Revealing driving factors of groundwater level depth changes in China using interpretable machine learning and the optimal GeoDetector model

Based on landform types, hydrogeological structure, and groundwater types across mainland China, the region was divided into five hydrogeological zones. We compiled GLD records from 1742 monitoring wells across China (2005–2016) and analyzed 12-year spatiotemporal variations using the Mann–Kendall test with Sen’s slope estimator. We further used interpretable machine learning (IML) techniques and the optimal GeoDetector model to explore the influences of human activities, climate, vegetation, soil, surface water, groundwater, and topography on GLD, and to identify the key drivers of GLD dynamics. At the national scale, groundwater systems generally show a trend in which degradation outweighs improvement, particularly in the Huang–Huai–Hai Plain. GLD in both unconfined and confined aquifers follows a “deep in the west and shallow in the east” pattern. Precipitation (PRE), temperature (TEM), and groundwater abstraction (GWA) are the key factors controlling GLD changes. Results from SHapley Additive Explanations (SHAP) and GeoDetector consistently indicate that groundwater system responses are primarily driven by the coupled effects of multiple factors rather than by any single driver. SHAP further reveals pronounced nonlinear and threshold responses among interacting factors. Overall, the coupled influences of GWA and climatic variables constitute the core mechanisms shaping GLD variations. These findings provide a scientific basis for developing region-specific groundwater management strategies in China. • Interpretable machine learning and GeoDetector disentangle drivers of GLD. • GLD across China shows a national trend in which degradation exceeds recovery. • Precipitation, temperature, and abstraction are identified as key GLD factors. • GeoDetector emphasizes spatial synergies, while SHAP captures nonlinear effects. • Coupled climate–abstraction effects drive groundwater dynamics across China.