Exploration of a method to validate surrogate safety measures with a focus on vulnerable road users

Abstract

Background. Traditional crash-based analysis of road safety at individual sites has its shortcomings due to low numbers and the random nature of crashes at individual sites and the related statistical issues, as well as the under-reporting of crashes and lack of information on contributing factors and the process preceding crashes. To get around the problem, road safety analysis based on surrogate measures of safety, i.e. not based on crashes, can be used. However, the question whether surrogate measures are valid indicators for safety remains unanswered and only a few attempts have actually been made to carry out proper large-scale validation studies. Aim. This work presents a methodological approach for a large-scale validation study of surrogate safety indicators focusing on vulnerable road users. With only one site analyzed so far, it presents the exploration of the data and of the performance of the technical tools used in the study. Method. Video-filming and consequent video analysis are used to measure the surrogate safety indicators. In the first step, the video is “condensed” using a watchdog software RUBA that selects situations with an encounter of a cyclist or pedestrian and a motor vehicle. At a later stage, the trajectories of the individual road users are produced using a semi-automated tool T-Analyst and several surrogate safety indicators are tested to set a severity score for an encounter. The performance of the surrogate indicators will be compared to the expected number of accidents at each site and availability of the data for developing a safety performance function (SPF) that is country-, manoeuvreand type of VRU-specific are explored. Results & Conclusion. From methodological perspective, limited accident data available seriously complicates building a reliable SPF (“ground truth”) against which the surrogate safety measures could be validated; some other, “indirect” methods of validation might be required. We present also the performance of the software tools and applicability of the various surrogate safety indicators that were tested.