Development of a Time-Dependent Disaggregate Hurricane Evacuation Destination Choice Model

Abstract

A time-dependent disaggregate destination choice model for evacuees from a hurricane is developed and tested in this paper. Using data from a survey conducted among South Carolina residents following Hurricane Floyd in 1999 as the basic information, dynamic information regarding the storm, the network, the destination zones, and decisions made by the emergency managers regarding the issuing of evacuation notices was added to the sample to provide a database that contained time-dependent data in 48 2-h time periods preceding landfall. Separate models were established for evacuees seeking shelter at the homes of friends and relatives and those who went to hotels and motels. The models describe destination choice in terms of time-dependent travel time between origin and destination, the amount of accommodation remaining in each zone in each time period, the estimated likelihood that the storm’s path takes it over a destination zone, the ethnic similarity between origin and destination zones, and the presence of a large metropolitan area and/or interstate highways in a destination zone. When estimating the performance of the models, their trip length frequency distributions are not significantly different to the observed trip length frequency distribution at the 95% level of significance. Applying the friend/relative model to the neighboring state of Georgia produced a trip length frequency diagram that had a similar pattern to the observed distribution but underestimated short-distance destinations and overestimated distant destinations.