Simulating Urban Population Activities under Extreme Events with Data-Driven Agent-Based Modeling

Knowing dynamic and fine-grained interactions between people and the built environments (e.g., facilities and infrastructure) during extreme events could assist in multiple emergency management operations. This study achieves the objective with agent-based modeling (ABM) simulation, which captures traveling patterns of affected population and reconstructs high-resolution intra-urban population movements under the perturbation of extreme events. Specifically, we extracted traveling patterns and activities referring to massive high-resolution, anonymized GPS traces and visited Point of Interests (POIs) (e.g., restaurants and grocery stores) collected from the city of Tallahassee in response to Hurricane Irma in 2017. We identified disparities in affected people's traveling patterns regarding visitation frequencies, activity durations, activity start time, and traveling distances under normal and extreme event circumstances. We then developed a data-driven ABM to simulate activity-based travels for normal and adverse conditions. The research outcomes could show the dynamic spatio-temporal patterns of human mobility and facility utilization under the perturbations of extreme events, which pinpoint locations with large throngs and critical facilities/infrastructure with more visitations and, therefore, informs future city and infrastructure planning projects for enhancing disaster resilience.