Search Content

The overall goal of this study was to use commercially-available travel speed data to develop arterial street mobility performance measures in the eight-county Twin Cities Metropolitan Area. The research team licensed 2011 historical traffic speed data from INRIX for 1,604 directional-miles of arterial streets, and conflated this speed data with MnDOT traffic volume data on the same street network. Based on prevailing practice, TTI recommended travel speed-based mobility performance measures that compare peak traffic speeds to speeds during light daytime traffic. However, it was recognized that light daytime traffic speeds are not necessarily the goal or target for the performance measures, but simply a convenient and easily-measured reference point. Instead, performance measure target values should be context-sensitive and based upon surrounding land use. Multiple measures should be used to quantify and monitor mobility on arterial streets, including delay per mile, travel time index, and the planning time index (a measure of reliability). The exact mobility performance measures and target values are likely to evolve and be refined as MnDOT and partner agencies gain experience in performance monitoring on arterial streets. At this time, TTI recommends calculating, tracking, and gaining experience with multiple measures, while also determining where these measures can be used to improve agency decisions.

In this study; the Texas A&M Transportation Institute (TTI) conducted an independent; follow-up evaluation of StreetLight Data's 2019 traffic count estimates using MnDOT sources of traffic count data. At 442 permanent benchmark locations; TTI found that average annual daily traffic ( AADT) estimation accuracy by StreetLight Data has improved significantly since the 2017 evaluation; especially in moderate- to high-volume categories (i.e.; more than 10;000 AADT). The mean absolute error ranged from 8% to 10% for locations greater than 10;000 AADT and gradually increased to 42% for sites with less than 1;000 AADT. TTI also found significant overestimation bias for low-volume roadways (i.e.; less than 2;500 to 5;000 AADT). This result was present in the permanent benchmark sites and more pronounced in the 265 short-duration count sites. Based on these findings; TTI recommends that MnDOT consider a phased approach to using probe-based traffic count estimates: 1) Continue to maintain MnDOT permanent counter sites; 2) start using probe-based counts for about 90% of the moderate- to high-volume roadways (20;000 or more AADT); 3) continue to use traditional short-duration counts at the remaining 10% of the moderate- to high-volume roadways as a spot check to ensure that probe-based AADT estimates remain within acceptable tolerances in the next five to ten years; 4) periodically monitor the error of AADT estimates on low- to moderate-volume roadways (less than 20;000 AADT); and 5) once acceptable error tolerances for these lower-volume categories are reached; repeat Step 2 for these lower-volume categories.

In this project, TTI worked with StreetLight Data to evaluate a beta version of its traffic volume estimates derived from global positioning system (GPS)-based mobile devices. TTI evaluated the accuracy of average annual daily traffic (AADT) volume estimates as well as average annual hourly volume (AAHV) estimates from Streetlight Data using actual volume counts from MnDOT traffic monitoring sites. Traffic volume estimation from mobile devices has potential, but analytic enhancements are needed to improve accuracy and granularity of estimated traffic volumes. Some of the AADT volume estimates from StreetLight Data were within acceptable error ranges (10% to 20% absolute percent error), but other estimates were significantly outside this acceptable error range (greater than 100% absolute percent error). Lower volume roadways had the highest errors, presumably due to lower mobile device sample sizes. The AAHV evaluation results at 12 non-public MnDOT sites reinforce the need for analytic improvements, as these results showed higher error (49% mean absolute percent error) than the results at the 69 public permanent sites (39% mean absolute percent error).