A social media driven model for evaluating coupled flood damage and resilience at a fine scale.

With the increasing frequency of extreme rainfall events due to climate change, enhancing flood resilience has become a critical focus for urban researchers and planners. However, existing studies rarely evaluate the relationship between flood resilience and damage at fine geospatial scales, limiting the development of site-specific enhancement strategies. This study employs a baseline-adaptive resilience framework that integrates machine learning techniques with multidimensional social media data analytics. The methodology implements a three-phase analytical approach-stepwise clustering analysis model, hierarchical partitioning analysis, and coupling coordination degree analysis-to identify the resilience determinants influencing both physical and psychological flood damage. This framework enables assessment of coordination between resilience dimensions and empirical disaster impacts at high spatial resolution. The analysis highlights the dominant role of fire station accessibility in enhancing resilience, alongside other critical factors such as medical facility accessibility, road network structure, population density, and public service provision. Our findings reveal significant spatial differentiation in flood risk patterns, with higher-risk areas concentrated in suburban regions, new urban districts, and densely populated city centers. By identifying dominant resilience factors and uncovering spatial heterogeneity, this study provides a valuable tool for policymakers to identify flood risk areas, while at the same time advancing knowledge within the broader framework of flood resilience research.