Landslide susceptibility under two decades of land use change (2000–2020)

To address the insufficient consideration of dynamic driving factors in landslide susceptibility mapping (LSM), this study investigates the contribution of land use and land cover change (LUCC) to the spatial distribution of landslides in Sichuan Province, China, a region highly prone to geological hazards. An XGBoost model was used to integrate multi-source spatial data within a unified 30 m grid to produce a province-wide susceptibility map. SHAP-based attribution analysis was employed to quantitatively assess the relative importance of predisposing factors. The results indicate that dynamic LUCC exerts a stronger explanatory influence on landslide susceptibility than several static geological and environmental factors. High susceptibility is concentrated in low-elevation are as (