Calibration of Predictive Models for Estimating Safety of Ramp Design Configurations

Abstract

The goal for the calibration process is to use predictive models developed with data collected from other jurisdictions and apply them to the jurisdiction of interest by adapting the models for local conditions and characteristics. Given the large costs associated with data collection, this process is often the only method available to transportation agencies for estimating the safety of different transportation facilities. Thus, recalibrating models produced from other jurisdictions allows agencies to produce their own models at relatively low costs. The objective for the research was to recalibrate a set of crash prediction models for different ramp design configurations. The ramp design configurations addressed included diagonal ramps, non-free-flow loop ramps, free-flow loop ramps, and outer connection ramps. A total of 44 ramps located in and around Austin, Texas, were used in the calibration process. The results of the study showed that more crashes occur on exit ramps than entrance ramps by a ratio of about 6 to 4. The results also showed that the non-free-flow ramp experiences twice as many crashes as other types of ramp. Similarly, more crashes occur on rural than urban ramps.