Floods and landslides frequently occur as compound hazards in tropical regions, causing severe socio-economic losses. In November 2025, Tropical Cyclone Senyar triggered widespread floods and landslides across Aceh Province, Indonesia. To identify hazard-prone areas and understand the underlying driving mechanisms, the interpretable susceptibility models using Extreme Gradient Boosting (XGBoost) and Shapley Additive Explanations (SHAP) are developed based on this compound hazard event. The optimized XGBoost models achieve excellent performance, with AUC values of 0.9712 and 0.9501 and AP values of 0.9589 and 0.8654 for flood and landslide susceptibility, respectively. The susceptibility maps reveal distinct spatial patterns, with flood-prone areas concentrated in low-lying coastal plains and landslide-prone areas mainly distributed in inland mountainous regions. SHAP analysis reveals that elevation dominates flood susceptibility, whereas landslide susceptibility is jointly controlled by topographic and hydrological factors. These findings provide valuable insights for disaster risk reduction and land-use planning in tropical regions.

Flood-landslide compound hazards,Susceptibility mapping,XGBoost-SHAP model