A Tale of Two Hurricanes: Assessing the Effectiveness of Evacuation Orders in the cases of Hurricane Ian and Hurricane Idalia

In the aftermath of Hurricane Ian in 2022, officials in Lee County, Florida who did not issue mandatory evacuation orders until only one day before the hurricane made the landfall were heavily scrutinized about their decision-making: Could issuing an order one day earlier have led to more evacuation? Would it have saved lives? Hurricane Ian killed at least 148 people in Florida, 61 of those were in Lee County. Almost one year later, Hurricane Idalia made landfall in Florida as a Category 3 hurricane. This time, the death toll was 4! What was different for Hurricane Idalia? While the two hurricanes shared key characteristics (e.g., major hurricanes with destructive storm surge; approached from the Gulf of Mexico and intensified rapidly as they neared landfall), Ian made landfall in a densely populated, developed barrier island area, while the slightly weaker Idalia made landfall in a sparsely populated region characterized by marshlands and woodlands. But is that the only reason? Out of the 61 people who died because of Ian in Lee County, 33 were in the Federal government’s documented storm surge danger zone. On the other hand, there was a mandatory evacuation order in place in much of the Florida Gulf Coast from Tampa Bay northwards at least two days prior to Idalia’s landfall. How did residents in areas that received evacuation orders during Hurricane Ian respond to the orders? Were the evacuation orders sent out during Idalia effective?

This Rapid Grant proposal builds upon the EI CoLab project "Is it time to go? Using mobility data to examine the timing of hurricane evacuation decisions in response to forecasts". In the CoLab project, a novel database was developed of hurricane evacuation orders in the U.S. (i.e., HEvOD, https://www.hurrevacorder.info/) and developed a robust statistical methodology to estimate the effective of evacuation orders. Additionally, extensive geospatial mapping was conducted to map hurricane evacuation zones onto CBGs. These assets were used in conjunction with a high-resolution mobility dataset from Spectus, Inc. to analyze evacuation order responses and evacuation patterns at census block group level (CBGs). Specifically, the causal effect of evacuation orders issued during each hurricane were determined. Machine learning techniques were then used to identify key socio-economic and demographic factors influencing compliance level at the CBG level, enabling a comparative analysis across the two hurricanes.