Accident prediction models (APMs) are mathematical models which express the safety performance in terms of traffic and road characteristics and estimate the road safety performance of a specific road type or intersection type. They allow road authorities the opportunity to monitor the safety performance of their roads and enhance the safety performance of their roads when necessary. This paper outlines a methodology for developing accident prediction models disaggregated to road characteristics. Instead of developing one accident prediction model for a general road category (e.g. all rural distributor roads), road links and intersections are divided into several categories. This will result in several homogeneous groups of road links and intersection types. An accident prediction model will be developed for each road link and intersection group. This study focuses on Dutch rural 80 km/h roads and intersections which are under provincial administration. The database used to develop the accident prediction models consists of 1200 kilometres of road length and 500 intersections from two Dutch provinces. The database contains geometric data of the provincial 80 km/h roads, traffic and road crash data. Variables that correlate with the number of crashes (e.g. traffic volume and road length) will be used to develop the accident prediction models. (Author/publisher)
Abstract