Modeling Flood Risk Communication and Evacuation Dynamics Using an Agent-Based Framework: A Cedar Rapids, Iowa Case Study

S M Samiul Islam , Ibrahim Demir

The benefits of effective flood-risk communication in urban river basins are reliant upon the timing of warnings with respect to flood evolution and challenges encountered by vulnerable populations. However, these benefits can reduce losses substantially. This study presents a geospatial agent-based model of flood evacuation for Cedar Rapids, Iowa. The model accounts for stage-specific flood inundation scenarios, road accessibility, evacuation locations and capacities, and heterogeneous household behavior. Evacuation outcomes were assessed across three flood stages, which represent onset (pre-peak), peak (crest), and recession (post-peak) conditions. The warning delays were 0, 2, 4, and 6 hours. The model generates a normalized economic loss index, final evacuation completion success, and evacuation trajectories over time. In the onset stage, the ultimate evacuation is reduced from approximately 57% to 39% (overall) and from 38% to 25% (vulnerable) when warnings are delayed from 0 to 6 hours. This delay results in a concurrent increase in loss from approximately 0.28 to 0.36. The results demonstrate a strong stage dependence. Although evacuation completion remains at nearly 99% across delays during the summit stage, losses increase from approximately 0.33 to 0.38, highlighting the significance of timing-dependent exposure even when evacuation is nearly complete. A persistent vulnerable deficiency of 10–12 percent is observed during the recession stage, with outcomes that are essentially delay-insensitive (overall 94–95%, vulnerable 82–83%, loss 0.30–0.31). These results underscore the necessity of tailored mobility and shelter-access interventions to mitigate persistent inequities in evacuation outcomes and support stage-aware warning strategies.