In 2004, the Federal Transit Administration (FTA) completed a research project that examined whether the traffic assignment components of MPO regional travel models appeared to be “robust enough” for use in calculation of the traffic congestion relief benefits of big transit projects. While the research noted assignment convergence is a step in the right direction, it was not possible to examine the actual accuracy of model-predicted auto travel times.
New research sponsored by U.S. DOT (with FTA as the lead) has re-examined whether MPO models can be relied upon to predict reasonably accurate auto travel times and volumes in different environments. The initial work reviewed the current practices used by 30 large MPOs and found widespread deficiencies. The work then focused on a more detailed assessment of assignment and feedback practice in five MPO models (from Atlanta, Dallas-Fort Worth, Phoenix, San Diego, and Seattle). The model-predicted “link” travel times were compared against a comprehensive set of observed “link” travel times obtained from the company known as HERE. The findings show there is “room for improvement” in regards to spatial rectification; zone sizes and centroid connector locations; intersection representation; the free flow link speed and capacity calculations; the volume-delay functions; and the origin-destination vehicle trip tables. The presenter will summarize what appears to be emerging as a “better” traffic assignment practice. While the initial research motivation focused on “congestion relief” predictions, the findings may have significant relevance to the accurate prediction of travel times for performance-based planning and traffic operations analysis.