Collision Prediction Models for Three-Dimensional Two-Lane Highways

Abstract

Collision prediction models for three-dimensional (3-D) alignments of two-lane rural highways are lacking in the literature. This paper presents such models with vehicle collision data on 5,760 km (3,600 mi) of two-lane rural highways in Washington State collected from 2002 through 2005. Five statistical models were developed for different combinations of 3-D alignments to establish the relationship between collision frequency and the relevant variables. The alignment combinations are (a) horizontal curves combined with crest vertical curves, (b) horizontal curves combined with sag vertical curves, (c) horizontal curves combined with multiple vertical curves, (d) horizontal curves combined with grades of less than 5%, and (e) horizontal curves combined with grades of more than 5%. For each combination, four different statistical techniques were explored: Poisson, negative binomial, zero-inflated Poisson, and zero-inflated negative binomial. For validation, two models were selected and estimated by using the first 3 years of collision data and validated with the last-year data. The results show that the most significant predictors for collisions on horizontal curves on 3-D alignments are the degree of curvature, roadway width (lanes plus shoulders), access density, product of grade value and grade length, road section length, and average annual daily traffic. The results of this study should be useful in evaluating road safety on 3-D alignments and optimizing their design based on the substantive safety approaches.