Metropolitan Council Travel Behavior Inventory

Report
Metropolitan Council Travel
Behavior Inventory
Study Overview
presented to
TRB Applications Conference
presented by
Cambridge Systematics, Inc.
Anurag Komanduri
May 8 2013
Transportation leadership you can trust.
Presentation Outline
What I did for the last three summers
Travel Behavior Inventory - Overview
Data Collection
Modeling Framework
Lessons Learned & Future Vision
TRAVEL BEHAVIOR INVENTORY
3
TBI Goals
Snapshot of personal travel in Minneapolis-St. Paul
Collect and provide quality data
» Stand-alone data products
» Regional initiatives + research
» Travel demand modeling
Build a fine-grained policy-sensitive model using data
» State of the practice activity-based model
“Create a lasting legacy for the region”
4
TBI Approach
Perform study in phases
» Phase I – Survey design
» Phase II – Data collection and processing
» Phase III – Model development and calibration
Set goal + allocate resources
» Be flexible – needs change
» Reset and reload
Regular updates
» Doses of (dis)agreement better than ONE shouting match
“Keep it simple – do it well”
» Innovate incrementally
5
TBI Challenges
Balance innovation with pragmatism
Big team
» Manage roles…budgets..schedules..
» Project management role - important
Data management – “where do pieces fit in”
Multi-year schedule
» 2010 – Ongoing
» Stay focused…pay attention
6
TEAM MEMBERS
7
Staff on Project
Metropolitan Council + PMT
» Jonathan Ehrlich, Mark Filipi (Met Council)
» David Levinson (U-Minn), Jim Henricksen (MnDOT)
CS Staff
»
»
»
»
Kimon Proussaloglou (Project Manager)
Anurag Komanduri (Deputy PM)
Thomas Rossi , David Kurth (Senior Advisors)
Brent Selby, Daniel Tempesta, Cemal Ayvalik, Sashank Musti,
Monique Urban, Jason Lemp, Ramesh Thammiraju
Partners
»
»
»
»
Laurie Wargelin, Jason Minser (Abt SRBI)
Evalynn Williams, Parani Palaniappan, Martin Wiggins (Dikita)
Angie Christo, Pat Coleman, Srikanth Neelisetty (AECOM)
Peter Stopher, Kevin Tierney, John Hourdos, NexPro
PHASE I
MODELING FRAMEWORK
9
Modeling Framework - Approach
Evolving process
» Conceived as a hybrid trip + tour model
» Upgraded to an activity-based model
Impact on data analysis
» Tour structures for “all” trips
» Greater emphasis on household activity survey
Budget + schedule
» Seek efficiencies
» Revise scope (always fun!)
Model estimation + validation
» Intricate modeling framework
» “Nuanced” validation
10
Modeling Framework – Key Features
Model design plan – during data collection
» Committee buy-off
Custom activity-based model
» Assess “forecastable” data
» Locally relevant models (toll transponder ownership)
Utilize efficiencies, wherever possible
» PopGen developed by ASU
» Benchmark against HGAC models
Modeling sequence
» Estimation order – application order
11
PHASE II
DATA COLLECTION
12
Data Collection Goals
Collect travel behavior data
»
»
»
»
Household travel surveys – year long effort, seasonality
On-board surveys
Special generators – Mall of America, Airport
External surveys
Update supply-side information
»
»
»
»
»
Highway counts and speed profiles
Transit ridership counts
Park-and-ride utilization
Parking lots – space and costs
Networks – highway, transit, bike-ped
Variety of collection methodologies
» Horses for courses
13
Data Collection Approach
Effort
Survey
Complete
Medium
Innovation
Household Activity
Survey
14,000+ HHs
Web
Mail-back
Telephone
GPS
Effect of incentives on
participation
Transit
On-board Survey
16,000 riders
Hand-outs
Counts
Combine 2005 and 2010
data
Special Generator
Surveys
330 MoA
550 Airport
Personal interview
Tablet-based surveys
External O-D
Survey
5,000 surveys
Counts
LP capture
Mail-back
Response Rate > 20%
Traffic Speeds
Year-long data
TomTom data
purchase
TransCAD routines for
instant analytics
14
Data Collection Challenges
Household survey
» “Hard to reach” population
» Lower participation from “working households”
» GPS assessment
On-board survey
» Limited budget
» Expand data to match “true” ridership patterns
Special generator survey
» Poor response rates
External O-D survey
» Time consuming – license plate capture, mail-back survey
15
PHASE III
DATA ANALYSIS & MODELING
16
Data Analysis – Approach
Data preparation – multiple steps
»
»
»
»
»
Data transfer protocols
Delivery dates… more delivery dates… yet more…
Geocoding
QA/QC routines
Expansion
Assign gate-keepers for “surveys”
» Version control
» Survey database experts
Data utilization approach
» Evolving process – model design plan
Dataset Utilization
Household activity survey
» Estimation dataset
» Primary validation dataset
Transit on-board survey
» No tours - not used in estimation
» CRITICAL validation component
Special Generator survey – validation
» O-D survey – external model
» Airport survey – visitor model
TomTom speeds + Traffic counts
» Free flow speeds
» BPR curve sensitivity testing
18
I-94: from TH 61 to I-35E
AM Shoulder
AM Peak
Mid-day
PM Early
PM Peak
Evening
late
Overnight
PHASE INFINITY
CONTINUOUS LEARNING
20
Things we picked up along the way…
Myth 1 – TRAVEL DATA CAN BE MADE PERFECT
» Travel surveys are complex…respondents “trip up”
» “Cleaning” is great, but impact tails off
Myth 2 – UNOBTRUSIVE DATA ARE PERFECT
» Still dependent on human behavior
» Cracking the GPS paradigm – close, but not 100%
Myth 3 - LOCAL EXPERTISE IS KEY
» Team from 9 states (including MN)
» “Open communication” channels key
Myth 4 – MIDWESTERNERS ARE POLITE
» Not a myth
» Fabulous response rates
» O-D mail-back had response rate of about 20 percent
Things we picked up along the way…
Collecting large data repositories is fabulous
» All data from the same timeframe
» Great for modeling
» Requires strong team working together
Travel behavior is changing
» Fewer overall trips
» Increased bike usage
Travel data are becoming ubiquitous – overwhelming!
» Highway - Speed data, counts
» Transit - Farebox, AVL and APC data
» Personal travel – cell phone data, GPS logs, smartcard
usage, toll transponder transactions
» Freight (not used) – GPS logs
23

similar documents