Geometric consistency is becoming a useful technique, with research being carried out in various countries to develop predictive models to improve rural road safety. The main aim of this research is to develop reliable accident prediction models for rural single carriageways in the UK using highway geometric design consistency measures. This paper describes a research project that is being carried out using data from two counties in the north east of England. In order to collect the relevant data, a novel method of estimating horizontal curvature details from digital mapping was developed for this study, as no construction plans were available for such roads. After that, roads under study were divided into sections according to certain criteria. Traffic flows and accident data for each road section were obtained for the 5 year period 2000-2004. Accidents were located in the correct element (e.g. curve or tangent) for each road section. As operating speed (V85th percentile) was needed in order to apply the some of the consistency measures, several single and multiple operating speed regression models were developed for both horizontal curves and tangents. As a step towards achieving the main goal of this project which is to develop several accident prediction models, a set of consistency measures over entire road sections were defined. A bivariate correlation analysis between accident rates and these measures have been carried out. Many consistency measures were found to have an association with accident rates especially in the case of single vehicle accidents. Applying univariate regression has shown that some relationships do exist, but that further work is required using multivariate analysis. This work is now being out, using appropriate statistical techniques including artificial neural networks (A). For the covering abstract of the conference see E217780.
Abstract