Selection of Representative Crossings Database for the Evaluation of Driver Behavior Over Highway-Rail Grade Crossings

While the literature suggests that driver behavior is the main cause of most of highway-rail grade crossing crashes, it has proven to be a challenging area for research. The SHRP2 Naturalistic Driving Study (NDS) opened a window of opportunity to make a systematic analysis of the phenomenon because it includes an in-vehicle direct observation of the drivers. The first step in the analysis was the selection process of approximately 300 representative crossings for analysis from over 1,000 crossings included in the NDS. In order to allow the analysis of driver behavior in various environments, the selected set was comprised of different types of crossings. Key parameters that were considered are the types of crossings based on the installed traffic control devices, the configuration of nearby intersections, and the number of accidents that took place at the crossing in recent years. From a statistical standpoint, each group must have a size large enough to generalize the observed conclusions across other crossings with similar characteristics. In addition to NDS, resources such as the FRA accident database, the FRA crossings inventory, and Google-Maps were used in order to determine the crossings that fit the selection criteria. In future steps of the project evaluation of driver behavior over selected crossings is expected to help identify patterns that carry high risk for highway-rail crossing accidents.