Special Issue on the Impact of Vehicle Movement on Exploitation Parameters of Roads and Runways TRANSPORT ISSN 1648-4142 / eISSN 1648-3480 2016 Volume 31(2): 221–232 doi:10.3846/16484142.2016.1193046 Corresponding author: Ahmad Shukri Yahaya E-mails: ceshukri@usm.my; shukriyahaya@gmail.com Copyright © 2016 Vilnius Gediminas Technical University (VGTU) Press http://www.tandfonline.com/TRAN EVALUATING THE EFFECTS OF ROAD GEOMETRY, ENVIRONMENT, AND TRAFFIC VOLUME ON ROLLOVER CRASHES Mehdi Hosseinpour 1 , Ahmad Shukri Yahaya 2 , Ahmad Farhan Sadullah 3 , Noriszura Ismail 4 , Seyed Mohammad Reza Ghadiri 5 1, 2, 3 School of Civil Engineering, University of Science, Malaysia 4 School of Mathematical Science, National University of Malaysia, Malaysia 5 Dept of Transportation Engineering, Malaysia University of Science and Technology, Malaysia Submitted 30 April 2014; resubmitted 11 January 2015; accepted 10 July 2015 Abstract. Tere are a number of factors that cause motor vehicles to rollover. However, the impacts of roadway charac- teristics on rollover crashes have rarely been addressed in the literature. Tis study aims to apply a set of crash predic- tion models in order to estimate the number of rollovers as a function of road geometry, the environment, and trafc conditions. To this end, seven count-data models, including Poisson (PM), negative binomial (NB), heterogeneous negative binomial (HTNB), zero-infated Poisson (ZIP), zero-infated negative binomial (ZINB), hurdle Poisson (HP), and hurdle negative binomial (HNB) models, were developed and compared using crash data collected on 448 seg- ments of Malaysian federal roads. Te results showed that the HTNB was the best-ft model among the others to model the frequency of rollovers. Te variables Light-Vehicle Trafc (LVT), horizontal curvature, access points, speed limit, and centreline median were positively associated with the crash frequency, while UnPaved Shoulder Width (UPSW) and Heavy-Vehicle Trafc (HVT) were found to have the opposite efect. Te fndings of this study suggest that rollo- vers could potentially be reduced by developing road safety countermeasures, such as access management of driveways, straightening sharp horizontal curves, widening shoulder width, better design of centreline medians, and posting lower speed limits and warning signs in areas with higher rollover tendency. Keywords: rollover; crash prediction models; over-dispersion; zero-altered models. Introduction Globally, over 1.2 million people are killed in trafc crashes every year, and as many as 50 millions are in- jured. Te global economic losses from road crashes are estimated to be more than US$ 500 billion annually (WHO 2009). In Malaysia, 414421 road crashes were re- ported in 2010, resulting in 6872 deaths and more than 9 billion ringgit of loss to the country’s economy (RMP 2011); of which, rollovers accounted for nearly 1.4% of the total fatal crashes (ITF 2012). Rollovers occur when a vehicle rotates at least one-quarter turn about its lateral or longitudinal axis (Conroy et al. 2006). According to the National Auto- motive Sampling System – Crashworthiness Data System (NASS–CDS), there are eight types of rollover crashes based on the cause and configuration of the collision (Fig.). For more details on each rollover type, the reader is referred to Thomson et al. (2006). As noted in prior studies, despite the relative rarity of rollovers in compar- ison to other collision types, they account for a consid- erable number of serious injuries and fatalities (Khattak et al. 2003; Pape et al. 2008; Keall, Newstead 2009; Funk et al. 2012). For example, Conroy et al. (2006) reported Flip over Trip-over Turn over Climb over Fall-over Bounce-over Collision with another vehicle End-over-end Fig. NASS-CDS classifcation of rollover crashes (Tomson et al. 2006) _________ Tis article has been corrected since frst published. Please see the statement of correct (DOI:10.3846/16484142.2016.1235833 of the erratum).