Estimation of Average Annual Daily Traffic (AADT), parameters indicating roads usage, for every segment in a highway network is critical for planning purposes. However, the collection of ground based traffic count data is expansive. Thus a very small proportion of the network can be sampled. Since the daily traffic counts on the network segments are spatially correlated, two sources of information can be exploited to improve these estimates and possibly decrease the ground based resources: (i) Satellite imagery offers the potential to sample many more segments than one can afford by using ground-based traffic crews only, but the corresponding sampling period is much less, so the estimates of daily counts are noisier, and (ii) Incorporating highway network structure in the statistical models to exploit the inherent underlying correlations between link flows, which exist because the statewide network has a large number of Origin-Destination zones, and a relatively smaller number of segments. We use loop detector data and satellite images to approximate errors in AADT estimates from imagery and ground counts obtained in current and previous years. We are also investigating Bayesian and classical approaches to estimate AADT on simulated networks. Simulation results and theoretical insights will be presented, pointing out some network scenarios, under which the traditional estimates can be improved upon.
Meet the speaker in Room 212 Cockins Hall at 4:30 p.m. Refreshments will be served.