Simulation of Hurricane Irma Evacuation Process

Abstract

Hurricane evacuation management is very challenging, due to the lack of experience/data for extreme events, the complexity of human decision making and travel behavior, and the large uncertainty in storm prediction. Recent Hurricane Irma (2017) created the largest scale of evacuation in Floridas history, involving about 7 million people on mandatory evacuation order and 10 million evacuation vehicles. Here we reorganize the traditional hurricane-prediction and governmental-orderbased traffic demand generation model into an agent-based mean-field model that considers also human decision making, to perform fast evacuation modeling, using Irma as a study case. With a fast simulation algorithm, the model can be evaluated and calibrated with (often very limited) traffic observation, partially overcoming the problem of data deficiency. The calibrated model is found to well capture the global features of the evacuation process in Irma. The analysis also reveals that people may have put more weight on the predicted hurricane category than the governmental order when making evacuation decisions during Irma, indicating possibly a higher panic level than that in previous storms. The developed model can also be used to help improve evacuation management.