Date of Award
Doctor of Philosophy
Civil, Construction, and Environmental Engineering
Shauna L. Hallmark
Traffic safety during winter seasons has been a serious concern in Iowa as hundreds of people are injured on Iowa's highways each winter. As the goal of the state transportation agency is to ensure the mobility of road users without compromising the safety during winter periods, it is important to understand the factors affecting winter-weather crash frequency and occupant injury risk through quantitative prediction models. It is of utmost importance to identify locations prone to winter-weather crashes to utilize the limited resources efficiently for improving safety during winter conditions. This research intended to develop a systematic prioritization technique to identify winter-weather crash hotspots by using Empirical Bayes technique that addresses the serious limitations of the traditional methods to screen road networks for identifying high crash locations. This research also addresses the issue of hierarchical structure in the crash data by developing quantitative models to predict occupant injury risk for crashes occurring during winter seasons to obtain unbiased and accurate estimation of the parameters for better management of road safety during winter seasons. Along with developing site prioritization techniques for identifying roadway segments with potential for safety improvement through traditional statistical methods using raw crash data, Empirical Bayes technique is used to screen roadway segment through developing safety performance functions for winter-weather crashes. A novel approach is adopted to extract weather data from information reported by winter maintenance crew members to incorporate weather related factors in developing safety performance functions at network level for three roadway types in Iowa. Weather factors such as visibility, wind velocity, air temperature are found to have statistically significant effects on winter-weather crash frequency. The ranking of roadway segments based on Potential for Safety Improvement (PSI) by employing Empirical Bayes technique differs from the ranking produced by simple crash frequency. Safety Performance Functions developed in this research can be used to produce ranking based on PSI by using crash observations made over a specific number of years for winter-weather crashes. Models predicting occupant injury risk with binomial logit formulation are developed considering the hierarchical structure of the crash data in a Bayesian framework in this research for weather-related crashes, non-weather related crashes, and all crashes occurring during the four winter seasons (2008/09 to 2011/12) in Iowa. These models are developed using disaggregate crash data with occupants nested within crashes. High values of between-crash variance for the three models underscore the justification of considering the hierarchical nature of the crash data due to the natural crash data collection process. Factors related to occupants (gender, seating position, trap status, ejection status, airbag deployment, safety equipment used) had statistically significant effects on occupant injury risk for all the models. Weather-related variables such as visibility and air temperature were found significant predictors of all crashes and weather-related crashes during the winter seasons. The variable representing road surface condition is also found to be a significant factor in all three models developed to predict occupant injury risk during the winter seasons.
Mohammad Saad Bin Shaheed
Shaheed, Mohammad Saad Bin, "Bayesian analysis of factors affecting crash frequency and severity during winter seasons in Iowa" (2014). Graduate Theses and Dissertations. 13810.