Quantifying Impacts of Forecast Errors on Evacuation Outcomes Using an Agent-Based Modeling Framework

In addition to measuring forecast skill in terms of errors in a hurricane’s track and other meteorological characteristics, there’s a need to more directly measure the impact of these forecast errors on societal outcomes such as evacuations. With this need in mind, an agent-based modeling framework is modified and used to explore connections between forecast errors and evacuations. Called FLEE (Forecasting Laboratory for Exploring the Evacuation-system), the framework integrates several empirically-informed models of the natural hazard (hurricane), the human system (information flow, evacuation decisions), the built environment (road infrastructure), and connections between systems (forecasts and warning information, traffic, impact zones). Here we discuss recent changes made to the modeling framework, including additional model verification using risk perception data collected during Hurricane Ian. We then discuss how the new and improved model can explore questions like Q1 – Which forecast elements (track, intensity, storm size) are most important to accurately predict for successful evacuation of real and hypothetical storms and does this vary with different measures of evacuation success? Q2 – Are there diminishing returns in forecast accuracy with respect to evacuations and how does that compare with physical limits of predictability of tropical systems? Q3 – How do answers 1–2 vary across future projected population-infrastructure-climate scenarios and risk reduction strategies? In beginning to answer these types of questions we demonstrate how modeling frameworks like FLEE can combine existing physical-social-computational research to study the hurricane forecast-evacuation system from a new perspective. This includes providing a societally-relevant alternative perspective to traditional metrics of forecast accuracy, which may ultimately help the weather-hazard enterprise make forecasts more useful to society.