Scholarlink

<p>Landslides pose significant threats to life, property and sustainable development in mountainous regions worldwide, with their occurrence increasingly influenced by climate change. This study addresses the critical need for accurate landslide susceptibility models in the Western Province of Rwanda, where traditional methods have shown limitations. It employed and compared three deep learning models: convolutional neural network (CNN), deep neural network (DNN), and multi-layer perceptron (MLP), to assess the landslide risks, incorporating climate change considerations. The study utilised 16 conditioning factors, carefully selected to avoid multicollinearity, with the digital surface model (DSM) showing the highest variance inflation factor (VIF) of 3.9730. The CNN model demonstrated superior performance, achieving the highest overall accuracy (93.7% for training, 91.5% for validation) and area under the curve (AUC) (98.8% for training, 96.2% for validation). Shapley Additive Explanations (SHAP) analysis revealed that slope degree, DSM, and rainfall are the most influential factors in determining landslide susceptibility in the province. The northern and central parts of the province, particularly the districts of Karongi, Rutsiro, and Ngororero, were identified as high-risk areas. This research provides a robust framework for integrating deep learning with climate change considerations in landslide risk assessment, offering valuable insights for sustainable development and disaster risk reduction strategies in Rwanda and similar regions globally.</p> Graphical abstract <p></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

Deep learning models for comprehensive landslide susceptibility study considering climate change and sustainable development implications in Western Province of Rwanda

  • 期刊论文

  • 发表日期：

    2026年5月7日

查看全文
  • Vincent E. Nwazelibe,
  • Mubarak Abdulkarim,
  • Johnson C. Agbasi,
  • Emmanuel C. Okpala,
  • Chaitanya B. Pande,
  • Sani I. Abba,
  • Johnbosco C. Egbueri

摘要

Landslides pose significant threats to life, property and sustainable development in mountainous regions worldwide, with their occurrence increasingly influenced by climate change. This study addresses the critical need for accurate landslide susceptibility models in the Western Province of Rwanda, where traditional methods have shown limitations. It employed and compared three deep learning models: convolutional neural network (CNN), deep neural network (DNN), and multi-layer perceptron (MLP), to assess the landslide risks, incorporating climate change considerations. The study utilised 16 conditioning factors, carefully selected to avoid multicollinearity, with the digital surface model (DSM) showing the highest variance inflation factor (VIF) of 3.9730. The CNN model demonstrated superior performance, achieving the highest overall accuracy (93.7% for training, 91.5% for validation) and area under the curve (AUC) (98.8% for training, 96.2% for validation). Shapley Additive Explanations (SHAP) analysis revealed that slope degree, DSM, and rainfall are the most influential factors in determining landslide susceptibility in the province. The northern and central parts of the province, particularly the districts of Karongi, Rutsiro, and Ngororero, were identified as high-risk areas. This research provides a robust framework for integrating deep learning with climate change considerations in landslide risk assessment, offering valuable insights for sustainable development and disaster risk reduction strategies in Rwanda and similar regions globally.

Graphical abstract