Groundwater Crisis in the Eastern Loess Plateau: Evolution of Storage, Linkages with the North China Plain, and Driving Mechanisms

Abstract

Understanding the dynamics and drivers of groundwater storage (GWS) is crucial for sustainable resource management. Most studies attribute GWS changes to climate change or human activities, often neglecting external hydrological influences. In this study, we categorize the driving factors influencing GWS changes into three groups: climate change, human activity, and regional hydrological pressure. We emphasize that the coupling effects and potential disturbances from adjacent hydrological systems may significantly affect local groundwater evolution. This perspective differs from conventional approaches that focus solely on local factors. This study analyzes the spatiotemporal evolution of GWS in Shanxi Province, located in the eastern Loess Plateau, from 2003 to 2023 using GRACE and GLDAS data. We examine the linkage between GWS in Shanxi and the North China Plain through correlation analysis, Engle–Granger cointegration tests, and Granger causality tests. The results show that GWS in Shanxi showed an average annual reduction of −17.27 ± 1.4 mm/yr, with the most severe depletion occurring in the southeastern region, which is geographically adjacent to the North China Plain. The results of the Engle–Granger cointegration test and Granger causality analysis reveal a bidirectional causal relationship between GWS changes in the two regions, indicating that changes in GWS in either region may have a significant impact on the other. The results of the contribution analysis indicate that the North China Plain’s groundwater decline contributes approximately −53.89% to the reduction of GWS in Shanxi, while human activities and external hydrological influences together explain over 98% of the change. This result suggests that relying solely on climatic and human activity factors to explain groundwater changes may lead to significant biases, as ignoring interregional hydrological linkages can amplify or obscure the attribution of local groundwater variations, resulting in distorted conclusions. These findings highlight the value of remote sensing in capturing regional hydrological interactions and underscore the need to integrate interregional groundwater connectivity into policy design for sustainable groundwater governance.

1. Introduction

The Loess Plateau is the largest area covered by loess in the world and has very sensitive ecosystems with extremely limited water supplies [1,2]. The Loess Plateau is in a zone that is seriously affected by drought, and under the combined impacts of climate change and human activities, drought frequency in the region has increased significantly [3,4], which has seriously restricted regional economic and social development. Meanwhile, the shortage of water resources has become one of the typical factors restricting the sustainable development of the regional ecosystem. Especially, the large reduction of GWS in the eastern Loess Plateau has resulted in a number of long-lasting and severe ecological degeneration and socio-economic disputes [5,6]. For example, in typical regions such as the Yuncheng Basin and Linfen Basin in Shanxi Province, long-term overexploitation of groundwater has led to land subsidence, water source depletion, and irrigation difficulties [7,8,9]. In this context, the study of GWS evolution and the causes for the above phenomena in the eastern Loess Plateau has important practical and scientific implications for the sustainable exploitation of local water resources. Meanwhile, scientifically clarifying the impact of regional hydrological interactions on groundwater evolution helps enhance the spatial coordination of current water resource management strategies and holds important value for promoting regional collaborative governance.

Research on groundwater problems in the Loess Plateau has already been carried out extensively, and scholars investigate issues from multiple perspectives and at different spatial scales [5,6,10,11,12,13]. For example, Li and Ma [5] investigated the temporal trends of the evolution of groundwater due to human activities and climate change; they also assessed sustainability of groundwater in the Loess Plateau based on Reliability–Resilience–Vulnerability (RRV) framework. Han et al. [6] analyzed the effects of large-scale vegetation restoration on groundwater in the Loess Plateau, and concluded that the restoration had remarkable effects on groundwater drought. Xie et al. [14] verified the reasonability of GRACE-based groundwater estimations associated with the Loess Plateau by combining GRACE data with data from 136 groundwater monitoring wells. These studies generally indicated that GWS decreased over Loess Plateau as a whole, with a more significant decline in the eastern region. Most previous studies are qualitative in nature, lack a systematic analysis of temporal and spatial variation in GWS change in the eastern Loess Plateau and lack clear explanation of the driving forces. Furthermore, most existing studies have only covered climate change and human activities in the study area, without accounting for external groundwater systems, which may result in a wrong local impact estimate and bias of local water resources management decision.

Although existing studies have addressed the overall groundwater evolution in the Loess Plateau, including some discussions on Shanxi Province, most have remained focused on natural and anthropogenic factors within the Loess Plateau, lacking attention to and quantification of cross-regional groundwater linkages with neighboring areas. The North China Plain, located to the east of Shanxi Province, is one of China’s most economically developed and densely populated regions. Due to long-term overexploitation of groundwater for agricultural, industrial, and domestic use, the North China Plain has formed extensive cone-shaped depressions. Given the geographical proximity and hydraulic connectivity between the North China Plain and the eastern Loess Plateau, it is essential to consider the potential cross-regional hydrological interactions. Therefore, a thorough exploration of the spatiotemporal patterns of GWS variation in Shanxi Province, its linkage with the North China Plain, and the identification of key driving factors will not only improve understanding of groundwater dynamics in the Loess Plateau but also provide a scientific basis for regional water resource management and sustainable groundwater governance. Although the two regions have diverse topographic features, groundwater systems in some areas may be interlinked, especially the deep groundwater, which can cross regional boundaries. Persistent overuse of groundwater in the North China Plain could have indirect impacts on groundwater systems in adjacent areas, including the eastern Loess Plateau, through hydraulic connections [15,16]. Nevertheless, the specific nature and extent of this interaction remain insufficiently understood. Shanxi Province, located in the eastern Loess Plateau and adjacent to the North China Plain [17,18], provides a natural laboratory for studying the impacts of climate change, human activities, and interregional groundwater linkages. Groundwater overexploitation is severe in Shanxi Province, and environmental and geological problems caused by coal mining are particularly prominent [19,20]. Meanwhile, the Taihang Mountains in eastern Shanxi serve as a critical water recharge area for the North China Plain, and there is a strong hydraulic connection between the groundwater systems of the two regions [16,21,22].

To address the insufficient attention paid in existing studies to the interference mechanisms of adjacent hydrological systems, this study categorizes the driving factors affecting changes in GWS into three groups: climate change, human activities, and hydrological pressures from neighboring regions—particularly the North China Plain. Furthermore, the concept of regional hydrological interaction is introduced to highlight the potential disturbances and coupling effects of adjacent hydrological systems on local GWS variations. This perspective helps move beyond the conventional analytical framework focused solely on local climate and human activities, providing a more comprehensive understanding of groundwater system dynamics in a cross-regional context.

In this study, the temporal and spatial evolution of GWS in Shanxi Province was analyzed using long-term time series of GRACE and GLDAS data. Correlation analysis, cointegration tests, and causality models were employed to investigate the interrelationship between GWS in Shanxi Province and that in the North China Plain. Meanwhile, the driving factors influencing GWS in Shanxi Province were quantitatively assessed by integrating ERA5 reanalysis data, statistical data, and high-resolution land use data. Shanxi Province, situated in the transitional zone between the Loess Plateau and the North China Plain, is subject to multiple influences on GWS dynamics. In addition to the strong influence of groundwater flow interactions with the North China Plain, previous studies have shown that changing climate conditions and human activities jointly exert pressure on the regional groundwater system [5,6,23,24]. To assess climatic impacts, precipitation, evaporation, and runoff were selected as indicators of natural recharge. For anthropogenic influences, water consumption, coal mining, and land cover (measured by NDVI) were used to represent human disturbances. In addition, considering the strong hydraulic connection between the North China Plain and Shanxi Province, this study further incorporates changes in GWS in the North China Plain as an indicator of external hydrological pressure, aiming to quantify the potential cross-regional impacts of neighboring areas on the groundwater system in the study region.

The main objectives of this study are: (1) to characterize the spatial and temporal evolution patterns of GWS in the eastern Loess Plateau; (2) investigate the cross-regional relationship between Shanxi and the North China Plain through cointegration and causality analysis; (3) to quantitatively evaluate the contributions of various potential driving factors to GWS changes in Shanxi Province. The overall research framework and methodological process are illustrated in Figure 1.

2. Study Area and Data

2.1. Study Area

Situated at the eastern margin of the Loess Plateau [17,18], Shanxi Province (Figure 2) features terrain that descends from northeast to southwest, with elevation spanning 55 to 1500 m and covering an area of 156,700 km² [25]. Separated from the North China Plain by the Taihang Mountains, Shanxi Province enjoys a unique geographic setting. The province experiences a temperate continental monsoon climate with low annual precipitation [26], averaging 622 mm over multiple years, with most rainfall concentrated between June and September. Shanxi Province is located in the middle reaches of the Yellow River Basin, with major rivers including the Yellow River, Fen River, Qin River, and Sanggan River. These rivers serve as important surface water sources in the region, supporting both agricultural activities and urban domestic water demands. The region faces a severe water shortage, with per capita water resources amounting to only 361.34 cubic meters, far below the internationally recognized threshold of 500 cubic meters for extreme water scarcity [27,28,29], placing it in a state of chronic water scarcity. Groundwater extraction accounts for 38.6% of the total water withdrawal, highlighting the stress on groundwater resources. According to the delineation results of groundwater overexploitation zones released by the water conservancy department in 2015, Shanxi Province contains 22 groundwater overexploitation zones, covering a total area of more than 10,609 square kilometers, accounting for 6.8% of the province’s total area. Among these, severely overexploited areas cover 1848 square kilometers, primarily distributed in river valley plains. The main types of aquifers in the region include Quaternary unconsolidated aquifers in river valley alluvial deposits and fractured rock aquifers in piedmont zones [30,31]. Groundwater recharge primarily depends on atmospheric precipitation infiltration and lateral seepage from rivers, while human activities such as agricultural irrigation and coal mining have led to a continuous decline in groundwater levels. The prolonged decline in GWS poses serious threats to the local ecological environment and the sustainable use of water resources, presenting major challenges for future regional water management and development.

The North China Plain, located to the east of Shanxi Province, spans approximately 400,000 km² and covers parts of Shandong, Hebei, Henan, Beijing, Tianjin, Anhui, and Jiangsu. It is the second-largest plain in China and one of the most densely populated regions, home to nearly a quarter of the national population. Annual precipitation ranges from 400 to 600 mm [32], primarily concentrated in the summer months. The plain features fertile soils formed by alluvial deposits from the Hai River, Huai River, and Yellow River [33], making it highly suitable for agricultural development. However, rapid urbanization and industrial expansion have led to long-term overexploitation of groundwater, resulting in continuous declines in groundwater levels. In some areas, the water table has dropped by more than 1 m per year, creating extensive groundwater depression cones and land subsidence, which have significantly altered the regional hydraulic gradient. The ongoing depletion of deep confined aquifers in the North China Plain exerts hydrological pressure on adjacent regions, such as Shanxi Province, through hydraulic connectivity and may induce lateral groundwater flow. This transboundary hydrological influence constitutes a key component of the “regional hydrological interaction” framework proposed in this study.

Figure 3 illustrates the spatial distribution of groundwater levels across the study region in August 2005, based on the 1 km resolution groundwater level dataset constructed by Wang et al. [34]. This dataset was developed using 4,275,506 observation records from 11,911 monitoring wells across China. To improve spatial coverage, the dataset employed inverse distance weighting (IDW) interpolation, resulting in a gridded groundwater level product with a spatial resolution of 1 km. The map shows a clear west-to-east decreasing trend in groundwater levels from Shanxi Province to the North China Plain, indicating a regional groundwater level gradient between the two regions.

2.2. Data

2.2.1. GRACE Data

The GRACE data used in this study were derived from the GRACE/GRACE-FO RL06 Mascon gravity field model released by the Center for Space Research (CSR) at the University of Texas (https://www2.csr.utexas.edu/grace/RL06_mascons.html, accessed on 17 March 2025). The dataset spans from January 2003 to December 2023 and is provided at a spatial resolution of 0.25° × 0.25°. Singular spectrum analysis (SSA) was applied to interpolate the missing data between the GRACE and GRACE-FO missions [35]. The GRACE-derived terrestrial water storage (TWS) comprises several components, including variations in canopy water content, soil moisture, and snow water equivalent, as well as GWS variations. Since GRACE alone cannot distinguish between different components of terrestrial water storage changes, the Global Land Data Assimilation System (GLDAS) was used to assist in separating above-ground water components and to indirectly estimate GWS changes.

2.2.2. GLDAS Data

The GLDAS dataset, accessed via NASA’s Goddard Earth Sciences Data Center, provided the basis for analysis (https://disc.gsfc.nasa.gov/datasets/GLDAS_NOAH025_M_2.1/summary, accessed on 18 February 2024). This research employed the Noah land surface model from GLDAS-2.1 to extract variables such as soil moisture and snow water equivalent. Monthly anomalies from 2003 to 2023 were calculated by subtracting the multi-year mean (January 2004–December 2009) from the corresponding monthly values of soil moisture, snow water equivalent, and canopy water storage derived from GLDAS. Monthly variations were represented by averaging the soil moisture across depths (0–10 cm, 10–40 cm, 40–100 cm, and 100–200 cm), along with canopy water content and snow water equivalent.

2.2.3. Groundwater Monitoring Data

Ground-based groundwater level measurements are considered a more accurate data source, although their application is limited in complex terrain areas. In this study, groundwater level data from 12 monitoring wells publicly released in Shanxi Province were selected for validation, including 2 submersible wells and 10 confined aquifer wells, provided by the China Institute of Geo-Environmental Monitoring. The accuracy of GRACE-derived GWS anomalies was verified by calculating the correlation coefficient between the GWS and observed groundwater level data.

2.2.4. Other Data

Annual environmental data used to analyze the drivers of GWS changes included precipitation, evapotranspiration, runoff, and NDVI. Annual precipitation, evapotranspiration, and runoff data for 2003–2023 were obtained from the ERA5-Land reanalysis product. These meteorological variables are used to characterize regional climate change processes and serve as important natural driving factors influencing changes in GWS. Annual maximum NDVI data for 2003–2023 were derived from the MODIS MOD13A3 product with a 1 km resolution [36]. NDVI is a commonly used indicator for measuring changes in vegetation cover and is widely applied to reflect the indirect impact of land use change on groundwater recharge processes. Therefore, it is included as one of the variables representing human activities. Given that the study area is an important energy base in China, water consumption and coal mining data were selected to reflect the impact of human activities on GWS. Water resource utilization data reflect the direct extraction pressure on regional water resources, while coal mining activities indirectly lead to groundwater leakage and water table decline by causing underground goafs, surface subsidence, and geological fractures. These disturbances exert structural impacts on the groundwater system. Coal mining has severely damaged the groundwater system, forming extensive goaf areas and causing subsidence and fissures in overlying strata, leading to groundwater leakage and water table decline. Data on water consumption and coal production were obtained from the Shanxi Statistical Yearbook. Considering the strong hydraulic connection between the study area and the North China Plain, this study introduces the GWS variation in the North China Plain as an external hydrological pressure variable to quantify the potential cross-regional impact of neighboring areas on the groundwater system in Shanxi Province.

3. Research Methods

3.1. Methods for Estimating Changes in GWS

Terrestrial water storage ($TWS$) primarily comprises surface water, groundwater, soil moisture, snow water, canopy water, and biomass water [37,38,39]. biomass water can generally be neglected [40,41,42]. Among these, GWS anomalies ($GWS$) can be estimated by deducting the contributions of other components from the total $TWS$ anomalies.

$$GWS=TWS-SMS-SWE-CWS-SWS$$

(1)

where $GWS$ indicates GWS anomalies, $TWS$ indicates terrestrial water storage anomaly, $SMS$ indicates soil moisture storage anomaly, $SWE$ indicates snow water equivalent anomaly, $CWS$ indicates canopy water storage anomaly, and $SWS$ indicates surface water storage anomaly.

3.2. Mann–Kendall Test

The Mann–Kendall (MK) test is a commonly applied non-parametric method for detecting trends and assessing their statistical significance in hydrological time series, meteorological, and other time series data [43,44,45]. It includes both trend and abrupt change (mutation) detection, offering the advantages of computational simplicity and conceptual clarity. In this study, the Mann–Kendall test was employed to detect potential abrupt shifts in the GWS time series of Shanxi Province during 2003–2023, using a significance level of 0.05. If a change point is identified, the analysis period is subsequently divided into two segments: one preceding and one following the detected point. If no mutation is observed, the entire period is treated as a whole for trend analysis. The test statistic for the MK trend test is defined as follows:

$$S=\sum _{i=1}^{n-1}\sum _{k=i+1}^{n}sgn(x_k-x_i)$$

(2)

$$sgn\left(x_k-x_i\right)=\left\{\begin{array}{c}\begin{array}{c}-1\hspace{1em}{x}_k<x_i\\ \hfill 1\hspace{1em}{x}_k>x_i\end{array}\\ \hfill 0\hspace{1em}{ x}_k=x_i\end{array} \right.$$

(3)

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

(4)

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

(5)

Here, $x_k$ and $x_i$ denote the observed values in the time series, and $n$ is the sample size. The standardized statistic $Z$ measures the strength and direction of the trend: a larger $Z$ indicates a more pronounced trend, with $Z$ > 0 signifying an increasing trend and $Z$ < 0 a decreasing trend. At a given significance level $\alpha$, the trend is deemed statistically significant if $Z$ exceeds the critical value. In this study, we adopt $\alpha$ = 0.05, which corresponds to a critical threshold of ±1.96.

3.3. Sen’s Slope Estimator

To robustly assess the trends in GWS for each grid cell, this study employs the Sen’s Slope Estimator method. This non-parametric statistical technique is widely used in hydrological and climate research. The formula is as follows:

$$Q=median\left({\displaystyle \frac{{GWS}_j-{GWS}_i}{j-i}}\right), for all i<j$$

(6)

where ${GWS}_i$ and ${GWS}_j$ represent the GWS values in year $i$ and $j$, respectively. The resulting slope $Q$ denotes the annual rate of GWS change for each pixel over the period 2003–2023.

In this study, GWS data derived from GRACE were used to construct regional-scale groundwater trend maps using the Sen’s slope method. The results were further classified using the natural breaks method to identify areas with the most pronounced negative trends, which were designated as groundwater depletion hotspots. These were used in subsequent analyses of spatial heterogeneity and regional comparison.

3.4. Engle–Granger Cointegration Test

To assess the presence of a long-term equilibrium link between GWS in Shanxi Province and the North China Plain, the Engle–Granger cointegration test was applied. The first step in conducting a cointegration test is to assess the stationarity of two or more time series variables by testing for the presence of a unit root. If the variables are already stationary, a cointegration test is unnecessary. However, if two or more non-stationary time series have a linear combination that is stationary. This indicates that the variables maintain a stable long-term equilibrium connection [46,47,48]. This long-run relationship can be identified by examining the stationarity of the residuals from a regression equation, which is specified as follows.

$$Y_t=\alpha +\beta X_t+{\mu }_t$$

(7)

In Equation (7), $Y_t$ denotes the response variable, $X_t$ stands for the predictor, $\alpha$ and $\beta$ are the estimated coefficients, and ${\mu }_t$ refers to the stochastic error term. If the residuals from the regression equation are stationary, this suggests that a stable long-term equilibrium connection exists between the time series variables. Otherwise, no cointegration is present. The decision criteria are as follows: a long-term equilibrium is considered to exist when the ADF (Augmented Dickey–Fuller) test statistic falls below its critical threshold or when the p-value is under 0.05, the residuals are considered stationary, implying that the two variables are cointegrated. Conversely, if these conditions are not met, the variables are considered not to be cointegrated.

3.5. Granger Causality Test

Following the Engle–Granger cointegration analysis, the Granger causality test was applied to identify the causal direction between GWS in Shanxi Province and that in the North China Plain. Given two time series, $X_t$ and $Y_t$, if past values of $X_t$ help improve the prediction of $Y_t$, then $X_t$ is said to “Granger cause” $Y_t$. To formally test whether $X_t$ Granger causes $Y_t$, a lagged regression model of $Y_t$ is constructed, typically of order $P$. This involves comparing a restricted model that includes only lagged values of $Y_t$ with an unrestricted model that also includes lagged values of $X_t$. If the inclusion of $X_t$’s lagged terms significantly improves the model, then $X_t$ is considered to be a Granger cause of $Y_t$.

$$Y_t={\beta }_0+\sum _{i=1}^P\alpha X_{t-i}+\sum _{j=1}^P\beta Y_{t-i}+{\mu }_t$$

(8)

$$X_t={\delta }_0+\sum _{i=1}^P\lambda Y_{t-i}+\sum _{j=1}^P\delta X_{t-i}+{\gamma }_t$$

(9)

In these equations, $X_t$ and $Y_t$ represent the original time series, while $X_{t-i}$ and $Y_{t-i}$ denote the values of the respective series lagged by $i$ periods. ${\beta }_0$ and ${\delta }_0$ are the constant terms, and the parameters $\alpha$, $\beta$, $\lambda$, and $\delta$ represent the regression coefficients of the corresponding time series. The variable $i$ refers to the lag order, $P$ is the maximum number of lags, ${\mu }_t$ and ${\gamma }_t$ are the residual terms. Equation (8) tests whether the time series $X_t$ Granger-causes $Y_{t-}$, while Equation (9) tests whether $Y_t$ Granger causes $X_t$. In this study, we use Stata 17 to conduct cointegration and Granger causality analyses on the time series data.

3.6. Relative Contribution

In the analysis of driving mechanisms behind changes in GWS, commonly used methods for identifying variable importance include multiple linear regression, ridge regression, principal component analysis, and ensemble learning models. Among these, ensemble learning methods have demonstrated strong performance in capturing complex nonlinear relationships, handling high-dimensional features, and improving model predictive capabilities [49]. However, these methods primarily emphasize prediction accuracy, while their complex model structures and relatively weak interpretability of variable contributions may hinder the understanding of physical mechanisms in groundwater systems.

Therefore, considering that the primary objective of this study is to quantify the directional contributions of various factors to changes in GWS while ensuring strong physical interpretability, we first conducted variance inflation factor (VIF) tests to identify potential multicollinearity issues. The results showed that only precipitation (PRE) and surface runoff (SR) exhibited strong multicollinearity (VIF > 10), while the VIF values of the remaining variables were below common thresholds. Accordingly, we adopted the ridge regression method to address multicollinearity and to enhance the stability and reliability of coefficient estimation.

Ridge regression introduces an L2 regularization term (i.e., the sum of squared regression coefficients) into the traditional least squares loss function. This effectively reduces the impact of multicollinearity among variables and improves model robustness. In this study, we used ridge regression to derive standardized regression coefficients (α_n) for each variable, which indicate both the direction (positive or negative) and the relative strength of their impacts on GWS variation. Based on these coefficients, we calculated the relative contribution rate for each factor to identify dominant drivers and quantify their influence. Unlike stepwise regression and other variable selection methods, ridge regression retains all predictors, making it well-suited for analyses involving multiple climatic and anthropogenic variables.

Moreover, because data related to human activities were only available at the annual scale, all variables were standardized to the same temporal resolution. This ensured comparability and consistency across variables. Ridge regression enabled us to include all variables without exclusion, thus preserving the completeness of the influencing factor set and maintaining scientific validity.

In this study, a ridge regression model is constructed with GWS as the dependent variable and its potential influencing factors as independent variables. Since data related to human activities are available only on an annual scale, all variables in the model were standardized to the same temporal resolution—that is, annual data. Therefore, annual-scale data from 2003 to 2023 were used to construct the regression model for estimating the contributions of climate change, human activities, and external hydrological influences beyond the region to changes in GWS. The regression equations are presented as follows [50,51]:

$$GWS={\alpha }_1{X}_1+{\alpha }_2{X}_2+\cdots +{\alpha }_n{X}_n+ϵ$$

(10)

The regression coefficients ${\alpha }_n$ corresponding to each independent variable $X_n$ are obtained from the multiple linear regression model. These coefficients represent the strength and direction of the relationship between each influencing factor and GWS.

To calculate the actual contribution of each variable, the normalized value of each independent variable is multiplied by its corresponding regression coefficient. This yields the relative contribution of each factor to the variation in GWS.

$$C_{\left(i\right)}={\alpha }_n\times {\displaystyle \frac{{dX}_i}{dt}}$$

(11)

${\displaystyle \frac{{dX}_i}{dt}}$ represents the annual rate of change of each variable, while $C_{\left(i\right)}$ denotes the actual contribution value of each factor, which can be either positive or negative. A positive $C_{\left(i\right)}$ indicates that the corresponding factor contributes positively to the increase in GWS, where as a negative $C_{\left(i\right)}$ suggests a suppressing effect.

The relative contribution of different drivers to GWS reflects their relative importance in explaining the variation in groundwater dynamics [50,51].

$$R_{\left(i\right)}={\displaystyle \frac{C_{\left(i\right)}}{\sum \left|C_{\left(i\right)}\right|}}\times 100\%$$

(12)

$R_{\left(i\right)}$ represents the relative contribution of influencing factor $i$ to changes in GWS.

4. Results

4.1. Reliability Test of GWS Data Based on GRACE Estimation

To evaluate the reliability of GWS estimates derived from GRACE-GLDAS inversion, correlation analysis was conducted between GRACE-based data and observational data from 12 groundwater monitoring wells in Shanxi Province. The purpose of this analysis is to evaluate the reliability and accuracy of the inversion results by comparing them with in situ groundwater measurements. The agreement between observed groundwater levels and GRACE-derived GWS values serves as a key indicator of inversion quality. The correlation analysis shows that most of the monitoring wells have positive correlations with the GRACE-based estimates. Among them, seven wells have strong correlation values. This means the measured data and the GRACE-based data match well and change in similar ways (

Table 1. Correlation between observed groundwater levels and GRACE-derived GWS estimates.

Well	Elevation (m)	Monitoring Depth (m)	Location	Groundwater Type	Correlation Coefficient
W1	797.15	52.00–104.00	Taiyuan	Unconfined Aquifer	−0.708
W2	779.03	139.67–250.55	Taiyuan	Confined Aquifer	−0.814
W3	346.73	92.20–194.00	Yuncheng	Confined Aquifer	0.463
W4	447.82	57.28–187.00	Linfen	Confined Aquifer	0.616
W5	1059.73	19.26–92.50	Datong	Confined Aquifer	0.104
W6	1096.4	0.00–439.94	Shuozhou	Confined Aquifer	0.929
W7	934.61	0.00–160.00	Changzhi	Confined Aquifer	0.763
W8	968.53	316.77–500.00	Changzhi	Confined Aquifer	0.693
W9	1052.11	37.00–93.00	Datong	Confined Aquifer	0.993
W10	1050.7	0.00–157.00	Datong	Confined Aquifer	0.917
W11	1040.22	0.00–75.00	Datong	Confined Aquifer	0.917
W12	1039.84	0.00–20.00	Datong	Unconfined Aquifer	0.402

). Three wells show weak but still positive correlations. Two wells have negative correlations. In general, the results show that there is good agreement between the GRACE-GLDAS estimates and the local well data. This proves that the GRACE-GLDAS method works well for estimating GWS in Shanxi Province. These results also show that this method is suitable for studying how groundwater changes in the region.

It is worth noting that two monitoring wells (W₁ and W₂) show significant negative correlations with the GRACE-derived GWS data, with correlation coefficients of −0.708 and −0.814, respectively. Both wells are located in the urban area of Taiyuan City, where human activity intensity is high. These areas are subject to considerable disturbances from urban water supply regulation, industrial groundwater extraction, and infrastructure development. Additionally, the region experiences frequent land use changes and complex groundwater recharge-discharge processes, further contributing to the inconsistency between well-based measurements and the regional-scale GRACE signals.

Overall, despite a few outlier wells, the majority of monitoring wells (10 out of 12) show positive correlations with the GRACE estimates, with 7 wells exhibiting significant correlation coefficients (R > 0.6), accounting for 58%. These results demonstrate that the GRACE-GLDAS method has good regional applicability and retrieval reliability in Shanxi Province and can be effectively used for monitoring and assessing groundwater changes at the regional scale.

4.2. Interannual Trend of GWS Changes

The time series of GWS change over Shanxi Province during 2003–2023 obtained from GRACE data presents a decreasing trend (Figure 4). Non-significant change points were detected by the Mann–Kendall (MK) mutation test, showing that a steady and continuous decrease occurred in the selected period. In this way, the full period 2003–2023 was considered as a single continuous time series for analysis. The MK trend test further identified a statistically significant decreasing trend of GWS (Z = −5.23). The annual average rate of decrease is calculated to be −17.27 ± 1.4 mm/yr, which is equivalent to a total amount of 56.58km³ GWS loss through 21 years. The continuous negative trend in the GWS anomaly clearly shows that the rate of GWS depletion is much higher than the natural process of GWS recharge in the region, which is a great concern for sustainable water resource management. The GWS historical minimum was recorded in June 2021 with a value of −419.5 mm, which suggests an extreme groundwater deficit. By contrast, it was the highest in Jan. 2004, with a value of 92.4mm, when GWS was relatively abundant. These two extremes are evidence of increasing volatility in GWS, especially the high negative extremes are more pronounced in the recent decades. This trend is likely affected by various factors, such as precipitation variation, climate variability, and human activity (such as agricultural irrigation and industrial water consumption). During the past 21 years, the annual precipitation in Shanxi Province is relatively stable, being about 500 to 870 mm. As a result, such precipitation has not been excessive enough to reasonably compensate for the continued GWS decline. The unbalance between precipitation and GWS sheds light on the intricate response of the groundwater system. Greater understanding of the forces behind GWS dynamics is urgently needed to delineate the most important causative factors and inform science-based water management and policy.

Based on the analysis of GWS changes in Shanxi Province from 2003 to 2023, we observed a significant declining trend in both annual averages and seasonal values. The seasonal rates of GWS decline were estimated at −15.9 ± 1.6 mm/yr in spring, −17.6 ± 1.6 mm/yr in summer, −16.1 ± 1.6 mm/yr in autumn, and −17.6 ± 1.9 mm/yr in winter, with the most pronounced decline occurring in summer and winter, and relatively smaller decreases in spring and fall (Figure 5), the Mann–Kendall (MK) trend test applied to the seasonal GWS time series indicated statistically significant downward trends across all four seasons. No abrupt change points were detected in the seasonal Mann–Kendall tests, suggesting that while the depletion of GWS is a long-term and continuous process, there were no statistically significant structural shifts during any particular season, and the changes may be attributed to gradual long-term influences rather than abrupt short-term fluctuations. As shown in the seasonal variation graphs, GWS depletion was most pronounced during summer and winter—particularly during periods of peak agricultural irrigation demand, when groundwater extraction tends to accelerate. In contrast, GWS declines were relatively moderate in spring and fall. The persistent depletion of groundwater has serious implications for ecosystems, irrigated agriculture, and local livelihoods. In particular, it poses risks to agricultural productivity and water supply security, potentially resulting in water shortages, increased production costs, and threats to food security. These findings underscore the urgent need for more scientific, sustainable, and adaptive water resource management strategies in Shanxi Province to address the worsening groundwater crisis.

4.3. Intra-Annual Variations in GWS

Based on annual precipitation data from 2003 to 2023 in Shanxi Province, this study classifies hydrological years using the quartile method: years with precipitation below the 25th percentile are defined as dry years, those above the 75th percentile as wet years, and the remainder as normal years. According to this classification, 2019—with the lowest precipitation (497 mm)—is identified as a dry year, 2007 (599 mm) as a normal year, and 2016 (699 mm) as a wet year. Figure 6 illustrates the intra-annual variation in GWS in Shanxi Province over different hydrological year types (dry, normal, and wet years), along with the multi-year average. Overall, GWS in Shanxi exhibits pronounced seasonal fluctuations. The multi-year average shows that GWS typically peaks in March and April, followed by a sharp decline from May to July, reaching the lowest levels of the year. Recovery begins gradually after August. This seasonal pattern reflects significant groundwater depletion driven by a surge in water use during late spring and early summer, combined with insufficient recharge.

However, both the timing and magnitude of groundwater recovery differ among hydrological year types. In the dry year (2019)—the year with the lowest precipitation—storage levels remained severely negative throughout the year, with the lowest point occurring in May. This indicates that recharge deficits and intensified human impacts placed a heavy burden on groundwater systems. In the normal year (2007), GWS showed a net transition from surplus to deficit. Storage was positive from January to April but began a steep decline in May. Some slight recovery occurred in August due to increased precipitation and runoff, but overall groundwater levels remained low after June. In the wet year (2016), GWS levels were relatively low overall, despite higher precipitation. Although there was a seasonal peak in January, groundwater recovery remained weak throughout the year, suggesting that excessive consumption and delayed recharge continued to limit the system’s ability to replenish.

The combined intra-annual patterns revealed by the multi-year average and different hydrological year types highlight the seasonal sensitivity and complex response of groundwater dynamics to both human activities and climate variability. These findings emphasize the need for enhanced efforts to safeguard groundwater sustainability in Shanxi Province.

4.4. Spatial Changes in GWS

Figure 7 illustrates the spatial evolution of GWS in Shanxi Province from 2003 to 2023. Overall, GWS across the province exhibits a declining trend during this period, particularly pronounced in the southeastern region. Based on spatial distribution patterns, the evolution of GWS can be divided into three distinct stages:

Stage I (2003–2006): During this period, GWS was generally abundant and evenly distributed across Shanxi Province. The southeastern region—especially near the junction with the North China Plain—showed particularly high GWS, likely due to relatively abundant precipitation and recharge conditions. Groundwater remained stable throughout this stage, with minimal decline, and the southeast did not exhibit any significant signs of depletion.

Stage II (2007–2021): From 2007, the steady fall of GWS started, particularly in Southeast of the province. Likewise, the southeast became a never-recovering low-storage area with storage declining annually. This decrease was most apparent in years 2017 and 2021 as GWS growth rates of such considerable magnitude are negative for many patches. In addition, the lateral infiltration by surrounding areas to the groundwater may have contributed to the aggravation of groundwater deficit in southeast Shanxi due to the fact that southeast Shanxi was closer to the water-stressed regions of the North China Plain.

Stage III (2022–2023): Following 2022, a gradual recovery in GWS levels has been observed. Spatial disparities in GWS began to lessen, and the southeastern region showed some rebound. However, the recovery was limited and did not restore groundwater levels to earlier conditions. This may reflect improved local precipitation and more effective implementation of water resource management measures. Nonetheless, the proximity of the southeast to the North China Plain continues to exert pressure on the region’s groundwater system, limiting the extent of recovery.

To investigate the spatial characteristics of GWS in Shanxi Province, Figure 8 illustrates the monthly variation in GWS across the region from 2003 to 2023. The monthly average GWS values represent the multi-year averages for each corresponding month over the period 2003–2023. Specifically, the average GWS values from January to December are −119.42, −118.08, −111.33, −110.27, −129.61, −145.16, −134.41, −123.08, −109.34, −118.42, −111.55, and −133.15 mm, respectively. Evidently, the spatial patterns of GWS closely align with these monthly averages: GWS reaches its lowest levels in May and June, reflecting the annual minimum. During these months, the southeastern part of Shanxi Province also displays the most pronounced groundwater depletion hotspots. from January to December, the southeastern region consistently exhibits low GWS, with this pattern being most pronounced during summer. The transitional zone between southeastern Shanxi and the North China Plain emerges as a key area of GWS variability. The North China Plain, one of the driest regions in China, has long suffered from groundwater stress, and chronic over-extraction has further exacerbated water scarcity. Although the southeastern region of Shanxi receives relatively high precipitation, the GWS there remains insufficiently recharged. This is primarily due to intensive groundwater withdrawal—particularly for irrigated agriculture—and elevated evapotranspiration rates, especially during the summer months.

4.5. Linkage Between GWS in the North China Plain and Shanxi Province

Groundwater storage has significantly declined in southeastern Shanxi Province, adjacent to the North China Plain, creating an extended belt of groundwater reduction. Shanxi Province is located in the transitional zone between the Loess Plateau and the North China Plain. Its groundwater dynamics are influenced not only by climate change and human activities [52,53] but also by lateral recharge, groundwater extraction, and potential funneling effects from adjacent regions. To examine the relationship between GWS in the eastern Loess Plateau and that in the North China Plain, we employed correlation analysis, covariance-based tests, and causality analysis models to assess potential linkages between the two regions.

Correlation analyses were conducted using both annual and monthly data for GWS in Shanxi Province and the North China Plain (Figure 9). The results indicate a strong relationship: the monthly correlation coefficient reaches 0.89, suggesting high synchrony in intra-annual variations between the two regions over the 21-year period. Furthermore, the annual correlation coefficient is as high as 0.97, demonstrating a remarkable consistency in long-term trends. These findings confirm that GWS in eastern Shanxi Province (eastern Loess Plateau) evolves in strong synchrony with that of the North China Plain. This indicates that the groundwater system in Shanxi is significantly influenced by the hydrological dynamics of the neighboring North China Plain.

We calculated the spatial distribution of GWS in the North China Plain from 2003 to 2023 (Figure 10). The results reveal a prominent zone of groundwater depletion in the western part of the North China Plain, adjacent to Shanxi Province. The core of this depletion zone is primarily concentrated near the border between Hebei and Henan provinces. Notably, the groundwater depletion hotspots in Shanxi Province and the North China Plain exhibit a high degree of temporal and spatial synchronization, predominantly occurring between 2014 and 2021. During this seven-year period, the average minimum GWS value in the hotspot region of the North China Plain reached −507.7 mm, while the corresponding average extreme value in Shanxi Province was −481 mm. Over the seven-year period, the lowest yearly GWS values in hotspot regions of the North China Plain were always lower than those found in Shanxi Province. This shows that, each year, the groundwater conditions in those areas were worse than in Shanxi. So, groundwater depletion in the North China Plain was more serious and lasted longer during this time.

This area of groundwater depletion seems to be connected to the groundwater overdraft zone in southeastern Shanxi. This spatial link suggests that there may be a trans-regional groundwater funnel or a connected area of overdraft between the two places. These results show that the groundwater systems in the two regions may be linked through hydrogeological processes. So, it is important to develop a joint groundwater management plan. This plan should go beyond provincial borders and focus on shared water resources.

Although Figure 9 presents the regional average trends of GWS in Shanxi Province and the North China Plain, showing that GWS in Shanxi is slightly lower in most years, regional averages often mask intense local variations. In contrast, depletion hotspots can more effectively reflect zones of concentrated stress within groundwater systems and are thus crucial for identifying areas of potential risk.

To further compare the spatial characteristics of GWS changes between Shanxi Province and the North China Plain, this study employed GRACE data and used the Sen’s slope method to calculate annual trends of GWS from 2003 to 2023. The natural breaks (Jenks) classification method was then applied to identify areas with significant declining trends. Figure 11 illustrates the spatial distribution of GWS trends in the two regions during 2003–2023. The trends were derived using the Sen’s slope estimator and categorized using the Jenks natural breaks method. The results reveal significant spatial heterogeneity in groundwater evolution. In Shanxi, GWS trends ranged from −33.67 to −2.02 mm/yr, with the steepest declines mainly occurring in the southeastern part of the province. In contrast, the North China Plain experienced a broader and more intense decline, with trend values ranging from −38.24 to 1.87 mm/yr. The most severe declines were concentrated in the western part of the plain, adjacent to the eastern border of Shanxi.

This area forms a “depletion hotspot zone” that spans both regions, primarily located in a geomorphic transition zone. Overall, the North China Plain exhibits not only a wider spatial extent of GWS decline but also a faster depletion rate, indicating more severe pressure on groundwater resources. Therefore, compared with regional averages, focusing on the spatial variations of depletion hotspots provides a more accurate means of identifying high-risk areas within groundwater systems. This approach offers more effective support for water resource management and regulation decisions, and further reinforces the conclusion that “groundwater depletion in the North China hotspot area is more severe.”

In order to deepen the study on the association of GWS in the NCP and Shanxi province, we conducted the Engle–Granger cointegration test and the Granger causality analysis model. We initially subjected the GWS time series on Shanxi Province and the North China Plain to Augmented Dickey–Fuller (ADF) unit root test. The test statistics of ADF were −2.724 in Shanxi and −2.049 in North China Plain, and both the values were larger than the 5% critical values, which suggested that the two variables were non-stationary and therefore cointegrated relation could be tested. The Engle–Granger test showed that the residuals from the cointegration regression were stationary (

Table 2. Engle–Granger cointegration test results.

Variable	ADF Statistic	5% Critical Value	p-Value	Conclusion
GWSSX	−2.724	−3.431	0.226	Non-stationary
GWSNCP	−2.049	−3.431	0.575	Non-stationary
Residuals	−4.139 ***	−1.950	<0.01	Cointegration exists

***: denotes significance at the 1% level.

). The ADF statistic of the residuals was −4.139, with a p-value less than 0.01. This means the two regions have a cointegration relationship. The result shows a stable long-term link between GWS in Shanxi and in the North China Plain, with their GWS changes showing clear synchronization during the study period.

The Granger causality test (

Table 3. Granger causality test results.

Causal Direction	χ2 Statistic	p-Value	Conclusion
GWSSX → GWSNCP	12.739 ***	0.002	Causal relationship exists
GWSNCP → GWSSX	33.664 ***	0.001	Causal relationship exists

***: denotes significance at the 1% level.

) shows a two-way causal link between GWS in Shanxi Province and the North China Plain. This means the two groundwater systems affect each other. Both directions of causality are significant at the 1% level (p < 0.01), which shows that the interaction is strong and meaningful. The result suggests that Shanxi, which is upstream, helps recharge the groundwater system in the North China Plain. At the same time, long-term overuse of groundwater in the North China Plain has caused a wide and deep cone of depression. This large cone makes the decline in Shanxi’s groundwater worse. The two-way link shows that the groundwater systems in both areas are not separate. The two regions are connected, and changes in one can lead to changes in the other. This shows that there is a strong link between the two groundwater systems through water movement. Because of this, it is important to use a joint method to manage groundwater. Managing groundwater together across regions can help us see how groundwater changes in the long run. It can also help improve future plans and rules for using water.

Multiple analytical methods consistently indicate that there is a significant and robust hydrological connection in the evolution of GWS between Shanxi Province and the North China Plain. This linkage is reflected not only in the high temporal synchronization between the two time series, but also in the spatial continuity of overexploited groundwater zones across regional boundaries. The cointegration test further confirms the existence of a long-term equilibrium relationship between the GWS systems of the two regions. The persistent decline in GWS in Shanxi Province is significantly influenced by long-term overexploitation and the traction effect exerted by the North China Plain. As such, groundwater depletion in the North China Plain has become an important exogenous driving force in the evolution of groundwater in Shanxi.

4.6. Influencing Factors of GWS Evolution in Shanxi Province

We analyzed the correlations between each of these factors and GWS in Shanxi to determine their relative importance (Figure 12). The results show a remarkably strong correlation (r = 0.97) between GWS in the North China Plain and that in Shanxi Province, indicating a high degree of temporal synchronization and suggesting a close interconnection in their groundwater evolution processes. In contrast, climatic variables—precipitation, evaporation, and runoff—exhibited lower correlation coefficients with GWS, indicating relatively limited influence. These variables remained relatively stable during 2003–2022 (Supplementary Figure S3a–c), suggesting that although climate plays a role in groundwater recharge, it cannot adequately explain the long-term decline in GWS in Shanxi Province. Among human factors, water consumption and coal mining showed significant negative correlations (r < −0.8). Although total water consumption is strongly negatively correlated with changes in GWS, the time series reveals noticeable fluctuations around 2011 (Supplementary Figure S3d), rather than a stable linear increase. However, these fluctuations are not directly reflected in the GWS trends, which may suggest that more dominant driving factors have masked the direct impact of water consumption changes. Coal mining activities also exhibit a significant negative correlation with GWS (r = −0.81). The coal production in Shanxi shows a clear upward trend during 2003–2022 (Supplementary Figure S3e), reflecting the province’s role as one of China’s key energy bases. Large-scale coal extraction can disrupt aquifer structures and alter natural groundwater flow paths through subsidence and the formation of goaf zones, which subsequently weakens recharge capacity. Moreover, coal mining operations are typically accompanied by artificial drainage, which further accelerates groundwater loss and serves as a dominant anthropogenic driver of GWS decline. NDVI also exhibits gradual increases until 2022, followed by a slight decline (Supplementary Figure S3f). These factors have exacerbated pressure on groundwater systems. NDVI, representing vegetation cover, showed a weak negative correlation (r = −0.53), indicating that vegetation change in the Loess Plateau has had only a limited influence on GWS. This analysis is supported by long-term variable trends (Supplementary Figure S3) and inter-factor correlation results (Figure 12), which together offer a mechanistic interpretation of how cross-regional groundwater linkages and anthropogenic stressors outweigh climatic influences.

Although the direct impact of climatic factors on GWS in Shanxi Province is relatively limited compared to human-driven factors, their potential synergistic or aggravating effects should not be overlooked. Climate change may indirectly intensify groundwater stress through mechanisms such as increased evapotranspiration, seasonal shifts in precipitation, and a higher frequency of extreme drought events. When these climatic pressures combine with intensive human extraction activities for agriculture, industry, and urban use, they can further accelerate the rate of groundwater depletion and weaken aquifer recovery capacity. In particular, under the increasingly evident warming and drying trend in northern China, the contradiction between rising water demand and declining surface water availability has become more severe, leading to increased reliance on groundwater. Therefore, even if the independent statistical contribution of climate variables appears limited, their interaction with human activities may still amplify long-term risks to groundwater systems. Under the context of climate change, understanding and addressing this coupled mechanism is essential for developing adaptive groundwater management strategies.

Overall, the evolution of GWS in Shanxi Province is driven by a combination of factors, with the most significant being the pumping funnel effect induced by long-term groundwater overexploitation in the North China Plain. While climatic influences remain relatively stable, among various drivers, intensified water usage and extensive coal mining caused by human activities have become one of the important contributors to the sustained reduction in GWS.

5. Discussion

5.1. Evaluating the Impacts of Climate Change, Human Activities, and Groundwater Evolution in the North China Plain on GWS Trends in Shanxi Province

To gain a deeper understanding of how climate change and human activities have influenced the evolution of GWS in Shanxi Province, we applied a contribution analysis model to quantify the impact of each factor (

Table 4. Relative contribution rates of climate change, human activities, and external groundwater linkages to GWS.

Category	Influencing Factor	Contribution Rate (%)
Climatic Factors	Precipitation	0.08
	Evaporation	−0.83
	Runoff	−0.22
Human Activities	Water Consumption	−18.49
	Coal Mining	−25.70
	NDVI	−0.78
External Factors	GWSNCP	−53.89

). The results show that GWS variation in the North China Plain exerts the strongest negative influence, with a relative contribution of −53.89%. This highlights a significant externality effect: regional groundwater is under continuous exploitation, and the decline in groundwater levels in the North China Plain can spatially propagate to Shanxi Province, indicating the presence of a clear cross-regional linkage mechanism.

Among the human-related factors, coal mining and water consumption contributed −25.7% and −18.49%, respectively. These findings suggest that, as a major coal-producing region in China, the intensification of coal extraction and water consumption in Shanxi has significantly accelerated the depletion of groundwater resources. Coal mining not only disrupts aquifer structures and alters groundwater flow paths, but can also sever the hydraulic connectivity between aquifers, thereby reducing recharge capacity [54,55,56]. Moreover, large-scale coal mining is typically accompanied by artificial dewatering, which exacerbates groundwater loss and leads to further declines in groundwater levels [57,58].

In contrast, natural climatic variables—precipitation, evaporation, and runoff—exerted relatively modest impacts on GWS in Shanxi. Specifically, the contribution of precipitation was only 0.08%, while evaporation had a slightly negative effect (−0.83%), and runoff also showed a minor negative contribution (−0.22%). These results indicate that although climatic factors such as precipitation and evaporation influence groundwater recharge to some extent—particularly in semi-arid regions like the Loess Plateau—human activities and external hydrological pressures remain the primary drivers of long-term groundwater decline in Shanxi Province.

As a whole, the combined contribution of anthropogenic factors and extra-regional influences exceeds −98%, underscoring their dominant role in driving the evolution of GWS in Shanxi Province. The North China Plain, as the most economically developed, densely populated [59,60,61], and water-scarce region in China, has long experienced large-scale groundwater overexploitation, resulting in the formation of extensive funnel-shaped depressions [62,63]. This sustained depletion not only directly reduces GWS within the North China Plain, but also significantly impacts adjacent areas such as Shanxi Province through cross-regional hydraulic connectivity.

Among human activities, the impact of coal mining is particularly pronounced. Large-scale coal extraction has disrupted the natural circulation of groundwater systems by altering flow paths and reducing recharge capacity. Existing research has demonstrated that coal mining activities can damage aquifer structures, form subsidence zones and voids, alter natural groundwater flow paths, and ultimately reduce the region’s capacity for groundwater recharge [64,65]. In addition, mine drainage operations require continuous groundwater pumping to keep the working areas dry, which leads to substantial groundwater loss. Regarding water consumption, both agricultural and industrial sectors—especially in regions lacking sufficient surface water—rely heavily on groundwater, thereby increasing extraction intensity. In many parts of Shanxi Province, groundwater has become the primary source for irrigation and industrial cooling, resulting in long-term pressure on aquifers. These processes, combined with low natural recharge in semi-arid regions, are among the key factors contributing to the persistent decline in GWS.

Overall, these findings emphasize the urgency of addressing both regional and inter-regional groundwater pressures. In particular, the cross-regional groundwater linkage effect and the intensifying impact of human activities on the groundwater system require targeted management and policy interventions.

Although this study mainly focuses on a retrospective analysis of GWS, the spatiotemporal evolution trends of GWS and the identified cross-regional hydrological linkages between Shanxi Province and the North China Plain provide important references for future groundwater resource management. The findings suggest that if current patterns of human activity and groundwater extraction continue—especially under the long-standing hydraulic pressure from the North China Plain—the risk of groundwater decline in southeastern Shanxi will further increase. This may expand the extent of local groundwater funnel areas and cause irreversible degradation of the regional water system.

Although climate-related studies have examined long-term patterns in northern China, the role of climate in groundwater dynamics remains complex and context-dependent. Our findings indicate that, at the provincial level, climatic factors show relatively limited direct influence on GWS in Shanxi. However, the indirect pathways through which climate may impact groundwater systems—such as changes in recharge timing, variability in soil moisture, and feedback between surface and subsurface processes—warrant further investigation. These interactions may not be easily captured in retrospective models but could become more pronounced under future climate scenarios. Incorporating these uncertain yet plausible effects into groundwater assessments will be critical for accurately identifying future risks and building more resilient water management frameworks across regions.

Therefore, future groundwater assessment and management should include the dynamic changes of the climate system in the causal analysis framework, and consider them together with regional hydrological linkages. This can help build a more complete understanding of the drivers behind groundwater evolution. To enhance the practical value of the research for water resource planning, future studies should also consider scenario-based simulations. For example, typical scenarios could be developed and combined with climate change projections to simulate possible future groundwater pathways. This approach can improve the ability to evaluate the effects of different policy interventions and provide more scientific and forward-looking guidance for regional water regulation.

5.2. Disregarding External Influences May Lead to Overestimation of Local Climate or Human Activities Impacts on GWS

Shanxi Province, located in the eastern Loess Plateau and adjacent to the North China Plain, has experienced a significant decline in GWS, particularly in its southeastern region. Through correlation analysis, Engle–Granger cointegration tests, and Granger causality models, we demonstrate a strong hydraulic connection between the groundwater systems of Shanxi Province and the western North China Plain. The persistent “pumping funnel effect” caused by long-term groundwater overexploitation in the North China Plain has emerged as a major external driver of groundwater depletion in Shanxi Province—one that cannot be ignored in regional assessments of groundwater dynamics.

However, conventional studies often focus only on climate change and human activities, while overlooking hydrological pressures from adjacent regions. For instance, previous research typically attributes groundwater depletion in solely to local climate change and anthropogenic pressures. To better understand the influence of external hydrological pressure, we compared the contribution of each factor under models with and without external variables (Figure 13). Yet, our contribution model analysis reveals that when the groundwater influence from the North China Plain is excluded, coal mining becomes the most dominant factor, with a magnified contribution of −52.11%. Water consumption follows closely behind. Notably, the contribution of water consumption is exaggerated from −18.49% to −37.69% when external influences are disregarded—indicating a serious overestimation of its actual impact. It is also noteworthy that NDVI exhibits a negative contribution to GWS in both model scenarios. When external groundwater influences from the North China Plain are excluded, the negative impact of NDVI is significantly amplified—from −0.78% to −7.91%—making it the third-largest negative contributor. This magnification effect suggests that disregarding external hydrological pressures may overstate the role of local land use changes in groundwater depletion, potentially leading to misinterpretations in attribution analysis and misguided decisions in water resource management and ecological restoration. NDVI, as a key indicator of vegetation coverage, may influence groundwater recharge and discharge indirectly by altering evapotranspiration, soil infiltration, and surface runoff. Several studies have indicated that vegetation restoration in semi-arid regions can increase evapotranspiration and reduce effective groundwater recharge, especially in water-scarce environments like the Loess Plateau [6,66]. In such ecologically fragile regions, large-scale vegetation restoration may improve ecosystem stability but can also intensify pressure on groundwater systems. Therefore, the role of NDVI must be interpreted in the context of local hydrological conditions.

Overall, it is essential to scientifically reveal and account for external influences—particularly those from neighboring regions—when analyzing the drivers of groundwater evolution. Such consideration is critical for developing accurate, regionally coordinated, and sustainable groundwater management policies.

5.3. Policy Suggestions

The results of this study show that there is a strong two-way Granger causality between GWS in Shanxi Province, which lies in the eastern part of the Loess Plateau, and the North China Plain. This means that the two areas are connected by a shared groundwater system. The connection happens in both space and time, and there are strong water-related links between the two regions. Because of this, the common way of managing groundwater by using administrative borders should be changed. It is important to promote regional cooperation and shared management. If we only depend on local control measures, the problem of falling groundwater levels will likely remain. Policies should not only pay attention to local overuse. They should also focus on how strong, flexible, and connected the full regional groundwater system is. This is important to make sure that groundwater can be used in a sustainable way in both the North China Plain and the whole Loess Plateau.

5.4. Limitations

This study systematically investigates the driving mechanisms of GWS variation in Shanxi Province by integrating GRACE-GLDAS data, cointegration tests, Granger causality analysis, and contribution models. It also incorporates cross-regional hydrological linkages into the analytical framework for the first time, offering new perspectives and complementary insights for groundwater attribution research. However, as with all quantitative analyses, the employed methods involve certain technical assumptions and application constraints that leave room for further refinement and extension.

The cointegration test and Granger causality analysis are widely used to identify long-term equilibrium relationships and temporal causality among time-series variables. In this study, they effectively reveal the interconnections between the groundwater systems of Shanxi Province and the North China Plain. Nevertheless, these methods rely on the stationarity of variables and sufficient time-series length, and are sensitive to lag length selection. Future research could enhance the robustness and explanatory power of the conclusions by incorporating longer time series and conducting comprehensive robustness checks.

In the process of multiple regression modeling, ridge regression was employed to address potential multicollinearity among independent variables, which improved model stability and interpretability. However, linear models remain limited in capturing potential nonlinear relationships and time-lag effects within complex systems. Future studies may consider incorporating nonlinear approaches or machine learning techniques—such as random forests or support vector machines—to enrich analytical dimensions and improve predictive performance, while maintaining a balance with model interpretability.

Moreover, although this study reveals a potential cross-regional hydraulic connection between Shanxi Province and the North China Plain using remote sensing data combined with statistical models, statistical correlations alone do not imply physical causation. We have included a spatial groundwater level distribution map, which shows a west-to-east decreasing trend in groundwater levels from Shanxi to the North China Plain (Figure 3). This spatial pattern substantially compensates for the lack of hydrological linkage evidence and provides hydrological background support for the statistical inference of the “pumping funnel effect.” However, the map remains an indirect form of evidence and is insufficient to fully verify the hydraulic connection between the regions. Future research should incorporate direct hydrogeological evidence, such as cross-sectional profiles, groundwater contour maps, or well-based monitoring sections, to further confirm the hydraulic gradient and flow direction across regions.

Overall, this study has systematically identified the main regional and cross-regional drivers of groundwater variation based on the available data and methods, and the findings provide meaningful scientific references. Further research could focus on expanding temporal scales, diversifying analytical methods, and deepening the understanding of spatial mechanisms to provide more robust technical support for the sustainable management of groundwater systems.

6. Conclusions

This study quantified the spatial and temporal variations in GWS in Shanxi Province, located in the eastern Loess Plateau, from 2003 to 2023, using GRACE satellite data combined with GLDAS datasets. The spatio-temporal linkage between GWS in Shanxi Province and the North China Plain was examined through correlation analysis, Engle-Granger cointegration tests, and Granger causality models. Additionally, a contribution analysis model was employed to quantitatively assess the driving forces behind groundwater changes. The main conclusions are as follows:

(1)

Based on GRACE–GLDAS data, GWS in Shanxi Province exhibited a clear declining trend from 2003 to 2023, with an average annual decrease of −17.27 ± 1.4 mm/yr, resulting in a cumulative loss of approximately 56.58 km³ over the study period. Spatially, the most significant groundwater depletion occurred in the southeastern region of the province.

(2)

Correlation analysis of annual and monthly GWS data between Shanxi Province and the North China Plain reveals strong synchronization between the two regions. The monthly correlation coefficient reached 0.89, while the annual coefficient was 0.97, indicating high consistency in long-term trends. A distinct groundwater decline belt has formed along the transitional zone between Shanxi and the North China Plain.

(3)

The Engle–Granger cointegration test and Granger causality test both show that there is a stable and long-term link between the GWS in Shanxi Province and the North China Plain. The strong and two-way causality suggests that changes in groundwater in one region can clearly affect the other. This mutual influence means that the two groundwater systems are closely connected and respond to each other. These findings point to a strong hydrological link between the two areas. They also show why it is important to include cross-regional interactions in the way groundwater resources are managed.

(4)

The groundwater decline in the North China Plain, as an external hydrological pressure, contributed −53.89% to GWS variation in Shanxi Province. Combined with human activities, the total contribution exceeded −98%, while the influence of climatic factors was comparatively minor. These results indicate that human activities and cross-regional hydrological pressure are the dominant drivers of GWS decline.

(5)

If cross-regional hydrological interactions are ignored and GWS variation is attributed solely to climatic or anthropogenic factors, it may lead to an overestimation of internal influences and undermine the scientific basis of water resource policy decisions. It is important to use a full framework. This framework should include hydrological linkages between different regions. For areas like Shanxi, which lie between different groundwater systems, such a method is very important for making correct groundwater management decisions.