Scholarlink

<p>The recent surge in landslides in Uttarakhand’s Himalayan region, driven by intense monsoonal precipitation and rapid infrastructure development, has gained global attention. The Rishikesh-Gangotri National Highway (RGNH) experiences frequent landslides, leading to substantial economic losses. This study employs Machine Learning (ML) and statistical algorithms such as Extreme Gradient Boosting (XGB), Multi-Layer Perceptron (MLP), Information Value Method (IVM) and Analytical Hierarchy Process (AHP) for a comprehensive landslide analysis in the RGNH utilising ten geo-environmental parameters. The generated susceptibility maps of the study area showcase critical portions of the highway as landslide prone. The performance validation for each model was conducted using the Receiver Operating Characteristics-Area Under the Curve (AUC-ROC) method. The results achieved demonstrates that the, IVM outperformed the AHP, achieving an AUC of 87% versus 72%. ML models excelled over statistical ones, with XGB achieving 93% AUC compared to MLP at 92%. Additionally, accuracy metrics F1-Score, recall and precision were utilised to assess the performance of the two ML models. Further, implementation of Shapely Additive Explanation (SHAP) for addressing explainability of the ML models suggest that slope, drainage density, lineament density, and, road proximity are prominent landslide influencing parameters of the study area. Slope stability analysis at selected landslide susceptible sites indicates a factor of safety (FOS) of 1.06 and 1.11 at Dharasu and 0.56 and 0.58 at Uttarkashi using Janbu’s and Bishop’s simplified methods, respectively. Maximum future landslide depths were estimated at approximately 9&#xa0;m and 4&#xa0;m at these potentially vulnerable locations. The study emphasizes the increasing risk of slope instabilities, highlighting the need for continuous monitoring and timely interventions to mitigate future landslide risks along the RGNH.</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 hazard assessment in parts of upper Bhagirathi Basin: a comparative study using earth observation and machine learning based initiatives in perspective of slope instability

  • 期刊论文

  • 发表日期：

    2025年11月5日

查看全文
  • P Danuta Mohan,
  • Rahul Das,
  • Shovan Lal Chattoraj,
  • Yateesh Ketholia

摘要

The recent surge in landslides in Uttarakhand’s Himalayan region, driven by intense monsoonal precipitation and rapid infrastructure development, has gained global attention. The Rishikesh-Gangotri National Highway (RGNH) experiences frequent landslides, leading to substantial economic losses. This study employs Machine Learning (ML) and statistical algorithms such as Extreme Gradient Boosting (XGB), Multi-Layer Perceptron (MLP), Information Value Method (IVM) and Analytical Hierarchy Process (AHP) for a comprehensive landslide analysis in the RGNH utilising ten geo-environmental parameters. The generated susceptibility maps of the study area showcase critical portions of the highway as landslide prone. The performance validation for each model was conducted using the Receiver Operating Characteristics-Area Under the Curve (AUC-ROC) method. The results achieved demonstrates that the, IVM outperformed the AHP, achieving an AUC of 87% versus 72%. ML models excelled over statistical ones, with XGB achieving 93% AUC compared to MLP at 92%. Additionally, accuracy metrics F1-Score, recall and precision were utilised to assess the performance of the two ML models. Further, implementation of Shapely Additive Explanation (SHAP) for addressing explainability of the ML models suggest that slope, drainage density, lineament density, and, road proximity are prominent landslide influencing parameters of the study area. Slope stability analysis at selected landslide susceptible sites indicates a factor of safety (FOS) of 1.06 and 1.11 at Dharasu and 0.56 and 0.58 at Uttarkashi using Janbu’s and Bishop’s simplified methods, respectively. Maximum future landslide depths were estimated at approximately 9 m and 4 m at these potentially vulnerable locations. The study emphasizes the increasing risk of slope instabilities, highlighting the need for continuous monitoring and timely interventions to mitigate future landslide risks along the RGNH.