Flood risk assessment and driver analysis in mining subsidence-affected high-groundwater watersheds: a coupled HEV and machine learning approach

Coal mining subsidence reshapes surface topography and hydrological connectivity, intensifying compound flood hazards in high-groundwater-level mining regions. Conventional flood risk assessments often treat mining areas as ordinary land-use units and insufficiently account for the coupled effects of subsidence-induced depressions, shallow groundwater and recurrent inundation. To address this gap, this study develops an integrated flood risk assessment and attribution framework for high-groundwater mining areas. The framework combines the hazard-exposure-vulnerability (HEV) model with subsidence-specific indicators, an integrated AHP–EWM weighting method, and LightGBM-based attribution analysis. Taking the South Four Lakes Basin as a case study, flood hazard, exposure, vulnerability, and integrated flood risk were mapped at watershed and sub-catchment scales. The results show that high flood risk is mainly concentrated in downstream lake districts, urbanized and riparian areas, and mining-affected regions. The proportion of high-risk zones is substantially higher in coal mining areas (46.63%) and subsidence areas (24.83%) than in non-mining (11.85%) and non-subsidence areas (15.93%). Validation using 381 historical high-frequency inundation points indicates that 70.34% of these points fall within high-risk zones in the AHP–EWM based risk map, including 37.80% in very high-risk zones, demonstrating good consistency with historical inundation evidence. LightGBM results reveal clear scale-dependent driving mechanisms. At the watershed scale, precipitation and hydrological factors dominate (44.09%) flood risk formation, whereas subsidence-related variables show limited independent contributions (4.27%). In contrast, within mining areas, the contribution of subsidence characteristics increases to 11.07%, and subsidence-induced topographic factors (6.50%) become important local drivers. These findings indicate that mining-induced geomorphic change modifies flood-formative environments by altering terrain gradients, and enhancing persistent waterlogging. The proposed framework improves the identification of localized flood risk amplification mechanisms and provides practical support for differentiated flood management in subsidence-affected regions.