In order to achieve a transportation system that is safe and sustainable, engineers have recognized the need to include traffic safety as an integral part of the traffic planning process. Knowledge of the safety levels resulting from the route choices made by drivers can be useful in incorporation of safety in route assignment plans and ultimately in safety-oriented road designs. This study is aimed at quantifying the relationship between accidents and traffic flow with respect to the manoeuvres made at intersections. The first part of this study involved the development of accident prediction models for each turning movement on intersections. The second part of the study tested the practical application of these models as tools for accident prediction by applying them to a different dataset. The results show that the process of modelling accidents is complex, with different factors influencing the results differently. This makes predictive modelling, and comparison of model performance a particularly difficult task. The study was able to quantify the safety levels associated with the different route choices made at intersections and found that the flow increases with increasing traffic flow for all route choices. The number of accidents per unit vehicle crossing the intersections is highest for left turning flow, followed by straight through flow. Right turning flow has the lowest accident rate. The results show that although accident prediction models are accurate at explaining the data used in their development, their predictive power is low when applied to a different dataset. Suggestions were made on how the models can be used and how the transferability of models can be improved. (Author/publisher)
Samenvatting