Secondary crash (SC) occurrences are major contributors to traffic delay and reduced safety, particularly in urban areas. National, state, and local agencies are investing substantial amount of resources to identify and mitigate secondary crashes to reduce congestion, related fatalities, injuries, and property damages. Though a relatively small portion of all crashes are secondary, determining the primary contributing factors for their occurrence is crucial. The non-recurring nature of SCs makes it imperative to predict their occurrences for effective incident management. In this context, the objective of this study is to develop prediction models to better understand causal factors inducing SCs. Given the count nature of secondary crash frequency data, the authors used count modeling methods including the standard Poisson and Negative Binomial (NB) models and their generalized variants to analyze secondary crash occurrences. Specifically, Generalized Ordered Response Probit (GORP) framework that subsumes standard count models as special cases and provides additional flexibility thus improving predictive accuracy were used in this study. The models developed account for possible effects of geometric design features, traffic composition and exposure, land use and other segment related attributes on frequency of SCs on freeways. The models were estimated using data from Shelby County, TN and results show that annual average daily traffic (AADT), traffic composition, land use, number of lanes, right side shoulder width, posted speed limits and ramp indicator are among key variables that effect SC occurrences. Also, the elasticity effects of these different factors were also computed to quantify their magnitude of impact.
Keywords: Count models; Freeways; GORP; Prediction model; Secondary crash frequency; Segment characteristics.
Copyright © 2016 Elsevier Ltd. All rights reserved.