| Summary: | Modern transportation planning considers issues such as road congestion, pollution and mobility proactively. However, road safety is usually evaluated in a reactive manner only after the implementation of transportation plans, and when safety problems arise. Although this reactive road safety approach has been very effective, it is associated with significant costs on communities. Therefore, several researchers advocate a more proactive approach to road safety analysis. Several studies developed macro-level Collision Prediction Models (CPMs) that could assess the road safety in a proactive manner, and provide a safety planning decision support tool to community planners and engineers. However, these models have limitations as they do not target the safety evaluation of different goals of a typical city transportation plan. Therefore, the motivation for this research arises from the necessity of developing tools that could predict the safety effect of a typical city transportation plan such as changes in the transportation and transit network configurations, and ultimately could evaluate the safety estimates among alternatives of different transportation plans and policies.
The main goal of this thesis is to develop a set of macro-level collision prediction models to investigate the relationships between various transportation and sociodemographic characteristics, and the overall roadway safety. The developed models consider the Poisson variations and the heterogeneity (extra-variation) on the occurrence of collisions and the spatial effects on the distribution of collisions caused by the similarity in environment and geography of the neighbouring sites. It has been shown that the goodness of fit improved with the incorporation of spatial effect.
In this study, a set of zonal-level transit reliant and application-based collision prediction models were developed. Data from Metro Vancouver, British Columbia were used to develop models using a generalized linear modelling approach with a negative binomial error structure. Different transit-related variables were found to be statistically significant, namely bus stop density, percentage of transit-km traveled with regard to total vehicle-km traveled, percentage of commuters walking, percentage of commuters biking, and percentage of commuters using transit. The CPMs related total, severe, and property damage only collisions to the implemental aspects related to the goals of long-term transportation plans.
|
|---|
