Stochastic Properties of Peak Short-Time Traffic Counts

Traffic flow past a point varies with time owing partly to statistical fluctuations and partly to a variation in the mean. Highway traffic surveys frequently include as a measure of this variation, peak short-time flows (1-, 5-, or 15-min flows) during a longer time (15, 30, or 60 min). Near the peak of the rush hour, the expected flow should be nearly stationary and variation in short-time flows should be due mostly to statistical fluctuations. This paper describes methods for estimation of these peak flows under the assumption that the short-time flows are approximately a stationary Gaussian process with a variance-to-mean ratio of about 1. The theory suggests that one can predict the distribution of peak short-time flows (5-min flows, for example) from the peak long-time flows (15 or 30 min flows). Experimental data seem to confirm this. The implication is that these short-time peaks have no obvious relation to highway capacity beyond that already inferred from the longer-time flows.