Societal impacts of human behavior–Aware evacuation modeling for urban flood resilience

Abstract

Climate-induced flood risks expose critical limitations in evacuation models that prioritize shortest-distance routing while neglecting heterogeneous human vulnerability. This study develops a behavior-aware evacuation framework that integrates cellular automata (CA) to simulate flood dynamics, agent-based modeling (ABM) to represent individual heterogeneity, and a depth-sensitivity-perception-enhanced A* (DSP-A*) algorithm. The proposed approach incorporates physiological safety constraints by explicitly linking floodwater depth to agent stature, thereby enabling the generation of safety-oriented evacuation routes rather than purely distance-minimizing paths. Validation using empirical data from Hurricane Harvey (2017) demonstrates the model’s ability to reproduce height-dependent evacuation behavior and consistently avoid high-risk inundation zones. In an urban grid environment, the DSP-A* algorithm reduces the number of explored nodes by 93% and improves computational efficiency by 88% compared with conventional A*-based methods. Moreover, it decreases hazardous immersion exposure for vulnerable populations by 44–77%. These findings underscore the value of embedding human-centric safety thresholds into routing heuristics, thereby enhancing both the safety and reliability of evacuation planning. The framework offers a scalable, policy-relevant tool to advance risk-informed emergency management and promote inclusive urban flood resilience.