Sub-problem 1a: Maxwell Drive PM Peak Hour - Existing Conditions

Sensitivity to Data
Quite often, traffic engineers collect intersecting volumes for a given site on a single day. The reason is simple: primary data collection is expensive. So the question is, how certain can we be that the assessment of facility performance on our “typical” day is really typical of the intersection? We need data for 20-30 days for the time period being studied to be reasonably confident of the interval on the average delay. But that doesn’t detract from the fact that it’s a significant issue.

For this particular intersection, we have traffic data from three different days for the PM peak hour volumes. How different are the delays and levels of service that their volumes predict? Exhibit 2-10 presents the delay estimates based on the three sets of data. You can view the input data for Dataset 1, Dataset 5 and Dataset 6 (the base case and two variations of input data, respectively). 

In this instance, the delays are similar for all three datasets. The largest differences arise on the southbound approach where the average delays for the southbound left range from 15.8 to 17.3 seconds and the delays for the southbound right range from 18.7 to 26.7 seconds. The message isn’t that you should expect to be this lucky every time you do an analysis, but to be sensitive to this issue and prepared to find that others have reached different conclusions for the same site based on different, equally defensible data. Each of the data sets in the table below include heavy vehicles, base signal timing, skipped phases, and has an eastbound arrival type 2.

Discussion:
What has been your experience with the variability in traffic flows? Do you see variations in flows in the field? In delays? Is there merit in looking at a high as well as an average set of volumes for a given condition? How do you communicate to others the differences in the results you obtain?