Scholarlink

<p>Due to the influence of complex geological structures and reservoir operations, geological disasters frequently occur in reservoir bank areas. Conducting susceptibility assessments in these areas is essential for ensuring the safe and stable operation of reservoirs.In susceptibility assessments of mountainous regions, traditional models often neglect the uncertainty inherent in dynamic environmental factors. The Interval Intuitionistic Fuzzy Set (IIFS) model, by introducing elastic interval representations, offers a more flexible means of characterizing the spatiotemporal variability and evolutionary patterns of such dynamic factors, thereby enhancing model adaptability and prediction accuracy. In this study, ascending and descending Sentinel-1 SAR data from September 2021 to September 2023 were utilized to derive ground surface deformation using time-series InSAR analysis. Key influencing factors of reservoir bank landslides in the Xiaowan Reservoir–Lancang River section—including topography, climate conditions, and geological characteristics—were incorporated into the IIFS model to conduct a comprehensive landslide susceptibility assessment. The results show that: (1) The IIFS-based model demonstrated superior performance in landslide susceptibility evaluation, achieving a ROC-AUC of 0.902, outperforming the BPNN (0.864), Random Forest (0.790), and Information Value model (0.680). Additionally, the IIFS model achieved an F1-score of 0.85, precision of 0.82, and recall of 0.88, indicating strong classification performance and balance. (2) High-susceptibility zones were primarily concentrated on the left bank of the upstream section of the Xiaowan Reservoir–Lancang River, with the extremely high susceptibility area accounting for 13.28% of the total, encompassing 21 historical landslide points. The landslide density in this zone was approximately 32% higher than that predicted by the BPNN model. (3) Sensitivity analysis with ± 5% perturbations applied to key input factors—such as DEM, annual rainfall, and InSAR deformation velocity—showed AUC fluctuations within 0.02. This indicates that the model maintains strong robustness and generalization capability when facing uncertainties in input data. Overall, the IIFS model effectively captures the uncertainty of environmental factors, enhances the prediction accuracy and spatial focus of reservoir bank landslide susceptibility, and provides scientific and practical support for geological hazard risk management.</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

Landslide susceptibility assessment based on fuzzy set theory: Xiaowan reservoir–Lancang river

  • 期刊论文

  • 发表日期：

    2025年9月6日

查看全文
  • Hong Wenyu,
  • Xi Wenfei,
  • Yang Zhiquan,
  • Gu Shixiang,
  • Huang Guangcai,
  • Jin Tingting,
  • Zhuang Yongzai,
  • Bai Shihan,
  • Ma Yijie

摘要

Due to the influence of complex geological structures and reservoir operations, geological disasters frequently occur in reservoir bank areas. Conducting susceptibility assessments in these areas is essential for ensuring the safe and stable operation of reservoirs.In susceptibility assessments of mountainous regions, traditional models often neglect the uncertainty inherent in dynamic environmental factors. The Interval Intuitionistic Fuzzy Set (IIFS) model, by introducing elastic interval representations, offers a more flexible means of characterizing the spatiotemporal variability and evolutionary patterns of such dynamic factors, thereby enhancing model adaptability and prediction accuracy. In this study, ascending and descending Sentinel-1 SAR data from September 2021 to September 2023 were utilized to derive ground surface deformation using time-series InSAR analysis. Key influencing factors of reservoir bank landslides in the Xiaowan Reservoir–Lancang River section—including topography, climate conditions, and geological characteristics—were incorporated into the IIFS model to conduct a comprehensive landslide susceptibility assessment. The results show that: (1) The IIFS-based model demonstrated superior performance in landslide susceptibility evaluation, achieving a ROC-AUC of 0.902, outperforming the BPNN (0.864), Random Forest (0.790), and Information Value model (0.680). Additionally, the IIFS model achieved an F1-score of 0.85, precision of 0.82, and recall of 0.88, indicating strong classification performance and balance. (2) High-susceptibility zones were primarily concentrated on the left bank of the upstream section of the Xiaowan Reservoir–Lancang River, with the extremely high susceptibility area accounting for 13.28% of the total, encompassing 21 historical landslide points. The landslide density in this zone was approximately 32% higher than that predicted by the BPNN model. (3) Sensitivity analysis with ± 5% perturbations applied to key input factors—such as DEM, annual rainfall, and InSAR deformation velocity—showed AUC fluctuations within 0.02. This indicates that the model maintains strong robustness and generalization capability when facing uncertainties in input data. Overall, the IIFS model effectively captures the uncertainty of environmental factors, enhances the prediction accuracy and spatial focus of reservoir bank landslide susceptibility, and provides scientific and practical support for geological hazard risk management.