Scholarlink

<p>Northern Kerala is vulnerable to landslides due to its undulated terrain, uneven rainfall, soil conditions, and geomorphic changes. This research investigates landslide-prone areas in North Kerala, referencing to the 2018 landslide and flood events. A landslide hotspot inventory map was created using extensive historical data from the Geological Survey of India. Logistic regression and random forest algorithms were employed to analyse landslide susceptibility in the region. While natural factors like slope and rainfall influence landslides, the impact of soil erosion was also included in both models. Soil erosion was found to significantly contribute to landslide occurrences, with a positive coefficient of 0.048 in the LR model. The RF and LR models created to assess landslide susceptibility showcase remarkable consistency, highlighted by an outstanding Area Under the Curve (AUC) score of 0.985 for the RF model and 0.812 for the LR model, as indicated by the Receiver Operating Characteristic (ROC) analysis. The RF model performed superior to the LR model, highlighting the importance of soil erosion in predicting landslide events. The RF probability results show 67.59% of the study area at very low risk, 5.28% at low risk, 6.26% at medium risk, 4.26% at high risk, and 16.59% at very high risk for landslides.</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

GIS-based Landslide Susceptibility Model for North Kerala Using Logistics Regression and Machine Learning Approaches

  • 期刊论文

  • 发表日期：

    2026年3月7日

查看全文
  • S. Babu,
  • RM. Narayanan

摘要

Northern Kerala is vulnerable to landslides due to its undulated terrain, uneven rainfall, soil conditions, and geomorphic changes. This research investigates landslide-prone areas in North Kerala, referencing to the 2018 landslide and flood events. A landslide hotspot inventory map was created using extensive historical data from the Geological Survey of India. Logistic regression and random forest algorithms were employed to analyse landslide susceptibility in the region. While natural factors like slope and rainfall influence landslides, the impact of soil erosion was also included in both models. Soil erosion was found to significantly contribute to landslide occurrences, with a positive coefficient of 0.048 in the LR model. The RF and LR models created to assess landslide susceptibility showcase remarkable consistency, highlighted by an outstanding Area Under the Curve (AUC) score of 0.985 for the RF model and 0.812 for the LR model, as indicated by the Receiver Operating Characteristic (ROC) analysis. The RF model performed superior to the LR model, highlighting the importance of soil erosion in predicting landslide events. The RF probability results show 67.59% of the study area at very low risk, 5.28% at low risk, 6.26% at medium risk, 4.26% at high risk, and 16.59% at very high risk for landslides.