Modeling Crash Frequency of A Typical Mountainous Freeway

Abstract

Mountainous freeways are a serious road hazard which is generally associated with high crash rates and serious human injuries. This study aimed to investigate the safety effects of various relevant risk factors associated with mountainous freeways by developing crash frequency prediction model. The data of analysis were obtained from a 20-mile-long mountainous section on Colorado Interstate-70, which is regarded as one of the most dangerous roads in the United States. As over-dispersion and zero-inflation problems were observed from the subject data, a zero-inflated negative binomial model was established. Estimation of the model coefficients indicates that features such as steep slopes, curves, seasonal factors, and tunnels significantly contribute to crash occurrence on mountainous freeways. The relative effect of each risk factor was also calculated to evaluate the relative significance to road safety. Notably, a new method was proposed to evaluate the relative significance of the variables under the zeroaccident state. The research results from this study are helpful in crash hotspot identification and safety countermeasure development.