Study on the Changing Trend of Terrestrial Water Storage in Inner Mongolia Based on GRACE Satellite and GLDAS Hydrological Model

Abstract

To address the challenges of water scarcity and the limited accuracy of terrestrial water storage (TWS) estimation in Inner Mongolia, this study integrates GRACE satellite observations, the GLDAS-Noah hydrological model, and ground-based precipitation records, in combination with Theil–Sen median trend analysis and the Mann–Kendall test, to systematically evaluate the spatiotemporal evolution of TWS from 2003 to 2016. The results demonstrate that: (1) GRACE data reliably capture regional water storage dynamics. Over the study period, TWS exhibited a significant overall decline, with an average rate of −5.2 × 10⁻⁴ cm/year, and seasonal variations were strongly coupled with precipitation patterns. (2) Spatially, TWS anomalies (TWSa) decreased from northeast to southwest, with values ranging from approximately +1.22 cm to −2.94 cm. The most pronounced decline was detected in the southern Ordos region. (3) Soil water changes were more substantial than those in canopy or snow water, with sharp reductions occurring during 2004–2007 and 2013–2015. Soil water exhibited clear stratification across different depths, and variations in deep soil water and groundwater were primarily influenced by non-precipitation factors. These findings provide a scientific basis for the sustainable utilization of water resources in Inner Mongolia and yield important insights for regional water management and policy formulation.

1. Introduction

Water resources are the core elements for maintaining the stability of ecosystems and ensuring the sustainable development of the natural environment [1]. In recent years, with the global population growth, the surge in industrial and agricultural water demand, and the intensification of water pollution, the contradiction between water supply and demand has become increasingly prominent and has become a key factor restricting regional development. In this context, the accurate estimation of water resources has become an important prerequisite for achieving scientific management and sustainable utilization of water resources. TWS is a key indicator for measuring the overall status of regional water resources. Its main components include groundwater, soil moisture, snow cover, lakes and rivers, and water contained in biomass [2]. It can reflect the comprehensive impact of regional precipitation, evapotranspiration, runoff and other factors on the total amount of water resources [3]. Traditional terrestrial water storage monitoring methods mainly rely on hydrological variables such as runoff, evaporation, soil moisture, as well as hydrological observation stations and hydrological models. However, these methods have obvious temporal and spatial limitations due to factors such as uneven site distribution, high observation costs, and uncertainty in model parameterization [4].

With the advancement of satellite remote sensing technology, the use of the Gravity Recovery and Climate Experiment (GRACE) satellite and the Gravity Recovery and Climate Experiment Follow-On (GRACE-FO) time-varying gravity field model to conduct dynamic monitoring of TWS in global and local areas has made it possible to accurately and efficiently estimate water resource changes [5]. At present, domestic and foreign scholars have used GRACE satellite data to explore the changing characteristics of terrestrial water storage and groundwater in the three northern regions [6], the northwest region [7], the Qinghai–Tibet Plateau [8], the entire Amazon basin [9], the Congo Basin [10] and other regions or basins around the world. GRACE(-FO) data have also been widely used by researchers both domestically and internationally to analyze changes in terrestrial water storage in various river basins, such as the Yangtze River Basin [11]. However, due to limitations in the performance of the GRACE(-FO) satellites, the spatial resolution of its products remains relatively coarse. Su et al. [12] proposed a combined modeling approach based on machine learning, but this approach is limited to certain geographical locations. In addition, there are studies that compare and contrast the results of GRACE inversion with measured data, hydrological models or precipitation data [9], verifying the reliability of this technology in regional and even global scale water storage monitoring, further promoting its application in global water cycle research.

The Inner Mongolia Autonomous Region is located in the arid and semi-arid region of northern China. It is controlled by a typical continental climate, and its regional water resources are characterized by insufficient total volume and uneven temporal and spatial distribution. At the same time, strong solar radiation and the continuous warming and drying trend have further aggravated the regional water deficit. The synergistic effect of these natural factors has led to a widespread water shortage in Inner Mongolia. In recent years, with the rapid economic development, ecological problems such as land desertification caused by excessive water resource development have seriously threatened the ecological security of the region. Therefore, scientifically assessing the water resource carrying capacity and optimizing water resource allocation are of great strategic significance for ensuring the sustainable economic and social development of Inner Mongolia and maintaining the stability of the ecosystem. Studies have shown that the TWS in northern China has shown a significant downward trend [13]. Feng et al. [14] by removing the simulated soil moisture changes from the GRACE-derived terrestrial water storage changes, revealing a continuous decline in groundwater in northern China from 2003 to 2010. Guo et al. [15] studied the spatiotemporal changes in TWS and groundwater storage (GWS) in Inner Mongolia from 2003 to 2021. Although these studies involve Inner Mongolia or parts of Inner Mongolia, there is still a lack of systematic research on the entire Inner Mongolia region that compares and contrasts the long-term changes in TWS with measured data, hydrological models, or precipitation data. Therefore, this study used GRACE satellite data and the Global Land Data Assimilation System (GLDAS) Noah hydrological model, combined with water resources bulletin data and precipitation data, to conduct a comparative study. Using the water balance equation, the authors determined the long-term spatiotemporal trends in TWSa and various hydrological components over Inner Mongolia. The results can provide a quantitative basis for scientific management of regional water resources and have important practical value for achieving the coordinated development of efficient water resource utilization and ecological protection.

2. Materials and Methods

2.1. Study Area

The Inner Mongolia Autonomous Region (97°100′ E–126°4′ E, 37°24′ N–53°20′ N) is located in the northern border of my country, bordering eight provinces and municipalities. It has a total land area of 1.183 million km² and an average elevation of approximately 1000 m (Figure 1). As a typical arid and semi-arid region, Inner Mongolia has a temperate continental climate, with an average annual temperature of approximately 6.2 °C and dramatic temperature fluctuations throughout the year. Winters are cold and long, with January average temperatures ranging from −28 to −14 °C. Summers are warm and brief, lasting only one or two months in most areas, with July being the hottest month, with average temperatures between 16 and 27 °C. Annual precipitation is scarce and unevenly distributed, concentrated primarily in June through August, accounting for 60–70% of annual precipitation. Average annual precipitation for the region is 300–400 mm, decreasing in a stepwise manner from northeast to southwest, exceeding 500 mm along the mountainous ridges and less than 100 mm in parts of the western Alxa League. In terms of hydrological characteristics, the outflowing river systems of Inner Mongolia Autonomous Region include the Yellow River, Hai River, Luan River, Liao River, Nenjiang River, and Ergun River. The outflowing river systems are mainly distributed in the southeastern part of the mountainous area from the Greater Khingan Range and the Yinshan Mountains to the Helan Mountains. The inland rivers include the Ulagail River, Qamdo River, Tabu River, Aibugai River, Huangqihai River System, Daihai River System, etc. [16].

2.2. Data

The multi-source dataset used in this study is centered on the Google Earth Engine (GEE) cloud platform [17], integrating GRACE satellite gravity observations, the GLDAS Noah v2.1 model assimilation dataset, and precipitation observations from ground-based meteorological stations. The data resources provided by GEE simplify data acquisition and enable a relatively straightforward and comprehensive analysis of Earth’s hydrological and energy cycles.

2.2.1. GRACE Satellite

GRACE data products include Level-1, Level-2, Level-3, and Mascon. This article uses the “GRACE Monthly Mass Grids-Land” Level-3 dataset, generated through a series of post-processing of Level-2 in the GEE platform. This dataset was generated by three different centers: the University of Texas at Austin Center for Space Research (CSR), the German GeoForschungsZentrum Potsdam (GFZ), and the Jet Propulsion Laboratory (JPL) of the National Aeronautics and Space Administration (NASA). The data were processed using Gaussian smoothing and harmonic filtering to enhance accuracy. The temporal resolution was monthly, and the spatial resolution was 1° × 1°. The temporal resolution of these datasets is monthly and the spatial resolution is 1°. All data are included in GEE. The values are expressed as Equivalent Water Height (EWH), which represents the depth of a uniform water layer distributed over the land surface, measured in centimeters (

Table 1. Data band properties of the GRACE monthly gridded land dataset.

Band Name	Unit	Range	Description
lwe_thickness_csr	cm	−5.13–4.44	Equivalent water height in centimeters calculated by CSR
lwe_thickness_gfz	cm	−6.49–3.53	Equivalent water height in centimeters calculated by GFZ
lwe_thickness_jpl	cm	−5.48–3.45	Equivalent water height in centimeters calculated by JPL

). For each GRACE product, the monthly TWSa is computed as the deviation from the mean TWS during 2004–2010 [18]. In this study, the ensemble mean of TWSa from the three processing centers (CSR, GFZ, and JPL) was adopted.

Furthermore, since this paper focuses on understanding changes in freshwater resources distributed over land, Mascon products were not used. GRACE Tellus Level-3 data are only available from April 2002 to June 2017. Since this study covers the period from January 2003 to December 2016, the “GRACE Monthly Mass Grids Release 06 Version 04-Land” released by NASA’s Jet Propulsion Laboratory was used as a reference. While the GRACE-FO satellite, a continuation of the GRACE mission, has been operational since its launch in 2018 and produces geodetic and gravity data, GRACE-FO data were not utilized or considered in this analysis.

2.2.2. GLDAS LSM Data

Global Land Data Assimilation System (GLDAS) was jointly established by NASA’s Goddard Space Flight Center (GSFC) and the National Centers for Environmental Prediction (NCEP). Its main purpose is to use advanced land surface models and data assimilation technology to extract satellite-based ground observation data products to generate optimal state variables and flux data for surface water system changes [19]. Currently, the most commonly used land surface models (LSM) include Catchment, Noah, CLM (Community Land Model), and VIC (Variable Infiltration Capacity). In this study, the data output by the GLDAS Noah 2.1 model (“GLDAS-2.1: Global Land Data Assimilation System”) was used. This model calculates water and energy storage fluxes on a global scale at a frequency of 3 h. It contains a variety of surface information, such as snow water equivalent, plant canopy water, and soil water. The soil water has four layers, with depths of 0–10 cm, 10–40 cm, 40–100 cm, and 100–200 cm below the surface, respectively. The spatial resolution is 1° and the temporal resolution is 1 month.

Table 2. GLDAS-2.1 Data Band Description.

Band Name	Unit	Range	Description
CanopInt_inst	kg·m−2	0–0.5	Plant canopy surface water
RootMoist_inst	kg·m−2	2–949.6	Root zone soil moisture
SWE_inst	kg·m−2	0–120,787	Snow depth water equivalent
SoilMoi0_10cm_inst	kg·m−2	1.99–47.59	Soil moisture at 0–10 cm
SoilMoi10_40cm_inst	kg·m−2	5.99–142.8	Soil moisture at 10–40 cm
SoilMoi40_100cm_inst	kg·m−2	11.99–285.6	Soil moisture at 40–100 cm
SoilMoi100_200cm_inst	kg·m−2	20–476	Soil moisture at 100–200 cm

lists the data variables used in the water balance analysis in this study. To maintain consistency with GRACE data units, data units (kg/m²) marked with the suffix “_inst” (representing instantaneous variables) need to be converted to equivalent water depth units of centimeters. Since GRACE product data are almost always monthly averages, this study subtracted the 2004–2010 average values from the terrestrial water storage data simulated using the 3 h resolution output of the GLDAS-NOAH component data to obtain monthly changes in terrestrial water storage. This was then used to estimate terrestrial water storage changes, ensuring temporal and spatial comparability between the GRACE data and the GLDAS model output.

2.2.3. Precipitation Data

This paper utilizes and collects annual precipitation data and compares it with the analyzed water storage changes to examine their relationship. Precipitation data from 2003 to 2016 with a spatial resolution of 0.0083333° (approximately 1 km) was used. This data was obtained by interpolating and cropping data from 2472 meteorological stations across China. This dataset features high accuracy, high resolution, and a long time series. The data is sourced from the Earth Resources Data Cloud Platform (www.gis5g.com).

2.3. Research Methods

2.3.1. Derivation of Water Balance Equations

TWSa reveals the monthly changes in terrestrial water storage. The GRACE dataset provides the monthly grid water storage equivalent water height relative to the time-averaged baseline terrestrial water storage from 2004 to 2010. The data contained in the monthly mass grid dataset is in the unit of “equivalent water height”, which represents the deviation of the mass in the vertical range of water. The changes in terrestrial water storage in the vertical direction include changes in vegetation moisture, changes in surface water bodies, changes in soil water and changes in groundwater [20]. For example, surface water (including rivers, lakes, reservoirs, etc.), groundwater, soil water, ice and snow, biological water, etc., among which biological water is difficult to measure, and its changes are negligible compared to other components in arid and semi-arid areas [21]. Therefore, TWSa includes the sum of surface water (SWa), groundwater storage (GWSa), soil moisture (SMa), snow water equivalent (SWEa), and canopy water storage (CWSa):

$$\mathrm{TWSa} = \mathrm{SWa} + \mathrm{SMa} + \mathrm{SWEa} + \mathrm{GWSa} + \mathrm{CWSa}$$

(1)

Among them, SWa is the surface water storage anomaly, SMa is the soil moisture anomaly, SWEa is the snow water equivalent anomaly, GWSa is the groundwater storage anomaly, and CWSa is the canopy water storage anomaly.

In Equation (1), the components CWSa, SMa, and SWEa can be obtained from the Canop, SM, and SWE bands in the GLDAS flux data variables, respectively, with anomalies calculated using the same temporal baseline (2004–2010). However, within the GLDAS-2.1 LSM, it is not possible to directly obtain SWa and GWSa. To address this limitation, a new variable—Accessible Water Storage Anomaly (AWa)—was introduced in Equation (2), representing the combined contribution of SWa and GWSa [22].

$$\mathrm{AWa}=\mathrm{SWa}+\mathrm{GWSa}=\mathrm{TWSa}-{\mathrm{CWS}}_{\mathrm{can}}\mathrm{a}-{\mathrm{SM}}_{\mathrm{root}}\mathrm{a}-{\mathrm{SWE}}_{\mathrm{snow}}\mathrm{a}$$

(2)

Among them, ${\mathrm{CWS}}_{\mathrm{can}}\mathrm{a}$ is the plant canopy surface water anomaly, ${\mathrm{SM}}_{\mathrm{root}}\mathrm{a}$ is the root zone soil moisture anomaly, and ${\mathrm{SWE}}_{\mathrm{snow}}\mathrm{a}$ is the snow water equivalent anomaly, all of which are derived from the GLDAS band.

In this study, Equation (2) assumes that the SMa derived from GLDAS represents only the root zone soil water (i.e., ${\mathrm{SM}}_{\mathrm{root}}\mathrm{a}$). The SM layers at four different depths below the root zone, 0–10, 10–40, 40–100, and 100–200 cm, are considered as groundwater, i.e., GWSa, without considering deeper aquifers. Therefore, AWa can be expressed using Equations (3) and (4):

$$\mathrm{GWSa} = {\mathrm{SM}}_{0\_10}\mathrm{a} +{ \mathrm{SM}}_{10\_40}\mathrm{a} +{ \mathrm{SM}}_{40\_100}\mathrm{a} + {\mathrm{SM}}_{100\_200}\mathrm{a}$$

(3)

$$\mathrm{SWa}=\mathrm{AWa}-\mathrm{GWSa}$$

(4)

Among them, ${\mathrm{SM}}_{0\_10}\mathrm{a}$ is the soil moisture anomaly of 0–10 cm below the root zone, ${\mathrm{SM}}_{10\_40}\mathrm{a}$ is the soil moisture anomaly of 10–40 cm below the root zone, ${\mathrm{SM}}_{40\_100}\mathrm{a}$ is the soil moisture anomaly of 40–100 cm below the root zone, and ${\mathrm{SM}}_{100\_200}\mathrm{a}$ is the soil moisture anomaly of 100–200 cm below the root zone.

2.3.2. Trend Analysis and Significance Test

This paper uses Theil–Sen trend analysis to calculate the trend of TWSa from January 2003 to December 2016, and then uses the Mann–Kendall method to test the significance of the trend. Theil–Sen trend analysis is usually used in combination with the Mann–Kendall significance test to evaluate the trend of long-term series data. The advantage of this combination is that it is not affected by outliers and does not need to obey a certain distribution [23]. Theil–Sen trend analysis is a non-parametric estimation method. The specific calculation formula is as follows:

$$\mathsf{\beta } = \mathrm{Median}\left({\displaystyle \frac{{\mathrm{x}}_{\mathrm{j}} -{\mathrm{x}}_{\mathrm{i}}}{\mathrm{j} - \mathrm{i}}}\right), \forall \mathrm{j} > \mathrm{i}$$

(5)

where i, j represent time series, ranging from 2003 ≤ i ≤ j ≤ 2016, and beta represents the median of the slope of all data pairs. When $\beta$ > 0, it shows an upward trend, otherwise it shows a downward trend.

The Mann–Kendall test method can be used to evaluate the significance of the Theil–Sen Median trend [24], and the calculation formula is as follows:

$$\mathrm{Z}=\left\{\begin{array}{c}{\displaystyle \frac{\mathrm{S}-1}{\sqrt{\mathrm{S}\left(\mathrm{s}\right)}}} \mathrm{S}>0\\ 0 \mathrm{S}=0\\ {\displaystyle \frac{\mathrm{S}+1}{\sqrt{\mathrm{S}\left(\mathrm{s}\right)}}} \mathrm{S}<0\end{array}\right.$$

(6)

$$\mathrm{S}=\sum _{\mathrm{j}=1}^{\mathrm{n}-1}\sum _{\mathrm{i}=\mathrm{j}+1}^{\mathrm{n}}\mathrm{sgn}({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}})$$

(7)

$$\mathrm{S}\left(\mathrm{s}\right)={\displaystyle \frac{\mathrm{n}(\mathrm{n}-1\left)\right(2\mathrm{n}-5)}{18}}$$

(8)

$$\mathrm{sgn}({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}})=\left\{\begin{array}{c}1 ({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}})>0\\ 0 ({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}})=0\\ -1 ({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}})<0\end{array}\right.$$

(9)

where n is the length of the time series; $\mathit{sgn}$ is the sign function. At a given significance level α, when $\left|\mathrm{Z}\right|> {\mathrm{u}}_{1-\mathrm{a}/2}$, the time series trend is significant. When the statistic $\left|\mathrm{Z}\right|$ ≥ 2.58, $\left|\mathrm{Z}\right|$ ≥ 1.96, $\left|\mathrm{Z}\right|$ ≥ 1.65, it is considered that there is a 99%, 95%, and 90% confidence, respectively, that the time series has undergone significant changes.

3. Results and Analysis

3.1. TWS Changes

3.1.1. Terrestrial Water Storage Anomaly (TWSa)

Figure 2 shows TWSa data from the CSR, JPL, and GFZ data sources (black, blue, cyan curves, respectively), as well as the average of the three TWSa data (red curve), from January 2003 to December 2016. The three data sources are highly correlated, exhibiting consistent trends with distinct peaks and troughs. Minor differences exist among the three data sources due to differences in processing methods and models. The minimum TWSa value is −6.485 cm in 2016, and the maximum is +4.436 cm in 2004. Minimal TWSa values occur in 2007, 2011, and 2016, while maxima occur in 2004 and 2013. During this period, terrestrial water storage in Inner Mongolia showed an overall downward trend, at a rate of −5.2 × 10⁻⁴ cm/year. Specifically, the region experienced an upward trend between 2003 and 2004, followed by a continuous downward trend from 2004 to 2012. After 2012, TWSa rebounded and then declined.

Table 3. GRACE TWSa (CSR, JPL, GFZ) monthly and annual average TWSa.

14-Year Monthly Average of TWSa (cm)
Jan	Feb	Mar	Apr	May	Jun
−1.239	−0.716	−0.292	−0.202	0.102	−0.578
July	Aug	Sep	Oct	Nov	Dec
0.126	0.244	−0.455	−1.012	−0.755	−0.663
Annual Average of TWSa (cm)
2003	2004	2005	2006	2007	2008
0.345	1.576	0.87	0.048	−0.87	−1.453
2009	2010	2011	2012	2013	2014
−0.227	−0.416	−2.342	−2.092	1.257	−0.042
2015	2016
−0.416	−2.342

presents the interannual and intermonth average TWSa values from the GRACE TWSa datasets (CSR, JPL, and GFZ) for the 14-year period from 2003 to 2016. The TWSa trend is strongly correlated with the climate of the study area. During the rainy season (May to August), the TWSa values in the region show a positive trend over time. Conversely, during the dry season (October to February), the TWSa values in the region show a negative trend.

3.1.2. Spatial Trends of TWSa

Figure 3 shows the spatial distribution of TWSa in Inner Mongolia from 2003 to 2016. TWSa across Inner Mongolia exhibits significant spatial heterogeneity, with a gradual decrease from northeast to southwest. The multi-year average TWSa amplitude is approximately 4.16 cm. High TWSa values are primarily concentrated in the northeastern region of the study area (Hulunbuir City and Hinggan League), with a maximum value of approximately 1.22 cm. Increased precipitation in mountainous areas and forests in the northeast has, to some extent, offset the impact of warming on snow and ice. The lowest values, at −2.94 cm, are primarily distributed in the southern Ordos region of Inner Mongolia. This region is significantly impacted by human activities, with excessive groundwater extraction leading to a decline in terrestrial water reserves. Human interference significantly outweighs the impact of natural factors on changes in terrestrial water reserves.

Figure 4 shows the spatial distribution of changing trends in terrestrial water storage in Inner Mongolia. Overall, terrestrial water storage in Inner Mongolia showed a downward trend from 2003 to 2016. Declining regions accounted for 59% of the total, primarily located in central and southern Inner Mongolia. A significant upward trend occurred in 29% of the total, radiating outward from the orange and red areas shown in the figure. No regions in Inner Mongolia experienced significant, extremely significant, or moderately significant increases.

3.2. Water Balance Components of TWSa

3.2.1. GLDAS-Based CWSa, SMa, and SWEa

Figure 5 shows the changing trends of CWSa, SMa, and SWEa from 2003 to 2016. Among the three groups of hydrological components, the anomaly variation in SMa is more obvious than that of the other two components, ranging from −1.2 to +3.004, and a clear downward trend was observed between 2004–2007 and 2013–2015. In contrast, the variation in CWSa is extremely weak, and the anomaly value is always stable within the range of ±0.001, indicating that the canopy water storage process is highly stable in this region. The maximum range of SWEa does not exceed 0.6 cm, reflecting that the contribution of seasonal snow to the regional hydrological system is relatively limited.

3.2.2. GRACE and GLDAS-Based AWa, GWSa, and SWa

Figure 6 shows the changing trends of AWa, SWa, and GWSa in Inner Mongolia from 2003 to 2016. As can be seen from the figure, AWa showed an overall downward trend. AWa is determined by both SWa and GWA, with SWa fluctuating most significantly and being the primary factor in AWa’s variation. In 2007, SWa reached a maximum of 1.783 cm but subsequently declined, reaching a minimum of −5.278 cm in 2013. In comparison, GWA remained relatively stable from 2003 to 2016 and continued to increase from 2011 to 2014, reaching a peak of 3.34 cm in 2013. AWa varied between −3.056 and 0.372 cm, with the minimum occurring in 2015 and the maximum in 2007.

4. Discussion

4.1. Reliability Evaluation of TWSa

This study systematically analyzed the evolution of water resources in Inner Mongolia from 2003 to 2016, using TWS data derived from the GRACE time-varying gravity field model. As a key indicator of the surface water cycle, TWS reflects the coordinated changes in multiple hydrological components, including surface water storage, canopy water storage, and groundwater storage. Comparison with statistical data from the Inner Mongolia Water Resources Bulletin (

Table 4. Statistical data of Inner Mongolia water resources bulletin from 2003 to 2016 (unit: 100 million m³).

Year	Total Precipitation	Water Consumption	Water Use	Water Supply	Land Water
					Sum	Surface Water	Groundwater
2003	3367	109.05	166.87	166.87	495.57	355.57	239.29
2004	2612.0	110.7	171.01	171.01	437.61	310.23	222.57
2005	2477.3	115.15	174.76	174.76	456.18	338.69	214.64
2006	2575.3	118.07	178.69	178.69	411.29	294.33	213.56
2007	2371.1	118.48	180.12	180.12	295.86	183.03	206.88
2008	3205.1	115.02	175.78	175.78	412.07	274.82	235.15
2009	2679	118.14	181.25	181.25	378.15	263.36	214.35
2010	3014.3	120.8	181.9	181.9	388.54	253.38	227.65
2011	2741.2	122.0	184.7	184.7	419.0	298.16	213.37
2012	3670.8	123.08	185.35	184.35	510.25	349.24	258.38
2013	3649.7	122.02	183.22	183.22	959.81	813.52	249.33
2014	3238.59	120.94	182.01	182.01	537.79	397.61	236.26
2015	3134.43	123.68	185.78	185.78	536.97	402.12	224.57
2016	3274.05	128.37	190.29	190.29	426.5	268.51	248.17

and

Table 5. Comparative data of Inner Mongolia Water Resources Bulletin over the years.

	Precipitation (100 Million m3)	Compared with Average	Surface Water (100 Million m3)	Compared with Average	Ground-Water (100 Million m3)	Compared with Average
2003	3367	3.20%	355.57	−12.50%	239.29	−0.90%
2004	2612	−20%	310.23	−23.70%	222.57	−7.80%
2005	2477.3	−24.10%	338.69	−16.50%	214.64	−11.10%
2006	2575.3	−21.10%	294.33	−27.60%	213.56	−10.40%
2007	2371.1	−27.40%	183.03	−55.00%	206.88	−13.20%
2008	3205.1	−1.80%	274.82	−32.40%	235.15	−1.40%
2009	2679	−17.90%	263.36	−35.20%	214.35	−10.10%
2010	3014.3	−7.60%	253.38	−37.70%	227.65	−4.50%
2011	2741.2	−16%	298.16	−26.70%	213.37	−10.50%
2012	3670.8	12.50%	349.24	14.80%	258.38	8.40%
2013	3649.7	11.80%	813.52	100.10%	249.33	4.60%
2014	3238.59	−0.80%	397.61	−2.20%	236.26	comparable
2015	3134.43	−4%	402.12	−1.10%	224.57	−4.90%
2016	3274.05	0.30%	268.51	−34.00%	248.17	5.10%

, and Figure 7) confirmed the high reliability of the GRACE inversion results. The results show that TWS in Inner Mongolia exhibits significant temporal and spatial variations. In terms of time, the TWSa decreased from 1.576 cm to −0.872 cm (an average annual decrease of 0.82 cm/yr) between 2004 and 2007, which is highly consistent with the trend of decreasing groundwater reserves (an average annual decrease of 523 million m³) during the same period; while the rapid recovery of TWSa from 2011 to 2013 (an average annual increase of 1.80 cm/yr) is closely related to the significant increase in surface water reserves (from 29.816 to 81.352 billion m³). This finding is similar to the research conclusion of Zhou Y [3], further confirming the trend of fluctuating and decreasing water resources in Inner Mongolia.

Human activities are the key factors driving changes in TWS. This study shows that TWSa in the central and southern parts of Inner Mongolia showed a very significant downward trend between 2003 and 2016, which was mainly attributed to large-scale agricultural irrigation in the region and coal mining in the Ordos region. Specifically, the main source of water for agricultural irrigation in the region is groundwater, which has led to a continuous decline in groundwater reserves, and the drainage measures implemented during coal mining (especially open-pit mining) have directly caused structural losses in the groundwater system. Existing research data show that agricultural irrigation, industrial production and coal mining are the three major sources of water resource consumption in Inner Mongolia, accounting for 60%, 16% and 18%, respectively [25]. Based on the above research results, it is recommended that in the future water resource management practices of Inner Mongolia, on the one hand, the efficiency of agricultural irrigation water use should be improved through technological innovation and model optimization, and at the same time, a sound water resource protection and supervision mechanism in mining activities should be established.

4.2. Relationship Between TWSa and Climate Change

A large number of studies have found that there is a significant positive correlation between TWSa and precipitation changes [26,27,28]. Figure 8 shows the rainfall distribution over the 14 years from 2003 to 2016, and Figure 9 shows the corresponding annual average of GRACE TWSa. Through direct comparison of the two datasets, it is obvious that the overall annual variation in GRACE TWSa is closely related to the rainfall distribution. For example, in the years of 2003 and 2004 with less rainfall, GRACE TWSa showed a significant decline. It is worth noting that the hysteresis response characteristics of the hydrological system are significantly manifested in the data. Although the impact of the relatively dry years of 2006 and 2007 was not fully reflected in the TWSa data until 2009–2010, TWSa still declined despite a significant increase in precipitation in 2012–2013. This phenomenon highlights the complexity of changes in terrestrial water storage: TWS, a comprehensive indicator that includes groundwater, surface water, and dam reservoir storage, is driven not only by precipitation input but also by human factors such as groundwater recharge lags, agricultural irrigation withdrawals, and surface water management. Therefore, despite a high statistical correlation between precipitation and TWSa, it is difficult to fully explain TWS trends based solely on rainfall changes; the combined effects of natural hydrological processes and human activities must be considered.

Water shortages are becoming increasingly prominent in Inner Mongolia due to the combined impacts of climate change and human activities. In particular, western Inner Mongolia experiences frequent and persistent hydrological droughts due to scarce precipitation and high potential evapotranspiration. Although water reserves in western Inner Mongolia have declined relatively little, and drought severity has been relatively mild, the combination of prolonged drought and high evapotranspiration has exacerbated the imbalance between water supply and demand. Once water reserves continue to decline, even a relatively small decrease will be difficult to recover from [15].

Based on the monthly mean observation data of water storage components in Inner Mongolia from 2003 to 2016 (Figure 10), combined with the differential characteristics of the three types of indicators in Section 3.2.1 and Section 3.2.2, this paper further reveals the contribution mechanism of different hydrological elements to regional water storage anomalies. The results indicate that SMa exhibit pronounced dominance in both spatial heterogeneity and interannual variability, whereas anomalies in CWSa and SWEa play only a secondary role in influencing the overall terrestrial water storage anomalies across Inner Mongolia. This result is similar to the research results of Zhao et al. [29]. In addition, the seasonal evolution of SMa has a significant vertical gradient feature. The peak of soil moisture in the 0–10 cm layer occurs in August (0.20 cm), while the maximum value of soil moisture in the 40–100 cm layer is reached in October (0.86 cm). The hysteresis phenomenon reflects the vertical transmission time of the precipitation infiltration process. This is consistent with the research results of Fu [30]. Precipitation is the main meteorological factor driving groundwater changes [31] and determines the amount of groundwater recharge. As the main source of recharge, precipitation plays a substantial role in maintaining groundwater balance [32]. Therefore, the shallow SMa is affected by rainfall and shows a significant increasing trend in June in summer. As the soil depth increases, the influence of precipitation becomes smaller and smaller, and it will not directly lead to changes in deep soil moisture and groundwater. Secondly, the absolute value of the monthly mean SWEa is always lower than 0.33 cm, and the annual variation is only 0.5 cm. The monthly mean CWSa is closer to the zero value range (−0.002–0.003 cm), indicating that the proportion of water intercepted by the canopy in arid and semi-arid ecosystems is extremely low. In contrast, soil moisture content at different depths showed significant differences: the SMa of the 0–10 cm layer ranged from −0.07 to 0.20 cm; the 10–40 cm layer was from −0.57 to 0.57 cm; and the 40–200 cm layer showed a larger fluctuation of −0.86 to 0.86 cm.

4.3. The Significance of TWSa for Ecological Construction

In arid and semi-arid regions, the continuous decline in water resources has become a critical bottleneck restricting sustainable regional development. Previous studies have shown that this downward trend poses unprecedented challenges to ecosystem restoration and reconstruction [33,34]. Of particular concern is that, although ecological restoration programs implemented in northern China have improved environmental conditions to some extent [35,36], the absence of rigorous assessments of regional water resource carrying capacity has led to excessive water consumption from large-scale afforestation. This high-consumption restoration model has directly contributed to the persistent decline in TWS and exacerbated the imbalance between water supply and demand. The results of this study reveal a similarly significant downward trend in TWS across Inner Mongolia. If previous vegetation restoration strategies continue to be applied–especially those emphasizing afforestation, this will likely create a vicious cycle of escalating water consumption, further intensifying the vulnerability of already fragile water resource systems [37]. In light of these findings, a systematic optimization of existing restoration strategies is essential. First, it is necessary to abandon the extensive development model that prioritizes quantitative targets, and instead establish a scientific evaluation framework grounded in water resource carrying capacity. Second, vegetation selection should prioritize drought-tolerant shrubs and grasses with lower water requirements, thereby enhancing water-use efficiency at the community level through optimized structural configurations. Finally, a multi-objective, synergistic ecosystem management framework should be developed, aiming to minimize dependence on water resources while maintaining ecological functionality. Overall, ecological restoration in Inner Mongolia should transition from scale expansion to efficiency improvement, establishing a resilient and adaptive management model for arid and semi-arid regions. Such an approach not only offers a pathway to alleviating current water resource pressures but also provides new directions for ensuring the sustainable development of ecological security barriers in arid and semi-arid regions.

5. Conclusions

The Google Earth Engine (GEE) cloud computing platform provides a crucial technical foundation for the rapid analysis of terrestrial water storage anomalies. This study, focusing on Inner Mongolia, utilizes GRACE satellite data and the GLDAS hydrological model, combined with ground-based precipitation data, to comprehensively monitor changes in terrestrial water storage anomalies over the study area from 2003 to 2016. Furthermore, based on this analysis of terrestrial water storage anomalies, we also discuss the trends of various components in water resource changes. The main conclusions of this study are as follows:

• The temporal trends in terrestrial water storage over the study area derived from data sources provided by the three GRACE institutions (JPL, CSR, and GFZ) are relatively consistent. The ensemble average of these three data sources was used to analyze the terrestrial water storage trends over the study area. From 2003 to 2016, TWSa over Inner Mongolia showed a significant overall downward trend, with a multi-year rate of change of 5.2 × 10⁻⁴ cm/year. The seasonal distribution of TWSa is highly correlated with precipitation, increasing during the rainy season and decreasing during the dry season.
• In terms of spatial trends, TWSa in Inner Mongolia gradually showed a decreasing trend from northeast to southwest. The areas with a more significant upward trend accounted for 29%, with a maximum value of about 1.22 cm, which appeared in the northeastern part of the study area; the areas with a downward trend accounted for 59%, and the minimum value of −2.94 cm appeared in the southern part of Ordos, Inner Mongolia.
• The component CWSa in the water balance equation based on the GLDAS/NOAH model output has the least impact on TWSa, followed by SWEa. SMa has a more significant anomaly than these two components, with a clear downward trend between 2004–2007 and 2013–2015. Soil water at different depths exhibits significant differences, and the hysteresis reflects the vertical transmission time of the precipitation infiltration process. Precipitation primarily affects shallow SM, but the correlation between precipitation and deep soil water and groundwater is relatively small.