The model requires various types of information, the first being about the storm—such as intensity and projected path—and the second being what officials are doing. For example, have they issued evacuation orders? Are they opening up a contraflow? Are the orders regarded as mandatory or voluntary? The third type of information concerns peoples’ living conditions—such as whether they live in mobile homes and whether they have cars.
Wilmot’s team uses that information to determine what the evacuation behavior will be. The data fed into the model is dynamic and changes over time as the storm and other conditions change.
Wilmot’s model also uses data from post-event surveys. He gets dynamic data about each storm from the National Hurricane Center, and that is used to build up databases that can feed the models.
Wilmot’s group breaks out models for separate types of behavior. For example, one model might address whether the household would evacuate. The next model would look at where the household would evacuate to. Then there would be a model looking at the route the household would use to evacuate. Other models would look at what type of vehicle the households would evacuate in.
The models are used in a daisy chain to attempt to reproduce the total behavior of people. The goal is to predict, based on certain conditions of a storm and certain actions taken by decision makers or emergency officials, which households are likely to evacuate when, and if they do, where they would be heading and what route they would choose. “That enables you to give sort of traffic estimates,” Wilmot says.