Predicting the impact of heavy groundwater pumping on groundwater and ecological environment in the Subei Lake basin, Ordos energy base, Northwestern China

Most of the world's large energy bases are located in arid and semi-arid areas where the ecological environment is very fragile, water shortages constrain local social-economic development and the construction of energy bases (Kahrl & Roland-Holst 2008; Siddiqi & Anadon 2011). Groundwater is essential for sustainable energy production in arid and semi-arid regions, but groundwater exploitation in arid areas will inevitably disrupt the natural equilibrium of groundwater and the ecological environment (Feng et al. 2011; Liu et al. 2016). With the development of energy bases, a large number of groundwater well fields have been built and put into use to meet increasing water demand. Groundwater overexploitation has changed dramatically the natural water cycle, causing tremendous changes in hydrochemical and hydrodynamic fields (Nakamura et al. 2017). Consequently, the overexploitation of groundwater has broken the balance between hydrology and ecosystems, causing a series of hydro-ecological problems such as lake atrophy, desertification, vegetation degradation, etc. (Zhao et al. 2005; Cai et al. 2015; Liu et al. 2017). Such adverse ecological impacts are extremely inconsistent with the development of energy bases. Thus, in order to coordinate the relationship among energy development, groundwater exploitation and the ecological environment, this research has very important scientific significance and practical needs.

The Subei Lake basin is a good example of such areas where the groundwater system and ecological environment are impacted by over-pumping of groundwater for industrial purposes. It is representative of more than 400 lake basins with diverse sizes distributed in the Ordos artesian basin that contains the second largest coal reserves in China (Dai et al. 2006). The Ordos energy base is a large-scale regional industrial zone integrated with coal mining, electrical power generation, and coal-based chemical industry. In energy bases, the industrial sector is the major user of water resources. In order to meet the water demand for energy production, several waterworks have been built in some lake basins, including Haolebaoji waterworks built in the Subei Lake basin (Hou et al. 2006). However, the aquifer systems in these inland lake basins are currently subject to increasing pressure from altered hydrodynamic conditions associated with groundwater abstraction due to lack of a reasonable groundwater management strategy. The geology and hydrogeology in Ordos energy base have been investigated by the China Geological Survey Bureau since the 1980s (Zhang et al. 1986; Hou et al. 2008). In previous studies, the impact of the Energy Base Water Project on the groundwater has been identified by utilizing high-frequency groundwater level data in this ecologically sensitive area (Liu et al. 2017). In addition, the impact of natural and anthropogenic factors on the geochemical evolution of groundwater in the Subei Lake basin has also been identified by using the techniques of hydrochemistry and stable isotopes (Liu et al. 2015a, 2015b). The above research on the study area laid a solid foundation for the development of the hydrogeological conceptual model and groundwater flow model in the present study. Therefore, in order to fully quantify the hydrological and anthropogenic impacts on the groundwater regime, a high-precision groundwater flow model will be required to determine the exchange capacity of groundwater between the different source/sink terms, which can provide valuable information for local authority and researchers to make sustainable groundwater management in these lake basins.

Groundwater models play an important role in the development and management of groundwater resources and in predicting the effects of management measures. With rapid increases in computation power and the wide availability of computers and model software, groundwater modeling has become a standard tool for professional hydrogeologists to effectively perform most tasks (Tamma Rao et al. 2011). Numerical modeling provides an efficient method of clarifying the flow path and groundwater regime, as well as quantifying the amounts recharged from each source (Liu et al. 2014; Moiwo & Tao 2014). Numerical models have been used successfully for simulating and predicting groundwater levels for many years (Wen et al. 2007; Xi et al. 2009; May 2014). More importantly, groundwater flow modeling has become an invaluable tool for proper management of groundwater systems (Sedki & Ouazar 2011; Lachaal et al. 2012; Manghi et al. 2012). In the practice of groundwater management, the technique of numerical simulation was used to assess the impact of existing and future activities on groundwater resources (Huang et al. 2008; Banks et al. 2011; Singh 2013; Sefelnasr et al. 2014), especially for simulating the impacts of groundwater pumping on groundwater–surface water interaction (Quinodoz et al. 2017) and vegetation change (Zhao et al. 2005; Abdalla 2009; May 2014). Thus, modeling groundwater levels' variation and the impacts of groundwater pumping on the ecological environment under various possible water resources management scenarios can help to determine measures for sustainable groundwater management.

The present study aims to: (1) develop a three-dimensional groundwater flow model for analyzing groundwater budgets and scenario simulations; (2) predict the impacts of two different water resources management scenarios on groundwater levels, the Subei Lake and vegetation. The findings of this research can contribute to the heated discussion about the coordinated development between water and energy production in arid regions, and it will advance necessary knowledge for the sustainable groundwater management in such arid areas.

The Subei Lake basin, covering an area of 400 km², extends between longitudes of 108°51′24″–109°08′40″E and latitudes of 39°13′30″–39°25′40″N and is located in the northern part of the Ordos artesian basin, Northwestern China (Figure 1). The Subei Lake basin has a continental semi-arid to arid monsoon climate, with average annual precipitation and evaporation values of 328.4 and 2,284.2 mm over the last 30 years, respectively. The topography of the west, east, and north sides in the study area is relatively higher with elevations between 1,370 and 1,415 m above sea level, while its south side is slightly lower with altitudes between 1,290 and 1,300 m above sea level. The main water bodies are Subei Lake and Kuisheng Lake (Figure 1). Subei Lake, covering an area of about 6 km², is located in the low-lying center of the study area, which is an inland alkaline lake; while Kuisheng Lake is also a perennial inland lake and is located in the northeastern corner of the study area, and covers 2 km². The two inland lakes are mainly replenished by groundwater (Hou et al. 2006; Wang et al. 2010).

In the area, the industrial sector is the major user of water resources, accounting for 69% of the total water consumption in 2009 according to the unpublished hydrogeological report from the Inner Mongolia Second Hydrogeology Engineering Geological Prospecting Institute. Haolebaoji waterworks has 22 production wells stretching northeast to southwest, which became operational for industrial purposes in 2006. The production wells were designed in the shape of a rectangular grid at an interval of 1.5 km to each other. According to the data obtained by this study, the total production rate climbed from 12,932 m³/d in 2006 to 24,746 m³/d in 2012 with an average annual growth rate of 1,969 m³/d; all the production wells are screened within the confined aquifer.

Two distinct geological units (i.e., Quaternary unconsolidated sediments and Cretaceous strata) can be observed in the study area. The Quaternary unconsolidated sediments are mainly distributed around Subei Lake with the thickness varying from 0 to 20 m. The Quaternary layer is chiefly composed of the interlaced layers of sand and mud. The Cretaceous strata mainly consist of sedimentary sandstones and generally outcrop in the areas with relatively higher elevation. The maximum thickness of Cretaceous rocks could be nearly 1,000 m in the Ordos Plateau (Yin et al. 2009).

The Subei Lake basin is a relatively closed hydrogeological unit given that a small quantity of lateral outflow occurs in a small part of the southern boundary (Wang et al. 2010). A phreatic aquifer and confined aquifer can be observed in the study area. According to Wang et al. (2010), the unconfined aquifer is composed of the Quaternary sediments and the upper part of the Cretaceous strata, with the thickness ranging from 10.52 to 63.54 m. The unconfined groundwater is mainly recharged by precipitation infiltration via vadose zone, and it can be also replenished by lateral inflow from groundwater outside the study area. In a natural state, irrigation return flow and upward leakage from the underlying confined aquifer can also contribute a small proportion to groundwater recharge. Evaporation is the major flux from the unconfined groundwater. Moreover, the unconfined groundwater is also discharged by lateral outflow, artificial exploitation, and leakage discharge. Hydraulic head measurements conducted in the groundwater wells (in September 2003) were contoured to illustrate the general flow field in the area (Figure 2). As shown in Figure 2, by analyzing the contours and flow direction of groundwater, lateral outflow occurs in a small part of the southern boundary. The groundwater flows predominantly from surrounding uplands to low lands, and is controlled by topography. Overall, groundwater in the phreatic aquifer flows toward the Subei Lake.

The unconfined and confined aquifers are separated by an uncontinuous aquitard composed of a mudstone layer, and discontinued mudstone lens can also be observed in Cretaceous strata (Figure 3). The lower part of Cretaceous strata can be viewed as confined or semi-confined in nature due to the presence of mudstone lens. The hydraulic properties of the confined aquifer are variable in space. The flow direction of confined groundwater was similar to that of unconfined groundwater (Figure 2). Given the huge thickness and high permeability of the confined aquifer, it is viewed as the most promising water-supply aquifer for domestic and industrial usages.

where Ω is the vadose zone; h = h(x,y,z) is the groundwater level (m); h_r is the water level of lakes (m); K_x, K_y, K_z is the hydraulic conductivity of the aquifer in x, y, z direction, respectively (m/d); K_n is the hydraulic conductivity in the direction of the outer normal line of the boundary face (m/d); S_s is the specific storage (1/m); μ is the specific yield in phreatic aquifer; σ is the resistance coefficient of aquitard in the lake bottom (1/d); ɛ is the source/sinks terms of aquifer (1/d); p is the intensities of evapotranspiration and rainfall recharge per unit area (m/d); h₀ is the initial head distribution (m), h₀ = h₀(x,y,z); Г₀ is the upper boundary of the vadose zone, i.e., groundwater free surface; Г₁ and Г₂ are the boundaries of the second type and the third type; n is the direction of the outer normal line of the boundary face; q(x,y,z,t) is the flow rate per unit area of the second type boundary conditions with inflow being positive and outflow being negative (m³/d·m²).

The Groundwater Modeling System (GMS) software, which includes the MODFLOW package for simulating flow, is not only a versatile and effective model for simulating groundwater flow but is also used by researchers worldwide (Markner-Jäger 2008). MODFLOW is a modular three-dimensional finite difference groundwater flow model. This deterministic numerical model based on Darcy's law was developed by McDonald & Harbaugh (1988) and Harbaugh et al. (2000). This model can simulate transient/steady groundwater flow in complex hydraulic conditions with various natural hydrological processes and anthropogenic activities, which can be readily incorporated into future studies for optimal groundwater management. Hydrogeological layers can be simulated as confined, unconfined, or a combination of the two. Boundary conditions include specified head, specified flux, and head-dependent flux. Thus, GMS software was selected in the present study, and solution of the governing equations was achieved by using GMS software.

The numerical model covers an area of about 400 km². The study area was discretized into a horizontal grid of 235 rows and 265 columns, with a cell size of 100 m × 100 m. For each cell, there is a single point, called a node, at which head is calculated (Mcdonald & Harbaugh 1988). Vertically, the model consists of two layers including one phreatic aquifer and one confined aquifer.

Based on the groundwater level data availability, the model simulates transient flow from 1st September 2003 to 31st August 2004 with a monthly time step. All hydrological stresses can be assumed to be constant within each time step. During the modeling period, the total volume of groundwater pumping in the study area was only 1.25 × 10⁴ m³/d which was used for agriculture and domestic use. Therefore, anthropogenic activities exerted less impact on the groundwater system and groundwater flowed in a natural state during the modeling period. For reasons of data availability, the model calibration has a start time of 1st July 2009 with the end time set to 1st December 2009 with a daily time step. During the calibration period, a large amount of groundwater monitoring data from in situ pumping tests were obtained, and can be used to verify the rationality of boundary conditions and hydrogeological parameters.

Groundwater flow in the aquifer is controlled by the boundary conditions of the regional system. According to the previous hydrogeological investigations, boundary conditions are shown in Figure 4. The groundwater flow boundary conditions were determined in terms of the natural topography and groundwater divide. No-flow boundaries were defined between points B–C, D–E, F–G, and H–A in the basin, because the contours of groundwater levels are perpendicular to these boundaries and thus they form the groundwater divide. Constant influx cells were defined between points A–B, C–D, and E–F where groundwater flows from outside the study area into the Subei Lake basin. Lateral outflow occurs along the boundary G–H by analyzing the contours and flow direction of groundwater (Figure 2). A constant outflux was defined at this boundary.

In the model, the source/sink factors represent the amounts of groundwater recharge and discharge. The sources term includes the recharge volume by percolation from precipitation, irrigation return flow and lateral inflow. Precipitation infiltration via the vadose zone is the major source of recharge. The precipitation was 399.6 mm during the modeling period and the coefficient of precipitation infiltration was affected by topography, geomorphy, and soil type. During the modeling period, the amount of irrigation return flow was estimated to be 0.22 × 10⁴ m³/d by the amount of agricultural irrigation multiplied by the coefficient of irrigation infiltration. The coefficient of irrigation infiltration was obtained from the local water conservancy department. The amount of lateral inflow was estimated to be 3.47 × 10⁴ m³/d using Darcy's law, and the parameters in Darcy's law were obtained from the contours of groundwater level and hydrogeological data. The sink term is mainly groundwater evapotranspiration. The effect of evapotranspiration is mainly affected by groundwater depth. In the Subei Lake basin, groundwater discharge from evaporation can be neglected when the groundwater depth is deeper than 5 m. Groundwater pumping for energy production, irrigation, and domestic use is the major anthropogenic stress affecting the groundwater regime. The volume of groundwater pumping was 12,590.11 m³/d for irrigation and domestic use during the modeling period. In addition, Subei Lake is replenished by groundwater discharge. The volume of lateral outflow was estimated to be 0.84 × 10⁴ m³/d using Darcy's law during the modeling period.

Before the model was used to simulate future groundwater levels, calibration was carried out using historical data for groundwater levels. Model calibration consists of changing values of model input parameters in an attempt to match field conditions within some acceptable criteria. A calibrated model uses selected values of hydrogeological parameters, sources and sinks, and boundary conditions to match historical field conditions. After the model has successfully reproduced measured changes in field conditions, it is ready for predictive simulations.

The hydraulic properties, such as permeability coefficient (K), specific yield (μ), and specific storage (S_s), were calibrated in this groundwater model. The study area was divided into seven sub-areas according to local hydrogeology, and these hydrogeological parameters were constant for each sub-area. The calibration makes the results calculated by the model match the measured groundwater level data as much as possible. The calibrated values of the hydrogeological parameters are presented in Figure 5 and Table 1.

Contour maps (Figures 6 and 7) of simulated groundwater levels versus observed groundwater levels indicate fairly good agreement, which confirmed that the calibrated values of hydrogeological parameters can reflect actual hydrogeological conditions in the Subei Lake basin.

To assess the performance of the calibrated model, simulated hydraulic heads were compared to observed groundwater levels at a total of 20 observation wells due to the limited field data. Statistical indices were used for quantitative evaluation of model performance in this study. The results of calibration and validation can be evaluated by mean absolute error (MAE) and root mean squared error (RMSE) (Rejani et al. 2007). The scatterplot of simulated and observed heads in the 20 observation wells during the modeling and calibration periods is shown in Figure 8. It can be seen that the model reasonably reproduces the observed values, with RMSE and MAE being 0.88 and 0.63. Hence, the obtained results indicated that the conceptual model, boundary conditions, and hydrological parameters used in the model are reliable, and it can be further used to simulate the effect of industrial activities on groundwater levels.

The groundwater budget during the modeling period (1st September 2003 to 31st August 2004) is presented in Table 2, which reflects the situation of water balance in a natural state without the impact of groundwater pumping from Haolebaoji waterworks. For the groundwater system irrespective of the internal exchange capacity (i.e., leakage flow) in the study area, the total recharge of groundwater was 15.24 × 10⁴ m³/d, of which about 75.8% was from infiltrating rainfall; the total discharge of groundwater was 14.30 × 10⁴ m³/d, mainly including groundwater evapotranspiration and groundwater discharge to the Subei Lake, accounting for 45.3% and 40.1%, respectively, of the total discharge; the balance error was 0.94 × 10⁴ m³/d. The hydraulic connection between the unconfined and confined aquifers is very close due to an uncontinuous aquitard between them. Overall, the exchange between the unconfined aquifer and confined aquifer was very close. The volume of the downward leakage from the upper aquifer to the confined aquifer was 1.02 × 10⁴ m³/d, while the volume of the upward leakage from the confined aquifer to unconfined aquifer was 0.98 × 10⁴ m³/d. In other words, in a natural state, rainfall infiltration and groundwater evapotranspiration were the major source/sink terms for the groundwater system in the study area, which played a primary role in the groundwater circulation. This is also consistent with the actual hydrogeological situation in the study area.

To control the decline of groundwater levels and protect the ecological environment in the Subei Lake basin, several measures have been proposed to reduce groundwater exploitation, including reducing the pumping rate of Haolebaoji waterworks and possible water diversion projects. The calibrated model was used to evaluate plans for potential exploitation of groundwater in the Subei Lake basin, with the objective of predicting the trends of groundwater evolution in the next 30 years and the impact of groundwater pumping from Haolebaoji waterworks on the ecological environment. Simulations were made for the period 2010–2039 with a monthly time step. The actual data of precipitation and evaporation from 1985 to 2014 were used as those of the precipitation and evaporation for the future years in the predicted model. A water diversion project has been operational since 2016; the volume of groundwater pumping from Haolebaoji waterworks was mostly replaced by the drainage water from a coal mine outside the study area which was used to meet the water demand for energy production. As a result, the volume of groundwater pumping from the waterworks was reduced to 0.5 × 10⁴ m³/d only for domestic use. Two different scenarios were considered in this study. Scenario 1: the waterworks maintain the present pumping rates (unfavorable condition); scenario 2: the water diversion project was put into use in 2016, and the volume of groundwater pumping from the waterworks was reduced to 0.5 × 10⁴ m³/d. The actual production rates from 2010 to 2012 were used as the volume of groundwater pumping from Haolebaoji waterworks during the same period. Due to lack of data of production rates from 2013 to 2015, the production rate of the waterworks remains 0.5 × 10⁴ m³/d from 2016 to 2039 in scenario 2, while the production rate in 2012 is used as that in other predicted years. The other terms of groundwater recharge and discharge were followed by the model identification stage. The data of groundwater levels which were investigated by the Inner Mongolia Second Hydrogeology Engineering Geological Prospecting Institute in December 2009 were used to interpolate the initial groundwater levels of the study area.

The contours of groundwater level in the unconfined and the confined aquifers at the end of the predicted period for the two scenarios are shown in Figures 9 and 10. The pattern of groundwater level contours in the unconfined aquifer by 2039 was similar for both scenarios (Figure 9). However, great changes will take place in groundwater flow field in the confined aquifer that was pumped from the production wells of Haolebaoji waterworks. For scenario 1, the obvious groundwater depression cone will be formed in the production field by 2039 (Figure 10(a)). According to the predictive hydrograph in the center of the groundwater depression cone, the maximum drawdown is 11.70 m and the average rate of decline during the predicted simulation 0.39 m/a (Figure 11). Groundwater level in the center of the groundwater depression cone dropped at the rate of 0.74 m/a significantly during the first half of the predicted period. For the rest of the years, the groundwater level reached a stable state with almost no change. For scenario 2, under the condition of reduction in groundwater pumping, the recovery of groundwater level is obvious and the groundwater depression cone will disappear as a result of the implementation of the water diversion project (Figure 10(b)).

Whether the water diversion project is put into use or not is the main difference between the two scenarios. The volume of groundwater pumping from the Haolebaoji waterworks is reduced to 0.5 × 10⁴ m³/d in scenario 2, while the waterworks maintains the present pumping rates (2.48 × 10⁴ m³/d) in scenario 1 under unfavorable conditions. The sources of the increased pumping volume were analyzed by the comparison of the change in groundwater budgets between the two scenarios (Table 3). Given that the water diversion project was put into use in 2016, the increased pumping rate of the waterworks between the two scenarios was 578.47 × 10⁴ m³/a. The increased volume of groundwater pumping was derived from storage depletion and capture, where capture includes both increases in recharge and decreases in discharge. About 33% of the increased pumpage was derived from a reduction in groundwater storage which reached 192.58 × 10⁴ m³/a. The reduction in groundwater evapotranspiration was 191.16 × 10⁴ m³/a, accounting for 33% of the increased pumpage. The volume of the additional lateral inflows and the reduction in outflows were 109.73 × 10⁴ m³/a and 47.11 × 10⁴ m³/a, respectively, contributing to 19% and 8% of the increment. In addition, the volume of decreased discharge into the Subei Lake was 37.89 × 10⁴ m³/a that was captured by the pumping from Haolebaoji waterworks, accounting for 7% of the increment.

For scenario 1, if the waterworks maintains the present pumping rates, the volume of groundwater discharge to the Subei Lake will decrease from 1,743.64 × 10⁴ m³/a in 2010 to 717.26 × 10⁴ m³/a at the end of 2039 (Figure 12). The amount of groundwater discharge to the Subei Lake will reduce sharply, which would result in the terrible situation that the area of the Subei Lake showed a declining trend year by year. For scenario 2, the variation of groundwater discharge to the Subei Lake can be divided into two stages. Before the year of 2016 when the water diversion project was put into use, the volume of groundwater discharge to the Subei Lake reduced from 1,743.64 × 10⁴ m³/a in 2010 to 615.90 × 10⁴ m³/a in 2016. After 2016, the amount of groundwater discharge to the Subei Lake will show an overall upward trend and increase to 785.01 × 10⁴ m³/a in 2039 (Figure 12). Although the volume of groundwater pumping from the Haolebaoji waterworks has been reduced to 0.5 × 10⁴ m³/d since 2016 in scenario 2, the average annual growth rate of groundwater discharge to Subei Lake will be only 7.05 × 10⁴ m³/a, indicating that the recovery of Subei Lake will be very slow. Therefore, groundwater pumping from Haolebaoji waterworks has caused a negative impact on Subei Lake to a certain degree and it will be very difficult for the area of Subei Lake to recover in a short time.

From the perspective of hydrogeology, the unconfined groundwater is the main recharge source for Subei Lake. The production wells of Haolebaoji waterworks are all distributed in the upstream of Subei Lake. The overexploitation of the confined aquifer has caused a general piezometric decline and further induced a large downward leakage from the unconfined aquifer. Therefore, the unconfined groundwater that originally flowed towards Subei Lake was captured by the production wells for industrial use, which caused a significant decrease in groundwater discharge to Subei Lake and affected the normal ecological function of the lake.

The eco-environment is very fragile in the Subei Lake basin due to its dry climate and the scarcity of surface water. Groundwater exploitation from Haolebaoji waterworks has inevitably affected the natural state of groundwater resources and groundwater-dependent vegetation. According to the unpublished hydrogeological report from the Inner Mongolia Second Hydrogeology Engineering Geological Prospecting Institute and Wang et al. (2010), the ecological groundwater depth of sandy semi-shrub herbs and trees is above 8 m, while that of phreatophyte ranges from 3 to 5 m. The spatial distribution of vegetation species in 2003 before the operation of Haolebaoji waterworks and depth to water table at the end of the predictive period for the two scenarios are shown in Figure 13.

Artemisia ordosica, a representative of sandy semi-shrub herbs, was the most widespread in the Subei Lake basin (Figure 13). As a dominant species, Artemisia ordosica grows in most areas except for sporadic distribution in low land such as Subei Lake and Kuisheng Lake. The vegetation is dominated by Artemisia ordosica, especially in highlands where the depth to water table exceeds 8 m. Although the depth to water table will be above 8 m in most of the northwestern areas, groundwater pumping from the Haolebaoji waterworks will exert little impact on the growth of Artemisia ordosica because it is a xerophyte and requires less water to maintain its growth.

In 2003, the trees including poplar, willow, and elm, were mostly distributed in southwestern areas. For both scenarios, the depth to water table at the end of the predictive period will exceed 8 m in some areas where the trees are distributed (Figure 13). However, these xerophytes, such as trees and sandy semi-shrub herbs, mainly depend on precipitation and soil water to sustain their growth. Therefore, groundwater pumping from the waterworks has little influence on these trees and the decline of groundwater level basically cannot result in the significant degradation of the trees.

The distributed area of phreatophytes, including Carex tristachya, Suaeda glauca Bge, Phragmites australis, and Achnatherum splendens, was second only to that of sandy semi-shrub herbs in 2003. These phreatophytes are mainly distributed around Subei Lake, Kuisheng Lake and their northern areas. Based on the predicted results for scenario 1 and scenario 2 at the end of 2039, the areas where the depth to water table exceeds 5 m will account for 65% and 61% of the distributed area where phreatophytes grew in 2003, respectively (Figure 13). The results indicate that the decline of groundwater level will have direct negative impacts on the growth of phreatophytes.

In other words, groundwater pumping from Haolebaoji waterworks will cause the noticeable degradation of phreatophytes and exert less impact on xerophytes for both scenarios. These results are consistent with remote sensing observations (Liu et al. 2016).

Groundwater resources in arid regions support a variety of human activities because they are frequently productive, dependable, and easily exploited. The groundwater is also critical to aquatic ecosystems because the inland lakes and some vegetation species are recharged by groundwater. Consequently, overexploitation or improper management jeopardizes the viability of lakes and vegetation that depend on groundwater. In this study, a three-dimensional groundwater flow model for the Subei Lake basin, an arid area of northwest China, was developed using GMS software to simulate regional groundwater changes under transient conditions. The results of calibration showed that the predicted results matched well with the observed data. The groundwater budget in a natural state showed that rainfall infiltration and groundwater evapotranspiration were the major source/sink terms for the groundwater system in the study area. The total recharge of groundwater was 15.24 × 10⁴ m³/d, of which about 75.8% was rainfall infiltration; the total discharge of groundwater was 14.30 × 10⁴ m³/d, mainly including groundwater evapotranspiration and groundwater discharge to Subei Lake, accounting for 45.3% and 40.1%, respectively, of the total discharge.

The calibrated model was used to predict the groundwater levels' variation and the impact of groundwater pumping from Haolebaoji waterworks on Subei Lake and vegetation for the period from 2010 to 2039 under two different pumping scenarios. For scenario 1, the obvious groundwater depression cone will be formed in the production field at the end of 30 years and the maximum drawdown will be 11.70 m if the waterworks maintains the present situation. For scenario 2, under the condition of reduction in groundwater pumping, the recovery of groundwater level will be obvious and the groundwater depression cone will disappear as a result of the implementation of the water diversion project. The increased volume of groundwater pumping between the two scenarios was derived from storage depletion (33%), captured groundwater evapotranspiration (33%), activated lateral inflow (19%), captured lateral outflow (8%), and discharge into the Subei Lake (7%). Groundwater pumping from Haolebaoji waterworks has caused negative impact on Subei Lake and it will be very difficult for the area of Subei Lake to recover in a short time. In addition, it will cause the obvious degradation of phreatophytes but exert less impact on xerophytes for both scenarios.

An important contribution of this study is that it provides a framework for quantifying the potential impacts of groundwater pumping on lake and vegetation. Integrative water resources management measures include water-diversion projects, restricting heavy groundwater pumping from the production wells close to the lakes, monitoring of water level in the groundwater depression cone periodically, and developing a long-term monitoring network for assessment of the viability of the groundwater-dependent ecosystem. It should be noted that the production wells should be scientifically designed away from ecological sensitive areas such as inland lakes and phreatophytes. The construction of well fields in the upstream of such ecological sensitive areas should be avoided. This is an important first step but more remains to be done to improve sustainable groundwater management in the world's arid but industrially important regions.

This research was supported by the State Basic Research Development Program (973 Program) of China (grant no. 2010CB428805). The authors are grateful to our colleagues for their assistance in data collection and field investigation. Special thanks go to the editor and the anonymous reviewers for their critical reviews and valuable suggestions.