Prototype Information System for Estimating Average Vehicle Occupancies from Traffic Accident Records

Abstract

Vehicle occupancy data are increasingly being used to assess and monitor congestion management strategies. They are traditionally collected in the field with the roadside windshield method and the carousel method. Although these traditional methods have the benefit of collecting up-to-date data and can be tailored to specific application needs, they are also much more costly to conduct, especially when repeated data collection is needed for monitoring purposes. An alternative to these traditional methods is to estimate average vehicle occupancies (AVOs) from vehicle occupancy information recorded in existing traffic accident records. This method is cost-effective because it involves no new data collection, and it can provide better spatial and temporal coverage. However, this method is also known to be affected by potential biases resulting from under-and overinvolvement of certain population sectors in traffic accidents. Also, for corridor studies, the number of accident records available may not meet the minimum required sample size for the desired accuracy. An effort was made to develop a prototype information system that can make use of vehicle occupancy information in accident data to estimate AVOs. The system was designed to account for driver's age and driver's gender biases. The system also includes a number of variables that can be used for spatial and temporal AVO monitoring, including district, county, hour, week, month, year, vehicle type, facility type, area type, and crash severity. A reasonableness check of the results from the system shows AVO estimates that are highly consistent with expectations.