Evacuation network design and planning under traffic congestion: An application to wildfires in Canada

Designing efficient evacuation networks is crucial for preparedness, as poorly planned and managed preventive evacuations can increase the time required for evacuees to reach safety zones or shelters, potentially resulting in adverse outcomes. While some studies have estimated traffic congestion, few have integrated multiple aspects of evacuation planning, within a unified framework. Yet, such integration is key to developing realistic and effective plans, despite the resulting complexity. In this study, we propose and solve a mixed-integer programming model to address the Evacuation Network Design and Planning Problem under Traffic Congestion (ENDPT). Our model jointly optimizes shelter locations, evacuation routes, and evacuee flows to minimize total evacuation time. Traffic congestion is captured using a formulation based on the Cell Transmission Model that incorporates time-varying road properties, multiple time periods, contraflow operations, and capacity constraints. To solve the ENDPT, we develop an exact method based on Benders decomposition and validate our approach through two complementary experiments. First, we generate instances from network topologies of communities that are threatened by wildfires to create standardized benchmarks. Results show that conventional models that ignore traffic dynamics can underestimate evacuation times by up to 60% compared to our integrated approach. Second, we develop a case study based on the 2023 wildfire evacuation of Chibougamau (Canada). This case study demonstrates the model’s practical value by generating tailored plans and assessing context-specific preparedness policies (e.g., showing that adding sheltering options with limited capacity can reduce traffic congestion by more than 95%) to support local decision-making.