Meta-heuristic Algorithms Applied to Urban Flood Evacuation Routes: A case study in Beijing, China

Abstract

Abstract: Climate change and urbanization have increased the occurrence of natural disasters, including floods, tsunamis and hurricanes. Among these disasters, floods occur with high frequency, impact a large number of people, cause high economic losses, and lead to high toll of deaths. Examples include floods in Indonesia in 2021 and Pakistan in 2022. The flood in Indonesia affected about 1 million people. The flood in Pakistan affected 33 million and killed 1,739 people and costed US$15 billion in economic damage. Flood risk assessment and evacuation are effective mitigation measures to create flood-resilient cities. Previous studies have focused on flood modelling and risk assessment, yet it is recently recognized that optimal evacuation routes are necessary and critical for social adaptation to flood risks. To date, there are limited research on evacuation route optimisation problem. There are two approaches for evacuation route optimisation: namely exact methods and meta-heuristic methods. The exact methods such as linear programming, weighted summation, and mixed integer programming have been widely applied. Nevertheless, meta-heuristic algorithms are gaining attention as flexible, non-problem-specific, and computationally efficient optimisation methods. The principle of Meta-heuristic algorithms is based on simulating the optimisations that occur naturally in biological or physicochemical processes. For example, there is a commonality between an animal herd searching for routes and a population searching for routes in a flood disaster. Commonly applied meta-heuristic algorithms are Genetic Algorithms, Ant Colony Algorithms, Particle Swarm Algorithms, and Sparrow’s Algorithms. This is because these algorithms have simple structures and high adaptability, desirable local and global convergence properties and require few parameters. In this study, the flood in Beijing, China, in late July and early August 2023 will be simulated. The flood claimed at least 33 lives, damaged 209,000 homes and more than 15,000 hectares of cropland and caused 127 thousand people to evacuate. The flood extent, water depth and flow velocity will be obtained from a two-dimensional hydrodynamic flood model. The flood risk for pedestrians or vehicles will be estimated with the hydrodynamic model result and a mechanic-based stability method. Optimal evacuation routes will be obtained with Genetic Algorithm, Ant Colony Algorithm, Particle Swarm Algorithm, and Sparrow’s Algorithm. The performance of the optimisation algorithms will be compared and evaluated.  This study contributes to the scientific planning of urban flood evacuation routes and provides insight for urban planners and managers to enhance urban resilience. Keywords: Urban floods, evacuation routes, Meta-heuristic algorithms.