Lake water storage changes and their cause analysis in Mongolia

Abstract

Lake is an important water resources in Mongolia, which has undergone a large variation in past decades. However, it is still challenging to monitor long-term changes in lake water storage (LWS) due to the lack of lake level monitoring and long-term satellite altimetry data for Mongolian lakes. According to the Advanced Land Observing Satellite Digital Elevation Model (ALOS DEM) and the Joint Research Centre (JRC) dataset, we estimated the LWS changes of 55 Mongolian lakes (> 10 km²) from 1991 to 2020. The results showed that the LWS increased by 40.24 km³ from 1991 to 1997, especially for northwestern Mongolia with 31.47 km³. However, the LWS decreased by 32.44 km³ from 1998 to 2010, and the lakes in the northwestern and southern decreased by 20.24 km³ (62%) and 7.38 km³ (23%), respectively, and then the LWS continued to decrease by 10.22 km³ from 2011 to 2020. The precipitation was the primary cause of lake change, which could explain 45.13% of LWS change based on Generalized Linear Model, followed by temperature (26.33%) and irrigated area (16.72%). Analyzing the changing characteristic and driving mechanisms of LWS can provide a scientific basis for local water resources management and planning.

Introduction

Mongolia is one of the largest inland countries in the world, which is located in the hinterland of Asia¹. The rate of warming in Mongolia is higher than the global^2–4, the average annual temperature in Mongolia showed a rapid upward trend during 1982–2015, with the highest annual temperature rising rate of 0.06 °C/a⁵. The average annual precipitation of Mongolia changed from 231.0 mm in 1961–1990 to 228.0 mm in 1991–2016, and about two-thirds of the annual precipitation falls between June and August⁶. Mongolia is particularly responsive to global climate change, with various land degradation symptoms associated with regional climate change^7,8. There was an imbalance in Mongolia’s ecosystem and a reduction in water resources, because of rising temperature and decreasing precipitation since the 1960s⁹. For another, changes in water resources can also feedback to regional microclimate and ecological environments, such as the degradation of grasslands and the deterioration of the regional ecological environment^10–13. Most of the surface water resources in Mongolia are contained in lakes (500 km³), which is the largest water resource, followed by glacier (19.4 km³) and groundwater (10.8 km³), respectively^14–18. Lakes play a significant role in regional water resources, nutrient cycling, human life, irrigated agriculture and economic development, and also a crucial component of wetlands, protecting some endangered species and playing a vital role in maintaining ecosystem diversity. Therefore, dynamic monitoring of lake change is essential to provide a scientific basis for water resources management^19–21.

Lakes are known as sensitive indicators of climate change, which increase or decrease in response to climate change^22–24. Lake monitoring provides important information on the seasonal or interannual water balance, which is helpful to research climate change. With the development of remote sensing technologies, high-quality DEM data supports the development of lake science effectively^25–27. Since 1990, numerous researchers have carried out extensive research on the lakes of Mongolia²⁸. Some researchers have explored the paleogeography of Mongolia and paleoclimatic changes through lake sediments^29–31. In addition, research on lake water conditions and trends in lake area and water level are the current research hotspot in Mongolian³². For instance, Sumiya et al. (2020) revealed that the most critical factors affected the Ugii Nurr Lake were the total annual evapotranspiration (r = –0.64) and the total annual river runoff (r = 0.67)³³. Dorjsuren et al. (2024) suggested that the warmer temperatures had led to a decrease in precipitation, which had caused a decrease in river runoff, and affected lake shrinkage in the Great Lakes Depression Basin in western Mongolia³⁴. Orkhonselenge et al. (2018) proposed that productive grazing activities had caused the decreased area of three large lakes (Ulaan Lake, Orog Lake, and Böön Tsagaan Lake) over the past 45 years in southern Mongolia³⁵.

At present, most of the studies on Mongolian lakes are focus on lake area changes with five-year or multi-year intervals. Meanwhile, the studies on LWS changes are only focus on a few large lakes, and it is still lack to analyze the continuous spatial and temporal changes of LWS of Mongolian lake groups. Because the topography of different lakes varies greatly, the change in lake area does not reflect the absolute change of lake, while the change in LWS can reflect the real change of lake. Therefore, in this study, we extracted the lake area using the JRC dataset, and estimated the changes of lake area and LWS from 1991 to 2020 using ALOS DEM and JRC data, and analyzed the spatial and temporal characteristics of the changes of lake area and LWS. We also conducted an attribution analysis of LWS changes in Mongolia based on historical meteorological data (precipitation, temperature, and evapotranspiration) and anthropogenic data (coal production, effective irrigated area, and grazing intensity). This research is conducive to explore the dynamics of the lake, providing a reference for hydrologists in scientific planning and allocation of water resources.

Study area

The topography of Mongolia gradually decreases from west to east, with mountains in the north-west, the Gobi Desert in the south-west, a flat area in the shape of an opening in the south-east, and relatively open terrain in the center and east, which shows large areas of hills and steppe landscapes¹. Mongolia is a vast country bordering China and Russia, with Russia to the north and China to the south. Most of Mongolia’s land is covered by steppe and desert Gobi. It is one of the largest landlocked countries in the world, spanning 41°35’ to 52°09’ N and 87°44’ to 119°56’ E, with an area of nearly 1,564,116 km². The longest distance from east to west is 2,392 km and the longest distance from north to south is 1,259 km. The average altitude is 1,580 m above sea level. The Gobi Desert is found in the south of Mongolia, and the steppes and desert steppes extend northwards to the central and eastern parts of the country⁶.

We analyzed the 55 lakes (larger than 10 km²) in Mongolia regarding the JRC dataset. To analyze the spatial variability of changes in LWS, we divided the study area into four parts based on basin boundaries (hydrographic basin boundaries are derived from the three levels of sub-basins in HydroBASINS)³⁶. As shown in Fig. 1, the four regions are located in the western (A), central (B), southern (C), and eastern (D) regions, respectively, the lakes are mainly distributed in the central and western regions of Mongolia (figures were plotted through ArcMap 10.0).

Fig. 1.

The distribution of lakes in Mongolia.

Data and method

Data

Lake area

The water body data are derived from the global water body dataset, which was published by the European Union Joint Research Centre (JRC) in 2016. The JRC dataset was obtained by extracting the global surface water changes over the last 32 years using all Landsat images, covering a total of 2 million images from the Landsat series, and the spatial resolution of the JRC dataset is 30 m. We used the JRC dataset to analyze the lake area change from 1991 to 2020.

ALOS DEM

The quality of digital elevation models (DEMs) and their spatial resolution are vital parameters. The DEM data was derived from ALOS (Advanced Land Observing Satellite-1). ALOS DEM data was derived from the PALSAR sensor observed from 2006 to 2011, and PALSAR’s L-band synthetic aperture radar (SAR) produced a large number of all-weather observations. Both the spatial resolution and vertical accuracy are 12.5 m, and ALOS DEM data are freely available through this website (https://search.asf.alaska.edu/).

The topography data around the lakes were obtained from ALOS DEM data. The lake area in Mongolia showed a decreasing trend since 1991, reaching a minimum and then a partial lake expansion around 2006–2009. The ALOS DEM data recorded the earth’s surface elevation data during the period of 2006–2011, and the lake surface area in Mongolia reached its minimum around 2006–2009. Therefore, the ALOS DEM records the topographic data when lakes of Mongolia at the minimum size. We established the relationship between lake area and lake water level based on ALOS DEM, and the relationship between lake area and lake water storage change using an empirical equation.

CRU dataset

CRU dataset is one of the most widely used climate datasets and is produced by the NERC Centers for Atmospheric Science (UK) (NCAS), and provides monthly data at 0.5° resolution covering the land surface globally from 1901 to 2020. The dataset is currently updated to have 10 sets of data based on near-surface measurements for the years of 1901–2020, included temperature (mean, minimum, maximum, and diurnal), precipitation (total, number of rainy days), humidity (e.g., vapor pressure), number of frosty days, cloudiness, and potential evapotranspiration³⁸. The dataset is generated using Angular Distance Weighted interpolation, based on monthly gridded data from the National Weather Service, and other external proxies based on monthly observations calculated from daily or sub-daily data. In this study, we used the CRU dataset to analyze the climate change.

The meteorological data used in our study included the annual mean temperature (AMT), annual precipitation (AP), and annual evapotranspiration (AE) to analyze climate change in Mongolia. The temperature and precipitation were obtained from the University of East Anglia Climate Research Unit (CRU TS V4.01) dataset (http://www.cru.uea.ac.uk/data/).

Terra climate dataset

Terra Climate is a gridded dataset of monthly climate on the global land surface from 1958 to 2022, which also produces monthly surface water balance datasets using a water balance model that incorporates reference evapotranspiration, precipitation, temperature, and interpolated plant extractable soil water capacity. These data provide important inputs for ecological and hydrological studies at global scales that require high spatial resolution and time-varying climate and climatic water balance data³⁹. The annual evapotranspiration data are acquired from the Terra Climate dataset, which uses climate-assisted interpolation to combine high spatial resolution climate normal from the World Climate dataset with more spatially resolved, but time-varying data from CRU Ts4.0 and the Japan Reanalysis of 55 years (JRA55).

Anthropogenic activity indicators

The anthropogenic indicators included coal production, effective irrigated area, and grazing intensity, and we analyzed the impact of anthropological activity on the LWS change. The effective irrigated area and grazing intensity were acquired from the Food and Agriculture Organization of the United Nations database (faostat.fao.org), and the coal production were acquired from the US Energy Information Administration (http://www.indexmundi.com/).

Method

Lake area from JRC

GEE is a cloud-based computing platform dedicated to satellite imagery and other Earth observation data, which hosts satellite imagery and stores it in public data archives, including historical Earth imagery dating back more than 40 years. The daily ingested images can be used for data mining on a global scale, including publicly available free image collections such as Landsat and Sentinel series, geospatial, and other environmental datasets. We obtained historical data for lake area from the JRC water body dataset during 1991–2020, and obtained the water body and classification bands (as shown in Fig. 2). After acquiring the water classification bands, the data was converted from raster to vector data, and the lake area was calculated based on the vector data for each year.

Fig. 2.

Flowchart for estimating the changes in LWS.

Estimating the change of LWS based on ALOS DEM

The lake surface area in Mongolia reached its lowest value during the period of 2006–2009, indicates the ALOS DEM could record the topographic data when the Mongolian lakes at the lowest water level. We obtained the topography data and the relationship between the lake level and lake area from ALOS DEM, and then estimated the lake level by combining with lake area. The specific steps of estimating the lake water level including: firstly, the lake boundary is obtained when the lake area is the largest, and then the buffer zone analysis ranging from 1 to 5 km is performed according to the largest lake boundary; then the ALOS DEM data is masked and converted to vector data; finally, the Grid code (elevation) field of the vector data are fused using the dissolve tool of ArcGIS to obtain the correspondences between lake water level and area. The results showed that the lake level in Mongolia showed a decreasing trend from 1991 until around 2009 when the lake level reached a minimum and then partially recovered.

Changes in LWS can be estimated using the relationship between lake area and lake water level, which is possible to estimate LWS changes from satellite image quickly and easily while avoiding challenging fieldwork. The ALOS DEM provides topographic data above the lake surface around 2006 to 2009, the increase in lake area could be estimated when the lake surface rises by 1 m or several meters, and the relationship between the lake area and the LWS change can be established. Therefore, the change in LWS can be estimated when the lake area is known. We estimated the lake waters storage from 1991 to 2020 by combining with annual lake area data from JRC dataset in Mongolia. This method to estimate lake water storage change is consistent with Yang et al. (2017) to estimate the LWS changes⁴⁰.

As shown in Table (1), we extracted the water level based on the ALOS DEM and obtained the lake area data based on the JRC water dataset. When the lake area and lake water level were known for two periods (second and third column of Table 1), LWS change with 1 m interval (fourth column of Table 1) was estimated based on the empirical equation (Eq. (1)); and by summing all the LWS change of 1 m interval, LWS change of this period relative to 2007 (1847 m) was also known (fifth column of Table 1). For example, LWS change of 1850 m to 1847 m (0.5288 km³) is to sum LWS changes with 1 m intervals of 1848 to 1847 (0.1743 km³), 1849 to 1848 (0.1762 km³) and 1850 to 1849 (0.1783 km³). After this determination, a linear relationship between the lake area and LWS change was established based on the third and fifth columns of Table 1. LWS change could be estimated relative to 2007 when the area was known based on the linear relationship.

1

Table 1.

The relationship between the lake area and water storage change from ALOS DEM.

Elevation	Area of each elevation value (km2)	Lake surface area(km2)	LWS changes with 1 m interval (km3)	LWS change relative to 1847 m(km3)
1847	173.5293	173.5293	0	0
1848	1.5839	175.1131	0.1743	0.1743
1849	2.2025	177.3156	0.1762	0.3505
1850	1.8889	179.2045	0.1783	0.5288
1851	1.9412	181.1457	0.1802	0.7090
1852	1.8577	183.0034	0.1821	0.8910

where S₁ and S₂ represent the lake areas of the two periods, △h represents the lake level change during the two periods, and △V represents the LWS change during the two periods.

Calculation of lake impact factor contribution

The generalized linear model (GLM) is an extension of the general linear model. The general linear model describes the linear dependence between one or more independent variables, requires the dependent variable to be continuous and obey normal distribution, and obtains the estimated value of the dependent variable through the linear predicted value of the independent variable. However, in practical applications, many data do not fully satisfy the application conditions of the general linear model. The GLM is a generalization of linear regression models, which emerges to overcome the shortcomings of linear regression models. Therefore, in order to further extend its application, the GLM method is proposed by transforming the expected value of the dependent variable Y into a nonlinear function. The GLM is used to quantify the relative contributions of climatic and anthropogenic factors to lake change. Since the study area covers a vast area, this study divides the study area into four parts and obtains the meteorological data (1991–2020), and the anthropological activity data. We quantified the influence of climatic factors as well as anthropological activity factors on lake water storage change in Mongolia, and we conducted multiple GLM analyses of six variables in each sub-region to quantify the relative contribution of each factor to LWS change.

Results

Lake area changes

In this study, the changes in the total lake area and LWS were estimated (as shown in Fig. 3), the results showed that the general trend of the lake area and LWS in Mongolia was approximately the same with an overall increase and then decrease. A total of 55 lakes (> 10 km²) were researched, there were 32 lakes in Region A, 12 in Region B, 2 in Region C, and 9 in Region D. The total lake area in Mongolia was 12610.6 km² in 1991, and the total lake area in regions A, B, C, and D was 8423.21 km², 3025.44 km², 349.98 km² and 811.98 km², respectively.

Fig. 3.

Total lake surface area and LWS changes from 1991 to 2020.

There are four large lakes, Uvs Lake (3465.80 km²), Kyrgyz Lake (1503.20 km²), Halusu Lake (1024.80 km²), and Khuvsgul Lake (2607.70 km²), which are larger than 1000 km², and the former three lakes are located in the northwestern part of the study area (region A), while Khuvsgul Lake is located in the northern part of Mongolia (region B). Most of the lakes in Mongolia range in size from 10 km² to 100 km² and most of these lakes are located in regions A and D. The area of the four large lakes varied within 100 km².

In Mongolia, the total lake area showed an increasing trend until 1997, and then decreased until 2020. The total lake area showed an increasing trend with a rate of 110.39 km²/yr during 1991–1997 with a fluctuating increasing trend. The total lake area decreased at a rate of 32.69 km²/y with a steady downward trend during 1998–2010, and then decreased by 180.73 km² from 2011 to 2020. The fastest of lake decrease was occurred in the period of 2007–2015 (27.30 km²/yr), and the change the lake area from 2015 to 2020 was smallest. The total lake area had increased by 89.33 km² in 1991 to 2020.

Time series characteristics of LWS change

We divided the three decades into three periods based on the general characteristics of lake changes, included 1991–1997, 1998–2010, and 2011–2020. The LWS changes for these three periods were shown in Fig. 4. 43 lakes showed an increasing trend in LWS change during 1991–1997. The LWS in Region A increased by 31.47 km³ (78%), with an increase of 8.37 km³ in Lake Gilchrist, one of the largest increases of LWS. Most of the lakes showed a trend of shrinkage and a consequent decrease in LWS during 1998–2010, with a greater decrease in regions A and C by 20.24 km³ (62%), 7.38 km³ (23%), respectively, and Region B and C decreased by 2.65 km³ and 2.17 km³, respectively. The largest decrease during this period was Khyargas Lake (region A) with 7.58 km³, followed by Böön Tsagaan Lake (region C) with 7.22 km³. The lakes continued to shrink, with a decrease of 5.57 km³ in Region A and 2.44 km³ in Region C during 2011–2020. Region A contained the largest lakes and the largest number of lakes, the change in lakes in Region A dominated the whole study area. The detailed information about the lake area and LWS change in each region is shown in Table 2.

Fig. 4.

LWS changes in the different regions from 1991 to 2020. (a) 1991–1997; (b) 1998–2010; (c) 2011–2020; (d) 1991–2020.

Table 2.

Lake area and LWS change in each region.

Period	Region A		Region B		Region C		Region D
	AC (km2)	WSC (km3)	AC (km2)	WSC (km3)	AC (km2)	WSC (km3)	AC (km2)	WSC (km3)
1991–1997	337.4	31.47	203.93	1.82	55.72	4.73	65.31	2.22
1998–2010	-228.2	-20.24	-59.43	-2.65	-45.09	-7.38	-59.57	-2.17
2011–2020	-67.26	-5.57	-10.51	-1.32	-82.94	-2.44	-20.02	-0.89
Sum	41.94	5.56	133.98	-2.14	-72.31	-5.09	-14.28	-0.84

Spatial characteristics of LWS change in each region

Region A is located in the northwestern part of Mongolia. The total lake area had increased from 1991 (8423.21 km²) to 2020 (8465.15 km²) with an increase of 41.94 km², and the total LWS had increased by 5.65 km³. The lake area and LWS increased by 337.40 km² and 31.47 km³ during 1991–1997, respectively. The lake area and LWS decreased by 125.82 km² and 12.42 km³ during 1997–1999, respectively. The LWS had shown a steady decline during 1999–2006 with a decrease of 0.60 km³, and decreased by 8.58 km³ during 2006–2010. The lake area and LWS had decreased by 28.45 km² and 6.87 km³ during 2010–2015, respectively, and decreased by 38.82 km² and 2.96 km³ during 2015–2020, respectively. The increasing lake area (337.4 km²) during 1991–2020 was mainly attributable to large lakes: Uvs Lake and Kyrgy Lake, which had increased by 149.74 km² (44%) and 89.72 km² (27%), respectively.

In Region A, the decreasing area and LWS of large lakes (> 1000 km²) contributed much more than that of small and medium-sized lakes. There were 19 lakes in region A showed an overall decrease in LWS, the largest shrinkage of those lakes was Kyrgyz Lake (1.86 km³), followed by Eugene Lake (1.15 km³). The largest increase in LWS was in Ubusu Lake, which had increased by 8.25 km³ during 1991–2020. 13 lakes were expanded, 9 lakes of them were located at the boundary of the basin, and 4 lakes were located inside region A.

Region B is located in the northern part of the study, and there are 12 lakes with a total area of 3025.44 km² (1991), and the area of Khuvsgul Lake was 2607.70 km² in 1991. The lake area in region B had increased rapidly by 203.93 km² and the LWS increased by 1.82 km³ during 1991–1997. The lake area and LWS had decreased by 59.43 km² and 2.65 km³ during 1998–2010, respectively. The lake area and LWS showed a decreasing trend during 2010–2020, and the lake area had decreased by 10.51 km² and 1.32 km³, respectively. The total LWS had decreased by 2.14 km³ during 1991–2020. 7 lakes in region B were shrinkage, which were located the southern part of region B and decreased by 2.72 km³. The largest decrease in LWS was Sanglin Dalai Lake (−0.07 km³ /yr). 5 lakes in region B increased in LWS, and the largest increase was Khuvsgul Lake with 0.50 km³.

Region C is located in the southern Gobi region of Mongolia, and the total lake area was 349.98 km² (1991). The number of lakes in Region C is small, there are only two lakes larger than 10 km², named Olog Lake and Böön Tsagaan Lake. The lake area and LWS had increased by 43.27 km² (12%) and 1.20 km³ during 1991–1993, respectively. The lake area and LWS had decreased by 62.62 km² and 1.31 km³ during 1993–2004, respectively. The total lake area showed a decreasing trend compared to 1991 during 2004–2008. The lake area and LWS had increased by 78.83 km² and 0.46 km³ during 2008–2011, respectively. The lake area and LWS had decreased by 82.94 km² and 2.44 km³ during 2011–2020, respectively. The lake area of Böön Tsagaan Lake ranged from 240.15 to 283.58 km², and the LWS fluctuated from 7.6 to 8.9% relative to 1991. Compared to the changes of Lake Böön Tsagaan, the Olog Lake was more variable. Both the two lakes in Region C were shrinkage, with a total decrease of 5.09 km³. Olog Lake decreasing by 3.56 km³, and Böön Tsagaan Lake decreased by 1.54 km³.

Region D is located in the eastern of Mongolia with relatively flat, and there are 10 lakes in this region. The total lake area was 811.98 km² in 1991 and increased to 877.29 km² in 1997, with an increase of 65.31 km² and 2.22 km³. The lake area had decreased by 39.47 km² during 1997–2003, and then the lake area decreased by 31.14 km² and the LWS decreased by 0.27 km³ during 2004–2012. The lake area continued to decrease, and the lake area and LWS had decreased by 21.8 km² and 0.05 km³ during 2012–2015, respectively. Three of the lakes in Region D were in a state of retreat and their LWS was decreasing, the largest decrease were Khukh Lake and Bel Lake in the eastern part of region D, with a decrease of 0.25 km³ and 0.71 km³, respectively. 6 lakes in region D showed an increasing trend in LWS.

The changing trend of LWS was as follows: 5.65 km³, −2.14 km³, −5.09 km³, and −0.84 km³ for regions A, B, C, and D (As shown in Table 2), respectively. It was obvious that the increasing LWS only occurred in Region A, while the LWS in the other three regions was decreasing, and the largest decrease in LWS was in Region C. The changing rate of LWS in was 0.19 km³/yr, −0.07 km³/yr, −0.17 km³/yr and −0.03 km³/yr for Region A, B, C and D, respectively.

Changes in climatic elements during the study period

Mongolia’s climate is characterized by long and cold winters, dry and hot summers, low precipitation, high temperature fluctuations, and a high number of sunny days per year. Over the last 20 years, Mongolia had experienced a rapid increase of temperature, and increasing rate was about three times of the global average rate of temperature increase (0.06 °C/a), and precipitation had been on a downward trend with a rate of −1.40 mm/a⁴¹.

As shown in Fig. 5, the AMT in regions C and D was much lower than that of regions A and B, reflected that the spatial variability of temperatures in Mongolia. The AMT increased in all regions during 1991–2000, which was consistent with the trend of increasing temperatures in the AIPP study program in Mongolia in the 1990s. The AMT in the study was negative from 1991 to 1994, with a value equal to 0.16 °C in 1995 and a rapid rise in AMT from 1997 to 2000, with a rate of 0.50 °C/yr, which was consistent with the IPCC report findings that the 10 hottest years measured globally in the last 120 years all occurred after 1980, with six of those years occurring after the 1 °C anomaly was observed for three consecutive years in 1997, 1998 and 1999. Moreover, the number and duration of hot weather events increased. The AMT was greater than 0 (0.11–0.82 °C), showed a small fluctuating upward trend during 1998–2002. The AMT showed a rapid upward trend during 2003–2008, reaching a maximum value in 2008 (1.80 °C), during which the AMT increased at a rate of 0.53 °C/yr. The AMT showed a decreasing trend after 2008, and AMT reached the minimum value (−1.14 °C) in 2012. During the period of 2013–2018, the AMT gradually increased and reached to 0.8 °C (2018). Overall, AMT values had shown an upward trend over the past three decades, with Mongolia’s temperature trends remaining consistent with global temperature trends, and we examined an overall rate of increase in AMT over the past three decades of 0.028 °C/yr. All of four regions showed varying degrees of temperature increase during 1991–2020, with 0.04 °C, 1.07 °C, 0.42 °C and 1.41 °C for region A, B, C and D, respectively. The average AMT was ranked in the four Regions: D > B > C > A, with the largest increase in temperature in Region D and the smallest in Region A.

Fig. 5.

Changes in climate factors and human activities in Mongolia during 1991–2020. (a) AMT (annual average temperature), (b) AP (annual precipitation), (c) PET (potential evapotranspiration), (d) area of irrigated, (e) number of sheep and goats, and (f) coal production.

Previous studies had shown an increase in temperature since the late 1980s and a general decrease in annual precipitation since the early 1990s. We analyzed the spatial distribution of annual precipitation for four regions, which showed a small decreasing trend. The average AP in the past 30 years for regions A, B, C, and D was 229.36 mm, 360.16 mm, 120.42 mm, and 250.83 mm, respectively. The maximum annual precipitation in Region A was in 1993 (334.22 mm) and the minimum annual precipitation was in 2010 (174.44 mm). For Region B, the maximum AP was 463.54 mm in 1994 and the minimum was 279.72 mm in 2002, the difference was 183.82 mm, and a change of 47% relative to the annual precipitation in 1991. For region C, the maximum and minimum AP were 172.99 mm (1994) and 80.53 mm (2002, 2010), respectively, and the difference was 92.46 mm, representing a 77% change relative to 1991. For region D, the maximum AP was 335.96 mm in 2014, and the minimum was 170.60 mm in 2008, with a difference of 165.36 mm and a 60% change relative to 1991. The magnitude of annual precipitation variation in the four Regions was C > A > D > C, and the highest annual precipitation was Region B.

The results revealed that the average annual evapotranspiration of Regions A, B, C, and D during 1991–2020 was 124.87 mm, 231.24 mm, 88.33 mm, 162.25 mm, respectively. The ratio of change relative to 1991 was 61%, 40%, 72%, and 59%, respectively, indicated that the average annual evapotranspiration in Region B was the largest and the ratio of change in Region B was the smallest. Although the annual evapotranspiration in Region C was small, the ratio of change in annual evapotranspiration in Region C was the largest and the fluctuation of change was also the largest.

We analyzed the zoning of meteorological factors in the study area, suggested that there were spatial differences in the climate of the study area (as shown in Fig. 6). There was a small downward trend in precipitation. Region C had the highest AMT and the least amount of precipitation with the greatest variability in precipitation. In addition, the evapotranspiration values fluctuated severely in region C. Kang et al. (2015) suggested that the average annual precipitation ranged from 110 to 329 mm with an average value of 203 mm during 1980–2008. Low latitudes showed a good negative correlation between average annual temperature and annual precipitation. The Gobi region was significantly hotter and drier than that of other regions (3.3 ± 1.70 °C, 150 ± 53 mm), and the climate in Region C was warm and dry. The warming was most pronounced in the high mountains and the valleys compared to the Gobi Desert. For region B, with an inter-annual average temperature of −2.27 °C, and average annual precipitation was 360.16 mm, and minimal variation in evapotranspiration, the climate in Region B was cold and wet. For region A, which is located in the northwestern part of the study area, the AMT was −1.95 °C, and the average annual precipitation was 229.36 mm, which was cold and humid. The AMT was 2.18 °C in Region D, and the average annual precipitation was 250.83 mm, and the evapotranspiration fluctuation rate was 59%, indicated that Region D was warm and humid.

Fig. 6.

Annual average climate distribution map of each region.

Changes in human activity

Coal mining is an extremely water intensive industry, cutting off rivers and destroying underground aquifers, consuming 2.54 m³ of water per ton of coal. We had obtained historical data on coal mining in Mongolia from the US Energy Information Administration (http://www.indexmundi.com/). There was an overall increasing trend in coal mining (as shown in Fig. 5). Mongolia’s coal production showed a rapid downward trend during 1991–1994, with coal mining declining by 2011 million tons (26%). Coal production showed a steady downward trend at a rate of 1 million tons per year from 1995 to 1998, with 54.96 million tons of coal produced in 1998. From 1999 to 2013, coal production began to recover, rising from 55.3 million tons in 1999 at a rate of 2.23 million tons per year, and 86.45 million tons in 2013. The coal production in 2013 increased by 9.38 million tons (12%) relative to 1991.

In recent years, some studies had shown a trend of expansion of arable land in Mongolia, and the use of groundwater and rivers for agricultural irrigation may be another important driving cause of lake shrinkage⁴². As shown in Fig. 5d, the effective irrigated area increased to 840,000 ha with a rate of 20,000 ha/yr during 1991–1993. The effective irrigated area showed a significant increasing trend, with the irrigated area expanding at a rate of 118,000 ha/yr during 1993–2012. The effective irrigated area showed a decreasing trend at a rate of 14,500 ha/yr during 2012–2017, and remained constant at 873,600 ha during 1997–2020. The irrigated area in 2020 increased by 73,600 ha relative to 1991, which was the same in 1999.

Overgrazing destroys the ability of grassland ecosystems to recover themselves, leading to the degradation of grasslands, which in turn leads to a decline in their soil function and water-holding function. The number of livestock increased at a rate of 580,000 head/yr and grazing intensity intensified during 1991–2020, putting pressure on Mongolia’s grassland recovery. The grazing intensity fluctuated upwards during 2000–2010, with livestock numbers increasing by 2,148,600 head/yr at a rate of 214,900 head/yr. According to the 2007 livestock census, livestock numbers had increased from 3.48 million to 4.03 million (15%) during 2010–2020. The total number of livestock was 57,794,200 in 2020, compared to 20,343,400 in 1991, an increase of 37,450,800 (184%) over the last three decades.

Driving factor analysis of lake change

According to the multiple GLM analyses, the overall attribution analysis of the LWS change showed that precipitation was the primary driving cause of the LWS change (As shown in Table 3), with a contribution rate of 45.13%. For region A, the contribution of precipitation to the LWS change was 46.44%, followed by evapotranspiration (18.93%) and temperature (15.76%), and the contribution of irrigated areas to LWS change was 12.44%, which was the most significant driving cause of the anthropogenic factors. For region B, the contribution of temperature to the change in LWS was 42.31%, followed by irrigated area (29.58%) and evapotranspiration (18.88%), irrigated area was the primary driving cause of the anthropogenic factor. For region C, temperature variation was the primary driving cause of LWS change (58.36%), followed by the irrigated area (22.6%), and evapotranspiration (8%). For region D, temperature (57.73%) was the primary driving cause of LWS change, followed by evapotranspiration (25.17%), precipitation (9.09%), grazing (4.9%)(Table 3.

Table 3.

Multiple GLM analyses of the LWS changes in Mongolia.

Region/(%)	AMT	AP	PET	Mining	irrigation	Grazing
A	15.76	46.44	18.93	3.86	12.44	2.58
B	42.31	4.29	18.88	2.18	29.58	2.75
C	58.36	3.17	8.00	1.36	2.26	7.03
D	57.73	9.09	25.17	2.09	1.03	4.9
Total	26.33	45.13	4.04	4.84	16.72	2.95

Discussion

Significance of research on LWS changes in Mongolia

The shrinking of lakes may bring a serious challenge to local ecosystems. The total lake area (Dalinor Lake, Gonzo Lake, and Doruno Lake) in the Dalinor Basin had decreased by 11.6% in the last 39 years (242.30 km², 1976; 214.30 km², 2015)⁴³. Lake shrinkage in the Dalinor Basin had caused significant ecological damage to Dalinor National Nature Reserve habitat quality, which was increasing the ecological risk to bird habitat. Lake shrinkage could lead to grassland degradation, and previous research had suggested that the area of Mongolia’s lakes had been significantly decreased from 1986 to 2015, and the lake area of Inner Mongolia had decreased by 34%, which larger than that of Mongolia (17.6%). The continuous dynamics of the lake area (> 1 km²) had been monitored on the Mongolian Plateau using the platform Google Earth Engine (GEE), revealing a significant shrinkage in the lake area from 1991 to 2009⁴⁴.

In recent years, many researchers had focus on the change of LWS by combining with multi-resources satellite data. The LWS fluctuation of Hulun Lake was estimated utilizing satellite altimetry data over 40 years, suggested that the LWS of Hulun Lake showed a rapid increase (1.67 km³/y) from 2009 to 2012⁴⁵. The LWS changes of six major lakes in Inner Mongolia were analyzed over the past 30 years, and showed that the LWS of these lakes decreased at different rates (Hulun Lake, 0.067 km³/yr; Dalinuoer Lake, 0.012 km³/yr; Daihai Lake, 0.021km³/yr; Hong jiannao Lake, 0.01 km³/yr; Hasuhai Lake and Wuliangsuhai Lake, 0.001 km³/yr for both)⁴⁶. Satellite data and water body datasets had been widely used in lake change research, and the combination of different satellite altimetry data and image was of great interest for monitoring the change of LWS^47–49.

However, previous researches had mainly focused on changes in the LWS of individual large lakes, and lacked analysis of LWS changes in a group of lakes in Mongolia. Research on changes in lake areas often cannot reflect the real changes in lakes, the estimation of LWS change combined with DEM data in this manuscript could reflect the real changes in lakes, which is useful to water balance analysis. This study combines DEM and Landsat image to estimate the LWS changes over the past decades, and provide basic data and scientific basis for water balance studies of lakes as well as local water resource management.

Difference analysis of surface topography

We research the variability of lake topography by analyzing the rate of change in lake area and the rate of change in LWS. The increase in lake area in 1991–1997 (662.36 km²) was about twice as large as the decrease in lake area in 1997–2002 (291.35 km²), but the LWS change in 1991–1997 (40.69 km³) was 10 times larger than that of the period in 1997–2002 (−3.77 km³). In addition, the decrease in lake area in 1997–2002 was 1.6 times of period of 2005–2008, but the change in LWS in 2005–2008 was 2.7 times of the period of 1997–2002. The results of the analysis showed that due to the differences of lake size as well as the topography around lake, the change in lake area and LWS showed different rates. Therefore, the lake area cannot reflect the “real lake change”, while the LWS change can reflect the real lake change. Estimating the changes of LWS and analyzing the spatial and temporal difference, which contribute to water resources management and development planning of local government.

ALOS DEM accuracy assessment

The use of bathymetry and altimetry data is an effective method of assessing the accuracy of water storage change estimation. However, given that the bathymetric data only covers part of the lake, there is some uncertainty in the estimation of lake water storage changes based on underwater topography. Therefore, the ICESat-2 data is able to provide high precision water level data, and the empirical equation constructed by combining ICESat-2 altimetry data with Sentinel-2 image to estimate water storage change with high accuracy, and compare the accuracy of lake water storage changes from ALOS DEM.

This research adopts a systematic approach to data comparison and analysis. Specifically, this research firstly acquired the ATL13 dataset from the ICESat-2 mission, which is specially designed for water body elevation measurements with high accuracy characteristics. In the data processing stage, the measurement data of six independent strips of ‘gt1l/‘, ‘gt1r/‘, ‘gt2l/‘, ‘gt2r/‘, ‘gt3l/‘, ‘gt3r/‘, including the precise latitude and longitude coordinates and the corresponding elevation values are extracted in a targeted manner. Subsequently, the above extracted ICESat-2 data were spatially filtered using prepared lake vector boundary data to ensure that only valid observations within the research lake area was retained. To further ensure the data quality, for the data of the same observation date, the Boxplot method was applied for outlier rejection, specifically retaining robust data points located between the 25% and 75% quantiles, thus effectively reducing the interference of noise and errors. Afterwards, an arithmetic mean was calculated for all the filtered elevation values in the six strips, and the result was used as a representative lake surface elevation for the area of the research lake. This step aims to improve the stability and reliability of lake surface elevation estimation through the comprehensive averaging of data from multiple sources.

Ultimately, this research fused the preprocessed ICESat-2 data with Sentinel-2 image to estimate the water storage changes of Dörgön Lake and Khyargas Nuur Lake, and evaluated the accuracy of water storage change estimation from ALOS DEM. We estimated the annual water storage change since 2018, and the results showed that the estimation error was relatively large for small changes in lake water storage and relatively small for large changes in lake water storage. Based on the comparison of ALOS DEM data with ICESat-2 results, the average error of lake water storage change estimation from ALOS DEM was about 15.7%.

Uncertainty analysis

Bendib et al. (2021) used the ALOS, SRTM1, and SRTM3 to estimate the water storage of Fountaine Des Gazelles Dam, and the results showed that ALOS DEM had the highest accuracy compared to SRTM DEM (NRMSE = 1.30%; MAPE = 1.5%)⁴⁹. However, some lakes in this research had small changes (< 1 km), which may impose large errors in the estimation of LWS changes. In this research, we used the ALOS DEM data in 2007 to estimate the LWS change, but the minimum area of some lakes was not in 2007, and we could not obtain the topographic elevation data corresponding to the minimum area of the lakes due to data limitation. The underwater topography of these special lakes was estimated using the principle of topographic similarity, and therefore, there were relative errors in the estimation of LWS changes.

Tao et al.‘s research on water resource changes in Mongolia showed that precipitation was the primary cause of lake changes in Mongolia, which changes in precipitation dramatically altered the size of the lakes⁹. In this research, we quantitatively analyzed the primary driving cause of LWS changes based on six factors, and the results were similar with the previous scholars, precipitation was the primary driving cause of LWS changes (45.13%). However, there were still uncertainties in the driving factors of LWS changes. There were many factors affecting LWS changes, and the changes in AMT, AP, PET, and the intensity of human activities will have different impact on LWS changes. If climate change is drastic in the future or the impact of human activities expand, the LWS will also change. Therefore, there is still uncertainty in the analysis of the driving factors.

Conclusion

In this study, we used the JRC water body dataset to obtain long-term lake area series, and applied ALOS DEM data to obtain the topographic data around the lake, and established the relationship between lake area and water storage change to estimate the LWS change from 1991 to 2020 in Mongolia. In addition, we combined climatic data and human activity data to analyze the causes of lake changes. The results indicated that the LWS in Mongolia showed an increasing trend with a total increase of 40.24 km³ during 1991–1997, and the largest increase was in the northwest part (region A) with 31.47 km³, and relatively small increase in regions B, C, and D (1.82 km³, 4.73 km³, and 2.22 km³). The LWS showed a significant decrease during 1998–2010, especially in the northwestern part (region A) of Mongolia and the southern Gobi region (region C), which decreased by 20.24 km³ and 7.38 km³, respectively. The LWS was still decreased, with a decrease of −5.57 km³, −1.32 km³, −2.44 km³, and −0.89 km³ in regions A, B, C, and D during 2011–2020, respectively. Based on the GLM model, combining with the data of precipitation, temperature, evapotranspiration, irrigation, coal mining, and grazing, the results showed that the primary cause of LWS change in Mongolia was precipitation (45.13%), followed by temperature (26.33%) and irrigated area (16.72%). The increase in LWS from 1991 to 2000 was directly related to precipitation, and the decline since 2000 was caused by an increase in average temperature.

Author contributions

H.H.Z. conceived the study together with B.J.Q. And H.H.Z. produced the figures and drafted the manuscript, while B.J.Q reviewed the manuscript. H.Y.L. and X.H.C. provided the fund support. Y.X. and G.F.W. Overall supervised the experiment. All authors read the manuscript and approved it.

Data availability

Data availabilityThe Climate Research Unit (CRU TS V4.01) dataset was freely available at http://www.cru.uea.ac.uk/data/. Data on the effective irrigated area and grazing intensity were from the Food and Agriculture Organization of the United Nations database (faostat.fao.org), and data on coal production were from the US Energy Information Administration (http://www.indexmundi.com/). The datasets used and/or analyzed related to lakes during the current study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.