Evaluation of the Highway Safety Manual Crash Prediction Model for Rural Two-Lane Highway Segments in Kansas

Abstract

While there have been numerous previous studies performed to develop the rural two-lane segment crash prediction models as part of the Highway Safety Manual (HSM), no previous study has been developed to validate the accuracy of the current model for states other than those the model was developed for. To address this gap the Kansas Department of Transportation (KDOT) commissioned this study to analyze both the accuracy and the practicality of using these crash prediction models on Kansas highways before deciding whether or not to implement the models as part of their normal project development process. To accomplish these goals this dissertation first determined gaps in KDOT data versus data requirements of the HSM. This effort identified an important inconsistency between the Kansas highway system and how the HSM recommends application of the model. Next, the model was calibrated using both the HSM procedure and new procedures that address specific qualities of the Kansas highway system. The calibration procedure derived through this dissertation outperformed the HSM procedure and shows promise as a model for calibration in other jurisdictions. Finally, the accuracy of the crash prediction models for Kansas highways was determined and a calibration procedure was recommended for implementation.