Field Validation of Speed Estimation Techniques for Air Quality Conformity Analysis

About the report:

The air quality conformity analysis process requires the estimation of speeds for a horizon year on a link-by-link basis where only a few future roadway characteristics, such as forecast volume and capacity, are known. Accordingly, the Virginia Department of Transportation (VDOT) has at its disposal a variety of techniques, known as "speed post processors" that estimate average travel speeds on each roadway link based on projected volumes from urban travel demand models. Using field data collected at 15 sites in Richmond and Charlottesville, the accuracy of three post-processing techniques was determined by comparing predicted average travel speed and measured average travel speed. On average, the mean absolute errors for the post processors were relatively similar, ranging between 8 and 12 mph. The post processors overpredicted speeds on some links and underpredicted speeds on others; the average of these positive and negative errors for the post processors was between 2 and 6 mph. Based on MOBILE6 simulation runs with Richmond area data, the differences in speed predictions from the speed post processors would have led to at most a 2.5 percent difference in estimated emissions of volatile organic compounds. All three post processors would have underestimated Richmond area emissions of nitrogen oxides by less than 2 percent. Although differences in national fleet data and Richmond fleet data hamper a direct comparison, additional MOBILE6 simulation results with national data suggest these Richmond results are indicative of the sensitivity of MOBILE6 emissions to changes in estimated vehicle speed. For a class of nine suburban arterial roadways, this study showed that the error associated with any of the post processors could be reduced through judicious altering of the default capacity. This reduction was effected relatively easily by modifying the group capacity rather than computing a capacity for each link. Therefore, although any of the three post processors can be used, this study recommends, in the short term, sampling a few links for each roadway category to determine the appropriate capacity for the category, following an approach similar to that presented in this study. For arterial facilities in particular, this study showed what has been anecdotally known in practice: average travel speeds are affected not just by volume but also by other factors such as signal timing. For this study, this proved to be both a curse and a blessing. On the one hand, the twin facts that the speed post processors are volume dependent and that volume explained only a small amount of the variation in travel speed meant that field results did not show the sensitivity to volume expressed in the literature. On the other hand, because average travel speeds tended to stay within a moderate range, this study showed how better calibration with simple post processors can lead to predictions that are within 5 mph of observed data.

Authors