This study investigates the effects of the traffic flow parameters related to individual drivers' lane-changing and car-following behavior on the occurrence of sideswipe and rear-end crashes on freeways. A total of 184 sideswipe and 605 rear-end crashes on I-4 freeway in Orlando with the loop detector data were used for the analysis. A set of the Bayesian logistic regression models were developed to estimate the likelihood of crashes in a specific lane compared to the adjacent lanes using the 5-min. average and lane-by-lane traffic flow parameters. The analysis results showed that thesignificant traffic flow parameters affecting crash occurrence in a specific lane are distinctively different between sideswipe and rear-end crashes. The flow-related variables were significant in the sideswipe crash models whereas the speed-related variables were significant in the rear-end crash models. The results also showed that it is important to observe the traffic flow parameters in the subject lane and the adjacent lanes, not onlythe parameters in the subject lane, to predict the crash occurrence in the subject lane. This suggests that the lane-by-lane traffic flow parameters are considered as surrogate measures of lane-changing and car-following behavior that are associated with the occurrence of sideswipe and rear-endcrashes. These parameters can potentially be used for real-time monitoring of the likelihood of crash occurrence by lane on instrumented freeways.
Abstract