A Tree-Based Ordered Probit Approach to Identify Factors Affecting Work Zone Weather-Related Crashes Severity in North Carolina Using the Highway Safety Information System Dataset

Work zone crashes are still on the rise due to the aging of US roads and the increase in traffic demand. Investigation of crash characteristics and determining contributing factors in work zones is one of the most important issues in many traffic safety studies. The effect of work zones on traffic safety can be exacerbated by weather conditions. A sudden reduction in visibility may intensify the severity of work zone crashes. Although many studies have investigated work zone crashes, research that investigates the impact of adverse weather conditions on work zone crashes is lacking. In this study, The Highway Safety Information System database for North Carolina was used to identify the characteristics of work zone weather-related crashes. A Tree-based Ordered Probit, a relatively recent and promising combination of nonparametric machine learning (decision tree) and classical statistics (ordered probit) techniques, was utilized to gain a better understanding about the effects of various factors on different work zone crash related injury and crash severity in adverse weather conditions. The results showed that Tree- based Ordered Probit model has a better performance compared to conventional Ordered Probit Model. Lighting conditions, number of vehicles involved in a crash, road characteristics, number of occupants, land use, presence of traffic control devices, and two types of crashes (sideswipe and rear-end crashes) were identified as the most important factors in work zone weather-related crash severity.