Safety impact of highway geometrics and pavement parameters on crashes along mountainous roads

Abstract

Having the capability of estimating both the number of crashes and their severity levels, crash prediction models are a precious tool in highway safety. However, there hasn't been any research on predicting traffic crashes on Indian mountainous rural highways. The primary objective of this research is to develop safety performance functions (SPFs) for traffic crashes occurring on rural roads located in the mountainous region of Uttarakhand, India.

For analysis, the study utilized five years of crash data collected from different types of rural roads. The road network was divided into constant segments of 500m each, and separate models were developed for single (TSV_C) and multi-vehicle (TMV_C) crashes using the negative binomial regression approach. These SPFs highlight important significant variables in terms of positive and negative association and a potential change in subject crash frequencies. The results concluded that different types of risk factors impact both types of crashes, with horizontal (H_C) and vertical curves (V_C) in common. For instance, spot speed increases TSV_C crashes by 3.87 %, whereas H_C and V_C tend to increase subject crashes by 8.32 % and 29.95 %, respectively.

Similarly, TMV_C is influenced by carriageway (C_W) and shoulder width (S_W). The result proposed that an increase in C_W and S_W can decrease frequencies by 0.668 times and 0.819, respectively. Additionally, the model highlighted the importance of rut-depth and the presence of pavement markings in the road safety analysis. At last, further research scope is suggested based on the limitations of this study.