||A Pilot Study for Developing the Sliding Window Screening Method using New Segmentation on Virginia Roadways
Hyun Woong Cho
The Virginia Department ofTransportation is currently using the simple ranking method as their primary screeningmethod for segments. While straightforward in implementation, the simpleranking method is prone to extreme segment length variation if applieddirectly. To be more specific, it was found that very short segments (< 0.01 miles) used innetwork screening resulted in high randomness in terms of the selection of high-risksites, and the use of very short segments induced the Safety Performance Factor(SPF) model to be more sensitive to errors in locations and may have a dramaticeffect on model performance. The Traffic Engineering Division acknowledged theshort-segment issue and grouped those short segments with neighboring ones tomitigate the issue. The Traffic Engineering Division also initiated a TechnicalAssistance study (Virginia Transportation Research Council TechnicalAssistance: Evaluation of Segment Length in Network Screening Safety Analysis),which determined that the slidingwindow screening method and the use of Average Annual Daily Traffic(AADT)-clustered links showed some promising properties in the development ofSPFs on Interstate networks. Following up the prior technical assistance study,the purpose of this project is to (1) develop the AADT -clustered segmentation method includingmultiple facility types on the Virginia roadway network, (2) develop theHighway Safety Manual’s recommended sliding window screening method with thedeveloped AADT-clustered roadway network and compare with the current simpleranking method’s result, and (3) document the data prerequisites, preprocessingprocedures, and the strengths and limitations of those methods.