A comparative analysis of identification of hazardous locations in regional rural road network

The first step of the road safety management cycle is the identification of hazardous road locations. Traditionally, the identification criterion in the Czech Republic has been recorded crash frequency; however, it has tended to omit the important influence of natural variations known as regression to the mean. Using the expected number of crashes and empirical Bayes adjustment is the recommended solution. This number is calculated with the use of a crash prediction model, taking into account available explanatory factors and controlling for potentially confounding variables at the same time. The aim of this study was to compare both traditional and empirical Bayes approaches to identification of hazardous road locations. A crash prediction model for the regional rural road network was developed for the purpose of the study. The whole 2nd class road network in Czech region of South Moravia was used. The resulting expected injury crash frequency was further adjusted by the empirical Bayes estimate. The empirical Bayes estimate was used as a criterion for the ranking of hazardous road locations list. At the same time, another list was produced using the Czech traditional criterion of recorded crash frequency. Three pairs of lists were developed for three time periods (2007 – 2009, 2008 – 2010 and 2009 – 2011) and compared. The results showed that the predictions models with EB adjustment offered the results which were more stable in time compared to the traditional approach. This proves the potential suitability of empirical Bayes approach identification in the road safety management practice.