Estimating accidents in a road network

This thesis describes the development of models for predicting accidents at the intersections in a road network, from the turning traffic volumes at each intersection. The accident prediction models were developed using Poisson and negative binomial regression for each of the major accident types, at each of the major intersection types. Countrywide models that predict the motor vehicle accidents (accidents involving motor vehicles only) in a five year period, from the product of the conflicting traffic volumes were developed for each accident type. Accident prediction models have also been developed to predict the number of motor vehicle accidents occurring in different periods of the day (eg. the morning peak, 7am to 9am) and in different urban centres. This thesis also describes three case studies, where the accident prediction models have been used to predict the total number of accidents (major accident types) at the intersections in three road networks. Case studies were performed on the Christchurch Southern Arterial network, Christchurch Central network and the Lower Hutt network. For the two latter networks the total number of intersection accidents predicted was quite similar to that observed. In the Southern Arterial network, however, the total number of intersection accidents was under-predicted considerably