How coal mining sites respond to extreme rainfall: Deep-learning-enhanced InSAR observations across the 2021 Henan 7.20 extreme rainfall

Coal mining-induced subsidence poses environmental hazards, particularly when interacting with extreme rainfalls that trigger flooding and hydrogeological imbalances. InSAR enables monitoring of mining-induced deformation, but its application across extreme rainfall remains limited, constraining both operational hazard mitigation and the understanding of hydrologically-controlled deformation in intensively-mined regions. Here, we integrate deep learning-enhanced InSAR to investigate how the 2021 Henan “7.20″ extreme rainfall event reshaped ground deformation patterns in coal mine goafs. This approach identified an additional 15.123 km 2 of rapid subsidence, effectively filling data gaps within mining-induced subsidence funnels that traditional methods missed. The time-series deformation and surface-water maps from Sentinel-1 SAR images (2017–2023) reveal that the post-event water accumulation closely aligns with InSAR-detected subsiding funnels. We observe that the extreme rainfall shortly reversed the ground subsidence to uplift in mined areas, lasting ∼4‒5 months. Notably, an area near a fault has been experiencing cumulative uplift of ∼30 cm without sign of chasing by the end of 2023, highlighting the role of rainfall-driven stress redistribution in inducing both short- and long-term uplift through hydraulic gradients. This work establishes a deep learning-enhanced InSAR framework that provides a critical quantitative basis for flood-risk assessment before extreme rainfall in ongoing mining sites. The revealed deformation patterns offer new insights into the interplay between climate extremes and anthropogenic subsidence.