Accident prediction models for rural roads

Following on earlier work, the development of additional models for estimating the accident potential of rural road sections is described. In the modelling procedure presented, Ontario data are used to develop regression model estimates of a section's accident potential based on its traffic and geometric characteristics, and an empirical Bayesian procedure is used for refining these estimates using the section's accident count. These refined estimates are shown in a validation exercise to be superior to estimates based on accident counts or regression models only. This method of estimating accident potential is recommended for use in the identification of accident blackspots and in the evaluation of safety treatments on rural roads. To facilitate this use, detailed example applications are presented.