Signalized Intersection Analysis and Design: Implications of Day-to-Day Variability in Peak-Hour Volumes on Delay

Traffic signal timing plans are typically developed on the basis of turning movement traffic and pedestrian volume counts aggregated to 15-min intervals and obtained over a 4 or $8\mathrm{h}$ period on a single day. These data are used to identify the peak hour and to compute the peak-hour turning movement traffic volumes. They may also be used to compute the peak-hour factor. These values are then used as input to the analysis methods defined in the Highway capacity manual or within popular signal timing optimization software to estimate signal performance in terms of expected average vehicle delay. Delay estimation methods explicitly consider several sources of variability (e.g., an assumed distribution of individual vehicle headways in the arrival traffic stream). However, these methods do not consider the day-to-day variability that exists within key delay estimation parameters such as the PHF and peak-hour traffic volume. The lack of consideration of this variability may be because: (1) it is assumed that the impact of the variability is small; and/or (2) methods have not been developed by which the variability can be considered. This paper presents findings of a study that quantifies the impact of day-to-day variability of intersection peak-hour approach volumes on intersection delay and demonstrates that this impact is not insignificant and therefore should not be ignored. Finally, the study explores the number of days for which intersection approach volumes should be counted in order to establish intersection delay within a desired level of confidence. The results indicate that for intersections operating near capacity 3 days of peak-hour volume observations are required to estimate the average intersection delay with an estimation error of 50% of the true mean and 7 days of traffic counts are required to estimate intersection delay with an error of 30% of the true mean.