Sampling and Analysis Methods for Estimation of Average Vehicle Occupancies

Abstract

Average vehicle occupancy (AVO) has been used in various applications, including conversion of person trips and vehicle trips in travel demand modeling, evaluation and monitoring of congestion management strategies, and computation of person delays in traffic analysis. AVO data are traditionally collected in the field using the roadside windshield method and the carousel method. The accuracy of AVO estimates is of particular importance because in traffic analysis a relatively small percentage of AVO will convert to the same percentage of vehicles. This paper describes a detailed process for estimating AVOs at the individual location, facility-type, and county levels. A detailed sampling process designed to select data collection locations and dates on three different facility-types (surface streets, freeways, and toll facilities) is presented. To ensure that all areas are covered, the sampling was performed in proportion to roadway vehicle miles traveled (VMT) within individual census tracts over a one-year period. A comprehensive statistical analysis of the collected data was then demonstrated. The results showed no significant difference in the AVO estimates among the three facility-types. Also, no significant difference was found among the AVO estimates derived from facility-weighted composite values and those derived from averaging the simple means. The process presented in this paper provides a rigorous example for the collection and analysis of vehicle occupancy data.