Excessive levels of heavy metals in soil can significantly impact human health and ecological safety. Evaluating heavy metal pollution and identifying its sources are crucial for environmental management. This study investigates the status of heavy metal pollution in a southeastern region of China and aims to identify its sources using data from the first national soil pollution survey, which includes 282 sampling points. Indicators such as the geoaccumulation index (Igeo), the potential ecological risk index (RI), the hazard index (HI), and the total lifetime cancer risk (TLCR) were utilized to assess contamination levels. Geographic information systems (GIS), positive matrix factorization (PMF) receptor modeling, cluster analysis (ClusA) and human health risk assessments were integrated to analyze the sources of heavy metals. The results indicate that agricultural pollution sources have a minor impact on overall heavy metal contamination, with low ecological risk levels in the eastern and western regions. In contrast, the central region exhibited moderate risk, with areas of extremely high risk distributed across the central-west and central-south regions. PMF analysis identified pollution sources including natural origins, coal combustion, industrial emissions, and traffic, with contributions of 17.62%, 18.50%, 28.35%, and 35.56%, respectively. Overall, the carcinogenic risk in the study area is not high. Targeted recommendations were made in response to the pollution situation in the study area. This research enhances our understanding of heavy metal pollution in the soil of the study area and provides a reference for pollution source delineation in other regions.

The ongoing global decline in groundwater levels poses significant challenges for sustainable water management. Satellite gravity missions, such as the Gravity Recovery and Climate Experiment Follow-On (GRACE-FO), provide valuable estimates of groundwater storage changes at regional scales. However, the relatively coarse spatial resolution of these satellite data limits their direct applicability to local groundwater management. In this study, we address this limitation for China by analyzing groundwater monitoring data from 108 cities with shallow groundwater use and 37 cities with deep groundwater use from the period 2019–2022, integrating in situ groundwater level records, official monitoring reports, monthly dynamic data, and GRACE-FO-derived groundwater storage estimates. Our findings reveal rapid groundwater depletion in northern China, especially in Xinjiang and Hebei Provinces. Fluctuations in shallow groundwater levels in Beijing and Jiangsu are closely related to precipitation variability. For deep aquifer regions, GRACE-FO-derived groundwater storage changes show a moderate Pearson correlation coefficient of 0.45 and groundwater level variations. Regional analysis for 2019–2021 in the Northeast Plain and the Huang–Huai–Hai Basin indicates better agreement between satellite-derived storage and groundwater levels, with a Pearson correlation coefficient of 0.58 in the Huang–Huai–Hai Basin. Groundwater level dynamics are strongly influenced by both precipitation and pumping, with an approximate three-month lag between precipitation events and groundwater storage responses. Overall, satellite gravity data are suitable for use in regional groundwater assessment and could serve as valuable indicators in areas with intensive deep groundwater exploitation. To enable fine-scale groundwater management, future work should focus on improving the spatial resolution through downscaling and other advanced techniques.

Accurate and rapid prediction of groundwater levels (GWL) is essential for effective groundwater management. Machine learning models are efficient tools for GWL prediction, but individual models often suffer from limited generalization due to inherent randomness. This study proposed a stacking-based GWL prediction framework suitable for arid regions in Northwest China. Feature variables affecting GWL were selected using variable importance in projection (VIP). Then, three machine learning models—artificial neural networks (ANN), random forests (RF), and Light Gradient Boosting Machine (LightGBM)—were developed, and their outputs were integrated using a support vector regression (SVR)-based stacking method to enhance the accuracy of GWL prediction. The results show that the factors of influencing GWL changes vary significantly across different regions, and selecting the most contributive feature variables is beneficial for model construction. Among the individual models, the RF model demonstrated higher accuracy and more stable performance, outperforming the ANN and LightGBM models. However, individual models exhibited poor generalization during validation. In contrast, the stacking model maintained high performance, demonstrating superior generalization. Compared to the best-performing individual model (RF) in validation period, the Nash–Sutcliffe efficiency ( NSE ) and Kling–Gupta efficiency ( KGE ) of stacking model improved by 0.11–0.66 and 0.05–0.41, the correlation coefficient ( R ² ) increased by 0.05–0.3, and root mean square error ( RMSE ) reduced by 0.01–0.1 m. In the stacking simulation, RF had the highest average contribution (80.2%), followed by ANN (13.9%) and LightGBM (5.9%). This study provides a stacking simulation framework based on machine learning methods for precise groundwater level simulation, which can serve as a reference for groundwater level simulation in other regions.

Intensive groundwater exploitation in the Wei River Basin (WRB) has caused severe depletion. While Gravity Recovery and Climate Experiment (GRACE) satellites monitor these changes, their coarse resolution fails to account for soil erosion-induced mass loss on the Loess Plateau limit basin-scale accuracy. To address this, we developed the groundwater storage model to dynamically downscale GRACE data to the resolution of 0.05° grid. This physically based model integrates Darcy’s law, water-balance principles, and an innovative correction for soil erosion mass migration. Validated against well data with maximum correlative coefficient of 0.73, the model reveals a severe groundwater decline of 55.89 × 10⁸ m³/year from 2003 to 2023. The high-resolution results successfully capture localized over-extraction hotspots in the Guanzhong Plain, showing a west-to-east decreasing gradient in groundwater storage along the Wei River channel. Furthermore, quantitative analysis indicates that human activities, including primarily agricultural and urban extraction, are the overwhelming drivers, accounting for over 80% of storage depletion. This framework provides a robust methodological reference for isolating groundwater signals in erosion-prone regions and supports refined water management in semi-arid basins.