Research on the evaluation of landslide susceptibility based on the integration of endogenic and exogenic driving forces: A case study of Diqing Tibetan Autonomous Prefecture in the Hengduan Mountains

Abstract

The Hengduan Mountains (HMs), located on the southeastern edge of the Tibetan Plateau, are marked by complex terrain and frequent natural hazards. Accurate delineation of landslide susceptibility zones in this region is essential for enhancing local safety. This research centers on the Diqing Tibetan Autonomous Prefecture (DTAP), the core area of the HMs, and analyzes landslide susceptibility influenced by both endogenic and exogenic factors. Nineteen conditioning factors were selected, including two novel indicators of endogenic forces: distance from hot springs and gravitational acceleration. A Random Forest (RF) model combined with Shapley Additive Explanations (SHAP) was used to assess susceptibility. The RF-SHAP model achieved an accuracy of 93.65%. Notably, gravitational acceleration and distance from hot springs ranked second and ninth, respectively, in SHAP-based importance. These findings demonstrate the model’s effectiveness for landslide susceptibility mapping in tectonically active regions and offer practical insights for disaster prevention and mitigation.