An Identification Method for Hazardous Roadway Locations Considering Varying Dispersion of Traffic Crashes

Abstract

Transportation safety has been recognized as one of the most urgent social issues considering that there are huge amounts of loss due to accidents on roadways. The identification of hazardous locations is a methodology that is used to decide the spatial scope for safety improvements by quantifying the level of hazard. The main purpose of this research was to improve the method for identifying hazardous locations and to validate the method based by simulation and by collecting actual data. Furthermore, in this research, the effects of varying dispersion, which is observed in actual crash data also was investigated. In order to clarify the mathematical relationship between the properties of safety performance functions and the identification of hazardous locations, an evaluation was performed using simulation data. The performances of the identified locations were compared using the crash-count method (existing system) and the potential safety improvement (PSI) method based on the predetermined true value. When the empirical Bayes method was applied, a generalized negative binomial regression model reflecting varying dispersion was evaluated in terms of estimation error for PSI. To overcome the limitations of the simulations, an evaluation was performed based on actual data. The statistical significance of the alternative model was evaluated. In addition, the practicality of its application was examined according by evaluating the level of hazard based on reproducibility. This research is applicable to the road safety audit process on expressways. The method suggested in this research provides an accurate spatial scope, which is definitely required for safety improvement projects. Also, this method improved the accuracy of the analysis and reduces associated costs.