Author(s): Hao-Ming Hsu; Chang-Ying Lee; Ming-Che Hu; Philippe Gourbesville

Linked Author(s): Philippe Gourbesville

Keywords: Agent-based modelling; Evacuation optimisation; Hydraulic simulation; Minimum-cost network flow analysis; Urban flooding

Abstract: Urban flooding induced by extreme rainfall and compound hazards poses significant challenges to evacuation planning, particularly in densely populated areas with complex road networks. This study proposes a phase-adaptive, multi-objective evacuation optimisation framework that integrates hydraulic simulation, minimum-cost network flow analysis, and agent-based modelling. Hydraulic simulations are first conducted to generate spatially and temporally explicit flood characteristics, including water depth, inundation extent, and arrival time. These outputs are translated into risk scores and accessibility constraints for road segments, forming dynamic risk maps. A minimum-cost network flow model is then employed to optimise evacuation routing by minimising total travel cost while considering network topology, road capacities, and shelter capacities. To evaluate the practical performance of the optimised strategies, an agent-based model is developed to simulate heterogeneous individual behaviours and congestion effects. Discrepancies between theoretical optima and simulated outcomes are iteratively used to refine evacuation strategies. Proof-ofconcept cases demonstrate the technical feasibility of the integrated framework and its capability to identify optimised evacuation routes, critical bottlenecks, and system cost variations under progressive network disruption. The proposed approach provides a foundation for supporting local governments in evacuation planning and emergency response under urban flooding scenarios.

DOI: https://doi.org/10.64697/iahr.proc.hic2026.91

Year: 2026