Household-Level Model for Hurricane Evacuation Destination Type Choice Using Hurricane Ivan Data Rodrigo Mesa-Arango, Samiul Hasan, Satish V. Ukkusuri, and Pamela Murray- Tuite Group 4

Background Hurricanes are one of the most dangerous and costly weather-related natural hazards in the United States. The average fatalities per year related to hurricanes increased Considering these devastating impacts and the role of evacuation on their mitigation, it is the responsibility of public agencies to understand all the dimensions of an evacuation process. Comprehensive evacuation plans must integrate transportation theory with evacuation behavior.

Previous Literature Models are classified into three major groups: trip generation, departure timing, and destination and/ or route choice. But the destination choice has only been studied by a small number of researchers. Three types of zonal-level models have been used : gravity model, intervening opportunity model and MNL. they focus on zonal trip distribution without incorporating the choice among destination types and considering the percentage of evacuees traveling to each destination type as a given input. Destination only include houses of friends and relatives and hotels.

Some definitions important for evacuation models Proximate destination- the point in the transportation network where the evacuee exits the risk area Ultimate destination refers to both the town and/or city and the type of accommodation where the evacuees will stay until they can return to their homes This paper focuses on the second part of the ultimate destination

In This Paper Destination: houses of friends and relatives, hotels, public shelters and churches; and other. Variables influencing this choice include: hurricane position at evacuation time, household geographic location, race, income, preparation time, changes in evacuation plans, previous experiences with major hurricanes, household members working during the evacuation, and evacuation notices. Using a nested logit model. Data from Hurricane Ivan 2004 is used to calibrate the model. Application: findings can be used to develop evacuation strategies

How does this research relate back to Policy making? Recognizing public shelter demand and improving their locations and settings Developing better evacuation notices per population segment, giving advice on what destination types to choose Developing cooperative programs with hotels guaranteeing some levels of demand Recognizing potential regions that are attractive for evacuees to anticipate traffic congestion

Modeling framework NL model ⇒ overcomes the IIA assumption problem by nesting alternatives and cancelling out their shared unobserved effects Fig.?? Nested logit structure for hurricane evacuation destination type choice ←First level : destination type ←Second level : each destination

Estimation flow Logsum value Unconditional Probability function of destination type Conditional probability of each destination The probability of household h choosing destination type j (1) (2) (3) (4)

variables 11 variables among 116 potential explanatory variables  Previous experience  Average distance between the hurricane and the centroid of the zip code where the house hold is located  Indicator variables for White race dummy  Indicator variables for Low income  Indicator variables for evacuation notice  Indicator variables for work during the evacuation Example of explanatory value which have high t value

Implication to Our Research Research flow should be followed: Data Analysis →Model Formulation (sorting out significant factors) →Model Estimation (including considering which model to use) →Elasticity Calculation (evaluating or drawing up policy) NL Model can be applied to our destination choice model. Destination choice model focusing on individuals seems difficult to apply to our research focusing on visitors, whose characteristics is difficult to assume to some extent.

MNL

NL

Implication to Our Research Research flow should be followed: Data Analysis →Model Formulation (sorting out significant factors) →Model Estimation (including considering which model to use) →Elasticity Calculation (evaluating or drawing up policy) NL Model can be applied to our destination choice model. Destination choice model focusing on individuals seems difficult to apply to our research focusing on visitors, whose characteristics is difficult to assume to some extent.