This paper presents collision prediction models for horizontal tangents combined with vertical alignments using vehicle collision data on 5,760 km (3,600 miles) of two-lane rural highways in the Washington State from 2002 2005. The paper complements a companion paper on collision prediction models for horizontal curves. Five statistical models were developed for the following combinations of horizontal tangents and vertical alignments: (1) horizontal tangents combined with crest vertical curves, (2) horizontal tangents combined with sag vertical curves, (3) horizontal tangents combined with multiple vertical curves, (4) horizontal tangent with constant grade |G| < 5%, and (5) horizontal tangent with constant grade |G| >= 5%. For each alignment combination, four different statistical techniques were explored: Poisson, negative binomial, zero-inflated Poisson, and zero-inflated negative binomial. For model validation, three years of data (2002 - 2004) were used for estimation and the last-year data (2005) were used for validation. The results confirmed the validity of the developed models and their ability to accurately predict collision frequency on horizontal tangents.
The results show that vertical curves have relatively little effect on collision occurrence on horizontal tangents. The gradient and grade length increase collision frequency. A larger horizontal curve radius before and after the tangent improves safety at the tangent. The results of this research should be useful for highway engineers and practitioners interested in promoting road safety on two-lane rural highways.