Scholarlink

<p>With the increasing frequency of extreme rainfall events, clustered shallow landslides have become a major geological hazard worldwide, particularly in the hilly regions of southern China. The June 2024 extreme rainfall event along the Guangdong–Fujian–Jiangxi tri-province boundary triggered the most extensive clustered shallow landslides in China over the past few decades. This study develops an integrated framework coupling YOLOv8s-based automated detection with CatBoost-SHapley Additive exPlanations (SHAP) interpretable modeling to decipher the spatiotemporal patterns of 35,407 landslides identified following the event. The characteristics of these landslides, including their area, geometric morphology, spatial distribution, and mobility, are analyzed. The explainability of machine learning (CatBoost-SHAP) further reveals the nonlinear synergistic driving effects of antecedent rainfall (&gt; 210&#xa0;mm), soil thickness (80–110&#xa0;cm), bedrock lithology (monzonitic granite &gt; potassic granite &gt; metasandstone), soil type (red earth &gt; red soil-like soil &gt; paddy soil), and slope gradient (20–40°). This research contributes to the application of automated landslide detection methods, enhances the understanding of the factors driving clustered shallow landslides in mountainous regions, and provides valuable insights to aid in optimizing emergency response and early risk warning under extreme rainfall conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Data-driven insights into the characteristics and drivers of the June 16, 2024 clustered shallow landslides in Southeastern China

  • 期刊论文

  • 发表日期：

    2025年7月14日

查看全文
  • Xiaoyu Yi,
  • Jiachen Zhao,
  • Wenkai Feng,
  • Chaoxu Guo,
  • Yanlong Zhao,
  • Zhenghai Xue,
  • Shuangquan Li

摘要

With the increasing frequency of extreme rainfall events, clustered shallow landslides have become a major geological hazard worldwide, particularly in the hilly regions of southern China. The June 2024 extreme rainfall event along the Guangdong–Fujian–Jiangxi tri-province boundary triggered the most extensive clustered shallow landslides in China over the past few decades. This study develops an integrated framework coupling YOLOv8s-based automated detection with CatBoost-SHapley Additive exPlanations (SHAP) interpretable modeling to decipher the spatiotemporal patterns of 35,407 landslides identified following the event. The characteristics of these landslides, including their area, geometric morphology, spatial distribution, and mobility, are analyzed. The explainability of machine learning (CatBoost-SHAP) further reveals the nonlinear synergistic driving effects of antecedent rainfall (> 210 mm), soil thickness (80–110 cm), bedrock lithology (monzonitic granite > potassic granite > metasandstone), soil type (red earth > red soil-like soil > paddy soil), and slope gradient (20–40°). This research contributes to the application of automated landslide detection methods, enhances the understanding of the factors driving clustered shallow landslides in mountainous regions, and provides valuable insights to aid in optimizing emergency response and early risk warning under extreme rainfall conditions.