Comprehensive analysis and assessment of extreme rainfall-induced clustered landslides: a case study of southern Qingyuan city, Guangdong province, China, in June 2020
Chenchen Xie, Chong‐Yu Xu, Xiwei Xu, Yuandong Huang, Zhi‐Qiang Zhang, Hao Li, Hourong Zhang, Dengjie Zhu
Abstract
Clustered landslide events often cause significant losses, making comprehensive analysis and evaluation of such events crucial. This process includes three steps: establishing a landslide database, analyzing distribution patterns, and constructing a landslide real-probability hazard assessment map. In this study, comprehensive analysis and evaluation were conducted based on clustered landslides induced by extreme rainfall in June 2020 in southern Qingyuan City, Guangdong Province, China. A total of 6,660 landslides with a total area of 12.79 km2 were identified in the 6,989.73 km2 study area. Landslides were densely distributed in the central, southwestern, and northeastern parts of the study area. Twelve influencing factors were integrated, including geomorphology, geological hydrology, accumulated rainfall, etc. Distribution patterns and triggering mechanisms were analyzed in detail. A real-probability hazard map was developed using the Fast and Lightweight AutoML (FLAML) framework and Random Forest (RF) model. The predicted landslide occurrence probabilities showed strong alignment between moderate-to-high risk areas and actual landslide distributions. The hazard map was classified into five levels, with “Extremely high” and “High” zones containing the majority of landslides. Through 100 different sampling analyses of factor weights, it was found that Accumulated rainfall, Elevation, Strata, Topographic relief, and Slope were the main influencing factors in this landslide event.