Scholarlink

<p>Intensifying climate change and urbanization have led to more frequent and widespread urban floods, threatening lives, property, and sustainable development. This study uses machine learning to assess flood risk across 41 cities in the Yangtze River Delta Region by integrating 14 geospatial indicators spanning hydrometeorological, geomorphological, and socioeconomic dimensions. Results demonstrate that Random Forest achieves superior accuracy, F1‑score, and AUC compared to alternative models, confirming its suitability for regional flood risk assessment. The distribution of flood risk within the study area exhibits significant spatial autocorrelation, with high-risk clusters predominantly located in central and southern Zhejiang Province and northern Jiangsu Province. Moreover, SHAP (SHapley Additive exPlanations) values identify typhoon frequency, GDP, runoff depth, maximum three-day precipitation, and DEM as the most influential drivers of flood risk, and their relative contributions differ across geographic areas that exhibit different watershed characteristics or observed flood risk levels. The findings offer a scientific basis for context-specific flood risk governance across the extended Yangtze River Delta region.</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

Evaluating flood risk in the Yangtze River Delta region using explainable machine learning

  • 期刊论文

  • 发表日期：

    2026年2月27日

查看全文
  • Yan Wang,
  • Mengya Li,
  • Beibei Hu,
  • Jun Wang,
  • Qian Yao

摘要

Intensifying climate change and urbanization have led to more frequent and widespread urban floods, threatening lives, property, and sustainable development. This study uses machine learning to assess flood risk across 41 cities in the Yangtze River Delta Region by integrating 14 geospatial indicators spanning hydrometeorological, geomorphological, and socioeconomic dimensions. Results demonstrate that Random Forest achieves superior accuracy, F1‑score, and AUC compared to alternative models, confirming its suitability for regional flood risk assessment. The distribution of flood risk within the study area exhibits significant spatial autocorrelation, with high-risk clusters predominantly located in central and southern Zhejiang Province and northern Jiangsu Province. Moreover, SHAP (SHapley Additive exPlanations) values identify typhoon frequency, GDP, runoff depth, maximum three-day precipitation, and DEM as the most influential drivers of flood risk, and their relative contributions differ across geographic areas that exhibit different watershed characteristics or observed flood risk levels. The findings offer a scientific basis for context-specific flood risk governance across the extended Yangtze River Delta region.