Impact of data resolution on peak hour factor estimation for transportation decisions

Inductance loop detection systems serve as a primary data source to contemporary traffic information systems. Measures like 20-second or 30-second average velocity, flow, and lane occupancy can be aggregated from individual loop detector actuation sampled at 60 Hz typically. Practically, these measures would sometimes be further aggregated into a much lower, e.g. 15-minute, resolution and then the raw data were lost. Valuable traffic information like flow variation may be distorted when the lower resolution aggregation is practiced. A biased conclusion could be drawn from a data integration system consisted of this kind of distortions. Three approaches estimating a peak hour factor based on traffic volume from loop detection systems are introduced in this paper to explore such a quality issue for data integration systems. Peak hour factor is commonly used in Highway Capacity Manual for determining and evaluating future system needs. By processing the raw data with the introduced approaches, different PHFs can be determined from a same traffic dataset. It is found that 2% to 5% (about one standard deviation from the mean) reduction in PHF may have 5 to 20 seconds increase in control delay estimation. The results suggest that distortion of control delay estimation at a signalized intersection exists due to an improper aggregation. That is, data quality might not be good enough for a right decision if the data were not processed appropriately.