Spatial distribution patterns and hazard assessment of co-seismic landslides triggered by the 2022 Luding earthquake (MS 6.8): An RF-AdaBoost model approach

Abstract

On September 5, 2022, a magnitude 6.8 earthquake struck Luding County, Sichuan Province, China, at a depth of 16 km, triggering numerous co-seismic landslides. Using multi-temporal satellite imagery, the authors identified 7463 co-seismic landslides with varying spatial distributions. Focusing on the differences between the western and eastern sections of the Xianshuihe fault zone, the influences of elevation, slope, aspect, topographic wetness index (TWI), stream power index (SPI), distance to river, distance to seismogenic fault, seismic intensity, lithology, and land surface temperature (LST) on the co-seismic landslide distribution were analyzed. Random Forest-Adaptive Boosting (RF-AdaBoost), Support Vector Machine (SVM) and Back Propagation Neural Network (BP) models were used to assess landslide hazards in the region. The RF-AdaBoost model performed the best, achieving an Area Under the Curve (AUC) of 0.954. The zones of high co-seismic landslide hazard were primarily concentrated along the Xianshuihe fault zone and Dadu River. These findings provide valuable insights into seismic landslide hazard assessments and offer scientific guidance for disaster prevention and mitigation strategies.