A correlated random parameter approach to investigate the effects of weather conditions on crash risk for a mountainous freeway. (January 2015)

• Record Type: Journal Article
• Authors: Yu, Rongjie; Xiong, Yingge; Abdel-Aty, Mohamed
• Is Part Of: Transportation research. Volume 50(2015)
• Page Start: 68
• Page End: 77
• Publication Date: 2015-01
• Subjects: Aggregate traffic safety -- Correlated random parameter model -- Tobit model -- Weather warning system; Transportation -- Periodicals; Transportation -- Technological innovations -- Periodicals; 388.011
• DOI: 10.1016/j.trc.2014.09.016
• Languages: English
• ISSNs: 0968-090X

Abstract

Highlights: Analyzing weather conditions' effects on crash risk for a mountainous freeway. Real-time weather information was utilized and treated as random parameters. Correlated random parameter Tobit model provided better modeling goodness-of-fit. Results from the models would shed lights on future applications of weather warning systems. Abstract: Freeway crashes are highly influenced by weather conditions, especially for a mountainous freeway affected by adverse weather conditions. In order to reduce crash occurrence, a variety of weather monitoring systems and Intelligent Transportation Systems (ITS) have been introduced to address the weather impact. However, the effects of weather conditions on crash occurrence have not been fully investigated and understood. With detailed weather information from weather monitoring stations, this study seeks to investigate the complex effects of weather factors, such as visibility and precipitation, on crash occurrence based on safety performance functions. Unlike conventional traffic safety studies which deal with crash frequency, crash rates per 100 million vehicle miles travelled were adopted as the dependent variable in this study. Three years of weather related crash data from a 15 mile mountainous freeway on I-70 in Colorado were utilized. First, a fixed parameter Tobit model was estimated to unveil the effects of explanatory variables on crash rates. Then, in order to characterize the heterogeneous effects of weather conditions across the homogeneous segments, a traditional random parameter Tobit model was developed. Furthermore, for the purpose of monitoring the intricate interactions between weather conditions and geometric characteristics, a multivariate structure for the distribution of random parameters was introduced; which result in a correlated random parameter Tobit model. Likelihood ratio test results demonstrated that the correlated random parameter Tobit model was superior to the uncorrelated random parameter and fixed parameter Tobit models. Moreover, visibility and precipitation variables were found to have substantial correlations with geometric characteristics like steep downgrade slopes and curve segments. Results from the models will shed lights on future applications of weather warning systems to improve traffic safety.