Scholarlink

<p>Floods are among the most destructive natural hazards, causing major losses to agriculture, infrastructure, housing and local livelihoods. The Dez and Karun river basins in Khuzestan Province are particularly vulnerable because of recurrent flooding and the downstream impacts of reservoir releases. In this study, Landsat 5, 7, and 8 imagery, historical dam-release records, and machine-learning methods were combined to identify flood-prone areas, map inundation extent, and evaluate exposed assets. Flood hazard zonation maps were produced using a Support Vector Machine (SVM) classifier and validated against field-observation data, achieving an overall accuracy of 75%. By integrating hazard maps with land-use information, vulnerable assets such as agricultural lands, orchards, and rural settlements were identified. In addition, a linear regression model was developed to quantify the relationship between river discharge and flooded areas, and the results showed that outflow discharge is a key control on flood severity and spatial distribution, with a predictive performance of 62.8%. The proposed framework demonstrates that combining remote sensing with machine learning provides an effective and rapid tool for flood risk assessment in regulated basins. The results can support reservoir-operation decisions, flood mitigation planning, land-use regulation, and risk-reduction strategies in flood-prone areas.</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

A data-driven framework for predicting downstream flood extent from Dez Dam releases using machine learning algorithms and remote sensing

  • 期刊论文

  • 发表日期：

    2026年6月23日

查看全文
  • Mohammad Abaforoushan,
  • Zareie Sajad,
  • Mohamad Azizipour,
  • Mohammad Farhadiyan

摘要

Floods are among the most destructive natural hazards, causing major losses to agriculture, infrastructure, housing and local livelihoods. The Dez and Karun river basins in Khuzestan Province are particularly vulnerable because of recurrent flooding and the downstream impacts of reservoir releases. In this study, Landsat 5, 7, and 8 imagery, historical dam-release records, and machine-learning methods were combined to identify flood-prone areas, map inundation extent, and evaluate exposed assets. Flood hazard zonation maps were produced using a Support Vector Machine (SVM) classifier and validated against field-observation data, achieving an overall accuracy of 75%. By integrating hazard maps with land-use information, vulnerable assets such as agricultural lands, orchards, and rural settlements were identified. In addition, a linear regression model was developed to quantify the relationship between river discharge and flooded areas, and the results showed that outflow discharge is a key control on flood severity and spatial distribution, with a predictive performance of 62.8%. The proposed framework demonstrates that combining remote sensing with machine learning provides an effective and rapid tool for flood risk assessment in regulated basins. The results can support reservoir-operation decisions, flood mitigation planning, land-use regulation, and risk-reduction strategies in flood-prone areas.