This report describes the further development of a road accident analysis tool, called the Road Accident Analyzer. In a previous report (RA-MOW-2011-011) it was already discussed how this tool is able to identify dangerous road segments on the main road network in Flanders. This report will firstly address some issues that came to light in the previous research, and how these are remediated in the current report. The most important problem arises in the accident registration process. It seems that approximately 90% of all the road accidents on the main road network in Flanders are registered in the same direction. Because of this registration error, the two driving directions are aggregated in the analyses carried out in this report. This restriction obviously results in less detailed results on the one hand, but provides a more reliable representation of traffic safety on the Flemish highways on the other hand. Thereafter, the theory behind the identification of dangerous road segments is being discussed. In studies concerning the identification of dangerous sections, one can usually distinguish two main phases. In the first phase, an accident indicator is calculated. Subsequently, it is examined whether this indicator significantly exceeds a certain threshold value in a second stage. After the theoretical discussion of these two phases, they are tested in practice. Three accident indicators are being calculated for the main roads in Flanders: the accident density, the accident risk and an expected number of accidents based on an accident prediction model (Empirical Bayes approach). Next, the most important and dangerous segments for each method can be identified based on average values and the accident reduction potential. Afterwards, the results of the different methods are being compared and critically discussed. Finally, the most dangerous segments that are identified by the Road Accident Analyzer are compared to the black spots determined by the 1-3-5-rule, as currently applied in Flanders. (Author/publisher)
Samenvatting