Scholarlink

<p>Saudi Arabia is vulnerable to various natural disasters, each of which can affect life and property significantly. As steep regions frequently experience seismic activity and significant precipitation, landslides have become widespread. Consequently, creating susceptibility maps by simulating these threats is of paramount importance. The potential landslide susceptibility in the NEOM area of northwestern Saudi Arabia was identified by incorporating topographic, geological, meteorological, and earthquake data from various sources. Landslide inventory data were evaluated to determine the probability of landslides in the NEOM area. Eighteen thematic layers were produced, each representing one of the key landslide factors. The relative importance of key landslide factors for landslide susceptibility mapping was determined using a random forest (RF) model, enabling us to evaluate the effectiveness and significance of each parameter in the mapping process. Three machine algorithms were utilized—“Logistic Regression (LR)”, “Random Forest (RF)”, and “eXtreme Gradient Boosting (XGB)” to integrate inventory data from field studies, historical documents, and high-resolution remote sensing images with key landslide factors. This approach allowed us to create landslide susceptibility maps that could effectively identify areas at risk for landslides. The landslide susceptibility indices (LSI) for the LR, RF, and XGB models were divided into five landslide susceptibility zones. Numerous statistical indices, including the kappa index (K), “mean absolute error (MAE)”, “root mean square error (RMSE)”, overall accuracy (OAc), hazard accuracy (HAc), non-hazard accuracy (NHAc), and “area under the curve (AUC), were used to evaluate the predictability of the landslide susceptibility models. Analysis of these indices indicated that the XGB model performed remarkably well in landslide prediction, with a high area under the curve (AUC) of 95.5% and a superior kappa score of 0.930. However, it provided the lowest MAE (0.064) and RMSE (0.213). The c map shows that high and very high landslide susceptibility covers 29.4%, moderate zone covers 16.9%, and low and very low susceptibility zones cover 53.7%. When implementing development projects in the NEOM region, the ultimate landslide susceptibility map can potentially increase decision-making effectiveness. Avoiding critical infrastructure and implementing geotechnical measures to prevent the negative consequences of landslides are possible with the assistance of this map.</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

Seismic-induced landslides susceptibility mapping of the NEOM area, northwestern Saudi Arabia using machine learning models

  • 期刊论文

  • 发表日期：

    2025年6月11日

查看全文
  • Bosy A. El-Haddad,
  • Ahmed M. Youssef,
  • Ali M. Mahdi,
  • Zeinab Karimi,
  • Hamid Reza Pourghasemi

摘要

Saudi Arabia is vulnerable to various natural disasters, each of which can affect life and property significantly. As steep regions frequently experience seismic activity and significant precipitation, landslides have become widespread. Consequently, creating susceptibility maps by simulating these threats is of paramount importance. The potential landslide susceptibility in the NEOM area of northwestern Saudi Arabia was identified by incorporating topographic, geological, meteorological, and earthquake data from various sources. Landslide inventory data were evaluated to determine the probability of landslides in the NEOM area. Eighteen thematic layers were produced, each representing one of the key landslide factors. The relative importance of key landslide factors for landslide susceptibility mapping was determined using a random forest (RF) model, enabling us to evaluate the effectiveness and significance of each parameter in the mapping process. Three machine algorithms were utilized—“Logistic Regression (LR)”, “Random Forest (RF)”, and “eXtreme Gradient Boosting (XGB)” to integrate inventory data from field studies, historical documents, and high-resolution remote sensing images with key landslide factors. This approach allowed us to create landslide susceptibility maps that could effectively identify areas at risk for landslides. The landslide susceptibility indices (LSI) for the LR, RF, and XGB models were divided into five landslide susceptibility zones. Numerous statistical indices, including the kappa index (K), “mean absolute error (MAE)”, “root mean square error (RMSE)”, overall accuracy (OAc), hazard accuracy (HAc), non-hazard accuracy (NHAc), and “area under the curve (AUC), were used to evaluate the predictability of the landslide susceptibility models. Analysis of these indices indicated that the XGB model performed remarkably well in landslide prediction, with a high area under the curve (AUC) of 95.5% and a superior kappa score of 0.930. However, it provided the lowest MAE (0.064) and RMSE (0.213). The c map shows that high and very high landslide susceptibility covers 29.4%, moderate zone covers 16.9%, and low and very low susceptibility zones cover 53.7%. When implementing development projects in the NEOM region, the ultimate landslide susceptibility map can potentially increase decision-making effectiveness. Avoiding critical infrastructure and implementing geotechnical measures to prevent the negative consequences of landslides are possible with the assistance of this map.