Weekday Peak Hour Mean Flow Estimation Using Two-Day Short-Count Data

Estimates of traffic volumes for particular periods within a day, such as the peak periods, are essential for capacity and level of service assessment, traffic control device selection and implementation, traffic impact analysis, and more refined exposure measures for traffic safety. In this paper, the authors propose a method to estimate the weekday peak hour mean flow using two consecutive weekday counts, which is able to characterize the estimation uncertainty and does not require creating and assigning factor groups. The authors ultimate goal is to use estimates of peak hour flow as an input to model-based estimates of arterial travel time. The authors model assumes that there exist similarities between a short-count site and an automatic traffic recorder (ATR) site in terms of temporal variation patterns and day-to-day flow covariance during peak hours. Given an ATR site and a pair of two-day counts for a short-count site, the posterior distribution of the weekday peak hour mean flow for the short-count site is derived. When a number of ATRs are available, the mean, variance, and 95% confidence interval of the weekday peak hour mean flow for the short-count site are calculated as the data-driven weighted averages of the estimates given individual ATRs. The proposed method is evaluated using actual Twin Cities 2005 ATR data and a leave-one-out cross-validation approach. It shows that more than 70% of short-count sites are able to capture the ground truth at a significance level of 0.05. The authors found that the estimation error is mainly caused by the temporal variation factor difference between a short-count site and the ATR sites providing it with adjustment factor information.