Identifying the impact of Energy Base Water Project on groundwater using high-frequency monitoring data in the Subei Lake basin, Ordos, Northwestern China

The coordinated development of water resources and energy infrastructure in arid regions is of great concern to hydrologists (Cai et al. 2014). Although the large-scale exploitation of energy resources has brought huge economic benefits, it has also often caused some negative impacts on groundwater systems and ecosystems (Kahrl & Roland-Holst 2008; Siddiqi & Anadon 2011). Groundwater resources play a very important role in semi-arid and arid regions (Moiwo & Tao 2014). Natural processes and aquifer properties vary at different spatio-temporal scales, and their variations affect the fluctuations of groundwater levels (Li & Zhang 2007). Over the last decades, groundwater has increasingly been extracted for energy production in arid regions around the world due to the fast growth in water demand for industrial use (Kahrl & Roland-Holst 2008; Siddiqi & Anadon 2011). The increasing concerns about potential threats to groundwater-dependent ecosystems in these areas should be taken into consideration in the framework of sustainable water resource management (Wang et al. 2012; Li et al. 2015; Cai et al. 2016).

There are various techniques for distinguishing the impact of anthropogenic and natural factors on groundwater systems with complicated hydrogeological settings. The techniques of hydrochemistry and stable isotopes have been widely used to qualitatively identify the impact of anthropogenic and natural factors on the groundwater (Baudron et al. 2014; Nyenje et al. 2014; Murgulet et al. 2015; Liu et al. 2015a; Liu et al. 2015b). However, groundwater monitoring is necessary because it is a direct measurement for quantitatively identifying the role of anthropogenic and natural factors in the dynamic behaviors of groundwater system (Rajmohan et al. 2007; Ebrahim et al. 2013; McCallum et al. 2013). Differences between groundwater hydrographs at different sites must be related to the distance from the source of influence and reflect the hydrogeological settings near the monitoring wells to a certain extent. Groundwater evapotranspiration (ET_GW) is an important component of the groundwater balance in arid regions. The White method estimates daily ET_GW using diurnal groundwater level fluctuations from a monitoring well. Moreover, the White method provides the basis for all subsequent ET_GW estimation algorithms from the hourly hydrographs (White 1932; Healy & Cook 2002; Loheide et al. 2005; Yin et al. 2013; Zhang et al. 2015). A principal application of the boxplots is to identify possible outliers, which are a minority of observations in a dataset that have different patterns from that of the majority of observations in the dataset (Hadi et al. 2009; Dovoedo & Chakraborti 2014). In addition, the technique of remote sensing was applied to understand the relative impacts of anthropogenic activities on the temporal variation of lake area (Seyoum et al. 2015). Although these methods individually are not new, their combined application provides a novel methodology to identify the impacts of anthropogenic and natural factors on the groundwater, which can be applied to catchments worldwide.

The main objective of the paper is to identify the impact of the Energy Base Water Project on the groundwater by utilizing the high-frequency groundwater level data in this ecologically sensitive area. The related findings will provide an important reference for the establishment of the hydrogeological conceptual model and groundwater flow model. In addition, this research can contribute to the ongoing discussion about the coordinated development between water and energy infrastructure in arid regions, and it will advance necessary knowledge for the sustainable groundwater management in such arid areas.

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). Unconfined and confined aquifers can be observed in the basin. The phreatic aquifer is composed of the Quaternary sediments and Cretaceous sandstones (Figure 3(a)), with its thickness ranging from 10.52 to 63.54 m. In terms of borehole data, the similar groundwater levels in the Quaternary and Cretaceous phreatic aquifers indicate a very close hydraulic connection between the Quaternary layer and Cretaceous phreatic aquifer, which can be viewed as an integrated unconfined aquifer in the area (Wang et al. 2010). The precipitation infiltration and lateral inflow are the major recharge sources for the unconfined aquifer. The locally upward leakage from the underlying confined aquifer and irrigation return flow can also provide a small percentage of groundwater recharge. Groundwater evapotranspiration is the main discharge means of the unconfined groundwater. In addition, lateral outflow, artificial exploitation and downward leakage are included in the main discharge patterns (Wang et al. 2010). Groundwater levels were contoured to display the groundwater flow direction in Figure 3(a), which were measured in the regional survey conducted by the Inner Mongolia Second Hydrogeology Engineering Geological Prospecting Institute during September 2003. The groundwater flowed predominantly from the surrounding highlands to lowlands, which is under the control of topography. Overall, groundwater in the phreatic aquifer flowed toward the Subei Lake (Figure 3(a)).

In the basin, groundwater was exploited for agricultural, industrial and domestic purposes. The proportions of agricultural, industrial and domestic water demand were almost 30%, 69% and 1%, respectively, in 2009, according to the unpublished hydrogeological report from the Inner Mongolia Second Hydrogeology Engineering Geological Prospecting Institute. The flood irrigation entirely relied on groundwater pumping from scattered irrigation wells due to the scarcity of surface water. Haolebaoji waterworks has 22 production wells stretching northeast to southwest, which have become operational for water supply in energy production since 2006. The production wells were designed in the shape of a rectangular grid at intervals of 1.5 km to each other. The production rate of individual wells was unknown throughout the monitoring period. However, 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, and all the production wells were screened within the confined aquifer due to the huge storage of confined groundwater.

The entire basin was covered by the monitoring network, composed of the eight monitoring wells (with well depths ranging from 6 to 300 m), which were well designed for the monitoring project in 2013 by the Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences (Figure 3 and Table 1). The length of screen pipes in all monitoring wells varies from 1 to 10 m, and every monitoring well has only one screen pipe rather than multiple screens. The distance between the bottom of the screen pipe and the total well depth ranges from 0 to 3 m in the study area. Special attention was paid to monitoring the dynamic variation of groundwater level at different aquifer units and identifying the potential observation points affected by large-scale groundwater exploitation. All sites were equipped with a data logger (KADEC-MIZU II, Japan) and the groundwater levels were recorded at 30 min intervals from December 2013 to December 2014. In addition, a rain gauge was installed in the basin for the collection of precipitation data.

Given that the confined aquifer was intensively exploited by 22 production wells for industrial use, except for wells JC01 and JC06 which were used to monitor the unconfined aquifer, other wells were used to monitor the confined aquifer (Figure 3). Well JC01 is located in the highland and was screened within the Quaternary sediments. Well JC06 is located in the lowland and was screened within the Cretaceous rocks. The two monitoring wells were used to quantify the daily ET_GW in the Quaternary and Cretaceous phreatic aquifers, respectively (Figure 3(a) and Table 1). The other six wells (wells JC02, JC03, JC04, JC05, JC07, and JC08) were designed to monitor the variation of the confining groundwater level so as to assess the impact of groundwater pumping from Haolebaoji waterworks on the confined aquifer. The six monitoring points were selected in consideration of various factors including the position in the groundwater flow system, the convenience of management and data collection, hydrogeological information, and the distances from 22 production wells (Figure 3(b) and Table 1). The missing values due to data logger malfunctions have been excluded.

In order to analyze the impact of groundwater pumping from Haolebaoji waterworks on the Subei Lake, the manually recorded water level of the Subei Lake and the rate of groundwater pumping from Haolebaoji waterworks from July to November in 2009 have also been collected from previous hydrogeological surveys conducted by the Inner Mongolia Second Hydrogeology Engineering Prospecting Institute.

A simple way widely used to identify outliers is based on the concept of the boxplot (Tukey 1977). This graphically-based method for outlier detection is especially attractive not only for its simplicity but, more importantly, because it does not employ the extreme potential outliers that can distort the computing of a measure of spread and lessen the sensitivity to outliers.

In this study, the temporal variation of lake areas was mapped using LANDSAT Thematic Mapper (TM) images obtained from the United States Geological Survey (http://glovis.usgs.gov/). The satellite images were pre-processed by the standard techniques such as atmospheric correction, radiometric calibration, and so on. NDWI was calculated from the green and near-infrared bands of LANDSAT TM by using processed satellite images. Finally, the images were processed in ArcGIS and were used to calculate the surface area of the lakes.

The variation of groundwater level in well JC01 can be classified into two different phases (Figure 5(a)). A stable trend with a groundwater level variation of −0.025 ∼ 0.028 m/d was observed from 2 December 2013 to 28 March 2014, indicating that groundwater inflow and outflow were almost equal. A fluctuating trend with a daily fluctuation variation of −0.24 ∼ 0.26 m was presented from 29 March 2014 to 14 May 2014, when the rainfall increased gradually and irrigation activities became frequent in the Subei Lake basin. However, the low rainfall from individual events may be subject to evapotranspiration before reaching the water table given that most of the rainfall amounts during this period were below 10 mm/d. The obvious decline and recovery of groundwater level was observed within a few hours, which was controlled by the combined effect of short-term groundwater exploitation from the nearby irrigation well and the rapid recharge from irrigation return flow due to the high permeability of the vadose zone, which was composed of the Quaternary unconsolidated sediments. Therefore, flood irrigation was mainly responsible for the groundwater regime during the fluctuating period.

Well JC06 exhibited a different pattern from the previous one (well JC01). The variation of groundwater level in well JC06 can be divided into two different phases (Figure 5(b)). The critical point for the two phases was 22 June 2014, when the groundwater level reached the its minimum value. Overall, the water table declined slowly before 22 June, indicating that a net groundwater outflow controlled the variation of groundwater level. Subsequently, in response to frequent rainfall events, the groundwater level rose quickly within a few days after 22 June. During the second period, precipitation infiltration predominated the hydrological processes and exerted a long-term effect on the groundwater regime. A fluctuation with a larger amplitude took place from 8 July to 17 July due to groundwater pumping from well JC06 itself. In the following days, the water table fluctuated with a relatively smaller amplitude but remained at a higher groundwater level.

It should be noted that the different dynamic patterns of the water table in these two wells may be explained by the local hydrogeological settings (such as lithology, hydraulic properties, the depth to water table, etc.), precipitation and anthropogenic activities (such as groundwater pumping and irrigation return flow caused by the traditional flood irrigation). In addition, the time series length of groundwater monitoring data in well JC01 was shorter than that in well JC06, so it is necessary to continue groundwater level monitoring in order to assess the temporal variations of groundwater level precisely in the next step.

Wells JC02 and JC04 were distributed in the upstream of the production field. However, there existed some differences between the groundwater regime in well JC02 and that in well JC04 during the monitoring period. The highest and lowest groundwater levels were 1,352.86 and 1,351.80 m with an average value of 1,352.59 m for well JC02, which is adjacent to the western boundary (Table 2). Before April, the groundwater level in well JC02 varied from 1,352.48 to 1,352.86 m and maintained a relatively equilibrium state, indicating that the groundwater regime was completely controlled by lateral flow and leakage without any human interference. However, after April, an obvious trough occurred in the daily hydrograph for well JC02, which was related to the disturbance from anthropogenic factors such as irrigation activities or groundwater pumping from the Haolebaoji waterworks (Figure 6(a)). A mean groundwater level of 1,334.10 m was observed in well JC04 close to the northern boundary, with a maximum value of 1,337.26 m and minimum value of 1,332.30 m (Table 2). As is shown in the daily hydrograph for well JC04, a general downward trend in groundwater level can be observed before April, showing that the groundwater inflow was less than the total outflow. Subsequently, the groundwater level exhibited a significant rising trend, indicating the aquifer in that area received more lateral inflow from the groundwater outside the study area or downward leakage from the overlying unconfined aquifer due to the advent of the monsoon season (Figure 6(c)).

The highest and lowest groundwater levels were 1,315.67 and 1,309.81 m for well JC03, the mean value was 1,314.34 m (Table 2). The daily hydrograph for well JC03 can be divided into three stages in the whole hydrologic year. Before April, the groundwater level showed a slow downward trend and varied from 1,314.18 to 1,315.67 m during the first stage. The groundwater level was subject to intense interference from human activities during the second stage from April to September. The groundwater level fluctuated with large amplitudes due to the rapidly increasing water demand with the advent of the irrigation season. The daily hydrograph for the second stage was composed of a series of sharp declines and rises. The sudden drop occurred when well JC03 was pumped for irrigation, while the groundwater level returned to the initial level quickly within a few days after the pumping stopped. After September, the impact of anthropogenic activities on the groundwater level can be neglected with the completion of agricultural activities. The groundwater level began to ascend slowly and reached a relatively steady state until December 2014. During the third stage, the groundwater regime was almost controlled by natural factors (Figure 6(b)).

The maximum and minimum groundwater levels for well JC05 were 1,302.96 and 1,296.00 m, averaging 1,301.22 m (Table 2). The daily fluctuation of groundwater level in well JC05 also showed three stages similar to those of well JC03. The major differences between wells JC05 and JC03 were the groundwater regime in the first and third stages. Unlike the general downward trend in the first stage for well JC03, the groundwater level oscillated with relatively larger amplitude before April. A series of sharp declines and rises occurred in well JC05 from April to September, when human activities were the most frequent of the whole year. It should be noted that after September, the groundwater level rose slowly from the trough until the peak in the middle of October, then decreased gradually until December 2014 (Figure 6(d)).

The groundwater level in well JC07 fluctuated from 1,295.34 to 1,300.47 m with an average value of 1,298.16 m (Table 2). The distinctive characteristic of pulse signals was observed in the daily hydrograph of well JC07. Well JC07 is ideal for monitoring the impact of groundwater pumping from the waterworks on the confined aquifer given that there is not any pumping from that well itself. In a natural state, the groundwater regime should be relatively stable and the annual amplitude would be relatively small. However, the maximum difference between the peak and the trough reached nearly 4 m, by the analysis of the daily hydrograph, indicating that well JC07 must be subject to strong interference from the waterworks. In addition, it is clear that the production wells close to well JC07 had been engaged in intermittent groundwater pumping from December 2013 to the end of October 2014 according to the hydrograph. The daily hydrograph during this period may be explained by the following statements. The on-going groundwater pumping from the waterworks was mainly responsible for the process from the peak to the trough. The original groundwater regime was disrupted by groundwater pumping from the waterworks, and the groundwater level declined continuously during the period from the peak to the trough. The trough in the hydrograph was formed when the production wells stopped pumping, then the groundwater level ascended gradually and formed the process from the trough to the peak. As a result, the cycle was repeated in the daily hydrograph of well JC07 before the end of October. After October, the production wells were not engaged in any pumping activities and groundwater level rose quickly during this period (Figure 6(e)).

In the present study, the monitoring data in unconfined aquifers were used to identify the hydrological processes controlling the WTFs and estimate groundwater evapotranspiration. Given that the missing values and the influence of groundwater pumping have to be excluded, the selected periods for the calculation of ET_GW in the two wells were the onset of the growing season and the peak of the growing season, respectively. During these specific periods, the fluctuation of the water table was only affected by recharge, plant transpiration and evaporation, so the results of the ET_GW calculation were reliable.

The hourly hydrographs were used to calculate the daily ET_GW values for wells JC01 and JC06 using the White method, as shown in Equation (2). In this study, the specific yield and the extinction depth were determined based on the results of groundwater flow modelling in the Subei Lake basin (Wang et al. 2010) and the hydrogeological survey conducted by the Inner Mongolia Second Hydrogeology Engineering Geological Prospecting Institute. The values of specific yield for wells JC01 and JC06 were 0.15 and 0.18, respectively. The extinction depth of ET_GW in the study area is 4 m, so the groundwater in wells JC01 and JC06 was subjected to evapotranspiration in different degrees given that the depth to water table was within 4 m. Table 3 shows the calculated ET_GW for wells JC01 and JC06. At the beginning of the rapid growing season for vegetation (April), the calculated ET_GW for well JC01 ranged from 5.40 to 6.75 mm/d, with an average value of 6.15 mm/d. However, during the summer season (July or August), the calculated ET_GW for well JC06 varied from 9.90 to 14.94 mm/d, with an average value of 12.48 mm/d. The results were similar to those obtained by Yin et al. (2011), in which the diurnal ET_GW values ranged from 10 to 22 mm/d. Hydrometeorological factors were mainly responsible for the fact that the daily ET_GW for well JC06 was clearly higher than that for well JC01. The groundwater level in well JC06 was shallower than that in well JC01, so the former evaporated more easily than the latter. In addition, air temperature played a very important role in groundwater evapotranspiration, and the mean temperature in July was obviously higher than that in April, which contributed to the higher ET_GW for well JC06.

Given that there was no groundwater withdrawal occurring at selected periods, by the comprehensive analysis of groundwater hydrographs, the calculated results of ET_GW were reliable. However, there are some uncertainties associated with the use of the White method. It is noteworthy that the White method is based on the assumption that diurnal WTFs are only caused by the evapotranspiration of plants (White 1932). However, lateral outflow and leakage discharge can also cause declines in the water table, which is a key source of error in the estimation of ET_GW. Another main uncertainty is the limited accuracy of the specific yield (Loheide et al. 2005), which depends on the sediment texture, the initial depth to water table and elapsed time of drainage (Nachabe 2002). In addition, a constant rate of groundwater inflow determined from the rate of change in the water table during a short time interval between 00:00 and 04:00 can result in some uncertainties in the calculation of ET_GW (Soylu et al. 2012).

The confined aquifer is sensitive to a change in aquifer stress such as earthquakes, changes of atmospheric pressure, groundwater withdrawal, etc. (Bras & Rodríguez 1985). Groundwater levels are inclined to display outliers with the impact of human factors, which can be explained by the micro-dynamics features reflected by the high-frequency groundwater level data. Therefore, it is necessary to identify outliers of groundwater monitoring data in the confined aquifer. A dataset of the rate of daily variation in groundwater level in these monitoring wells can be constructed in order to capture the unusual fluctuations of groundwater levels caused by groundwater pumping from the Haolebaoji waterworks.

However, Pattern 1(b) can only be observed in well JC08 and exhibited distinctly different dynamics for groundwater level, which can be explained by the impact of intensive groundwater withdrawal from the nearby production wells No. 21 and 22 at a distance of 763 m and 680 m from well JC08, respectively. Groundwater level declined instantaneously with an amplitude of 0.676 m at 13:00, but subsequently the rate of groundwater decline reduced slowly from 0.176 to 0.013 m/h until 07:30 the next day, when groundwater inflow was almost equal to groundwater outflow. It showed that the groundwater level regained a relatively steady state at 07:30 h on 2 April and the obvious recovery, with an amplitude of 0.726 m, was observed within just half an hour when groundwater pumping from the adjacent two production wells ceased. The groundwater level almost returned to the initial water level within a few hours, indicating groundwater pumping for industrial use can be satisfied due to the sufficient groundwater recharge in the vicinity of well JC08. However, it should be noted that the groundwater level in well JC08 took a longer time to regain its original water level than that in well JC03, which implied that well JC08 may be at least partially connected to a mudstone lens.

Pattern 2 (irregular) was characterized by erratic variation and can only be observed in well JC04. Given that the confined aquifer in the northern recharge area was only composed of the fissured Cretaceous rocks, the strong irregular variation in the groundwater level was likely to be caused by the heterogeneity of the confined aquifer (the density and orientation of the fissures, and the connectivity of the groundwater flow path). The strong irregularity was a distinctive characteristic of well JC04, which was located in the northern recharge area.

Groundwater pumping from the Haolebaoji waterworks was primarily responsible for the gradual hydrograph (Pattern 3), which was characterized by a slow continuous decline within a few hours or some successive days in groundwater level. The monitoring wells (JC02, JC05 and JC07) were disturbed by the pumping behaviors from the nearby production wells, which was evidenced by the unusual hydrographs (Figure 9). The three monitoring wells exhibited slightly different groundwater dynamics, due to the significant differences in the pumping patterns and the distance from the adjacent production wells. Well JC02 is close to the western boundary and is a long distance from the nearby production wells. It seemed that well JC02 would not be affected by the production wells, but surprisingly groundwater level gradually declined with an amplitude of 0.395 m from 08:00 to 20:30 h on 3 April, and the daily amplitude was unusual during the monitoring period (Pattern 3(a)). The impacts of natural factors and other human factors (i.e. pumping from well JC02 itself) were excluded, given the abnormally large drawdown and gradual changes in water level. Well JC02 was likely to be close to the margin of the groundwater depression cone. Given that well JC05 is surrounded by the production wells, the groundwater regime must be disturbed as long as the nearby production wells conducted groundwater pumping. It has been validated that the adjacent production wells indeed did carry out groundwater pumping by an analysis of the hourly hydrograph (Pattern 3(b)). For example, the unusual decline in groundwater level, with an amplitude of 1.175 m, was detected from 09:30 to 19:00 h on 9 March as a result of short-term interference from the nearby production wells. Pattern 3(c) was characterized by a steady decline within 24 hours. Well JC07 is the most appropriate monitoring well, given that there is no groundwater withdrawal in the well itself and it is surrounded by the production wells. The groundwater regime in well JC07 will be entirely controlled by natural factors if the nearby production wells do not carry out groundwater exploitation. So it is ideal for monitoring the pumping behaviors of the Haolebaoji waterworks. The hydrograph of well JC07 exhibited distinctly different dynamics (Figure 6) and showed the outliers during the monitoring period. The surprising result that the groundwater level showed a steady, continuous decline within 24 hours has been summarized. As shown in Pattern 3(c), the continuous decline in groundwater level, with an amplitude of 2.644 m, was observed from 3 to 6 March as a result of continuous groundwater extraction from the nearby production wells during this period.

By a comprehensive analysis of hourly hydrographs, the characteristic of groundwater pumping from the Haolebaoji waterworks is that only some individual wells had been engaged in the intermittent groundwater pumping at random times, which can be validated in the hydrogeological survey report from the Inner Mongolia Second Hydrogeology Engineering Geological Prospecting Institute.

Satellite-based temporal mapping of the lake surface area was calculated using Landsat TM imagery from 2000 to 2011 for the Subei Lake basin. Given that the Haolebaoji waterworks became operational for industrial use in 2006, the temporal variation of the two lake areas can be divided into two phases: before the operation of the Haolebaoji waterworks (2000–2006) and after the operation of the Haolebaoji waterworks (2006–2011).

The present study identified the impact of the Energy Base Water Project on the groundwater in the Subei Lake basin, which is subjected to strong interference from human activities, with various methods such as statistical analysis, the WTF method, hydrograph analysis and remote sensing. The combination of both the daily and hourly hydrographs has provided a comprehensive understanding of the natural processes and anthropogenic activities that control the groundwater regime of the whole water system. For natural factors, precipitation infiltration and groundwater evapotranspiration have relatively short response times, but may exert a long-term effect on the groundwater regime. However, the impact of human activities (such as groundwater pumping and flood irrigation) on the groundwater is relatively instantaneous. The groundwater regime was subject to intense interference from human activities from April to September. Human activities (groundwater pumping for energy production and irrigation) are the major anthropogenic stresses affecting the groundwater regime. The mean evapotranspiration rates in the Quaternary phreatic aquifer and the Cretaceous phreatic aquifer were 6.15 mm/d and 12.48 mm/d, which can be explained by the significant differences in the hydrometeorological factors and local hydrogeologic conditions (aqueous medium, soil texture, depth to water table).

The unusual hourly hydrographs in the confined aquifers exhibited three different patterns, namely mutational, irregular and gradual hydrographs. Different recovery times after being influenced by pumping may be related to the presence of mudstone lenses. The extent of the groundwater depression cone was qualitatively identified by gradual hydrographs, which may spread from the center area (wells JC05 and JC07) to the western boundary (well JC02). It can be validated that only some individual wells had been engaged in the intermittent pumping activities at random times. Groundwater pumping from the Haolebaoji waterworks has caused the Subei Lake to shrink more seriously than the Kuisheng Lake.

Further field investigation and high-precision groundwater flow models will be required to determine the extent of the aquitard and the exchange capacity of groundwater between the different source/sink terms in order to fully quantify the hydrological and anthropogenic impacts on the groundwater regime. Similar integrated approaches can be applied to understand the impact of anthropogenic activities on the groundwater in other data sparse regions. To achieve the objectives of protecting the vital groundwater resources and vulnerable ecosystems, government agencies should adopt the following strategies: restricting heavy groundwater pumping from the production wells close to the lakes; monitoring water levels in the groundwater depression cone periodically; and developing a long-term monitoring network for assessment of the viability of the groundwater-dependent ecosystem.

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.