Graduation Year


Document Type




Degree Granting Department

Civil and Environmental Engineering

Major Professor

J. John Lu, Ph.D.

Committee Member

Jayajit Chakraborty, Ph.D.

Committee Member

Manjriker Gunaratne, Ph.D.

Committee Member

Pei-Sung Lin, Ph.D.

Committee Member

Abdul Pinjari, Ph.D.

Committee Member

Yu Zhang, Ph.D.


Accidents, RSAP, Run-Off-Road, GIS, Cost-Benefit


Roadside crashes account for a large portion of total fatal crashes that occur annually in the United States. About 30% of those fatalities are the result of single vehicle run-off-road crashes. A large proportion of these fatal crashes occur in rural roads when vehicles depart from the travel lane and collide with trees or other roadside safety hazards. Many of these run-off-road accidents occur in local roads that carry traffic volumes between 1,000 and 20,000 vehicles per day. Many of these roads are part of the jurisdiction of county authorities faced with the dilemma of having too many "potentially dangerous" sites and lacking a methodology for assessing their risk to rank them accordingly; and to apply the limited resources to the ones that will bring the greatest benefit to society. This situation describes the case in Hillsborough County, Florida, in 2004 when they contracted a study with the Transportation Program of the Department of Civil and Environmental Engineering of the University of South Florida. The initial scope was to develop a methodology to assess the potential risk for each of 19 sites in a given list to prioritize further studies. The project was sponsored by the Engineering Division, Public Works Department, of Hillsborough County. The methodology developed considered the roadside safety hazards at each location and it was based on the use of the Road Safety Analysis Program (RSAP) software distributed as part of the 2002 AASHTO's Roadside Design Guide. This dissertation presents a further development of this approach: it continues to use the probabilistic approach built into RSAP to calculate the annual crash cost of each roadside safety hazard at 45 study segments. It then obtains regression models to predict that annual crash cost, as computed by RSAP, based on roadway and traffic characteristics as well as on the nature, location and physical dimensions of the roadside safety hazard. For each study segment, the annual crash cost of each feature (as estimated with the models developed) is added for a final comparison with the RSAP Annual Crash Cost. A coefficient of determination (R2) of 0.80 was obtained. The models developed were finally used to replicate the original 2005 study for Hillsborough County. Although there were minor variations on the risk index originally computed, the ranking of the 19 study sites remained basically the same with a clear cut indication of the sites that should be considered for further engineering studies.