Abstract: The surface deformation,water accumulation expansion and vegetation degradation caused by coal mining subsidence
have become the key problems restricting the ecological security and sustainable development of mining areas. The existing
research focuses on the change of single environmental factor,and lacks systematic analysis of the nonlinear coupling
mechanism of subsidence-hydrology-ecosystem. Taking Huainan mining area as a typical case,the analysis framework of multisource
remote sensing data and machine learning model (XGBoost-SHAP) was constructed,and the evolution of surface subsidence,
dynamic monitoring of water accumulation area,land use change and ecological response threshold identification were
systematically carried out. The study reveals that the subsidence pattern has changed from wide-area continuous subsidence
from 2000 to 2015 to the coexistence of scattered subsidence and local uplift from 2015 to 2020. The spatial coupling strength
of water accumulation-subsidence was significantly enhanced,and the area of water accumulation area expanded by 37. 6% in
20 years. There was a significant threshold response of the ecosystem to environmental factors. NDVI showed a significant nonlinear
threshold effect on factors such as slope (1. 37°) and temperature (16. 8 ℃). This study innovatively realized the systematic
analysis of the three elements of ′subsidence-water-vegetation′ in the mining area,and provided a scientific basis for
ecological restoration zoning,risk early warning and sustainable management of coal mining subsidence areas.

Key words: Huainan mining area,coal mining subsidence,dynamics of waterlogged areas,ecological restoration,threshold
effect,XGBoost model