Road safety performance measures and AADT uncertainty from short-term counts

Fireseeds North Infrastructure’s Craig Milligan and Ekow Ghanney, alongside the University of Manitoba’s Jeannette Montufar and Jonathan Regehr, have published a paper titled Road Safety Performance Measures and AADT Uncertainty from Short Term Counts in the December 2016 volume of the Accident Analysis and Prevention (AAP) journal.

While there is generally good awareness of the uncertainty in annual average daily traffic (AADT) data, the uncertainty is not well known or accounted for. Given that AADT data is the primary input in safety performance measures and performance models, this presents an opportunity for improved risk analysis of road safety measures with better AADT data.

The paper uses a data-driven approach to provide insight into the uncertainty that surrounds AADT estimates, particularly when the estimates are derived from expanded short term counts using individual permanent counter data. The approach samples data from 69 permanent automatic traffic recorders in Manitoba, Canada, from which almost two million short-term counts were simulated over a five-year period. Short term counts were expanded to AADT estimates by transferring temporal information from nearby permanent control stations. The resulting AADT values were compared to a known reference AADT to compute error. The paper also analyzed five factors likely to impact AADT error, including the length of the short-term count, the number of short-term counts, use of weekday versus weekend counts, the distance from a count to its expansion control station, and the AADT at the count site.

The paper demonstrates that mean absolute transfer error for expanded AADT estimates is 6.7%. This value varied by traffic pattern group from 5% to 10.5%. In terms of the reference percentiles of the error distribution, the paper shows that almost all errors are between -20% and +30%. The errors decrease substantially when using a 48-hour count instead of a 24-hour count, and only slightly when using two counts instead of one. Weekday counts were found to be superior to weekend counts, especially if the count is only 24 hours. It was also demonstrated that mean absolute transfer error increases with the distance to the control station (elasticity = 0.121, p = .001), and increases with AADT (elasticity = 0.857, p < .001).

These results can support evidence-based risk analysis of road safety performance measures that use AADT as inputs. Analytical frameworks for such analysis exist but are infrequently used in road safety because the evidence base on AADT uncertainty is not well developed.

The article can be obtained here: http://www.sciencedirect.com/science/article/pii/S0001457516303438. If you do not have access to the journal but would like a copy of the paper please contact craig.milligan@fireseedsnorth.ca.