Answers are grounded in this paper and cite their sources.
The combination of extreme rainfall and rapid urbanisation significantly escalates flood risks in major cities, underscoring the importance of accurately identifying causative factors and formulating differentiated management strategies to enhance urban resilience. This study focuses on the central urban areas of Wuhan, Changsha, and Zhengzhou, identifying flood locations outside the core area by comparing remote sensing images from multiple time periods. Use the BERT-BiLSTM-CRF model to extract flood information from social media data, enabling the rapid and accurate capture of urban flood location information within the core area. A multi-year, high-timeliness flooding disaster map identified 1008 flooding points in Changsha, 960 in Wuhan, and 675 in Zhengzhou. Interpretable machine learning methods (XGboots + SHAP) were utilised to analyse flood-inducing factors, revealing that the three cities share six primary factors, including topography, vegetation index (NDVI), and impervious surface, though their degrees of influence vary significantly. Topography predominantly affects Changsha, while Wuhan experiences significant impacts from topography, NDVI, and impervious surfaces. In Zhengzhou, NDVI and construction intensity emerge as the dominant factors. Additionally, the K-means dimension reduction clustering method provides a comprehensive analysis of the characteristics of flood-related disaster factors and their contributions. This enables the accurate identification of the causes of disasters in urban areas and their varying degrees of impact. Results highlighted the critical role of topography in flood occurrences, NDVI's mitigation potential within built-up areas, and the relatively limited flooding influence of certain anthropogenic construction activities. These findings recommend that urban flood management strategies account for spatial heterogeneity. • Integrated social media, remote sensing, and official data for flood mapping. • Applied XGBoost and SHAP to interpret nonlinear effects of flood drivers. • Used PCA and K-means to reveal intra-urban heterogeneity and zoning. • Identified city-specific flood mechanisms in Changsha, Wuhan, and Zhengzhou. • Proposed a “data–model–zoning” framework for flood risk governance.
No takes yet. Share an insight, caveat, or question.
Guo et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: