Scholarlink

<p>Identifying and predicting flood-prone areas is a crucial aspect of effective flood management. Over the past decades, various numerical and statistical techniques have been employed to assess inundation vulnerability across different regions. However, these traditional methods often suffer from limitations such as high computational costs, reliance on assumptions, and extensive simplifications. In contrast, artificial intelligence (AI) approaches have been widely adopted over the last twenty years to enhance flood susceptibility prediction in diverse contexts. This study aims to comprehensively review prior global research in this field. It collects and evaluates multiple facets, including prediction parameters, performance metrics, study areas, and data sources. Consequently, the most promising flood susceptibility methodologies are presented, alongside an analysis of key trends in their advancement. Effective strategies identified include hybrid modeling, data decomposition, model optimization, ensemble algorithms, the specificity of the study area, the type and quantity of input data, data source and temporal coverage, as well as evaluation criteria. This review revealed that use of ML models has increased significantly since 2018, while DL models have shown notable growth since 2020. The majority of research on flood susceptibility prediction has come from Asian nations, including Iran, China, India and Bangladesh. This indicates the region’s significant emphasis on managing water resources and reducing the risk of flooding. Additionally, satellite data has served as the main information source for many investigations. In terms of assessing model performance, it should be noted that because of its high discrimination ability and adaptability in classification tasks, the AUC-ROC evaluation index has been widely regarded as a crucial evaluation criterion in the majority of research. This research serves as a valuable reference for hydrologists in selecting the most suitable methods or models tailored to specific regions and flood prediction strategies. Additionally, it highlights existing research gaps and proposes directions for future investigations.</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 systematic review and comprehensive evaluation of artificial intelligence approaches for prediction flood susceptibility

  • 期刊论文

  • 发表日期：

    2025年12月16日

查看全文
  • Reza Farzad,
  • Ahmad Sharafati,
  • Yusef Kheyruri

摘要

Identifying and predicting flood-prone areas is a crucial aspect of effective flood management. Over the past decades, various numerical and statistical techniques have been employed to assess inundation vulnerability across different regions. However, these traditional methods often suffer from limitations such as high computational costs, reliance on assumptions, and extensive simplifications. In contrast, artificial intelligence (AI) approaches have been widely adopted over the last twenty years to enhance flood susceptibility prediction in diverse contexts. This study aims to comprehensively review prior global research in this field. It collects and evaluates multiple facets, including prediction parameters, performance metrics, study areas, and data sources. Consequently, the most promising flood susceptibility methodologies are presented, alongside an analysis of key trends in their advancement. Effective strategies identified include hybrid modeling, data decomposition, model optimization, ensemble algorithms, the specificity of the study area, the type and quantity of input data, data source and temporal coverage, as well as evaluation criteria. This review revealed that use of ML models has increased significantly since 2018, while DL models have shown notable growth since 2020. The majority of research on flood susceptibility prediction has come from Asian nations, including Iran, China, India and Bangladesh. This indicates the region’s significant emphasis on managing water resources and reducing the risk of flooding. Additionally, satellite data has served as the main information source for many investigations. In terms of assessing model performance, it should be noted that because of its high discrimination ability and adaptability in classification tasks, the AUC-ROC evaluation index has been widely regarded as a crucial evaluation criterion in the majority of research. This research serves as a valuable reference for hydrologists in selecting the most suitable methods or models tailored to specific regions and flood prediction strategies. Additionally, it highlights existing research gaps and proposes directions for future investigations.