Scholarlink

<p>Floods are among the most destructive natural hazards, causing significant casualties and financial losses worldwide. This study aims to analyze flood susceptibility in the data-scarce arid Bahukalat basin in southeastern Iran, focusing on the devastating floods in January 2020 and March 2024. The flood inventory maps were based on MNDWI (Modified Normalized Difference Water Index) derived from Sentinel-2 images (2020 and 2024) using the Otsu thresholding technique. The Random Forest (RF) algorithm was implemented and trained with 8 flood conditioning factors (FCFs), including the total scattering power index (SPAN) from Sentinel 1 synthetic aperture radar (SAR) images. The SPAN index was utilized as an innovative indirect proxy for soil parameters to address the lack of in-situ soil data. The model demonstrated high predictive performance with an area under the curve (AUC) of 0.965. To enhance the model’s interpretability, the Shapley additive explanation (SHAP) method was used. SHAP analyses indicated that the elevation is the most influential predictor of flooding, followed by SPAN, slope, valley depth, and stream density.</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

Flood Risk Assessment Using Multitemporal Satellite Data and Explainable Artificial Intelligence (XAI): Analyzing Susceptibility and Driving Factors

  • 期刊论文

  • 发表日期：

    2026年5月16日

查看全文
  • Pouya Mahmoudnia,
  • Jalal Karami

摘要

Floods are among the most destructive natural hazards, causing significant casualties and financial losses worldwide. This study aims to analyze flood susceptibility in the data-scarce arid Bahukalat basin in southeastern Iran, focusing on the devastating floods in January 2020 and March 2024. The flood inventory maps were based on MNDWI (Modified Normalized Difference Water Index) derived from Sentinel-2 images (2020 and 2024) using the Otsu thresholding technique. The Random Forest (RF) algorithm was implemented and trained with 8 flood conditioning factors (FCFs), including the total scattering power index (SPAN) from Sentinel 1 synthetic aperture radar (SAR) images. The SPAN index was utilized as an innovative indirect proxy for soil parameters to address the lack of in-situ soil data. The model demonstrated high predictive performance with an area under the curve (AUC) of 0.965. To enhance the model’s interpretability, the Shapley additive explanation (SHAP) method was used. SHAP analyses indicated that the elevation is the most influential predictor of flooding, followed by SPAN, slope, valley depth, and stream density.