R. Mussa, D. Chimba

Pages: 43-56

Abstract
The review of crashes occurring on the Florida state highway system revealed that non-limited access roadways with six or more lanes had fatality rates higher than 4-lane roadways. A number of factors including the variation of roadway geometrics and traffic characteristics on six-lane highways compared to four and two-lane highways are suspected to be the source of the difference. The purpose of this study was to build crash prediction models that would reveal significant variables that influence crashes on six-lane roadways and compare the design, maintenance and operational characteristics of these variables to those found on four-lane roadways. A Zero-Inflated Negative Binomial (ZINB) crash prediction model was developed based on crash frequency as a response variable and a number of roadway geometrics and traffic characteristics as independent variables. The model indicated that section AADT, number of access points, percentage of trucks in the traffic stream, and the width of the sidewalk had positive correlation with the crash frequency. In addition, the model showed that median width, shoulder width, surface width, roadway curvature, and posted speed limit were negatively correlated with the crash frequency. The additive change in the expected crash frequency caused by each independent variable was also determined while holding all other variables in the model at their mean values. The results showed that section AADT was the most positively correlated variable with crash frequency while the posted speed limit was the most negatively correlated based on the variable levels that were analyzed.

Keywords: road safety; crash occurrence; six-lane roadways; average annual daily traffic