New Jersey Department of Transportation NATMEC 2014 Improving Traffic Data Collection, Analysis and Use New Jersey Pilot Study
What We ll Cover Today Project Approach Pilot Corridors Data Sources Performance Measure Calculations Annual Hours of Delay Reliability Indexes Variations Thresholds Measurement Unit Aggregation Results Wrap-up Issues & Challenges Other Considerations Possible Next Steps/Recommendations 2
Project Approach Test system-level performance measure methodologies For two NJ corridors (Interstate, arterial) Test AASHTO recommended measures for delay and reliability Propose/test alternate formulations Test alternate thresholds Evaluate aggregation methods Assess results what the values are telling us Note barriers, challenges, assumptions, workarounds Develop a Summary Report Outline step-by-step procedures Present results, lessons learned Provide recommendations 3
Pilot Corridors Chosen based on regional importance, unique features and familiarity 78-A 78-B 78-C 78-D Interstate; rural to urban; portion with local & express lanes; toll & free Entire length 67.8 miles Divided into 4 sub-corridors: - PA border to I-287 (30.8 miles) - I-287 to GSP (22.6 miles) - GSP to NJ Tpk (5.4 miles) - NJ Tpk to Holland Tunnel (9.0 miles) 18-D 18-C NHS freeway & arterial; limited access & traffic signals; urban, commercial, semi-rural 18 -B 18-A Entire length 45.3 miles Divided into 4 sub-corridors: - NJ 138 to GSP (14.3 miles) - GSP to US 9 (16.1 miles) - US 9 to NJTPK (9.5 miles) - NJTPK to Hoes Lane (5.4 miles) 4
Data Sources Travel Time (Speed) Data Vehicle Probe Project (VPP) Suite (Massive Raw Data Downloader) Five-minute average -- reliability Hourly average -- delay Vehicle Probe Project Suite Vehicle/Truck/Bus NJDOT Congestion Management System (NJCMS) NJ TRANSIT General Transit Feed Specifications (GTFS) NJDOT Weigh-in-motion (WIM) Vehicle Occupancy Plan4Safety (Crash database) NJ TRANSIT (Bus load factors) 5
Performance Measure Calculations Annual Hours of Delay (AHD) & Reliability 6
Annual Hours of Delay (AHD) Compare hourly travel times with threshold travel time Consider options for Agency-specified thresholds Delay Time = Travel Time in excess of threshold Delay = Delay Time x Volume (vehicles/persons/trucks/buses) AASHTO Method Hourly travel times for an Average Week Calculate delay for an average week, multiply by 52 TMC delays can be aggregated to sub-corridors and corridors by adding Alternate Method Concern: Average Week data underestimates annual delay If particular Avg TT < Threshold TT, Delay Time = 0 But individual day TT for that time of day, day of week Proposal: Use hourly data for each day of the year 7
Reliability NATMEC 2014 AASHTO Reliability Index (RI 80 ) RI 80 = Maximum TT 80 /Threshold TT In words: TT multiplier to ensure on-time arrival 80% of the time if you happen to be traveling in the most congested 5-minute period, compared to a threshold TT for that segment, regardless of time of day Alternate Method (TTR Max ) Concerns: RI 80 threshold TT does not reflect congested conditions; most unreliable time of day may not be most congested time of day; extreme values for RI 80 (>7, <1) Proposal: TTR Max = max (TT 80 /TT 50 ) In words: TT multiplier to ensure on-time arrival 80% of the time, if you happen to be traveling in the most unreliable 5-minute period of the day, compared to a median travel time for that segment for that five-minute period Denominator doesn t have to be median, but should vary with time (otherwise is the same as RI 80 ) This method give less extreme values 8
TTR = TT80/TT50 Travel Time, minutes NATMEC 2014 Comparative Example RI 80 vs. TTR Max 4.5 4 3.5 3 TT80 TT50 TMC 120P04419 Max TT 80 = 4.19 min (7:50 AM) Free-flow TT = 0.56 min Annual Median TT = 0.59 min RI 80 = 7.1 or 7.5 2.5 2 1.5 Max TT 50 = 1.73 min (7:55 AM) 1 0.5 0 6 5 4 3 2 2:00 AM 4:00 AM 6:00 AM 8:00 AM 10:00 AM 12:00 PM 2:00 PM 4:00 PM 6:00 PM 8:00 PM 10:00 PM 12:00 AM TTR Time of Day TTR Max = 4.8 (8:55 AM) TT 80 = 3.32 min TT 50 = 0.69 min Maximum unreliability is not at the same time as maximum 80 th percentile travel time 1 0 12:00 AM 2:00 AM 4:00 AM 6:00 AM 8:00 AM 10:00 AM 12:00 PM 2:00 PM 4:00 PM 6:00 PM 8:00 PM 10:00 PM 12:00 AM Time of Day 9
Variations Thresholds Measurement Unit Aggregation 10
Threshold Variations Free-flow travel time Use calculated 15 th percentile travel time (using 5-minute data) Same as travel time at 85 th percentile speed Implies that freely flowing roadways are goal Median travel time over all days Single value, but similar to free-flow travel time, unless always congested Median travel time by day of week and hour of day Reflective of expected conditions by time of day & day of week May be more appropriate for reliability threshold than delay threshold Maximum throughput travel time (at 85% of the posted speed) Based on Washington State DOT practice Acceptable travel time Percentage of free-flow travel time that varies based on area type and time of day Context-sensitive All thresholds capped at speed limit Percent of Free-flow Travel Time Area Type Peak Off-peak Urban 167% 133% Suburban 133% 118% Rural 111% 105% 11
Travel time, min NATMEC 2014 Threshold Comparison For Average Week 3 2.8 2.6 2.4 Example Link A (TMC 120+04411) "Acceptable" TT Maximum Throughput TT Median TT, all days Freeflow TT Median TT, day/hr Average TT 2.2 2 1.8 1.6 1.4 Sun 12 AM Mon 12 AM Tue 12 AM Wed 12 AM Thu 12 AM Fri 12 AM Sat 12 AM Sun 12 AM Day of week and hour of day 12
Measurement Unit Variations Vehicles (vehicle-hours, vehicle-miles) Measures only the number of vehicles moved Person (person-hours, person-miles) Measures the total people moved Buses Auto passengers Bus passengers Where PMs impact buses Trucks Where PMs impact trucks Sub corridor 2008-2010 Vehicles 2008-2010 Occupants AVO 78A 4,202 5,958 1.42 78B 4,987 6,542 1.31 78C 3,075 4,248 1.38 78D 2,280 3,198 1.40 18A 781 990 1.27 18B 755 965 1.28 18C 3,424 4,673 1.36 18D 1,546 2,015 1.30 13
Aggregation Variations Delay (Units of Hours) Simple addition Reliability (Indices) TMC Sub-Corridor Corridor Reliability measures are unit-less, cannot be added together Option 1: Weighted Averages Many possibilities for TMC weights: Link length Daily vehicle miles traveled (VMT) Median travel time Daily person miles traveled (PMT) Daily vehicle volume Daily vehicle hours traveled (VHT) Daily person volume Daily person hours traveled (PHT) Option 2: Calculate reliability directly at sub-corridor/corridor level (by direction) Add up travel times for entire sub-corridor/corridor Calculate statistics and index values similar to TMC level 14
Results Corridor/Sub-Corridor for AHD and Reliability 15
AHD Results Example: I78 (TMC-level) Shown in Person-Hours of Delay/mile Hourly Person Volume x Travel Time above threshold Person Volume = Vehicle Volume x AVO Subcorr + Bus Passengers Over all days Varying Threshold Travel Time Creates Different AHD Results Free-flow travel time Maximum thru-put travel time (85% of posted speed) Additional thresholds not shown Yearly median Day/Hour median Acceptable travel time 0 0.75 1.5 3 Miles Free-flow Threshold 78-C Max. Thru-put Threshold 78-C 0 0.75 1.5 3 Miles 78-D Legend APHD/mile (Freeflow Threshold) 47-25,000 25,000-50,000 50,000-100,000 100,000-654,000 78-D N N Legend APHD/mile (Max Thruput) 0-25,000 25,000-50,000 50,000-100,000 100,000-1,430,000 16
Annual Person-Hours of Delay per mile (thousands) NATMEC 2014 AHD Threshold Variation Results Varying threshold travel time results in different outcomes No consistent pattern Each threshold option has unique policy implications for delay 140 120 100 80 60 40 20 0 18 Corridor 18A 18B 18C 18D 78 Corridor 78A 78B 78C 78D Free-flow Yearly Median Max Throughput Acceptable Speed Day/Hr Median 17
RI 80 Threshold Variation Results Varying Threshold Travel Time Creates Different RI 80 Results Free-flow travel time Yearly Median travel time Maximum Throughput travel time (85% of posted speed) Extreme values Values under 1.0 Worst 80 th percentile TT < threshold TT Not a logical multiplier Set to 1.0 Values above 7.0 Worst 80 th percentile TT > seven times threshold TT! Legend 1.00 1.01-1.20 1.21-1.50 1.51-2.00 2.01 < 18
Alternate Formulation (RI 80 vs. TTR Max ) TTR Max gives more moderate (realistic?) values than RI 80 on a TMC basis AASHTO Formula Alternate Formula 19
Alternate Formulation (RI80 vs. TTR Max ) but differences are less pronounced on a corridor or subcorridor level 2.8 2.6 2.4 2.2 2.0 1.8 1.6 1.4 1.2 1.0 Reliability Alternatives Median Threshold - Aggregated with PHT Weights RI80 TTRMax 20
Wrap-up Issues & Challenges Other Considerations Possible Next Steps/Recommendations 21
Issues & Challenges General MS Access has 2 GB file size limitation Use multiple databases with linked tables Alternatively, use an enterprise DBMS (e.g., SQL, Oracle) or statistics package (e.g., SPSS, SAS) Manual conflation of TMCs (for volumes) tedious and timeconsuming Hopefully a one-time process NJCMS missing hourly volumes by day of week, especially weekends Currently assuming average weekday hourly volumes apply to all 7 days Determining how to aggregate up TMCs for telling the story, particularly for reliability 22
Issues & Challenges NJ Specific Lacking traffic volume data in NJCMS for 128 miles of the enhanced NHS (need to find out coverage of INRIX and NPMRDS/here data) Accounting for transit ridership is problematic: Tedious process to use GTFS tables to get number of NJ TRANSIT buses on each TMC during each hour for weekday and weekends Applying typical peak/off-peak loading factors to get number of passengers Missing private and university bus data NJDOT Specific IT/OIT (Hardware/Software/Bandwidth) significant tech issues Staff Knowledge/Skill Sets rapidly evolving tech/data difficult to keep up with; using AOD creatively to tell the story challenging for most Time current workload/staffing makes taking on new, complex work difficult; department-wide attrition also a factor 23
Other Considerations Why PM changes occur (may be difficult to quantify and could affect target-setting) Project effectiveness? Change in background traffic growth? Economy boom or bust? Gas prices? Complex issues re signalized arterials Signal spacing/side friction Multiple speed curves Algorithms are evolving (makes trending tricky) Functional class More critical thinking needed re: agency-determined threshold travel times Across performance measures Across agencies Analytical tools HIGHLY DESIRABLE Automates the process Minimizes errors Speeds production Creates consistent summary output (tables, graphs, visualizations) 24
Possible Next Steps/Recommendations Look at multiple years to see what changes in PMs occur Temporally Spatially Test corridors/years where there have been recently completed projects Telling the story about project effectiveness Have other States and MPOs run through similar tests Varying levels of resources, tools, expertise, etc. Test the PM methodologies using NPMRDS datasets Data completeness, integrity, etc. 25
Thank you! NATMEC 2014 Questions Comments Discussion Keith Miller, Principal Planner Data Analysis & Forecasting kmiller@njtpa.org John C. Allen, Section Chief Bureau of Commuter,/Mobility Strategies john.allen@dot.state.nj.us 26