SIMERSAD: An Agent-Based Simulation Approach to Flood Evacuation Routing in Santo Domingo

Abstract Climatic events, such as the arrival of hurricanes or typhoons, or geomorphological events, like earthquakes, can also necessitate the evacuation of cities or, at the very least, parts of them. Additionally, situations of social alarm caused by terrorist attacks may also require the evacuation of cities or, at the very least, parts of them. In this context, efforts should not only focus on facilitating the relocation of potentially affected individuals. They should focus on carrying out these processes in coordination with authorities in the shortest possible time, while avoiding congestion on evacuation routes as much as possible. To this end, we present SIMERSAD, an agent-based co-model implemented in GAMA that couples a simplified, diffusion-based flood propagation model – omitting hydrostatic pressure, flow velocity, momentum, and energy losses, and therefore best suited to slow-onset riverine inundation rather than high-velocity flood scenarios – with pedestrian evacuation dynamics in the urban core of Santo Domingo, Dominican Republic. The modular architecture of GAMA facilitates adaptation to other cities, though application-specific recalibration of demographic, hydrological, and infrastructure parameters would be required. Simulations reveal that demographic heterogeneity in walking speed increases worst-case clearance time by over 60%. Dynamic congestion with real-time rerouting triples mean evacuation time, while increasing shelter count may reduce median evacuation significantly, but paradoxically worsens worst-case performance due to peripheral placement, as we have observed comparing with the unlimited shelter capacity scenario. Based on these results, three actionable policy recommendations are proposed for emergency managers, including a staged alert protocol and a hierarchical shelter routing strategy. Finally, we also discuss the limitations of our modeling approach.