COMPARISON OF SMALL-AREA OD ESTIMATION TECHNIQUES

Three different techniques for estimating or forecasting the vehicular origin-destination (OD) patterns within small areas such as central business districts are evaluated. The proportional technique estimates the OD table based solely on cordon gate traffic counts and internal zone trip generation estimates within the small study area. The LINKOD model technique estimates the OD table based on cordon counts, trip generation estimates, and traffic counts made within the study area. The regional technique employs a regional travel behavior model to identify regional travel patterns and extract a vehicular OD table for the small study area within the region. The regional technique was found to be the most accurate and the proportional technique the least accurate at estimating OD tables that, when assigned to the small study area street network, reproduced observed traffic volumes within the small area. However, the tradeoff between accuracy and data requirements was not linear. The regional technique required significantly more data and yet was only moderately more accurate than the proportional technique. The three techniques yielded very different estimates of the small-area OD table and yet when the three different OD tables were assigned to the street network, much of the difference was masked out by the assignment process. It was concluded that large errors in the OD estimation process could be tolerated at the expense of a small sacrifice in accuracy in the estimated traffic volumes. Based on this result, the potential of simplifying the OD estimation problem was then investigated. The internal zones were aggregated and all internal-internal cells in the OD table were deleted from the matrix. The result was a minor loss in accuracy when estimating traffic volume. The conclusion of this research is that the small-area vehicular OD table estimate does not have to be very precise to yield useful traffic volume estimates. One can simplify the OD estimation problem through aggregation and the use of simple OD estimation techniques. Although simpler OD estimation techniques are less accurate, the loss in accuracy is small compared with the savings in effort.