Water balance and influence mechanism analysis: a case study of Hongjiannao Lake, China

Hongjiannao groundwater exchange was the largest desert freshwater lake in China (57.25 km2 in 1986). However, it shrank sharply over the past 34a (1986–2019), with the smallest lake area 31.41 km2 in 2015. The objective of this study was to use the Landsat images, ASTER GDEM V2 data, and meteorology and statistics data, in combination with the water balance model to calculate the dynamics of water balance elements, quantify and characterize the interannual variations in lake-groundwater exchanges, and analyze its influencing factors by using the geographical detector. The results showed that in the stable stage (1986–1997), the average rate of the lake area, water level, and lake volume change was −0.26 km2/a, −0.0483 m/a, and −0.0009 km3/a, respectively. Precipitation, river inflow, and groundwater were 0.0203 km3, 0.0485 km3, and 0.0098 km3, which accounts for the whole input were 25.83%, 61.70%, and 12.47%, respectively; evaporation was 0.0786 km3. In the reduction stage (1998–2015), the average rate of the lake area, water level, and lake volume change was −1.21 km2/a, −0.2422 m/a, and −0.0101 km3/a, respectively. Before 2006, precipitation, river inflow, and groundwater were 0.0154 km3, 0.0475 km3, and −0.0025 km3, respectively; from 2006 to 2009, precipitation, river inflow, and groundwater were 0.0143 km3, 0.0334 km3, and 0.0058 km3, respectively; after 2009, precipitation, river inflow, and groundwater were 0.0139 km3, 0.0199 km3, and 0.0085 km3, respectively. Evaporation decreased from 0.0714 to 0.0480 km3 from 1998 to 2015. In the growth stage (2016–2019), the average rate of the lake area, water level, and lake volume change were 1.38 km2/a, 0.27 m/a, and 0.0088 km3/a, respectively. Precipitation, river inflow, and groundwater were 0.0209 km3, 0.0005 km3, and 0.0373 km3, which accounts for the whole input were 46.63%, 52.12%, and 1.25%, respectively; evaporation was 0.0187 km3. Compared with the stable stage, groundwater in the growth stage reduced from 12.47% (0.0098 km3) to only 1.25% (0.0005 km3). From 1998 to 2004, Hongjiannao Lake experienced continuous losing conditions (discharge from the lake to groundwater), with a variable exchange volume of up to −0.01582 km3 in 1999. Through geographical detector analysis, it was found that temperature was the dominant factor from 1988 to 1997, while human factors were the dominant factors from 1998 to 2015.


Introduction
Inland lakes are an important part of water resources in arid or semi-arid regions, which has an important impact on the balance of the local ecosystem and human production and life (Sheng et al., 2016;Gal et al., 2016;Zhang et al., 2017;Yang et al., 2020). In recent years, due to the impact of climate change and human activities, the area of some lakes in northern China has shrunk or disappeared (Ma et al., 2011;Liu et al., 2013;Yang & Lu, 2015;Liu & Yue, 2017). It is a very important and forward-looking work to analyze the evolution and genesis of inland lakes in arid and semi-arid regions and to evaluate the impact of climate change and human activities on lakes, which has guiding significance for the rational development and utilization of lake resources in arid and semi-arid areas.
Hongjiannao Lake is located at the junction of Shaanxi and Inner Mongolia in the arid and semiarid regions of Northwest China. It is currently the largest desert freshwater lake in China, which plays an indispensable role in regulating the local climate, maintaining the balance of water resources, and biodiversity. Hongjiannao Lake is formed by wind erosion depression. Before lake formation, it is a marsh with less water. Around 1929, due to the increase of water accumulation, the depression became a lake. In the 1930s, the lake area was only 1.3 km 2 and gradually expanded to 20 km 2 in 1947. In the middle and late 1950s, due to the dredging of surrounding wetlands for reconstruction, the lake area continued to expand. From the 1970s to the 1990s, the lake area was stable at about 60 km 2 (Tang et al., 2003;Shen et al., 2005;Yin et al., 2008). Since the 1990s, Northern Shaanxi has suffered from continuous drought and lake evaporation has increased. Meanwhile, Shendong Mining Area, China's largest coal production base, started large-scale coal mining in the northeast of Hongjiannao Lake, which is only 20-30 km away from the lake. The continuous development of the coal mining project has a great impact on the groundwater supply for Hongjiannao Lake. In 2015, the area of Hongjiannao Lake has been reduced to its lowest level of 31.41 km 2 , and the continuous reduction of the lake has resulted in the continuous deterioration of the ecological environment (Yue & Liu, 2019a, b). Therefore, it is necessary to monitor the variations of Hongjiannao Lake and to analyze its responses to climatic and anthropogenic factors, to provide a scientific reference for rational coal mining exploitation and protection of water resources, as well as regional sustainable development in the arid and semi-arid area of Northwest China.
Various studies have been put forward to analyze the changes of the lake area, water level, and lake volume and the dynamics to climate change and human activities by using remote sensing techniques and hydrological models (Ma et al., 2014;Liang, 2019;Liang & Yan, 2017;Yan et al., 2018;Wang et al., 2018a;Zhao et al., 2018;Liang & Li, 2019;Yue & Liu, 2019a, b). Indeed, Hongjiannao Lake has been largely neglected in hydrological research, as it receives less attention and lacks management compared with large lakes and reservoirs which have traditionally been managed and monitored. Hongjiannao Lake does not have enough hydrological information to assess and to understand its hydrological processes, because it does not have any hydrometeorological gauging stations surrounded by. Therefore, estimating the water balance and its components is a necessary precondition for this challenge. Many of the water balance components (surface runoff and lake-groundwater exchanges) are difficult to estimate directly even at the local scale and are often assumed to be in the residuals of the water balance equation (Kampf & Burges, 2010;Fawe et al., 2015;Huang et al., 2018;Guevara-Ochoa et al., 2020).
Previous studies have focused primarily on variations in the lake area and water volume by using remote sensing data. Comprehensive and systematic analysis of long-term continuous variations hydrological process of Hongjiannao Lake on quantitative analysis based on water balance is still lacking. Therefore, the objectives of this study were as follows: (1) based on the hydrogeological model of Hongjiannao Lake, we calculated the dynamics of water balance elements, including precipitation, evaporation, lake volume change, river inflow, and runoff change; (2) to quantify and characterize the interannual variations in lake-groundwater exchanges based on water balance model; and (3) to comprehensively explore the dominant factors affecting the lake area, water level, and water volume change of Hongjiannao Lake in terms of NDVI (Normalized Difference Vegetation Index), sheep population (SP), human population (HP), the industrial output value (GVIO), annual average temperature (AT), and precipitation (AP).

Study area
Hongjiannao Lake Hongjiannao Lake is located at the junction of Shenmu County, Shaanxi Province, and Yijinholo Banner of Inner Mongolia, and adjacent to the Mu Us Desert in the West (Fig. 1). The geographical coordinates are 109° 42′-110°54′ E and 39° 13′-39° 27′ N. Hongjiannao Lake is the largest desert freshwater lake in China, known as the "Pearl of the desert." Hongjiannao Lake is also a National Nature Reserve of China and the world's largest breeding ground for relict gulls and black storks under the national level I key protection.

Hongjiannao Basin
Hongjiannao Basin is located in the lake basin with high around and low in the middle. The northwest and northeast are undulating mountains and gentle hills. In the southwest and southeast, the surface watershed is composed of aeolian dunes. The highest point is located in the western watershed, with an altitude of 1510 m, and the lowest point is located in Hongjiannao Lake, with an altitude of 1223 m. In the north, the terrain fluctuates greatly, and in the East, there are relatively gentle dunes and beaches. The area of Hongjiannao Basin is about 1400 km 2 , including Xinjie Town, Taigesumu Town, Erlintu town, and Zhongji town. Xinjie town and Taigesumu town were merged into Zhasake town in 2005.

Watershed system
Hongjiannao Lake is a closed inland lake, which is mainly supplied by precipitation, river runoff, and groundwater. Before 2000, the runoff supply of Hongjiannao Lake came from Donghulusu River, Zhasake River, Qibosu River, Erlintu River, Malian River, Haolai River, and Songdaogou River, among which Manggaitu River, Zhasake River, and Qibosu River were the three rivers with the largest flow in the basin. In 2000, Malian River had no surface runoff and flows into Hongjiannao Lake. In 2005, the establishment of Zhasak reservoir was 6 km away from Hongjiannao Lake blocked the Zhasak River. Since then, Zhasak River had no water supply to Hongjiannao Lake. In 2009, the Inner Mongolia Autonomous Fig. 1 The spatial schematic diagram for Hongjiannao Lake and Hongjiannao Basin Region intercepted Manggaitu River and established a reservoir, which interrupted the runoff supplement of Manggaitu River to Hongjiannao Lake. At present, the largest runoff supply source of Hongjiannao lake is Qibusu River (Table 1).

Landsat images
In this paper, 359 Landsat images including Landsat-5 Thematic Mapper (TM), Landsat-7 Enhanced Thematic Mapper (ETM+), and Landsat-8 Operational Land Imager (OLI) were used to extract the lake area of Hongjiannao Lake. The path/row was 127/033 and the time was from 1986 to 2020. The images were obtained from the US Geological Survey (USGS) Earth Explorer (https:// earth explo rer. usgs. gov/).

ASTER GDEM V2 data
ASTER GDEM V2 data was released by the National Aeronautics and Space Administration (NASA) and the Ministry of Economy, Trade, and Industry (METI) on October 17, 2011. The spatial resolution is 30 m (confidence 95%), and the vertical accuracy is 20 m (95% confidence level). ASTER GDEM V2 data covered the global latitude of 83° north to 83° south, which covered more than 99% of the Earth's land surface. Compared with aster GDEM V1, which can recognize more than 12 km 2 lakes at most, GDEM V2 can recognize more than 1 km 2 lakes. In this study, ASTER GDEM V2 data was used to obtain the water level of Hongjiannao Lake from 1986 to 2020.

MODIS data
The MODIS product used in this paper was MODIS NDVI product MOD13Q1, which was often used in large-scale vegetation coverage study. The temporal resolution is 16 days; the spatial resolution is 250 m × 250 m. The path/row was h26v05, and the time series was from 2000 to 2018. The images were obtained from the National Aeronautics and Space Administration (NASA) (https:// search. earth data. nasa. gov/ search). We used the MODIS Projection Tool (MRT) to preprocess MOD13Q1 products in batches and the Maximum Value Composite (MVC) method to get the monthly data set of MOD13Q1 products.
The four datasets  were used for geographical detector analysis: the normalized vegetation index NDVI (Normalized Difference Vegetation Index), sheep population (SP), human population (HP), and the industrial output value (GVIO) come from the papers of Yue and Liu (2019a) and Liang and Yan (2017). The water used for irrigated livestock husbandry used to calculate the changes in the water volume of the Hongjiannao Basin was

Water balance model
The hydrology of the water balance model is controlled by an input-storage-output process, which can be expressed by the following water balance model (all components in km 3 /annual): where ΔV is the volume change of the lake, Q River is surface runoff into the lake, P Lake is lake precipitation, E Lake is lake evaporation, and Q Groundwater is groundwater recharge, which indicates the difference between the inflow and outflow of groundwater. This paper studies the relationship between groundwater recharge and lake volume change. Therefore, the formula of Hongjiannao water balance for many years can be expressed as

Lake area extraction: LA
We used the Deeply Clear Water Extraction Index (DCWEI) (Yue & Liu, 2019b) to extract the lake area; DCWDI can be expressed as where Red and NIR are the reflectance of the red and NIR bands, respectively.

Water-level derivation: WL
We used the waterline method combined with the ASTER GDEM V2 image and the Landsat image to obtain the Hongjiannao Lake boundary to calculate the multi-year water level of Hongjiannao Lake. Firstly, we used Landsat Image and DCWEI to extract the boundary (waterline) of Hongjiannao Lake from 1986 to 2020 (Fig. 2a). Secondly, the DEM within the lake boundary is extracted by superimposing the obtained Lake boundary in different periods on ASTER GDEM v2. Thirdly, calculate the arithmetic average value of the DEM within the lake boundary, which was the lake water level. Finally, the water level of Hongjiannao Lake from 1986 to 2011 and 2017 to 2020 was obtained through the waterline method and the DEM data ( Fig. 2b). Hongjiannao Lake continued to shrink after 2011. By January 2016, the area had reached the minimum level, and it had been rising since then. In January 2017, it reached the level of 2011. However, due to the inability to find ASTER GDEM data of the minimum elevation period in 2016, the water level of Hongjiannao Lake from 2012 to 2016 was obtained by fitting the relationship model between area and water level. We used a scatter plot to fit the area water level data from 1986 to 2011 (the equation is y = 0.2078x + 1199.4, R 2 = 0.9947, where x is the area and y is the water level). Then, substituting the area data from 2012 to 2016 into the equation, the water level data for this period was calculated.

Lake volume variation: ΔV
The volume algorithm proposed by Taube (2000) was used to calculate the lake volume change. The calculation formula can be expressed as where ΔV is the volume change of the lake, WL 0 and LA 0 are the initial water level and area of each lake during the study period, and WL 1 and LA 1 are the water level and area data during the study period. In the calculation of volume, the area data closest to the time of each period water level data acquisition was selected.
Lake precipitation and evaporation: P Lake and E Lake Annual precipitation and evaporation over the lake water surface area were based on the pan data (Ikebuchi et al., 1988;Abtew, 2001). The ratio of lake evaporation to pan evaporation of Φ 20 cm small evaporating dish to E601 large evaporating dish is 0.85; the conversion coefficient of evaporation from E601 evaporating dish to large water area (lake, etc.) is 0.9 and 0.91. Therefore, this study assumes that the annual pan coefficient is 0.905 as the conversion coefficient (Fu et al., 2004;Lowe et al., 2009). Therefore, the conversion coefficient of Φ 20 cm evaporating dish to large area water area is 0.85 × 0.905 = 0.769.

Geographical detector
The geographical detector is a statistical tool used to detect the spatial separation of relevant elements and reveal the driving force behind it (Wang et al., 2010. Based on the theory of spatial differentiation, the dependent variable and different discrete independent variables are detected on the same spatial scale. If the independent variable has an important influence on the dependent variable, the spatial distribution of the two variables is similar. The geographical detector contains four different geographical detectors: risk, factor, ecological, and interaction. In this paper, we used the factor detector and the ecological detector.
The evaluation index of the factor detector is q, which can evaluate the heterogeneity of spatial elements and the explanatory power of detection factors, analyze the interaction between two variables, and find the explanatory power of two variables. The range of q value is 0-1, which indicates the relationship between the independent variable x and dependent variable y. The larger the value, the greater the impact.
where i = 1,2,3……m refers to the classification of independent variable x, n i is the number of units corresponding to layer i, n is the total number of Environ Monit Assess (2021) 193: 219 219 Page 6 of 17 units in the study area, and i 2 and 2 are the variance of Y values in the i layer and the whole region, respectively.
The ecological detector is used to compare the influence of two factors X 1 and X 2 on the spatial distribution of attribute y, which is measured by F-statistics: where N X1 and N X2 represent the sample size of X1 and X2, respectively; SSW x1 and SSW x2 represent the sum of the intralayer variance of the stratification formed by X1 and X2, respectively; and L1 and L2 represent the number of variables X1 and X2, respectively. Among them, if H 0 : SSW X1 = SSW X2 is rejected at the significance level of α, it indicates that there is a significant difference between the two factors X1 and X2 on the spatial distribution of attribute y. In this paper, α is 95%, "Y" means there is a significant difference, and "N" means there is no significant difference.
In this paper, the lake area (LA), water level (WL), and lake volume change (ΔV) of Hongjiannao Lake from 1988 to 2015 are taken as dependent variables "Y." NDVI, human population (HP), sheep population (SP), the industrial output value (GVIO), annual precipitation (AP), and annual average temperature (AT) are taken as independent variables "X," and the main driving factors of Hongjiannao Lake are calculated.

Hydrogeological model
The isohydrograph of groundwater shows that the Hongjiannao Lake groundwater system is a closed independent groundwater system with high surrounding and low center (Fig. 3). After the groundwater is replenished by the atmospheric precipitation, it moves from the watershed to the lake along the terrain, and finally discharges into the Hongjiannao Lake, the infiltration of atmospheric precipitation forms phreatic water and shallow confined water, and supplies surface water (Yin et al., 2008). Therefore, Hongjiannao Lake had water all year-round, with weak seasonal change of water area and high correlation with dynamic change of groundwater (Ma et al., 2015). As a result of this closed hydrogeological structure, Hongjiannao Lake becomes a concentrated drainage area of groundwater.

Precipitation and evaporation
The calculated annual average precipitation and evaporation on the lake water surface showed high variability from 1986 to 2019, but represent similar behavior (Fig. 4). The precipitation ranged from 0.00948 km 3 in 2000 to approximately 0.02733 km 3 in 1995 with the mean value It can be seen that the whole change trend process of Hongjiannao Lake can be divided into different stages. The calculated annual average lake area, water level, and lake volume change showed similar behavior and high variability from 1986 to 2019. To more intuitively represent the stage changes, we divided the whole period of 1986-2019 into three different stages: stable (1986-1997), reduction (1998-2015), and growth (2016-2019). In the stable stage (1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997), the lake area changed from 57.25 to 54.13 km 2 with the average rate of change was −0.26 km 2 /a, the water level changed from 1211.25 to 1210.67 m with the average rate of change was −0.0483 m/a, and lake volume changed from −0.02158 to −0.03173 km 3 with the average rate of change was −0.0009 km 3 /a. In the reduction stage (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015), the lake area changed from 53.13 to 31.41 km 2 with the average rate of change was −1.21 km 2 /a, the water level changed from 1210.47 to 1206.11 m with the average rate of change was −0.2422 m/a, lake volume changed from −0.04285 to −0.22425 km 3 with the average rate of change was −0.0101 km 3 /a. In the growth stage (2016-2019), the lake area changed from 31.75 to 37.28 km 2 with the average rate of change was 1.38 km 2 /a, the water level changed from 1206.18 to 1207.26 m with the average rate of change was 0.27 m/a, and lake volume changed from −0.22227 to −0.18694 km 3 with the average rate of change was 0.0088 km 3 /a.

River inflow and runoff change
From the water balance formula of Hongjiannao Lake, the input included precipitation, river inflow, and groundwater recharge, while the river inflow all come from the seven rivers runoff into the lake (Table 1). In "Lake area, water-level, and lake volume," we divided the changing trend of the lake area, water level, and lake volume changes into three stages; here we undertake the different stages to analyze the spatio-temporal variation of the seven rivers change in Hongjiannao Basin. In the stable stage (1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997) (Fig. 6(1)), seven rivers including Zhasake River, Donghulusu River, Qibosu River, Songdaogou River, Erlintu River, Haolai River, and Malian River were all inflow into the lake. In the reduction stage (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015), we can further divide it into three periods according to the situation of river cutoff. In 2006, the Zhasake Reservoir was built in the upper reaches of the Zhasake River, which blocked the runoff supply for Hongjiannao Lake. Since then, Zasak River disappeared from the satellite image. In 2009, an underground reservoir was built in the upper reaches of Donghulusu River to intercept another supply runoff source for Hongjiannao Lake.
After 2009, Donghulusu River gradually disappeared on satellite images. Therefore, the reduction stage can be further divided into three periods: (1) 1998-2005 ( Fig. 6(2)), (2) 2006-2009 ( Fig. 6(3)), (3) 2010-2015 and (Fig. 6(4)). In 2016, Hongjiannao Lake was promoted to a National Nature Reserve of China; the local government agreed to make an ecological water supplement for Hongjiannao Lake by Zhasake Reservoir each year (Fig. 6(5)). That was the main reason that the curve of the lake area, water level, and lake volume raised when it reached the bottom of the valley in 2015; this trend was consistent with the growth period of the lake in 2016-2019.
Lake-groundwater exchanges Figure 7 shows the annual variation trend of the lake area and the surface runoff by the river inflow. The simulations indicated that the annual average lake area and surface runoff had high consistency, which consistent with the trend of different stages change. Table 2 shows the calculated components of the water balance model for Hongjiannao Lake in three different stages. In the stable stage (1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997), the average lake area was 54.34 km 2 , which kept at a high level, the water balance described as lake volume change (ΔV) = 0, which means that precipitation (P lake ) + river inflow (Q River ) + groundwater (Q Groundwater )-evaporation (E lake ) = 0. In this stage, the precipitation accounts for 25.83% (0.0203 km 3 ) of the input, river inflow accounts for 61.70% (0.0485 km 3 ) of the input, and groundwater accounts for 12.47% (0.0098 km 3 ) of the input. Groundwater recharge lake in the stable stage with the annual average Q Groundwater = 0.0098 km 3 . In the reduction stage, we based on the river inflow situation to analyze the water balance, from 1998 to 2005; the average lake area was 46.00 km 2 ; and we calculated the lake volume change (ΔV) = −0.0109 km 3 , according to the water balance: groundwater (Q Groundwater ) = lake volume change (ΔV) ± evaporation (E lake )-precipitation (P lake )-river inflow (Q River ), lake recharge groundwater with the annual average Q Groundwater = −0.0025 km 3 . From 2006 to 2009, the average lake area was 39.89 km 2 ; due to the block of the Zhasake River in 2006, the Q River reduced from 0.0475 km 3 in 1998-2005 to 0.0334 km 3 in 2006-2009, groundwater recharge lake with the annual average Q Groundwater = 0.0058 km 3 ; from 2010 to 2015, the average lake area was 1 3 33.41 km 2 ; due to block of Donghulusu River in 2009, the Q River further reduced to 0.0199 km 3 ; and the Q Groundwater = 0.0085 km 3 . In the growth stage (2016-2019), the average lake area increased to 35.18 km 2 , groundwater recharge lake with the annual average Q Groundwater = 0.0005 km 3 , the precipitation accounts for 46.63% (0.0187 km 3 ), river inflow accounts for 52.12% (0.0209 km 3 ) of the input, and groundwater accounts for 1.25% (0.0005 km 3 ) of the input. Compared with the stable stage, the groundwater supply in the growth stage reduced from 12.47% (0.0098 km 3 ) to only 1.25% (0.0005 km 3 ).
Annual lake water balance Figure 8 shows the annually estimated lake-groundwater exchange volumes across Hongjiannao Lake from 1986 to 2019. Most of the time in this period, the lake showed obvious gaining conditions (i.e., Q Groundwater > 0), discharge from groundwater to the lake, with a variable exchange volume of up to 0.03364 km 3 in 1993. The annual exchange volumes indicated that the Hongjiannao Lake experienced continuous losing conditions (1998)(1999)(2000)(2001)(2002)(2003)(2004), (i.e., Q Groundwater < 0), discharge from the lake to groundwater, with a variable exchange volume of up to −0.01582 km 3 in 1999. Unlike the freshwater lake system, (i.e., the Poyang Lake) , the losing condition was not much different from the gaining condition, suggesting that the groundwater supply for the lake input is a noticeable factor in affecting the lake water balance. Hongjiannao Lake is located in the arid and semi-arid areas in the northwest of China; with scarce rainfall and huge evaporation, groundwater recharge is essential to maintain the lake area.
Causes of changes in Hongjiannao Lake by using geographical detector From Tables 3, 4, and 5, it can be seen that the single dominant factor affecting the lake area, water level, and water volume change of Hongjiannao Lake from 1988 to 1997 was AT, and q values were 0.973, 0.969, and 0.964 respectively. At this stage, the single dominant factor was temperature. There were significant differences between the temperature and other factors in the changes in the lake area, water level, and water volume change, while there were no significant differences between other factors. Therefore, the change of Hongjiannao Lake from 1988 to 1997 was mainly affected by the natural factor temperature.
From 1998 to 2015, the single dominant factor of lake area had the strongest impact on GVIO followed by NDVI and SP, with q values of 0.985, 0.96, and 0.944, respectively (Table 6). NDVI and SP had a great influence on the area, and there was no significant difference between them. Among the single dominant factors of water level, GVIO was the strongest, NDVI and SP were the seconds, and q values were 0.979, 0.972, and 0.897 respectively (Table 7). There were significant differences between GVIO and HP, AP, AT, and SP, but no significant differences between GVIO and NDVI. NDVI was the most influential factor among the single leading factors of lake volume change, followed by GVIO and SP, and q values were 0.979, 0.97, and 0.897, respectively (Table 8). GVIO and NDVI had the greatest influence on the change of water quantity, and there was no significant difference between GVIO and NDVI, followed by SP and AT, and there was no significant difference between SP and AT. There were significant differences between GVIO and HP, AP, AT, and SP, but there were no significant differences between GVIO and NDVI. Therefore, the change of Hongjiannao Lake from 1998 to 2015 was mainly affected by human factors.
An understanding of the magnitude of the dynamics of Hongjiannao Lake's water balance is important for sustainable development not only in the local government management but in the energy policy implementation. Hongjiannao Lake is a typical lake as it is located in arid and semi-arid areas, and the lake-groundwater exchanges cannot observe directly. The volume of river inflow and as well as the volume of runoff change depends on the watershed system of the Hongjiannao Basin. Thus, the uncertainties associated with estimating river inflow and runoff change could contribute to the overestimation or underestimation of these components in water balance.
The results also showed that the calculation of water balance components can be used to better understand the situation of lake-groundwater exchange volumes. Catchment upstream of Zhasake Reservoir in 2006, the hydrological balance varied substantially before 2006 and after 2006, discharged from the lake to groundwater (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005), and discharged from groundwater to the lake (2006)(2007)(2008)(2009). In terms of inter-annual hydrological variability, there was a decrease of evaporation by 22.29% and a decrease of groundwater by 41.18% in 2016-2019 as compared to 2010-2015 and a 34.53% increase in precipitation and 5.02% increase in river inflow.
Besides, the industrial output value (GVIO) was considered to be the dominant factor of water level and water quantity change during 1998-2015. Liang and Yan (2017) pointed out that the total industrial output value (GIOV) of Hongjiannao Lake showed an upward trend from 1988 to 2015 and increased rapidly from 1999 to 2015. Hongjiannao is located in an area rich in coal resources and a key development zone, and coal mining is an important factor affecting its ecological environment. The coal chemical industry destroys groundwater resources, leading to a threat to groundwater recharge in Hongjiannao Lake (Liu et al., 2019).
Most of the time in 1986-2019, the lake showed obvious gaining conditions (discharge from groundwater to the lake), with a variable exchange volume of up to 0.03364 km 3 in 1993. From 1998 to 2004, Hongjiannao Lake exhibited continuous losing conditions (discharge from the lake to groundwater), with a variable exchange volume of up to −0.01582 km 3 in 1999.
The geographical detector showed that Hongjiannao Lake was affected by the natural factor temperature from 1988 to 1997, while it was affected by human factors from 1998 to 2015.
Funding The research is supported by the project of Key Laboratory of Mine Geological Hazards Mechanism and Control (KF2018-04), the Natural Science Basic Research Program of Shaanxi (2020JM-514), the National Natural Science Foundation of China (41401496), and Xi'an University of Science and Technology (2019YQ3-04).

Declarations
Competing interest The authors declare no competing interests.