Hurricane Evacuation Behavior Analysis Based on Mobile Location Data: A Case Study of Hurricane Florence

30 pagesHurricane Florence in 2018 was one of the worst in North Carolina history. The catastrophic flooding and infrastructure damage endangered millions of residents and visitors. The focus of this paper is the development of methods to understand evacuation behavior to aid in the development of future evacuation plans and emergency responses. While existing hurricane evacuation studies primarily use statistical models based on survey data, we develop a new approach using location data passively collected by cellphones. By analyzing location information from about 18,000 anonymous and opted-in users before and during the hurricane, we discern their home census area and whether they evacuated, and if they evacuated, where they evacuated too. About 91% of inferred home areas are found to be located on residential land parcels. North Carolina has 100 counties. Of those 100 counties, about 21 counties were in areas where the condition warranted the consideration of evacuation. In those 21 counites the evacuation rate was about 56%. In about six coastal counties, conditions were most severe and for those six counties, the evacuation rate was found to be about 67%, which is consistent with survey data collected from 1,421 respondents