Assessing the impact of geo-targeted warning messages on residents’ evacuation decisions before a hurricane using agent-based modeling

The increasing frequency and intensity of hurricane hazards have raised the urgency of improving hurricane warning effectiveness, especially in terms of motivating the evacuation of people living in high-risk areas. Traditional warnings for hurricanes have limitations of sending a general message for coarse spatial scales (e.g., county level) and do not include specific risks and orders for residents in distinct areas of finer scales. To overcome these limitations, geo-targeted hurricane warning systems have been proposed, but in practice, the existing systems have low accuracy because they neglect environmental factors when defining warning zones. Extant literature has focused on optimizing the geo-delivering process of warnings with limited efforts on geo-defining warning zones. It is still unclear to what extent the geo-targeted warnings motivate residents to evacuate from high-risk areas before a hurricane. Therefore, we developed an agent-based model (ABM) to simulate residents’ evacuation decision-making under geo-targeted warnings, which were generated based on characteristics of both hurricane hazards and the built environment. We used forecasted information of Hurricane Dorian as a case study; then conducted the ABMs under geo-targeted warnings, a general warning, and warnings based on storm surge planning zones; then we compared the three outcomes. The research finds an effective way to geo-define warning zones using the built environment data. The result suggests that geo-targeted warnings can motivate more residents in high-risk areas to evacuate. These findings contribute to the understanding of the effect of geo-targeted warning on evacuation and suggest the importance of warnings with more specific contents for finer spatial scales.