1. Lessons learned from Metro Vancouver
An exploration of ‘Big Data’ sources to inform best practice travel time studies:
Lessons learned from Metro Vancouver
Fearghal Kinga, Mohamed Mahmouda, Clark Limb
aTransLink; bAcuere Consulting (Vancouver, British Columbia, Canada)

2. Phased Approach Phase 1: Phase 2: Understand and Define Congestion
Measure and Report on Congestion
Literature Review
Literature Review
Definition of Congestion
Definition of Congestion
Identify Preferred Measures
Identify Preferred Measures
Recommend data sources, locations, timing
Recommend data sources, locations, timing
Confirm data availability, quality, validation
Confirm data availability, quality, validation
Develop and conduct congestion analysis plan
Develop and conduct congestion analysis plan
Generate results and maps
Generate results and maps
Produce Regional Congestion Report
Produce Regional Congestion Report

3. Presentation Outline Phase 2 Data Sources Congestion Analysis Plan
Field Validation
Travel Time Metrics
Preliminary Output
Lessons Learned and Next Steps

4. Phase 2 Data Sources

5. Phase 2 Data Source Data Source:
Probe data: smart phone GPS & probe vehicles
Commercial vehicles
Limitations:
Historical data – field validation is not an option
Aggregated annually/monthly data
Aggregated by day of week for a given month
Limited road coverage (road segments are ‘set’ with no flexibility)
Network link lengths
Network links with missing data
Network completeness
In-depth analysis revealed several inconsistencies in speed measures
This means that each 15 minute record contains samples encompassing 4 or 5 days of the week (of the same day of the week).
No flexibility to review or analyze a specific dates.
Fig 1 – long segments in downtown (Many significant roadways are represented by links with lengths that are relatively quite large)
Fig 2 - Aggregation of Samples by Link with No Samples, 7:30-8:30AM Wednesdays (5 hour total duration) January 2014
In a given month (Jan 2014): 45% of the segments had no data. (table 2.15) – to be fair mostly where outside peak periods
A baseline review of the average link speed data resulted in minimum speeds of 3 mph and maximum of 185 mph
Overall, the results were not consistent with our expectations, 2003 TTS results, or google maps data.
We have a report on this analysis.

6. Phase 2.2 Data Source Data Source: GPS data from smart phones Pros:
Data parameters can be determined by the analyst
Flexibility to collect and analyse by date/time/OD pattern/route segments
Data is collected in real-time
Limitations:
No historical data
Heavy data processing, management and cleaning required
Historic data used in the absence of real time
Black box
The main advantage of google maps data (other than it is actually valid) is the great flexibility it provides to develop the work plan
Real-time so it can be valid

7. Phase 2.2 Data Source Inputs:
Origin/Destination (or route segment start/end points)
Route or waypoints (optional)
Outputs:
Geo-coded origin/destination
Distance
Duration (typical)
Duration in traffic (real-time)
Route
Time stamp
The main advantage of google maps data (other than it is actually valid) is the great flexibility it provides to develop the work plan

8. Congestion Analysis Plan

9. Congestion Analysis Plan
Added value:
Ability to compare output with 2003 Travel Time Study
Sub-Centre Analysis
14 sub-centres
Sub-Regional Analysis
14 sub-regions, with 5-6 ODs
Road Network Analysis
Google Road Analysis Network (gRAN)
~1,500 one-way road segments (~1km each)
Added value:
Ability to analyse and report on localised travel times
Added value:
Ability to analyse and report on travel times by road segment

10. 1. Sub-Centre Analysis
(14 x 14 ODs)

11. 2. Sub-Regional Analysis

12. 2. Sub-Regional Analysis
(14 sub-regions; 5-6 ODs)

13. 3. Road Network Analysis
(~1,500 road segments)

14. Field Validation

15. Field Validation ‘Ground-truthing’ exercise (Nov 2016):
3 OD pairs; 6 routes; 6 drivers; 4 time periods; 6 days
Drivers departed start points in 10 min intervals
GPS data collected every second
Google maps data collected every minute
Data was matched spatially and temporally
Total of 4,016 samples (representing road segments)
Routes with predefined segments (not all on the gRAN) – each route is about 30 segments
They were defined based on the most likely shortest path (to be comparable to the API activity centers data)
The three routes cover different types of roads (urban/rural) and different classes (collectors/arterials/highways)
10 mins interval to spread data over the peak period
Coquitlam - UBC
Downtown Vancouver - Surrey
Langley - Richmond

16. Field Validation
Google = * GPS

17. Field Validation

18. Data Verification (Temporal verification)
Network-wide Average Speeds, Nov, 2016

19. Data Verification (Spatial verification)
Average Speed, working day, PM peak, Nov, 2016

20. Travel Time Metrics

21. Travel Time Metrics Standard Deviation of Travel Time
Measures variability in travel time from average travel time (norm)
Useful for understanding reliability of travel time
Travel Time Index
Measures the ratio between reference and congested travel times
Well known congestion index, easily understood
Widely used index (TomTom and INRIX reports)
Standard Deviation of Travel Time
Measures variability in travel time from average travel time (norm)
Useful for understanding reliability of travel time
Travel Time Index
Measures the ratio between reference and congested travel times
Well known congestion index, easily understood
Widely used index (TomTom and INRIX reports)

22. The importance of choosing the right Reference speed
𝑇𝑟𝑎𝑣𝑒𝑙 𝑇𝑖𝑚𝑒 𝐼𝑛𝑑𝑒𝑥= 𝐴𝑐𝑡𝑢𝑎𝑙 𝑡𝑟𝑎𝑣𝑒𝑙 𝑡𝑖𝑚𝑒 𝑅𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑠𝑝𝑒𝑒𝑑
𝟑𝟎 𝒎𝒊𝒏𝒔 𝟐𝟎 𝒎𝒊𝒏𝒔
1.50
𝟑𝟎 𝒎𝒊𝒏𝒔 𝟐𝟑 𝒎𝒊𝒏𝒔
1.𝟑𝟎
𝟑𝟎 𝒎𝒊𝒏𝒔 𝟐𝟖 𝒎𝒊𝒏𝒔
1.𝟎𝟕

23. Preliminary Output

24. TTI: Early Morning (12am – 6am)
Ref. speed: 24hr max
Downtown

25. TTI: AM Peak (6am – 9am)
Ref. speed: 24hr max
Downtown

26. TTI: Mid-day (9am – 3pm)
Ref. speed: 24hr max
Downtown

27. TTI: PM Peak (3pm – 7pm)
Ref. speed: 24hr max
Downtown

28. TTI: Evening Period (7pm – 12am)
Ref. speed: 24hr max
Downtown

29. Lessons Learned and Next Steps

30. Lessons learned and next steps
Passive (Big) data sources can offer a cost effective means to conduct travel time analyses
Data validation is necessary and important
True value lies in flexibility of setting parameters and granularity of data analysis & output (across space and time)
Developing a road network for analysis is complex and time consuming
Traffic volume data is still required to develop congestion metrics

32. EXTRA SLIDES

33. Why is Congestion Measurement important?
Monitor Improvements
Implement New Policies
We are here
Contribute towards fact-based dialogue
Set Goals, Targets, & Objectives
First 3 bubbles:
Performance-based management
Accountability
Goal-setting
Next 2 bubbles:
Stakeholder engagement
Structured decision-making
Policy options
Implementation
Report on Trends
Monitor Baseline Conditions