Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)


Civil and Environmental Engineering

First Advisor

Natacha Thomas


Hurricane landfalls have great potential to cause human injuries, loss of lives and loss or damage of properties. Currently, the prediction of a hurricane hit at a given location has significantly improved owing to the advancements in meteorology and other contributing technologies such as satellite sensing systems among others. In spite of the developments in hurricane track forecast, the most common risk aversion strategy in response to hurricanes still remains the emergency evacuation of the flood zones under the direction and coordination of government officials.

Rhode Island, known as the Ocean State, is the smallest state within the continental United States of America. Nonetheless, it boasts about 384 coastal miles along the Atlantic Ocean. Past history indicates a non-negligible risk posed by hurricanes to the coastal regions of Rhode Island with yearly frequencies of a hurricane hit within 75 nautical miles of central Providence, RI, at 5%, 6% and 2% for categories 1, 2, and 3, respectively as derived from the “Tropical Cyclones of the North Atlantic Basin from 1851 to 2001” database from the National Oceanic and Atmospheric Administration.

Given the devastating effect of hurricane Katrina, 2005, Campbell et al, 2007, conducted a hurricane risk assessment study for the state of Rhode Island using selected socio-economic factors, which pointed to Warwick, Newport, Barrington, Narragansett and Providence as the towns most potentially vulnerable to storm surges. To further the previous work, hypotheses are put forward in this study to query the association between the household socioeconomic and demographic attributes, the decision to evacuate, the behavior at evacuation and the evacuation preparedness level for a sample of earlier mentioned towns as well as Jamestown. The aim is to apprehend the data necessary to the calibration of a hurricane evacuation model rooted in the anticipated behavior of evacuating households. To this end, the study 1) probes heads of households using a survey instrument 2) conducts statistical analyses of the gathered data 3) compares the behavioral data obtained with the generic ones derived for southern states, and 4) develops a behavioral evacuation model of the RI flood zones. Based on the insights gained, it further provides in conclusion some suggestions on the desirable modifications to the survey that may promote further evacuation model enhancements given the simplifying assumptions made.

The findings show that about 80% of Rhode Islanders are willing to comply with evacuation notices when issued by government officials. Head of household’s age, education, household income, and prior hurricane evacuation experience do not display any association with the decision to evacuate. There seems to be a relationship between race and the decision to evacuate, chi-square p-value of 0.025, but the percentage of minority in the sample size is too low to reach a conclusive result. The findings highlight the positive association between hurricane workshops/meetings/classes participation and hurricane risk preparedness.

The survey analysis further confirmed that not all the owned household vehicles would be used at evacuation. It provided the basis for deriving a cross-classification table for owned household vehicles versus evacuating household vehicles, which enables the conversion of evacuating households into evacuating vehicles. The attractions, as obtained from the survey, were 68% for friend and family homes, 17% for shelters, and 15% for hotels/inns/bed-and-breakfasts. It is worthy to note that the total person-occupancy of 8,448 at 18 Red Cross-designated shelters in Rhode Island will not suffice to satisfy the evacuee demand of about 36,700 persons in a major hurricane scenario.

A gravity model was used for the purpose of trip distribution to friend and family homes outside of the flood zones. The friction factor was modeled by a Gamma function calibrated using the percentage of evacuating households willing to travel within given distances at evacuation. The resulting gamma function parameters a, b and c obtained equaled 50.057, 0.047, 0.008, respectively. A traffic loading curve, generated from the survey results, points to about 65% of the evacuating households as willing to leave the evacuation zone within the first three hours following a mandatory evacuation notice by government officials.

The traffic assignment results were discussed by color coding the network to present the volume to capacity ratios for all links over the entire Rhode Island Statewide Travel Demand Model network. The interstate freeways within Rhode Island and its neighboring states operate at steady states with volume to capacity ratios lesser than 0.9. However, some arterials and major local roadways in the evacuation area municipalities operate at or over capacity. Finally, future studies could modify the model to 1) simulate a dynamic evacuation that accounts for the delays in reaching evacuation orders on a township basis. They could also 2) deploy a revised survey instrument as proposed with a potential to enhance evacuation planning model accuracies.