A four-module integrated RIWP model was developed in order to simulate the complex system, including water supply from irrigation open canals, field crop water consumption, groundwater drainage into open ditches, and groundwater lateral flow.

The RIWP model developed in this study couples an irrigation system module, a drainage system module, a groundwater module, and a field-scale IWP evaluation module (Fig. 1). The irrigation system module simulates the water flow along canals and the canal seepage to groundwater (the recharge of the groundwater module), and it provides the amount of water available for field-scale irrigation. The drainage system module simulates the drainage to the main drainage ditches from groundwater, and this is the discharge of the groundwater module. The groundwater module is used to simulate the groundwater lateral movement, the groundwater boundary for the field-scale water–salt balance processes, and the groundwater level dynamics for the drainage module. In the field-scale IWP module, the vertical movement of water and salt in the soil profile is simulated in order to obtain the soil moisture and salinity of the crop root zone and to calculate the field-scale irrigation water productivity. This module provides deep percolation to the groundwater module and obtains capillary rise to soil from the groundwater module. The four abovementioned modules will be described comprehensively in Sect. 2.1.2–2.1.5.

The irrigation area is spatially heterogeneous in terms of soil, land use, meteorology, and groundwater. To include the spatial heterogeneities in the simulation of regional water and salt dynamics and its impact on crop growth, the irrigation district was divided into hydrological response units (HRUs; Kalcic et al., 2015). The HRU is an abstract artifact created by a hydrological developer and is like the smallest spatial unit of the model, which provides an efficient way to discretize large watersheds where simulation at the field scale may not be computationally feasible. In each HRU, soil texture and groundwater conditions are assumed to be homogeneous, but different cropping patterns can exist, e.g., sunflower fields, wheat fields, maize fields, and uncultivated lands. As the irrigation quota is different for different cropping patterns, the model first runs the field IWP model for each cropping pattern independently in each HRU in order to obtain the soil water and salt dynamics, the IWP, and groundwater recharge. The groundwater levels and salinity of each HRU can then be updated according to the area proportions of different cropping patterns in each HRU. The groundwater flow is determined by the pressure head gradient between adjacent HRUs.

The upper boundary of the model is the atmospheric boundary layer above the plant canopy, which determines reference ET and precipitation. The main irrigation canals and drainage ditches directly connect with groundwater and can be considered as the side boundaries in the model. With the canal conveyance water loss deducted from the gross water supplied, the amount of water diverted into the field can be calculated as the actual amount of irrigation. The local irrigation schedules of different crops and the actual time of canal water supply are both considered to determine the actual irrigation time and irrigation amounts. The lower boundary is the confining bed at the bottom of phreatic layer. The phreatic layer is vitally important due to its vertical exchange with the unsaturated soil zone in each HRU and its lateral exchange with adjacent HRUs to bond the whole region together.

When irrigation water passes through canals, regardless of if they are lined or unlined, seepage loss occurs and recharges groundwater. In a large irrigation area, there are many main, submain, lateral, and field canals, which are categorized as the first-, second-, third-, and fourth-order canals, respectively. During the water allocation period, canal seepage loss from different levels of canals can be divided into two parts: the seepage loss from the main and submain canals, which are permanently filled with water and recharge directly into groundwater along the route; and the seepage loss from lateral and field canals, which are intermittently filled with water and only recharge the groundwater units within their control area. Each HRU has its corresponding groundwater unit, which is used when calculating lateral exchange of groundwater between adjacent HRUs.

We calculated the decreasing water flow along the canal and water losses in the main and submain canals as follows (Men, 2000): 1σ=A100Qm2σ=dQQdl, where σ represents the water loss coefficient per unit length per unit flow in the canal (m-1); A is the soil permeability coefficient of the canal bed (m3m-1 d-m), m is the soil permeability exponent of the canal bed (–), and their values depend on the soil type of the canal bed (please refer to Guo, 1997, for the values). Q represents the daily net flow in the canal (m3 d-1), and dQ represents the daily flow loss of the water conveyance within dl distance in the canal (m3 d-1).

Thus, Eq. (1) is equal to Eq. (2), and they can be transformed into 3Qm-1dQ=Adl. Integrations of both sides of Eq. (3) give 4∫QLQgQm-1dQ=∫0LAdl5QL=Qgm-ALm1/m, where Qg is the daily gross flow in the head of the canal (m3 d-1), and QL is the daily net flow in the canal at L distance from the canal head (m3 d-1). Thus, flow loss in water conveyance process can be calculated as follows: 6QLs=A100Qgm-ALm(1-m)/m7Wls=Qls/n1×Asu, where QLs is the daily groundwater recharge due to water conveyance loss in the main and submain canals (m3 d-1), Wls is the daily groundwater recharge per unit area due to water conveyance loss in the main and submain canals (m d-1), n represents the total number of HRUs along selected main and submain canals (–), and AHRU is the area of each HRU (m2).

Lateral and field canals are densely distributed in the irrigated area, and they are intermittently filled with low water flow. Thus, it is assumed that seepage from these canals uniformly recharges groundwater units within their control area. The canal seepage is estimated by the following empirical formula: 8Was=In⋅ηmc⋅1-ηsbmc+In⋅ηmc⋅ηsbmc⋅1-ηlc+In⋅ηmc⋅ηsbmc⋅ηlc⋅1-ηfc, where Was represents daily groundwater recharge per unit area due to water conveyance loss in the lateral and field canals (m d-1), and In is the daily irrigation water depth applied per unit area (m d-1). ηmc, ηsbmc, ηlc, and ηfc are the utilization coefficients of the main, submain, lateral, and field canals, respectively (–).

In the drainage system module, only the groundwater draining into ditches is considered, as the precipitation directly on ditches is negligible in arid and semiarid areas. The drainage processes are simulated based on the spatial distributions of the main, submain, and lateral ditches, which are grouped into the first-, second-, and third-order ditches, respectively. Drainage is estimated by comparing local groundwater levels and ditch bottom elevation. According to Tang et al. (2007), the groundwater drainage was calculated as follows: 9Dg=γd×hdb-hg;hdb>hg0;hdb<hg, where Dg is the daily groundwater drainage per unit area (m d-1). γd is the drainage coefficient (–), which describes the groundwater table decline caused by the elevation difference between the groundwater table and the streambed of the drainage ditch, and it depends on the underlying soil conductivity and the average distance between the drainage ditches. hg represents the daily groundwater table depth (m d-1), and hdb is the daily streambed depth of the drainage ditch (m d-1).

For a plain irrigation area, groundwater levels are usually relatively flat on the large scale. In our model, it is assumed that groundwater lateral flow exists between one HRU and its four adjacent HRUs (Fig. 2). Using the water table gradient, groundwater flow between the current HRU and its adjacent HRUs can be calculated as follows: 10Wgr=K×h×BLga-LgD/B2, where Wgr is the daily groundwater inflow of the current HRU from adjacent HRUs (m d-1), and K is the daily permeability coefficient of unconfined aquifers in the current HRU (m d-1). h represents the thickness of the unconfined aquifers, which is the difference between the water table and the upper confined bed and varies with water table changes (m). B is the length of the groundwater unit (m) – here the value is 1 km. Lga and Lg represent the water table level of adjacent HRUs and the current HRU, respectively (m). D is the distance between the center of the current HRU and the centers of its adjacent HRUs (m). There are three types of groundwater boundary conditions: river head (when the boundary HRU, including the irrigation canal and the daily river flux, equals to the daily canal flux), river flux (when the boundary HRU, including the drainage ditches and the water heads in ditches, is assumed constant and equal to the river head), and constant flux (when the boundary HRU is mainly barren area and no irrigation is applied); thus, in our study, zero flux is assumed.

Based on the field-scale simulation, groundwater lateral exchange, canal seepage, and groundwater drainage are added in the daily water and salt balance calculations of each groundwater unit at the regional scale: 11hgi=hgi-1-1/SyPwgi-1-Gwgi-1-exti-1+Wgrupi-1+Wgrdowni-1+Wgrlefti-1+Wgrrighti-1+Wlsi-1+Wasi-1-Dgi-112SCai=Za×Sai-1+Wgrupi-1×Saupi-1+Wgrdowni-1×Sadowni-1+Wgrlefti-1×Salefti-1+Wgrrighti-1×Sarighti-1+Wlsi-1+Wasi-1×Isi-1-Dgi-1×Sai-1+Psgi-1-Gsgi-1, where Wgrup, Wgrdown, Wgrleft, and Wgrright represent the daily groundwater lateral runoff per unit area into the current groundwater unit from the up, down, left, and right adjacent groundwater units, respectively (m d-1). SCa is the daily soluble salt content in the saturated zone below the transmission soil profile (mg m-2 d-1). Za is the thickness of the saturated zone, which is the difference between the groundwater table depth and the depth that groundwater table fluctuations largely cannot reach (m). Za only affects the soluble salt concentration in the groundwater salt balance, while it has no effect on the water balance and groundwater fluctuation simulation. Sa, Saup, Sadown, Saleft, and Saright represent the salt concentration of the current groundwater unit and the up, down, left, and right adjacent groundwater units, respectively (mg m-3). “Is” represents the salt concentration of the irrigation water (mg m-3). Sy represents the specific yield (–), which is the ratio of the volume of water that can be drained by gravity to the total volume of the saturated soil/aquifer. “ext” is the daily groundwater extraction per unit area (m d-1). Pwg is the daily percolation water depth to groundwater from the potential root zone (m d-1), and Gwg is the daily water depth supplied to the potential root zone from shallow groundwater due to the rising capillary action (m d-1). Psg and Gsg are the quantity of soluble salt in Pwg and Gwg, respectively (mg m-2 d-1). The detailed calculations of the water and salt exchange components between unsaturated soil and groundwater, such as Pwg and Gwg, were described in our previous field-scale water productivity model (Xue et al., 2018).

Cropping patterns are complex for each HRU, and sometimes HRUs include uncultivated land, forest land, and other nonagricultural land. In our model, using a high-resolution land use map, different cropping patterns can be separated to simulate soil water and salt processes as well as the responses of ET and crop yields to the water and salt content of root zone. Here, we employed our previously developed field IWP model to simulate field water, salt, ET, and crop yield under shallow groundwater conditions (Xue et al., 2018). The soil profile is vertically divided into four soil zones: the current root zone, the potential root zone, the transmission zone, and the saturated zone. In each HRU, the soil water and salt balance processes and the water productivity are independently simulated for each cropping pattern under its corresponding groundwater unit condition. For uncultivated lands, only water and salt balance are simulated, and the IWP is zero. The water and salt exchange between unsaturated soil and groundwater of different cropping patterns are then weight-averaged by area proportion. Finally, the weighted averages are used to update the daily groundwater table and salinity (Fig. 3).

As a typical subdistrict of the Hetao Irrigation District, the Jiefangzha Irrigation District (JFID) is a typical arid irrigated area with shallow groundwater that has resulted from its arid continental climate, years of flood irrigation, and poor drainage systems (Fig. 5). Located on the Hetao Plain, the JFID is very flat with an average slope of 0.02 % from southeast to northwest (Xu, 2011). The mean annual precipitation is only 155 mm, 70 % of which occurs between July to September, and the mean annual potential evaporation is 1938 mm. The mean annual temperature is 7∘, with the respective lowest and highest monthly average being -10.1 and 23.8∘ in January and July. The JFID covers an area of 22×104 ha, 66 % of which is irrigated farmland area. Wheat, maize, and sunflower are the main crops in this region, comprising more than 90 % of the irrigated farmland area. The 12×108 m3 yr-1 of irrigation water is diverted from the Yellow River. Due to the poor maintenance of drainage ditches, it is quite common for poor drainage situations to exist in this area. Therefore, the annual average groundwater table depth ranges from 1.5 to 3.0 m during the crop growing season. Soils in the JFID are spatially heterogeneous and are primarily composed of silt loam in the northern region and sandy loam in the southern region. The shallow groundwater table and strong evaporation makes soil salinization a very serious problem in this area, and salinization is becoming the main constraint on crop production.

An irrigation and drainage network includes 4 main irrigation canals, 16 submain irrigation canals, 5 main drainage ditches, and 12 submain drainage ditches that control the water movement in the JFID (Fig. 5). The streambed depths of the regional main, submain, and lateral ditches were collected by a regional survey in 2016. Daily water flow data from the main and submain irrigation canals and monthly data from the five main drainage ditches were obtained from the local irrigation administration bureau. A total of 55 groundwater observation wells are installed in the JFID (Fig. 5). The groundwater level was measured on the 1st, 6th, 11th, 16th, 21st, and 26th of each month, and groundwater salinity was measured three times each month. Near the groundwater observation wells, soil moisture was measured four times, and soil electrical conductivity was measured once before wheat sowing and once before autumn irrigation. Due to the spatially homogeneous climate in the JFID, daily meteorological data (air temperature, humidity, wind speed, and precipitation) were obtained from the Hangjinghouqi weather station for the calculation of the regional reference ET.

HJ-1A, HJ-1B, and Landsat NDVI images from 2006 to 2013 with a 30 m resolution were downloaded from the official website of the CERSDA (2013) and the USGS (2013) in order to determine the annual cropping pattern distributions. Due to the lack of measured ET values, the ET values estimated by the SEBAL model using MODIS images from NASA (2013) were utilized as a reference for comparison with the simulated ET values (Bastiaanssen et al., 2003b).

The JFID was divided into 2485 1 km × 1 km HRUs (Fig. S1a in the Supplement). In terms of boundary conditions, the upper Quaternary aquifer layer was regarded as the phreatic layer in the model. It was modeled as an aquitard with loamy soil. From north to south, the thickness of the aquifer in the JFID varied from 2 to 20 m with an average of 7.4 m (Bai and Xu, 2008). Thus, the initial value of the average thickness of unconfined aquifer was set as 7.4 m. The water level contour maps of the JFID from 1997 to 2002 from Bai (2008) were used to determine the direction of water flow near the groundwater boundary. Based on the topography conditions, land use types, the locations of the main canals and ditches, and the directions of water flow, the regional phreatic layer was divided into five zones with river, drainage, and impervious boundary conditions (Fig. S1b).

The JFID was divided into four irrigation control sections and five drainage control sections, and each section was controlled by one main irrigation canal or one main drainage ditch. These sections were further divided into 48 irrigation control subareas and 17 drainage control subareas, and each subarea was controlled by one submain irrigation canal or one submain drainage ditch (Fig. S2). Sunflower fields, wheat fields, maize fields, and uncultivated lands were the four cropping patterns, i.e., land use types, in the RIWP model. When considering the irrigation schedule applied, many other studies on distributed hydrological models simply set the sowing and irrigation of a particular crop as occurring on the same day over the whole study area, which may be a simplification of actual conditions (Singh, 2005). In our study, the irrigation time and the amount of irrigation water for each HRU were co-determined by both the local irrigation schedule of the three main crops and the actual water amount flowing into the fields.

The simulation period was from 1 April to 20 September, which covers the growing seasons of all of the three main crops. The initial crop parameters were set as the default values suggested for sunflower, wheat, and maize by Allen et al. (1998). The empirical values of the regional canal utilization and the ditch drainage coefficient were obtained from the Jiefangzha administration.

To comprehensively evaluate the accuracy and reliability of the model, the data from 2010 to 2013 and from 2006 to 2009 were used as the calibration and validation datasets, respectively. The daily measured soil moisture content of the crop root zone (θ), the electrical conductivity of soil water (EC), and the groundwater table depth (hg) and groundwater salinity, were calibrated using measured data from the 22 soil water and salt observation sites and 55 groundwater observation sites (Fig. 5) mentioned in Sect.  2.3.1. The regional ET simulated by RIWP for each HRU was calibrated using the remote-sensing-based ET images obtained once every 8 d. The regional drainage processes was calibrated using the monthly groundwater drainage data from main ditches, in which the simulated drainage of each main ditch was the sum of the drainage of its controlling HRUs. Overall, the soil hydraulic parameters, the crop water productivity-related coefficient, and the canal conveyance and ditch drainage parameters were all calibrated using observed data from 2010 to 2013 and then validated with observed data from 2006 to 2009.

To quantify the model performance, the root-mean-square error (RMSE), the Nash and Sutcliffe model efficiency (NSE), and the coefficient of determination (R2) were used as the indicators. The RMSE was used to measure the deviation of simulated values from measured values, the NSE is commonly used to verify the credibility of the hydrological model, and R2 represented the degree of linear correlation. The indicators were calculated as follows: 13RMSE=∑i=1nOutputs-Outputo2n0.514NSE=1-∑i=1nOutputs-Outputo2∑i=1nOutputo-Outputm215R2=1-∑i=1nOutputo-Outputo‾Outputs-Outputs‾∑i=1nOutputo-Outputo‾2∑i=1nOutputs-Outputs‾2, where n is the number of simulations; Outputs and Outputo are the simulated and observed values of model outputs, respectively; and Outputs‾ and Outputo‾ are the average values of the simulated and observed model outputs, respectively. The RMSE indicates a perfect match between observation and simulation values when it equals 0, and increasing RMSE values indicate an increasingly poor match. Singh et al. (2005) stated that RMSE values less than 50 % of the standard deviation of the observed data could be considered low enough to be an indicator of a good model prediction. With a range between -∞ and 1, the NSE indicates a perfect match between observed and predicted values when it is equal to 1. Values between 0 and 1 are generally considered to be an acceptable level of performance, whereas values less than 0.0 indicate that the simulation is worse than taking an average of the observation values, which indicates unacceptable performance. The R2, which ranges between 0 and 1, describes the proportion of the variance in the observed data, and higher values indicate less error variance. Typically, an R2 value higher than 0.5 is considered acceptable (Santhi et al., 2001).

To find the key parameters significantly impacting the model output, a global sensitivity analysis was conducted. The analysis related the changes in three output variables – RIWP, the groundwater table depth, and the groundwater salinity – to eight parameters in the RIWP model. Latin hypercube sampling (LHS; please see Mckay et al., 1979; Muleta and Nicklow, 2005; Wang et al., 2008 for detailed descriptions of the sampling method), which is a typical sampling method for sensitivity and uncertainty analysis, was used to sample the parameter space. Following Dai (2011), the test number was set as 20, more than double the parameter number (which was 8), in order to ensure that the test points were evenly distributed in space and to guarantee the accuracy of the test. For uniform distributions, the parameter range was subdivided into 20 equal intervals. Each interval was sampled only once to generate random values of the possible parameter sets. The possible parameter value ranges referred to the local measurements, the survey data, and relevant research papers. Additionally, considering the spatial heterogeneity of the three output variables, 22 evenly distributed groundwater observation sites in the JFID were selected for the global sensitivity analysis. Based on the LHS method, 20 groups of parameter combinations were obtained, and the simulation was run 20 times. Finally, the sensitivity of the three output variables to the eight parameters were determined in SPSS Statistics. The absolute values of the standardized regression coefficients (SRCs) obtained quantified the significance of each parameter to each output variable (Table 1; Cannavó, 2012), and the plus or minus sign of the SRCs indicated the positive or negative correlations between the corresponding parameter and output variable pairs.

Validated by the measured soil moisture and salinity, the groundwater table depth and salinity, and the drainage water depth and ET, specifically the 2006–2013 time series of the groundwater table depth under the four cropping patterns, the RIWP model developed can be used to estimate the spatial distribution of IWP for the three main crops over the period from 2006 to 2013 (Fig. 9). Note that these IWP values were based on the simulated water balance and crop yields of individual HRUs, which may deviate to a certain extent from the real values. However, the values can still represent the utilization of water resources at the regional scale. We note that there are “red HRUs” in Fig. 9 that change with time and space due to the different irrigation water depths applied under different groundwater conditions. Even different crop species can result in a big difference in the IWP. As we mentioned previously, the spatial distribution of these three crops is very complex in the JFID, and the field plot is small; thus, we use remote sensing data to obtain the cropping pattern map with a 30 m × 30 m resolution. Every HRU has these three crops; therefore, we can simulate the IWP for each main crop in every HRU. The RIWP of the three main crops showed a declining trend during the period from 2006 to 2010 (Fig. 9a, b, c, d, e).This was mainly attributed to the increasing irrigation quota, as the excess water lowered the IWP. During the period from 2011 to 2013 (Fig. 9f, g, h), in contrast, the RIWP of the three main crops showed an increasing trend. This was because the irrigation quota was reduced over this period, and the contribution of groundwater compensated for the crop yield losses. When less irrigation water is applied, the number of red HRUs will increase.

Under a given irrigation water distribution, the spatial distribution of ET was the key factor controlling the RIWP distribution. The spatial distribution of ET was also fundamentally determined by the solar energy as well as the water and salt dynamics of soil. Recall that the climate and, therefore, the solar energy, was homogeneous in the JFID. The spatial heterogeneity of RIWP must then be attributed to the water and salt heterogeneity caused by the spatial heterogeneity of the cropping pattern, the groundwater table depth, and the irrigation and drainage networks. Particularly, when the farmlands had a limited supply of irrigation water, the groundwater table depth and salinity played an important role with respect to the IWP. Groundwater could drain both water and salt out of the field through the drainage ditches; thus, groundwater evapotranspiration decreased, and the soluble salt content involved in groundwater evapotranspiration also decreased as well. Despite the negative effect of draining water on the IWP, draining salt out of the field will positively affect the IWP. As we see from Fig. 9, the simulated IWP values for three crops were lower in the south, west, north, and northwest of the JFID than in the other regions. The south of the JFID is the main canal for water diversion, which provides a higher irrigation quota than other regions and results in a lower IWP. The west of the JFID, in contrast, is mainly uncultivated area; thus, the IWP is lower than in the other regions. In the northwest of the JFID, the main drainage ditch received the drainage water with a high saline content from the four submain ditches and drained all the way to the north of the region. Ditch seepage water with high salinity resulted in severe soil salinization in the north and northwest of the JFID, which will restrict the crop growth and lower the IWP. Thus, appropriate groundwater drainage management and dealing with salt accumulation at the end of main drainage ditches in irrigated areas is also a pressing and unsolved problem with respect to increasing the red HRUs, which needs to be solved by irrigation managers.

As the JFID is the major food-producing region of China, improving water productivity means greater food crop yields with less water usage, based on local or regional potential. With declining access to water resources, farmers will need to grow different crops to maintain or increase crop production profitability in the future. A comparison of the RIWP values of different crops (comparison of the three columns in Fig. 9) showed that maize had the highest IWP, wheat had the lowest IWP, and the IWP of sunflower was in between. Therefore, modestly increasing the planting area of maize will improve the crop production per unit of irrigation water. In addition, the RIWP of sunflower is a little higher than that of wheat, and the salt tolerance of sunflower is much higher than that of wheat. Thus, planting sunflowers should be promoted in the JFID if the amount of irrigation water available declines in the future; this practice will definitely increase the red HRUs.

In arid shallow groundwater areas, the irrigation water productivity (IWP) is affected by the irrigation water depth (IWD) applied and the groundwater table depth (hg). In all of the four simulated hg ranges, the IWP decreased when the IWD increased (Fig. 10a), which are findings consistent with Huang et al. (2005). Moreover, the magnitude of the IWP decrease per unit increase of IWD was different for different hg ranges. The magnitude of the IWP decrease for a shallower hg was smaller than that for a deeper hg. This effect of increasing hg on the relationship between IWP and IWD was consistent with Gao et al. (2017). The above results indicate that groundwater can help meet the crop water demand when irrigation water is insufficient. However, when irrigation water is excessive, a large proportion will eventually drain through the drainage ditches, and the IWP will decrease. Additionally, among the four hg ranges, the highest IWP was obtained between 2 and 3 m (Fig. 10b), which was consistent with Xue et al. (2018). This indicates that a hg deeper than the abovementioned depth provides insufficient water for crop growth, whereas a hg shallower than 2–3 m will increase the root zone soil salinity and the salt stress of crops. The negative effect of shallow groundwater salinity can also be seen in Fig. 10a when hg is less than 2 m; this indicates that when the amount of irrigation applied decreases from 300 < IWD< 400 mm to 200 < IWD < 300 mm, it leads to decreases in the IWP, which is caused by the faster reduction of ET than the irrigation applied. The shallow groundwater contribution will make up for the ET reduction when a smaller amount of irrigation water is applied; thus, another reason exists to accelerate the reduction of ET. We deduced that less irrigation water will weaken the role of irrigation on salt leaching and result in more severe salinization in the crop root zone. The negative effect of salt stress on the crop water use is greater than the positive effect of the shallow groundwater contribution on the crop water use in this situation. Thus, keeping the groundwater table depth within the optimal range and sustainable is of great importance to reach a higher crop IWP at the regional scale, and irrigation managers may need to reasonably determine the irrigation quota and constantly maintain the drainage system. Groundwater sustainability includes spacing withdrawals to avoid excessive depletion and taking measures to safeguard or improve groundwater quality. To achieve this, regional irrigation managers may need to undertake monitoring efforts to establish historic and current conditions, carry out research to model groundwater systems, forecast future variation, and create policy to manage activities influencing the groundwater table and groundwater quality.