Effects of Temporal Data Aggregation on Performance Measures and Other Intelligent Transportation Systems Applications

Abstract

Intelligent transportation systems (ITSs) data are a valuable resource for traffic operations, transportation systems management, performance measurement, and transportation research. Historically, these data are time-aggregated for collection, transmission, and storage, with only mean values saved for traffic parameters for each arbitrary time interval. This convention of aggregation discards valuable information that is necessary for some applications. To understand whether systems should continue the practice of aggregation, this paper investigates how temporal aggregation can affect performance measures and other data applications. The investigation uses disaggregate speed data from loop detectors on a London freeway and vehicle trajectories from video imaging on a California freeway. Aggregating measured speed data greatly reduces the spread in reported vehicle speeds, which will distort estimates of emissions, fuel consumption, and travel delay. Using aggregate data for travel time estimates from sampled speeds results in errors attributable to the constant-speed assumption, group-averaged travel times, and using the arithmetic mean speed (as opposed to the harmonic mean speed) to estimate average travel time. Arithmetic mean speeds consistently underestimate aggregate delay, although estimating a harmonic mean speed from the arithmetic mean speed and speed variance can partially mitigate this effect. Temporal aggregation also affects the identification of traffic state transitions times, the estimation of shockwave speed and shockwave travel times, and the construction of fundamental diagrams. The results of this research will help increase understanding of the ability of ITS data to describe transportation systems, and improve forthcoming sustainability performance measures in the Portland Oregon Regional Transportation Archive Listing data archive at Portland State University.