The Province of New Brunswick currently lacks the capability to identify road segments requiring safety remediation using state-of-the-art methodologies. This empirical research project enables collision locations to be objectively prioritized for cost-effective treatments. It is vital that only the most problematic locations be identified for improvement due to finite maintenance budgets. Collision history data were analyzed, and roadway, roadside, and traffic data were compiled for a sample of rural locations with the goal of developing collision prediction models (CPMs). Two collision prediction models were developed for divided and undivided road segments using negative binomial regression. Variables found to have the most significant effect on collisions include AADT, segment length, grade, posted speed, lane width, plant hardiness zone, surface age, surface type, and wildlife fencing. Predicted collision frequencies derived from the models can be compared to actual collision data to calculate potential for improvement (PFI) values to identify locations experiencing more collisions than expected. This will enable road authorities to make appropriate decisions to prioritize road locations that can benefit from remediation. A case study was completed to illustrate how the models can be applied to road segments outside of the model development dataset. A demonstration was provided to show how road segments can be ranked using different collision performance techniques. The case study was also used to determine how the proposed models would perform in comparison with the default procedure in the Highway Safety Manual (HSM). It was determined that the newly developed CPMs resulted in smaller PFI values than the HSM SPFs.
Zircon - This is a contributing Drupal Theme Design by WeebPal.
