Improved Processes for Meeting the Data Requirements for Implementing the Highway Safety Manual (HSM) and SafetyAnalyst in Florida.

摘要

Recent research in highway safety has focused on the more advanced and statistically proven techniques of highway safety analysis. This project focuses on the two most recent safety analysis tools, the Highway Safety Manual (HSM) and SafetyAnalyst. Meeting the data requirements is considered the most challenging task in implementing these tools. In the case of HSM, many of the data variables needed for deriving the HSM calibration factors are currently unavailable in Floridas roadway characteristics inventory (RCI) database. This project attempts to identify and prioritize influential calibration variables for data collection and to determine the minimum sample sizes to estimate reliable calibration factors. For each facility type included in the HSM, this project applied the random forest technique to rank both the required and desired variables based on their importance. The variables were categorized into three groups: variables of primary importance; variables of secondary importance; and variables of lesser importance. The minimum sample sizes to estimate reliable calibration factors for different facility types were also determined. It was found that the minimum sample size of 30-50 sites with at least 100 crashes per year, as recommended by HSM, is insufficient to achieve the desired accuracy for nearly all facility types. Compared to HSM, SafetyAnalyst has lesser but different data requirements. Two major efforts in applying SafetyAnalyst involve the conversion of local data into the strict data format required by SafetyAnalyst and the development of jurisdiction-specific safety performance functions (SPFs). This project developed a software program to convert the crash and roadway data for Floridas state roads to import files used by SafetyAnalyst. In addition, this project also developed the SPFs for unsignalized intersections to supplement those of other facilities which were developed under a separate project. For demonstration, SafetyAnalyst was applied using Floridas data to identify high crash locations. Recommendations for deploying SafetyAnalyst are also provided.