In this paper an attempt has been made to develop correlation between highway geometric parameters, traffic parameters and crash rate for two lane non urban highways. To simulate the stochastic nature of crash occurrence, Poisson regression and negative binomial regression has been employed. A detailed study has been carried out on a two lane non urban highway (National Highway No 58) in India. The study included collecting of traffic and geometric features of the highway along with past crash data for 8 years. Variables included in the model are access density per km, curve density per km, intersection density per km, space mean speed, shoulder width, volume of heavy vehicles, volume of non-motorized and animal drawn vehicles and total volume. The dependent variable is crash rate per year per km. Strong positive relationships were observed between volume of heavy vehicle and crashes and speed and crash rate. Shoulder width and lane width show negative relationship with crash rate. (a) For the covering entry of this conference, please see ITRD abstract no. 0612AR242E.
Samenvatting