Explainable multi-sensor geospatial modeling of forest biomass dynamics under compound climate extremes in subtropical monsoon regions

Climate extremes are intensifying worldwide, yet the mechanisms by which forest biomass responds to compound climatic and anthropogenic pressures remain poorly resolved. Here, we integrate multi-sensor remote sensing with explainable machine learning to quantify the interactions between multiple drivers and changes in aboveground biomass density (ΔAGBD) across a subtropical monsoon region of China during 2000–2019. Annual ΔAGBD maps were derived from Landsat and GEDI, and hotspots of climatic extremes were delineated using ETCCDI(Expert Team on Climate Change Detection and Indices). Across 92 predictors, attribution indicates that—even within areas exposed to extremes—AGBD is not driven by extreme events alone. Bioclimatic and extreme-climate variables dominate overall variability, whereas topography and human disturbance strongly modulate their effects. Gradient analyses further show that climatic influences intensify with elevation and heat load, whereas anthropogenic impacts remain pronounced at low population densities and in areas distant from impervious surfaces. Collectively, these findings demonstrate that biomass dynamics under climatic extremes do not arise from a single climatic forcing but from the mutual regulation of climate, topography, and human pressure. This nonlinear, compound-mechanism framework provides a transferable basis for assessing ecosystem vulnerability and designing adaptive, resilience-oriented management strategies under intensifying climate extremes.