Over the past few years, the transportation planning industry has seen a tremendous growth in the amount of roadway traffic data being collected and analyzed by different agencies. But, because of differences in data collection methodologies and data processing, each of these sources bring unique challenges in differentiating data “noise” from “signal”. Hence, planners need to evaluate and choose the best data source for evaluating the auto speeds in travel demand models or utilizing them during planning studies.
In this presentation, the authors compare arterial speed data collected from three different sources (October 2010 INRIX, September 2012/2013 Bluetooth and October 2013 HERE data) in southeast Florida. INRIX data is continuously collected from both passenger and freight vehicles and processed using proprietary software before it is made available at a granularity level chosen by the user. Florida DOT collected segment level data on different corridors at a 15-minute interval using Bluetooth receivers. HERE is the FHWA’s National Performance Management Research Data Set. Passenger probe data from numerous sources (cell phones, navigation devices etc.) and freight probe data from the American Transportation Research Institute are reported at a 5-minute interval.
The authors analyzed travel speed data on five heavily used arterial roadways in Broward County, Florida. Preliminary analyses show that:
• All three data sources estimate largely similar speed profiles both diurnally and along the roadway segments.
• Bluetooth and HERE data sets estimate remarkably similar “average” time of day travel speeds even at a segment-level.
• There is a greater variation in the night/early morning travel speeds in the three data sets than the day speeds.
• Bluetooth and HERE travel speeds are in general 5 to 10 miles per hour lower than INRIX speeds during the day.
• The travel speeds estimates from the local demand model were generally higher than all three data sources, especially for the mid-day period.
The presentation will outline the data processing efforts, detailed results, lessons learned and authors’ recommendations from this data comparison. The presentation will conclude with the ongoing speed data mining efforts.