Geospatial analysis of emergency shelter accessibility under flood-induced road disruptions: a case study of central Guangzhou, China

Geospatial modeling offers critical insights for assessing emergency shelter accessibility in the context of disaster-induced disruptions in flood-prone cities. This study develops a GIS-based framework that integrates hydrodynamic flood simulation and accessibility analysis to evaluate flood-induced road disruptions in central Guangzhou, China. Using the Cellular Automata Dual-DraInagE Simulation (CADDIES)-2D model, flooding scenarios with 10-, 20-, and 50-year return periods were simulated, and impassable roads were incorporated into a network-based analysis. Community-level accessibility to shelters was measured with the Gaussian Two-Step Floating Catchment Area (G2SFCA) method, which accounts for shelter capacity, distance decay, and disrupted travel paths. Results show a progressive decline in accessibility as flood severity increases: 46.77%, 47.44%, and 52.56% of communities experienced reduced access under the three scenarios. The number of high-risk communities, those exposed to both severe flooding and high vulnerability (measured by accessibility and socioeconomic conditions), rose from 6 (10-year) to 34 (20-year) and 98 (50-year). These communities are mainly concentrated in peripheral districts with sparse shelter coverage. These findings underscore the need for spatially optimized shelter allocation, resilient evacuation routes, and consideration of social vulnerability in emergency planning. The proposed framework offers transferable insights for urban resilience strategies, supporting equitable disaster preparedness in other flood-prone megacities.