This study modeled the location of crashes within work zones as a function of the lengths of the different work-zone segments, traffic volume, weather, and other exogenous factors. A multinomial-logit model was developed that can be used to assess the relative safeness of the different work-zone segments in terms of the crash probabilities per lane-mile. Data from crash reports from a work zone in Florida were augmented with spatial attributes using geographic information systems and used in model estimation. For the work-zone studied, the results indicate that the “Advance Warning” area is particularly unsafe during times of peak traffic flow and bad weather. In contrast, the “Exit” area has a larger probability of crash (per lane-mile) during uncongested times. The probability of a crash per lane-mile in the “Work Area” was generally smaller compared to the other work-zone segments. However, “Work Area” crashes were also found to more severe interms of the injuries sustained by the vehicle occupants. Policy implications of these results are discussed. While the empirical results are basedon data from a specific work zone in Florida, the methodology is generic and can be applied to any region if appropriate spatial data are available.
Abstract