Spatiotemporal dynamics and drivers of groundwater pollution risk: insight from multi-source remote sensing and...

The escalating threat of groundwater pollution, driven by rapid urbanization and industrialisation, requires a shift from static to dynamic risk-assessment frameworks. Traditional methods such as DRASTIC effectively assess intrinsic aquifer vulnerability; However, they largely overlook the critical dimension of anthropogenic pollution sources and their spatiotemporal dynamics. To address this gap, we introduce a novel integrated framework. This framework synergizes multi-source remote sensing, geographic detection, and machine learning for the dynamic assessment of groundwater pollution risk (GPR). The 22-year (2001-2022) spatiotemporal evolution of GPR in Handan City, China, was innovatively reconstructed. A Long Short-Term Memory (LSTM) model is employed for this reconstruction. The LSTM model outperformed other algorithms, including random forest (RF), support vector machine (SVM), and convolutional neural network (CNN). And its performance was validated against recent field data. Results indicate that high-risk areas are concentrated predominantly in western mining zones and central industrial regions. These areas show a strong spatial correlation with anthropogenic pollution loads rather than with intrinsic vulnerability alone. Importantly, the Geodetector analysis demonstrates that two-factor interactions, such as precipitation × gross domestic product (GDP) and precipitation × compounded night light index (CNLI), possess significantly greater explanatory power than any single factor. This reveals a multi-source synergistic mechanism driving GPR. Temporal analysis further reveals pronounced cyclical fluctuations in GPR, strongly linked to precipitation patterns. A significant long-term upward trend is also observed, driven by urban expansion as indicated by CNLI. The primary novelty of this work lies in the successful transition from a static snapshot to a dynamic, high-resolution reconstruction of GPR. These findings provide a powerful tool for predictive hotspot identification and offer practical guidance for targeted monitoring, sustainable groundwater management, and early-warning systems, ultimately supporting ecological security in rapidly developing regions.