Evaluating the feasibility of predicting safety at alternative intersections

The safety at alternative intersections is largely unknown. Alternative intersections typically relocate the left-turn movement from the primary intersection with the objective of achieving operational and safety benefits. Although there have been a handful of studies which provide information on the safety impact of a few select alternative intersections, these studies are limited in number and tend to be regionally specific. The work herein seeks to evaluate the feasibility of predicting safety at alternative intersections by building an intersection from a microscopic perspective. Four models were developed: a binary logit model was developed to predict the likelihood of a merge crash in a month at a merge conflict point, the likelihood of a diverge crash in a month at a diverge conflict point, and the likelihood of a crossing crash in a month at a crossing conflict point. A multivariate model was developed to better understand the interplay between merge, diverge and crossing conflict points at the intersection level. The models indicate that variables representing weather, exposure, geometry, and complexity are present in all of the microscopic models. Finally, the results of the aggregated model indicate that there is a relationship between the likelihoods of a merge, diverge and crossing crash.