Rural Road Safety Monitoring Using Crash Severity Predictive Models, Case Study Khorasan Razavi

Fatality, injury, disability, and medical expense, road and vehicle damage, and mental aspects are the main disadvantages of crashes in a community. This study tries to find the influential factors, which are predictable on the severity of crashes in the rural highway using the descriptive logit model family. First, crash data are integrated with traffic data obtained from loop detectors in rural road. Second, the Ordered logit (OL) and Multinomial logit (MNL) models are developed to identify the influential factors on crash severity. Third, the models are evaluated in terms of goodness of fit. They are applied for the short-term prediction of crash severity on rural road network in Khorasan Razavi, one of the most populated province in the northeast of Iran. Results show that the overall goodness of fit for both OL and MNL models are 0.017 and 0.019, respectively that expectedly indicates the MNL model is more accurate. The independent traffic variables "heavy vehicle flow" and "speed above 85 km/h" are significant in the suggested OL model. In the MNL model, the independent variable "minimum headway" is also significant for accidents with serious damage, in addition to the significant variables "heavy vehicle flow" and "speed above 85 km/h" for the severity levels of injury and fatality.