be included in irrigation impact studies.
The primary objective of this study is to incorporate a groundwater pumping scheme in a land surface model and to characterize groundwater depletion over the conterminous United States (CONUS). The irrigation scheme in the model was first calibrated to reproduce the irrigation amounts from agriculture census data. The calibrated irrigation amounts were further divided into surface water and groundwater withdrawals based on the census data. The impacts of groundwater exploitation for irrigation on water and energy budgets over CONUS were then investigated.
2. Data and methodology
a. Data
The Moderate Resolution Imaging Spectroradiometer (MODIS)-based irrigated fractional area map at the 500-m resolution over CONUS (Ozdogan and Gutman 2008; Ozdogan et al. 2010) was obtained and aggregated to the same resolution of the model grid cell. This map represents the percentage of potential irrigation area in each grid cell around 2001. The new map, which has been comprehensively validated against agricultural statistics, has been used in land surface modeling studies (Ozdogan et al. 2010; Leng et al. 2013) and climate modeling studies (Qian et al. 2013). The satellite-based ET product from Tang et al. (2009a) was aggregated to the county scale for model evaluation. This dataset was based entirely on satellite data and has been used in irrigation impact studies over western United States (Tang et al. 2009a; Sorooshian et al. 2012; Leng et al. 2013) and in hydrological applications (Gao et al. 2010; Tang et al. 2009b).
b. Hydrologic and irrigation schemes in CLM4
This study used the Community Land Model, version 4.0 (CLM4), developed by the National Center for Atmospheric Research (NCAR). In the framework of the Community Earth System (Collins et al. 2006; Gent et al. 2010; Lawrence et al. 2011b), CLM4 can be used in coupled and offline modes. Compared to its earlier versions, CLM4 represents extensive enhancements in representing hydrological processes such as runoff generation, soil hydrology, and groundwater dynamics (Lawrence et al. 2011a).
The hydrologic parameterizations relevant to the irrigation and groundwater schemes are briefly summarized in the appendix to help the readers understand the implementation of the irrigation and ground pumping schemes, which will be described in the following sections. Briefly, CLM4 represents the vertical soil moisture profile through solving the one-dimensional Richards equation, with surface and subsurface runoff generation processes parameterized following the simplified TOPMODEL concept (Niu et al. 2005). At each time step, a total liquid runoff term qrunoff leaves the soil column and is assumed to enter local streams that reside within the same grid cell. An unconfined aquifer is assumed to lie beneath the 3.8-m soil column, with storages in the aquifer and total storage under the groundwater table of depth 
CLM4 includes an irrigation scheme that computes water and energy processes for irrigated croplands (www.cesm.ucar.edu/models/cesm1.0/clm/CLMcropANDirrigTechDescriptions.pdf; see also Leng et al. 2013). Irrigation has been implemented for the C3 generic crop only, as it is the only crop type available in CLM4 and it responds dynamically to climate. The fraction of irrigated croplands in each grid cell is given by the smaller of 1) the grid cell’s total cropland area and 2) the grid cell’s area equipped for irrigation from a predefined input dataset (e.g., Siebert et al. 2005; Ozdogan and Gutman 2008). The remainder of the grid cell’s cropland area (if any) is then assigned as unirrigated croplands. Irrigated and unirrigated crops are placed on separate soil columns so that water and energy balance can be tracked separately.
c. A groundwater pumping scheme and its implementation in CLM4
d. Experimental design
We performed offline CLM4 simulations driven by the North America Land Data Assimilation System phase II (NLDAS2) meteorological forcings (Xia et al. 2012) at an hourly time step from 1979 to 2011. Vegetation, soil, and land cover characteristics for the model grids were derived from the high-resolution (i.e., 0.05°) CLM input dataset developed by Ke et al. (2012).
The numerical experiments in this study are summarized in Table 1. For all simulations listed in Table 1, we used the same initial condition derived by repeating the simulation with the NLDAS2 forcing and the default setting over the period of 1979–2007 for 36 cycles (i.e., 1044 yr) until all state variables reach equilibrium. Following Leng et al. (2013), two short simulations driven by NLDAS2 forcing in year 2000 were first conducted for the purpose of calibrating Firrig in Eq. (3), listed as IRRIGnocal and IRRIGcal in Table 1. IRRIGnocal serves as a reference case for comparison with IRRIGcal. In IRRIGcal, the value of Firrig for each irrigated grid cell over CONUS was calibrated by performing simulations of the year 2000 using the irrigation scheme without groundwater pumping by varying Firrig between (0, 1) at a regular interval of 0.05. By comparing the simulated irrigation amounts with the USGS census data at the county level in 2000, the best value of Firrig was chosen for each irrigated county. Grid cells within each irrigation county were then assigned the same calibrated Firrig value to simulate realistic irrigation amount comparable to observations at the county scale. Because of inconsistencies between the grid-based irrigation map and the USGS reported irrigation area, grid cells where USGS indicates no irrigation for the county are assigned the default value of Firrig and are excluded from the validation process, as will be described later. We note that by calibrating Firrig in Eq. (2), the simulated irrigation amount is forced to match the county-level statistics. Since the observed irrigation amount already reflects the availability of water for irrigation, our estimated irrigation amount is more realistic compared to previous studies and indirectly accounted for irrigation water availability. Furthermore, halting groundwater pumping before the groundwater storage becomes negative, as discussed in section 2c, adds further constraint to the amount of water that could be administered to the land surface.
Summary of numerical experiments.
Three offline simulations from 1979 to 2011 were then conducted to explore the effects of irrigation on surface and subsurface processes induced by groundwater exploitation. The control simulation without irrigation (i.e., CTRL) was conducted to simulate the natural state, ignoring human impacts. Two 29-yr simulations for 1979–2011 with the calibrated parameters in the irrigation scheme were performed as sensitivity experiments. These two simulations (hereafter denoted as IRRIG and PUMP) adopted the original irrigation scheme in CLM4 and the proposed irrigation scheme enhanced with groundwater pumping, respectively. The differences between these two simulations and CTRL were investigated to evaluate the potential responses of land surface hydrology to irrigation, while the difference between the IRRIG and PUMP provides a comparison of the characteristics using different irrigation schemes so that the effects of groundwater withdrawals for irrigation can be evaluated. Before the simulations, an offline spinup was performed by cycling the NLDAS2 forcing until soil moisture, temperature, and groundwater table depth in CLM4 reached a balanced state, consistent with the model setup in Leng et al. (2013). The simulation results for 2000–11 were used in the following analysis.
3. Results
a. Spatial distribution of calibrated Firrig and estimated rsrf and rgrd
Based on results in Leng et al. (2013), a large discrepancy and uncertainties exist between the irrigation fraction datasets from Siebert et al. (2005) and Ozdogan and Gutman (2008), which are important sources of uncertainty in modeling irrigation effects. However, when used as inputs to CLM4, the smaller irrigation fractional area in Ozdogan and Gutman (2008) produces smaller interannual variability in simulated irrigation impacts compared to simulations based on Siebert et al. (2005). Hence, the MODIS-based irrigation fraction dataset from Ozdogan and Gutman (2008) was used in this study to represent an estimate of the lower bound for irrigation impacts over CONUS.
Figure 1 shows the spatial distribution of irrigated fractional area from Ozdogan and Gutman (2008) aggregated to the county scale. The four boxes in this figure are defined as lower Mississippi (LM), southern Great Plains (SGP), California (CA), and Pacific Northwest (PNW). Major irrigated lands are located in the semiarid western United States, compared to the small fractions scattered over the eastern part of the country, where rain-fed crops dominate. East of the Continental Divide, irrigated areas are mostly found along the Mississippi valleys and agricultural regions in the eastern coastal plain and southwestern Georgia, where high-water-demand crops (e.g., cotton) are planted. In the central United States, irrigation areas are mostly found in the semiarid areas of western Kansas, Oklahoma, Texas, and Nebraska. In the western United States, major irrigated croplands are distributed along the Central Valley, Columbia River basin, Idaho’s Snake River basin, and the high mountains of Colorado.
Spatial distributions of irrigated fractional area from Ozdogan and Gutman (2008) aggregated at the county scale. The four boxes represent the LM, SGP, CA, and PNW analysis areas.
Citation: Journal of Hydrometeorology 15, 3; 10.1175/JHM-D-13-049.1
Figure 2 shows the spatial distribution of the calibrated Firrig values at the county scale. High values of Firrig are found in the western United States, which may be partly due to the smaller irrigated area used in the model compared to the USGS data. Conversely, low values of Firrig are distributed over the central and eastern United States, as the model irrigation areas are larger than those reported by USGS. We note that the pattern of the calibrated Firrig values reported here is consistent with that in Leng et al. (2013), in which calibration was performed at a water resources region (WRR) level. The calibration in this study is better constrained by observations (i.e., census) at a higher resolution (i.e., WRRs versus counties) and is more suitable for studying highly localized phenomena such as groundwater pumping, which is a focus of this study.
Spatial distribution of the calibrated Firrig at the county scale.
Citation: Journal of Hydrometeorology 15, 3; 10.1175/JHM-D-13-049.1
Figure 3 shows the spatial distribution of simulated irrigation amounts (km3 yr−1) in the year 2000 before and after calibration (i.e., IRRIGnocal and IRRIGcal, respectively), aggregated to the county level for comparison against those reported by USGS. Both IRRIGnocal and IRRIGcal capture the spatial pattern of irrigation amounts reasonably well at the county scale. Relatively high irrigation amounts were simulated by IRRIGnocal over major irrigated regions in CA, PNW, SGP, and LM. After calibration, CLM4 produced results generally closer to USGS than that before calibration in terms of both spatial distribution and magnitude.
Spatial distributions of annual irrigation amounts (km3 yr−1) simulated by (a) CLM4 before calibration (i.e., IRRIGnocal), (b) after calibration (i.e., IRRIGcal), (c) USGS estimates in 2000, and (d) the difference between IRRIGcal and USGS. Note that the simulation results were aggregated at the county scale for comparison with the USGS data.
Citation: Journal of Hydrometeorology 15, 3; 10.1175/JHM-D-13-049.1
To evaluate the impacts of irrigation on land surface water and energy budgets, we compared simulated ET against the MODIS-based ET product in Tang et al. (2009a) for all irrigated counties (see Fig. 4). The simulated ET before and after calibration both matches the MODIS ET estimates reasonably well. However, after calibration, notable improvements are found over the eastern, central, and western United States, resulting in smaller errors and higher correlations, as shown by the values of the root-mean-square error (RMSE) and bias.
Simulated ET (a) before calibration and (b) after calibration vs MODIS-based ET in the irrigated counties over three regions: western (124°–110°W, black), central (110°–95°W, red), and eastern (95°–67°W, green). Also listed in blue at the top of each panel is the RMSE and bias of the model for the entire CONUS.
Citation: Journal of Hydrometeorology 15, 3; 10.1175/JHM-D-13-049.1
Figure 5 shows the spatial distribution of rgrd and rsrf at the county scale based on the agricultural census in 2000. Surface water is the primary source for irrigation in the arid West and the mountainous regions (e.g., the Pacific Northwest and Idaho), with rsrf up to 90% in most counties, while groundwater serves as the primary source for irrigation in the central states, with rgrd up to 90% in most counties. For the four regions defined in Fig. 1, most counties within SGP and LM are dominated by groundwater withdrawals with rgrd up to 90%, while PNW and CA rely both on groundwater and surface water for irrigation, with rgrd up to 60% in some counties. From the spatial distributions of total, surface water, and groundwater withdrawals for irrigation (not shown), California, Idaho, Nebraska, and Colorado combined account for 50% of the total irrigation withdrawals. Specifically, California and Idaho account for 40% of surface water withdrawals, while California and Nebraska account for 30% of groundwater withdrawals. At the regional scale, in the Ogallala aquifer extending from Nebraska to Texas (Sophocleous 2012), where 27% of irrigated lands in the United States are located, 30% of all groundwater for irrigation in the United States is used.
Spatial distributions of (a) rgrd and (b) rsrf in percentages at the county scale.
Citation: Journal of Hydrometeorology 15, 3; 10.1175/JHM-D-13-049.1
b. Effects of irrigation and groundwater withdrawal on terrestrial water and energy budgets
Groundwater pumping for irrigation can perturb the surface water and energy balances through changes in ET, groundwater recharge, soil moisture, and other budget terms. Figure 6 shows the spatial distributions of annual mean water table depth [
Spatial distributions of (a)–(c) annual mean water table depth (m), (d)–(f) potential recharge (mm month−1), and (g)–(i) annual mean latent heat flux (W m−2) for (top) the CTRL simulation, (middle) PUMP − CTRL, and (bottom) IRRIG − CTRL.
Citation: Journal of Hydrometeorology 15, 3; 10.1175/JHM-D-13-049.1
From the IRRIG − CTRL panels (i.e., Figs. 6c,f,i), irrigation has significant effects on LH in the Central Valley, the Columbia Basin, and the western side of the central Great Plains, where irrigation amounts are high (Fig. 3) and the atmosphere during the growing season is relatively dry. For example, irrigation in the Central Valley can increase LH by more than 18 W m−2 in the summer of dry years. We note that these estimates obtained using offline land surface simulations assumed that land–atmosphere interactions would not significantly affect the simulated responses of the land surface. In reality, by adding water to the surface, the irrigation-induced increase in ET would lead to increases in near-surface specific humidity, but decreases in the sensible heat and surface temperature, and therefore could potentially enhance shallow and deep convections and increase cloud formation and precipitation (e.g., Qian et al. 2013; DeAngelis et al. 2010), which may then influence the ET response to irrigation.
Because irrigation water withdrawn from the total runoff is applied to the soil, PE increases in areas where irrigation amounts are large, which indicates a net increase of water recharged to deep soil layers as the difference in applied irrigation amounts and enhanced ET due to irrigation. The increase in PE, however, does not lead to notable changes in 
When groundwater pumping is included in the irrigation scheme (i.e., the PUMP − CTRL panels in Figs. 6b,e,h), similar changes in LH and PE to IRRIG − CTRL were found, because the total irrigation amounts are similar between the IRRIG and PUMP simulations when the same Firrig values were used in the irrigation scheme. However, as the rate of groundwater extraction through pumping exceeds the rate of recharge, the water table depth becomes deeper in the four regions. The maximum difference in annual mean 
Figure 7 shows the temporal variations of the simulated irrigation amounts and the irrigation effects on groundwater storage [Wt (mm)] and ET (mm) averaged spatially over each of the four regions in Fig. 1. The simulated irrigation amounts by IRRIG and PUMP over LM, SGP, CA, and PNW exhibit the same order of magnitude but distinct interannual variability. For example, in relatively dry years such as 2007 for SGP and 2011 for CA, when more irrigation water was required to meet the crop demands, PUMP simulated lower irrigation amounts compared to IRRIG because of the constraints by the low groundwater storage Wt, as shown in Fig. 7f.
Temporal variations in the (top to bottom) four regions studied: (left) the simulated irrigation amounts by IRRIG (blue dashed line) and PUMP (red solid line); and the differences in (middle) Wt and (right) ET for IRRIG − CTRL (blue dashed line) and PUMP − CTRL (red solid line). Note that all variables are area-weighted averages over each of the four regions indicated in Fig. 1.
Citation: Journal of Hydrometeorology 15, 3; 10.1175/JHM-D-13-049.1
In the IRRIG scenario, recharge from upper to deep soil layers and therefore to the groundwater storage was small over SGP, CA, and PNW, resulting in negligible change in Wt. However, because of the deep water table depth in LM and thus a large gradient in soil moisture content between upper and lower soil layers (Fig. 6a), irrigation water applied to the soil column could potentially infiltrate and recharge groundwater storage more effectively, leading to an increase of Wt in IRRIG, with a larger increase in Wt corresponding to larger irrigation amounts in dry years. In the PUMP scenario, water withdrawals from groundwater in LM could outweigh the irrigation-induced increase in recharge from the soil, resulting in a decrease of Wt. Consequently, groundwater pumping reduces groundwater storage throughout the year in PUMP, with larger decreases in dry years when more groundwater is withdrawn to meet irrigation demand. This leads to opposite interannual fluctuations in irrigation effects on Wt between IRRIG and PUMP, depending on the options in the irrigation scheme (i.e., with or without groundwater pumping). Therefore, as clearly demonstrated in Fig. 7, high groundwater pumping rate and low recharge combined resulted in large depletion of groundwater over SGP, CA, and PNW, whereas high recharge and a shallow groundwater table in LM resulted in sustainable groundwater use.
Irrigation, with or without groundwater pumping, leads to increases in ET throughout the water year because moisture availability in the root zone to supply ET is increased by irrigation. Moreover, differences in the regionally averaged ET between IRRIG and PUMP are negligible, except during the growing season in SGP and CA. Overall, the impacts of irrigation in IRRIG and PUMP exhibited similar magnitude in PE and ET, but groundwater exploitation tends to have much larger impacts on the subsurface water balance (e.g., 
Figure 8 shows the difference in volumetric soil water content (mm3 mm−3) between IRRIG (blue dashed line), PUMP (red solid line), and CTRL over SGP where irrigation practice heavily relied on groundwater exploitation. In the growing season of May–September, water added by irrigation increases the soil water contents throughout the soil column, but larger increase is found in the upper soils than the lower layers. With sufficient water applied through irrigation, the upper soil water content (i.e., from the top one to three soil layers, extending from 0–0.09 m from the surface) responded similarly between IRRIG and PUMP because they simulate about the same irrigation amounts (Fig. 8). As the groundwater and surface water systems are dynamically connected, reduction in groundwater storage by groundwater pumping in PUMP facilitates percolation and recharge from shallow to deep soil layers and to the underlying aquifer. This resulted in a smaller increase in the soil water contents in PUMP compared to IRRIG, particularly after the irrigation period. Moreover, ground pumping affects deeper-layer soil moisture more significantly, as evidenced by the bigger difference between IRRIG and PUMP in the deeper soil layers. As the groundwater table deepens each year, shallow soil water percolates more to recharge groundwater so the differences between IRRIG and PUMP increase over the years. Over the decadal scale analyzed in this study (i.e., 2000–11), recharge is more significant in a dry year when more groundwater is withdrawn for irrigation, leading to a more significant decrease in the water table depth and groundwater storage.
Differences in volumetric soil water content (mm3 mm−3) for IRRIG − CTRL (blue dashed line) and PUMP − CTRL (red solid line) over the irrigated areas (irrigation fraction > 0) in the SGP region. Shown are the average for (a) the top three soil layers (0–0.09 m), (b) the fourth to seventh soil layers (0.09–0.5 m), and (c) the eighth to tenth soil layers (0.5–3.8 m).
Citation: Journal of Hydrometeorology 15, 3; 10.1175/JHM-D-13-049.1
For this study, the most recent (release 05) Gravity Recovery and Climate Experiment (GRACE) monthly gridded total water storage (TWS) data at 1° developed by three data centers (Center for Space Research, Geo Forschungs Zentrum, and Jet Propulsion Laboratory) were acquired from the GRACE Tellus website (http://grace.jpl.nasa.gov/). To correct the GRACE TWS bias from signal modification (such as filtering and truncation), we also applied the GRACE TWS gain factors from the same website (Landerer and Swenson 2012). In addition, we used the gridded fields of leakage and the GRACE measurement errors to rigorously estimate the associated regional TWS uncertainties. As shown in Fig. 9, the monthly anomalies in TWS from the simulations compare reasonably well with the GRACE estimates. For regions such as LM and PNW, where the contributions of irrigation or groundwater withdrawal to the total water balance are small, the three simulations overlap with each other. For California and SGP, however, including both irrigation and groundwater withdrawals brings the model simulated anomalies slightly closer to the GRACE estimated variations, especially during the summertime.
Comparison of monthly anomalies (dS/dt) in total water storage for CTRL (green), IRRIG (blue), and PUMP (red) against GRACE (black). The gray shaded regions indicate the uncertainty range of the GRACE estimates.
Citation: Journal of Hydrometeorology 15, 3; 10.1175/JHM-D-13-049.1
To further evaluate the performance of the simulations, the simulated area-weighted water-level change of the Ogallaga aquifer (i.e., boundary shown as green in Fig. 1) over the simulation period is retrieved and compared to observations (http://ne.water.usgs.gov/ogw/hpwlms/tablewlpre.html), as shown in Fig. 10. Because of substantial groundwater-fed irrigation, water level decline has been reported since 1950 over the region. The USGS, in collaboration with federal, state, and local entities, has been monitoring water levels in more than 7000 wells since 1987. For a detailed description of the monitoring program and observations, readers are referred to McGuire (2013). It is evident that by including groundwater pumping, the declining trend of groundwater depth could be better captured by the model. However, the discrepancy between the simulations and observations becomes larger after 2007, at the beginning of which a significant drop of groundwater depth was observed due to the drought in 2006, and the drop in groundwater level has not recovered ever since. The model’s inability to capture the magnitude of depletion could be attributed to limitations in the subsurface parameterizations of the model. For example, the depth of water table over the region ranges from 0 to 100 m below the surface (McGuire 2013), but a uniform subsurface depth of 8.8 m, including the soil column and aquifer, is assumed in CLM4. As the simulated water level already reached the bottom of the aquifer in 2006, the model could not simulate further declines in the water depth, despite the drought conditions and increased groundwater pumping in 2006.
Comparison of simulated water level changes for CTRL (green), IRRIG (blue), and PUMP (red) against USGS observations (circles).The CTRL and IRRIG values overlap making it difficult to see the green CTRL values.
Citation: Journal of Hydrometeorology 15, 3; 10.1175/JHM-D-13-049.1
4. Conclusions and discussion
Irrigation is a common practice to support agricultural production in semiarid regions. To complement surface water supply for irrigation, groundwater pumping can lead to significant depletion of groundwater and perturb the surface and subsurface water balance. Using the GRACE satellite data, Famiglietti et al. (2011) estimated that between 2003 and 2010, the Central Valley lost groundwater at a rate of 31 mm yr−1, largely because of groundwater use for irrigation. To date, few modeling studies on irrigation impacts have accounted for groundwater pumping and the interactions between groundwater dynamics, shallow soil moisture, and land surface fluxes (Pokhrel et al. 2012; Zou et al. 2013; Ferguson and Maxwell 2011, 2012). Moreover, irrigation amounts were arbitrarily prescribed or estimated based on default parameters in most modeling studies, which may overestimate or underestimate the effects of irrigation on land surface fluxes and states in offline simulations and may also lead to misrepresentation of irrigation effects on local and regional climates in coupled simulations, as discussed in Leng et al. (2013).
In this study, we incorporated a groundwater pumping scheme to be used in combination with the existing irrigation module in CLM4, which extracts water from the groundwater storage based on a ratio between surface water and groundwater withdrawals, as suggested by historical agricultural census. Using the census data, we first calibrated the parameter in the CLM4 irrigation scheme following Leng et al. (2013), but at finer spatial resolutions (i.e., at the county level) to realistically simulate irrigation amounts to capture the spatial variability, and we then partitioned the simulated irrigation amounts into groundwater and surface water withdrawals at the county scale. Finally, we applied CLM4 with and without irrigation and/or groundwater pumping schemes to evaluate and compare the impacts of irrigation practices by types of sources of water on local- and regional-scale water and energy budgets over CONUS.
Our results show that changes in soil moisture content induced by irrigation have significantly altered the water availability and distribution, with large effects on land surface fluxes. The effects of irrigation on surface and subsurface water and energy budgets are more evident in irrigated cropland areas, most notably in the Central Valley and southern Great Plains where annual mean latent heat fluxes may increase by over 10 W m−2 at a similar magnitude in simulations with and without groundwater pumping. Of the four regions analyzed in more details, only the lower Mississippi (LM) can support sustainable groundwater use because of the high recharge rate and shallower groundwater table. In contrast, low recharge rate combined with high groundwater exploitation rate in the southern Great Plains (SGP), California (CA), and Pacific Northwest (PNW) lead to fast depletion of groundwater. As the groundwater table deepens, the increase in soil water due to irrigation gradually reduces because more water administrated to the root zone percolates and recharges the groundwater, particularly after the irrigation season (Fig. 8). This suggests that irrigation may become less effective in increasing soil moisture as the groundwater table is lowered because of pumping.
We acknowledge that neither the irrigation module nor the groundwater pumping scheme is perfect, so model development is important and is being pursued. As discussed in Leng et al. (2013), a more complete parameterization in calculating irrigation demands to incorporate phenological development of roots and crop specific parameters has been incorporated into later versions of CLM (e.g., CLM4.5, Oleson et al. 2013). Significant efforts using an Earth system modeling approach have also been made to relax the assumption that water resources are freely available for irrigation by coupling a physically based river routing module, the Model for Scale Adaptive River Transport (MOSART; Li et al. 2013), and a water management module that simulates reservoir operations for multiple water uses (Voisin et al. 2013a,b) with CLM. Furthermore, the model assumes 100% irrigation efficiency, which could be relaxed by considering transport loss as well as efficiency of different irrigation techniques such as flooding, drip, and sprinkler irrigation systems (e.g., Evans and Zaitchik 2008).
Our results suggest that large-scale pumping, currently being omitted from most irrigation modeling studies, might greatly alter the hydrologic and land energy fluxes in agricultural regions, which may further feedback to influence the local and regional climate. However, comparisons between our simulations against observations from the Ogallaga aquifer and GRACE suggest that further development of the groundwater parameterization in CLM4 to effectively represent the two-way interactions between surface water and groundwater is needed. Ferguson and Maxwell (2011) demonstrated that the magnitude of pumping and irrigation impacts at a given location within an agricultural watershed is governed by local water table depth through groundwater–land surface feedbacks. Our results are consistent with their findings and call for incorporating more realistic parameterizations of groundwater dynamics and their interactions with the shallow surface and vegetation processes in land surface models and climate models (e.g., Fan et al. 2007; Leung et al. 2011) to allow more detailed investigation of feedbacks between groundwater pumping and irrigation efficiency and feedbacks between groundwater use and climate extremes such as droughts. For example, by using surface and subsurface runoff parameterizations in a 1D soil column to represent lateral fluxes leaving the soil column, CLM4 and other macroscale land surface models excluded the exchange between river water and groundwater. The latter is an important process in regions such as the lower Mississippi, where lateral flow of groundwater contributes to the formation of shallow groundwater tables in the floodplain. The assumption of a uniform soil depth of 3.8 m with a shallow unconfined aquifer should also be relaxed to better represent realistic geologic conditions. This has great implications for developing adaptation and mitigation strategies for water resources under climate change. In addition to including water management practices in current land surface and climate models, our study also underscores the importance of including observed irrigation amounts and irrigation water source for more realistically simulating hydrologic cycle dynamics.
Acknowledgments
This study was supported by the Integrated Earth System Modeling (iESM) project funded by Department of Energy Earth System Modeling and Integrated Assessment Research programs. The Pacific Northwest National Laboratory (PNNL) Platform for Regional Integrated Modeling and Analysis (PRIMA) Initiative provided support for the model configuration and datasets used in the numerical experiments. PNNL is operated by Battelle Memorial Institute for the U.S. Department of Energy under Contract DE-AC05-76RLO1830. This work was also partly funded by the National Natural Science Foundation of China (Grant 41171031) and the National Basic Research Program of China (Grant 2012CB955403). The authors would like to acknowledge Drs. Nathalie Voisin, Hongyi Li, Mohamad Hejazi, and Huimin Lei for their insightful discussions and suggestions that inspired this work.
APPENDIX
Hydrologic Parameterizations in CLM4
REFERENCES
Adegoke, J. O., Pielke R. A. Sr., Eastman J. , Mahmood R. , and Hubbard K. G. , 2003: Impact of irrigation on midsummer surface fluxes and temperature under dry synoptic conditions: A regional atmospheric model study of the US High Plains. Mon. Wea. Rev.,131, 556–564, doi:10.1175/1520-0493(2003)131<0556:IOIOMS>2.0.CO;2.
Bonfils, C., and Lobell D. , 2007: Empirical evidence for a recent slowdown in irrigation-induced cooling. Proc. Natl. Acad. Sci. USA, 104, 13 582–13 587, doi:10.1073/pnas.0700144104.
Boucher, O., Myhre G. , and Myhre A. , 2004: Direct human influence of irrigation on atmospheric water vapour and climate. Climate Dyn., 22, 597–603, doi:10.1007/s00382-004-0402-4.
Clapp, R. B., and Hornberger G. M. , 1978: Empirical equations for some soil hydraulic properties. Water Resour. Res., 14, 601–604, doi:10.1029/WR014i004p00601.
Collins, W. D., and Coauthors, 2006: The formulation and atmospheric simulation of the Community Atmosphere Model version 3 (CAM3). J. Climate, 19, 2144–2161, doi:10.1175/JCLI3760.1.
DeAngelis, A., Dominguez F. , Fan Y. , Robock A. , Kustu M. D. , and Robinson D. , 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.
Diffenbaugh, N. S., 2009: Influence of modern land cover on the climate of the United States. Climate Dyn., 33, 945–958, doi:10.1007/s00382-009-0566-z.
Douglas, E., Beltrán-Przekurat A. , Niyogi D. , Pielke R. Sr., and Vörösmarty C. , 2009: The impact of agricultural intensification and irrigation on land–atmosphere interactions and Indian monsoon precipitation—A mesoscale modeling perspective. Global Planet. Change, 67, 117–128, doi:10.1016/j.gloplacha.2008.12.007.
Evans, J. P., and Zaitchik B. F. , 2008: Modeling the large scale water balance impact of different irrigation systems. Water Resour. Res., 44, W08448, doi:10.1029/2007WR006671.
Famiglietti, J. S., and Coauthors, 2011: Satellites measure recent rates of groundwater depletion in California’s Central Valley. Geophys. Res. Lett., 38, L03403, doi:10.1029/2010GL046442.
Fan, Y., Miguez-Macho G. , Weaver G. P. , Walko R. , and Robock A. , 2007: Incorporating water table dynamics in climate modeling: 1. Water table observations and equilibrium water table simulations. J. Geophys. Res., 112, D10125, doi:10.1029/2006JD008111.
Ferguson, I., and Maxwell R. M. , 2011: Hydrologic and land–energy feedbacks of agricultural water management practices. Environ. Res. Lett., 6, 014006, doi:10.1088/1748-9326/6/1/014006.
Ferguson, I., and Maxwell R. M. , 2012: Human impacts on terrestrial hydrology: Climate change versus pumping and irrigation. Environ. Res. Lett., 7, 044022, doi:10.1088/1748-9326/7/4/044022.
Gao, H., Tang Q. , Ferguson C. R. , Wood E. F. , and Lettenmaier D. P. , 2010: Estimating the water budget of major US river basins via remote sensing. Int. J. Remote Sens., 31, 3955–3978, doi:10.1080/01431161.2010.483488.
Gent, P., Yeager S. , Neale R. , Levis S. , and Bailey D. , 2010: Improvements in a half degree atmosphere/land version of the CCSM. Climate Dyn., 34, 819–833, doi:10.1007/s00382-009-0614-8.
Gordon, L. J., Steffen W. , Jonsson B. F. , Folke C. , Falkenmark M. , and Johannessen A. , 2005: Human modification of global water vapor flows from the land surface. Proc. Natl. Acad. Sci. USA, 102, 7612–7617, doi:10.1073/pnas.0500208102.
Haddeland, I., Lettenmaier D. P. , and Skaugen T. , 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.
Harding, K. J., and Snyder P. K. , 2012: Modeling the atmospheric response to irrigation in the Great Plains. Part I: General impacts on precipitation and the energy budget. J. Hydrometeor., 13, 1667–1686, doi:10.1175/JHM-D-11-098.1.
Kawase, H., Yoshikane T. , Hara M. , Kimura F. , Sato T. , and Ohsawa S. , 2008: Impact of extensive irrigation on the formation of cumulus clouds. Geophys. Res. Lett., 35, L01806, doi:10.1029/2007GL032435.
Ke, Y., Leung R. L. , Huang M. , Coleman A. M. , Li H. , and Wigmosta M. S. , 2012: Developing high resolution land surface parameters for Community Land Model. Geosci. Model Dev., 5, 1435–1481, doi:10.5194/gmdd-5-1435-2012.
Koster, R. D., and Coauthors, 2004: Regions of strong coupling between soil moisture and precipitation. Science, 305, 1138–1140, doi:10.1126/science.1100217.
Kueppers, L. M., and Snyder M. A. , 2011: Influence of irrigated agriculture on diurnal surface energy and water fluxes, surface climate, and atmospheric circulation in California. Climate Dyn., 38, 1017–1029, doi:10.1007/s00382-011-1123-0.
Kueppers, L. M., Snyder M. A. , and Sloan L. C. , 2007: Irrigation cooling effect: Regional climate forcing by land-use change. Geophys. Res. Lett., 34, L03703, doi:10.1029/2006GL028679.
Kustu, M. D., Fan Y. , and Robock A. , 2010: Large-scale water cycle perturbation due to irrigation pumping in the US High Plains: A synthesis of observed streamflow changes. J. Hydrol.,390, 222–244, doi:10.1016/j.jhydrol.2010.06.045.
Landerer, F. W., and Swenson S. C. , 2012: Accuracy of scaled GRACE terrestrial water storage estimates. Water Resour. Res., 48, W04531, doi:10.1029/2011WR011453.
Lawrence, D., and Coauthors, 2011a: Parameterization improvements and functional and structural advances in version 4 of the Community Land Model. J. Adv. Model. Earth Syst., 3, M03001, doi:10.1029/2011MS000045.
Lawrence, D., Oleson K. W. , Flanner M. G. , Fletcher C. G. , Lawrence P. J. , Levis S. , Swenson S. C. , and Bonan G. B. , 2011b: The CCSM4 Land Simulation, 1850–2005: Assessment of surface climate and new capabilities. J. Climate,25, 2240–2260, doi:10.1175/JCLI-D-11-00103.1.
Lee, E., Chase T. N. , Rajagopalan B. , Barry R. G. , Biggs T. W. , and Lawrence P. J. , 2009: Effects of irrigation and vegetation activity on early Indian summer monsoon variability. Int. J. Climatol.,29, 573–581, doi:10.1002/joc.1721.
Leng, G., Huang M. , Tang Q. , Sacks W. J. , Lei H. , and Leung L. R. , 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.
Leung, L. R., Huang M. , Qian Y. , and Liang X. , 2011: Climate-soil-vegetation control on groundwater table dynamics and its feedbacks in a climate model. Climate Dyn., 36, 57–81, doi:10.1007/s00382-010-0746-x.
Li, H., Wigmosta M. S. , Wu H. , Huang M. , Ke Y. , Coleman A. M. , and Leung L. R. , 2013: A physically based runoff routing model for land surface and Earth system models. J. Hydrometeor., 14, 808–828, doi:10.1175/JHM-D-12-015.1.
Liang, X., Xie Z. , and Huang M. , 2003: A new parameterization for surface and groundwater interactions and its impact on water budgets with the variable infiltration capacity (VIC) land surface model. J. Geophys. Res.,108, 8613, doi:10.1029/2002JD003090.
Lo, M.-H., and Famiglietti J. S. , 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.
Lobell, D., Bonfils C. J. , Kueppers L. M. , and Snyder M. A. , 2008: Irrigation cooling effect on temperature and heat index extremes. Geophys. Res. Lett., 35, L09705, doi:10.1029/2008GL034145.
Lobell, D., Bala G. , Mirin A. , Phillips T. , Maxwell R. , and Rotman D. , 2009: Regional differences in the influence of irrigation on climate. J. Climate,22, 2248–2255, doi:10.1175/2008JCLI2703.1.
McGuire, V. L., 2013: Water-level and storage changes in the high plains aquifer, predevelopment to 2011 and 2009–11. U.S. Geological Survey Scientific Investigations Rep. 2012–5291, 15 pp. [Available online at http://pubs.usgs.gov/sir/2012/5291/.]
Niu, G. Y., Yang Z. L. , Dickinson R. E. , and Gulden L. E. , 2005: A simple TOPMODEL-based runoff parameterization (SIMTOP) for use in global climate models. J. Geophys. Res., 110, D21106, doi:10.1029/2005JD006111.
Niu, G. Y., Yang Z. L. , Dickinson R. E. , Gulden L. E. , and Su H. , 2007: Development of a simple groundwater model for use in climate models and evaluation with Gravity Recovery and Climate Experiment data. J. Geophys. Res., 112, D07103, doi:10.1029/2006JD007522.
Oleson, K. W., and Coauthors, 2010: Technical description of version 4.0 of the Community Land Model (CLM). NCAR Tech. Note NCAR/TN-478+STR, 257 pp., doi:10.5065/D6FB50WZ.
Oleson, K. W., and Coauthors, 2013: Technical description of version 4.5 of the Community Land Model (CLM). NCAR Tech. Note NCAR/TN-503+STR, 420 pp., doi:10.5065/D6RR1W7M.
Ozdogan, M., and Gutman G. , 2008: A new methodology to map irrigated areas using multi-temporal MODIS and ancillary data: An application example in the continental US. Remote Sens. Environ., 112, 3520–3537, doi:10.1016/j.rse.2008.04.010.
Ozdogan, M., Rodell M. , Beaudoing H. K. , and Toll D. L. , 2010: Simulating the effects of irrigation over the United States in a land surface model based on satellite-derived agricultural data. J. Hydrometeor.,11, 171–184, doi:10.1175/2009JHM1116.1.
Pielke, R., Sr., Adegoke J. , Beltran-Przekurat A. , Hiemstra C. A. , Lin J. , Nair U. S. , Niyogi D. , and Nobis T. E. , 2007: An overview of regional land-use and land-cover impacts on rainfall. Tellus, 59B, 587–601, doi:10.1111/j.1600-0889.2007.00251.x.
Pokhrel, Y. N., Hanasaki N. , Yeh P. J.-F. , Yamada T. J. , Kanae S. , and Oki T. , 2012: Model estimates of sea-level change due to anthropogenic impacts on terrestrial water storage. Nat. Geosci., 5, 389–392, doi:10.1038/ngeo1476.
Puma, M. J., and Cook B. I. , 2010: Effects of irrigation on global climate during the 20th century. J. Geophys. Res., 115, D16120, doi:10.1029/2010JD014122.
Qian, Y., Huang M. , Yang B. , and Berg L. K. , 2013: A modeling study of irrigation effects on surface fluxes and land–air–cloud interactions in the southern Great Plains. J. Hydrometeor.,14, 700–721, doi:10.1175/JHM-D-12-0134.1.
Sacks, W. J., Cook B. I. , Buenning N. , Levis S. , and Helkowski J. H. , 2009: Effects of global irrigation on the near-surface climate. Climate Dyn., 33, 159–175, doi:10.1007/s00382-008-0445-z.
Scanlon, B. R., Fauntb C. C. , Longuevergne L. , Reedy R. C. , Alley W. M. , McGuire V. L. , and McMahon P. B. , 2012: Groundwater depletion and sustainability of irrigation in the US High Plains and Central Valley. Proc. Natl. Acad. Sci. USA, 109, 9320–9325, doi:10.1073/pnas.1200311109.
Segal, M., Pan Z. , Turner R. , and Takle E. , 1998: On the potential impact of irrigated areas in North America on summer rainfall caused by large-scale systems. J. Appl. Meteor.,37, 325–331, doi:10.1175/1520-0450-37.3.325.
Shibuo, Y., Jarsjö J. , and Destouni G. , 2007: Hydrological responses to climate change and irrigation in the Aral Sea drainage basin. Geophys. Res. Lett., 34, L21406, doi:10.1029/2007GL031465.
Siebert, S., Döll P. , Hoogeveen J. , Faures J. M. , Frenken K. , and Feick S. , 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.
Sophocleous, M., 2012: Conserving and extending the useful life of the largest aquifer in North America: The future of the High Plains/Ogallala Aquifer. Ground Water, 50, 831–839, doi:10.1111/j.1745-6584.2012.00965.x.
Sorooshian, S., Li J. , Hsu K. , and Gao X. , 2012: Influence of irrigation schemes used in regional climate models on evapotranspiration estimation: Results and comparative studies from California’s Central Valley agricultural regions. J. Geophys. Res., 117, D06107, doi:10.1029/2011JD016978.
Tang, Q., Oki T. , Kanae S. , and Hu H. , 2007: The influence of precipitation variability and partial irrigation within grid cells on a hydrological simulation. J. Hydrometeor.,8, 499–512, doi:10.1175/JHM589.1.
Tang, Q., Peterson S. , Cuenca R. H. , Hagimoto Y. , and Lettenmaier D. P. , 2009a: Satellite-based near-real-time estimation of irrigated crop water consumption. J. Geophys. Res., 114, D05114, doi:10.1029/2008JD010854.
Tang, Q., Rosenberg E. , and Lettenmaier D. , 2009b: Use of satellite data to assess the impacts of irrigation withdrawals on Upper Klamath Lake, Oregon. Hydrol. Earth Syst. Sci., 13, 617–627, doi:10.5194/hess-13-617-2009.
Voisin, N., Li H. , Ward D. , Huang M. , Wigmosta M. , and Leung L. R. , 2013a: On an improved sub-regional water resources management representation for integration into earth system models. Hydrol. Earth Syst. Sci., 17, 3605–3622, doi:10.5194/hess-17-3605-2013.
Voisin, N., Liu L. , Hejazi M. , Tesfa T. , Li H. , Huang M. , Liu Y. , and Leung R. L. , 2013b: One-way coupling of an integrated assessment model and a water resources model: Evaluation and implications of future changes over the US Midwest. Hydrol. Earth Syst. Sci., 17, 4555–4575, doi:10.5194/hess-17-4555-2013.
Xia, Y., Ek M. , Wei H. , and Meng J. , 2012: Comparative analysis of relationships between NLDAS-2 forcings and model outputs. Hydrol. Processes, 26, 467–474, doi:10.1002/hyp.8240.
Zou, J., Xie Z. , Yu Y. , Zhan C. , and Sun Q. , 2013: Climatic responses to anthropogenic groundwater exploitation: A case study of the Haihe River basin, Northern China. Climate Dyn.,doi:10.1007/s00382-013-1995-2.