AN EMPIRICAL-STUDY OF VEHICULAR HEADWAYS IN URBAN AREAS

Conventional statistical distribution models (e.g., Poisson) are known to perform satisfactorily in predicting the number of crossing opport unities in a traffic stream of certain characteristics. However, in ma ny traffic studies, one is interested in estimating the number of cros sing opportunities under traffic and roadway conditions that are not t ypical of any known statistical distribution. A typical example is the number of crossing opportunities across a two-way, multilane traffic stream within a signalized corridor. There does not appear to be a sin gle distribution capable of adequately describing the vehicular headwa y distribution under a variety of traffic and roadway conditions. This paper demonstrates that carefully developed regression models can be used to accurately predict the number of crossing opportunities in a t raffic stream under various roadway and traffic conditions. Compared w ith conventional statistical functions, the regression models are easy to use, transferable, and volume-based, thus allowing the user to pre dict the number of crossing opportunities using readily available traf fic count data. It was found that, depending on the traffic conditions under consideration, conventional statistical distributions, such as the Poisson function, can underestimate the number of crossing opportu nities by approximately one third. Typical applications of these model s are in the area of development of warrants for vehicular and pedestr ian traffic control devices.