Work Zone Mobility and Safety Program

Using Private Sector Probe Data to Examine Work Zone Performance: The Virginia DOT Experience

slide 1: Using Private Sector Probe Data to Examine Work Zone Performance: The Virginia DOT Experience

Mike Fontaine, Ph.D., P.E.
Virginia Center for Transportation Innovation & Research

Logo for Virginia Center for Transportation Innovation and Research



slide 2: Motivation

  • FHWA Work Zone Self Assessment Questions:
    • Has the agency established measures to track work zone congestion and delay?
    • Has the agency established work zone performance guidance on maximum queue lengths, maximum traveler delay, etc?
  • VDOT has allowable work hours but no established program to track


slide 3: Approach

  • Work zone safety coordinators met to discuss "ideal" measures
  • Desired measures:
    • Queue Length
    • Speed/delay/travel time
    • Reliability
  • Finishing research project to examine measures and define data sources and management tools


slide 4: Constraints


  • Manpower/cost
  • Data abailability
    • Are sensors available?
    • Are they functional?
  • Projuect duration
    • Monitoring short term projects
  • Project phasing impacts
    • Need to reposition sensors


slide 5: Private Sector Sources of Travel Time Data: INRIX


  • INRIX dervies travel times from a variety of sources, including fleet and passenget vehicle probes
  • Sells travel time data, does not install sensors
  • Speed and travel time only, no volume

Inrix travel time maps, Inrix logo, and a bar graph showing how Inrix derives travel times from various sources



slide 6: INRIX Use by VDOT


  • VDOT has statewide real time data from INRIX
  • Feeds real time traveler information (VMSs in DC, SE Virginia, soon Richmond; 511 app)
  • Very well received by public

Image of a roadway with a changable message sign displaying travel timemounted on an overhead gantry.



slide 7: INRIX Data Quality


  • VDOT has conducted internal validation using Bluetooth benchmark on over 340 miles of freeway
    • 95% within 10mph
    • 75% within 5mph

Line graph showing the predicted travel time against bench mark data



slide 8: Work Zone Metrics Being Examined Using INRIX Data


  • Research project to examine viability of using INRIX data for WZ performance measurement:
    • Queue length
    • Use INRIX bottleneck definition
    • 60% of speed typically observed at that time of day
  • speed/delay
  • Reliability measures (95th percentile speed, buffer index, planning time index)


slide 9: Case Study: I-95 SB, MP 158-162


  • Remove cantilever sign structure at 9PM on 2/17/12
  • 2 of 3 SB lanes closed
  • Impacts for about 6 hours over 3 miles

Image of a map displaying the Case Study zone of I-95 SB, MP 158-162.



slide 10: Long Term Work Zone Tracking


Time Historic Speed (mph) Queue Threshold Speed (mph) Observed Speed (mph) Delay (min) Queue (mi)
9-10 PM 52.6 31.6 22.1 9.7 1.90
10-11 PM 61.8 37.1 10.1 21.2 3.03
11- Mid 61.8 37.1 10.5 20.4 2.26
Mid - 1AM 61.8 37.1 9.5 22.5 3.03
1-2 AM 61.8 37.1 8.5 25.1 3.03
2-3 AM 61.8 37.1 22.6 9.5 0.84

Line graph displaying the 5minute vs hourly queue data



slide 11: Long Term Work Zone Tracking


  • 8 Freeway Work Zones:
    • Lengths: 1.97 to 11.9 miles long
    • 55 to 65 mph speed limit
    • Directional AADT between 4,800 and 25,230
    • Interchange density between 0 to 0.5 interchanges/mile
  • 7 Arterial Work Zones:
    • Lengths: 1.04 to 7.2 miles long
    • 35 to 60 mph speed limits
    • Directional AADT between 1,757 and 9,005
    • Between 0 and 4.6 signals/mile


slide 12: Work Zone Impacts on Mean Speed and 95th % Travel Rate


  • Used entire 24 hours
  • Significant degradations in both measures, especially on freeways
Facility Type Mean Speed (mph) 95th% Travel Rate (sec/mi)
Base WZ Δ Base WZ Δ
Freeway 60.08 56.58 -3.50* 79.01 104.72 +25.72*
Arterial 38.31 37.68 -0.63 131.16 133.19 +2.03
Combined 51.12 48.80 -2.32** 100.48 116.45 +15.96**
*Significant at α=0.10
**Significant at α=0.05


slide 13: Work Zone Impacts on Buffer and Planning Time Index


  • Used entire 24hours
  • Significant degradations in both measures, especially on freeways
Facility Type Buffer Index (%) Planning Time Index
Base WZ Δ Base WZ Δ
Freeway 26.00 46.33 +20.33 1.386 1.804 +0.417*
Arterial 25.02 26.01 +0.99 1.527 1.545 +0.019
Combined 25.60 37.96 +12.37** 1.444 1.697 +0.253**
*Significant at α=0.10
**Significant at α=0.05


slide 14: Lessons Learned


  • INRIX real time confidence scores and confidence values
  • Data availability
  • Issues with spatial match to work zones
  • Issues with temporal aggregation
  • Treatment around threhold values
  • Full road closures


slide 15: INRIX Confidence Scores/Values


  • INRIX indicates 3 levels of confidence:
    • "10"– Historic data
    • "20"– Blend of historic and real time data
    • "30"– Purely real time data
  • Scores > 10 also have a confidence value from 0-100 indicating degree of agreement with past trends and recent data
  • Callenge is to weigh responsiveness with accuracy


slide 16:


Scatter plot shows that INRIX data confidence score threshold changed on 8/12 from 85 to 30. Before 8/12, delay was at 61.8 min/day. After 8/12 delay was reduced to 16.0 min/day.

slide 17: Data Availability


  • Real time data availability of 98% or more during daytime periods on freeways
  • Data availability sometimes suffers overnight (Midnight to 4 or 5 AM)
  • Some data issues on arterial system outside of high colume NHS routes
  • Data availability is moving target


slide 18: Spatial Mismatch Issues


  • Data is reported using Traffic Message Channel (TMC) links
  • Typically located between major intersections
Macroscopic model of the major interstates and significant highways in Virginia

slide 19: Matching Private Data to Work Zones


  • TMC boundaries often do not precisely align with work zone boundaries/impacts (or DOT roadway inventory links)
  • Differences in lengths (TMC-Work Zone, 18 work zones)
    • Freeways: Mean of +1.16mi
    • Arterials: Mean of +2.42 mi


slide 20: Temporal Aggregation


  • Long aggregation intervals can "wash out" localized impacts
  • Short time intervals require more resources to analyze
  • Project level vs. programmatic tradeoffs


slide 21: Case Study: 1-81 and US 460/11


Map displaying the detour from Exit 132 to Exit 118 on US 460/US 11 from blasting on I-81SB making it a closed road.

slide 22: Speeds and Queuing Threshold


Line graph showing the speeds and time of the queuing threshold.

slide 23:


Line chart showing the 5minute vs. hourly queue data.

slide 24: Full Road Closures


  • If there is a full road closure, INRIX will report scores of "10" since they have no data unless they have been notified.
  • Need to account for this in performance measure calculations


slide 25: Summary Private Sector Probe Data


  • Advantages:
    • No infrastructure to install/maintain
    • Good data quality on freeways
    • Large coverae area
    • Useful for many other purposes
  • Limitations
    • Spatial granularity
    • Arterial coverage
    • Overnight coverage
    • Does not directly measure queuing


slide 26: Next Steps


  • TMC granularity
    • INRIX indicates improvements are coming to overcome long TMC lengths in rural areas
  • VDOT is reviewing data now to try to determind where to set performance threshold values
  • Parallel effort to develop system that integrates probe data with detector data in user friendly archive


slide 27: Questions?


  • Mike Fontaine, P.E., Ph.D.
  • 434-293-1980
  • Michael Fontaine@VDOT.Virginia.Gov

Return to List of Presentations
Office of Operations