Steve Polzin

1
Lessons Learned in Transit Stop Level Ridership
Model Deployment
12th TRB National transportation Planning
Applications Conference May 20, 2009

• Steve Polzin Mark Mistretta, Center for Urban
• Transportation ResearchRodney Brunner, Gannett
  Fleming, Inc.

2
Outline

• Description and History of TBEST
• User Needs
• Lessons Learned
• Future Development

3
What is TBEST

• Transit Boardings Estimation and Simulation Tool
  (TBEST) is 3rd generation
• 1st generation Integrated Transit Demand and
  Supply Model (ITSUP)
• 2nd generation Regional Transit Feasibility
  Analysis and Simulation Tool (RTFAST)

4
TBEST has been Around a While
5
Nature of TBEST

• Direct Demand Model
• Not mode choice
• Not interactive with auto travel
• Models at stop level but most useful at route
  segment level
• Designed specifically for transit
• Walk access scale
• Captures transit network accessibility

6
Responsive to Service Changes

• Operational Changes
• Short turning
• Route splitting
• Through routing
• Schedule Changes
• Running time adjustments
• Headway adjustments
• Schedule coordination between routes
• Span of service adjustments

• Alignment Changes
• Extending routes
• Shortening routes
• New routes
• System Changes
• Service redistribution
• Network structure
• Fare Changes
• First-boarding fare
• Transfer fare

7
Modeling and Prediction Framework

• Stop-Level Analysis
• Physical location
• Route
• Direction
• Multiple Analysis Periods
• 4 on weekdays
• Saturday
• Sunday
• Relationships in a Transit Network
• Neighboring stops
• Accessible stops
• Accessibility measures/transfer potential
• Separate Direct and Transfer Boardings

8
Variables Used in Boarding Equations

• Origin Buffer Characteristics (direct boarding)
• Accessibility Measures
• Impedance from origin to accessible stops
• First fare and transfer fare
• In-vehicle time
• First and transfer waiting
• Number of transfers
• Transfer walk time
• Population/employment in accessible stop buffers
• Transfer Potential (transfer boarding)
• Route and Stop Service Characteristics

9
Provided Data Inputs for Florida Agencies

• 2000 Census Data with Pre-Formatted SF1 and SF3
  Variables
• 2000 InfoUSA Employment Data Grouped by
  Commercial, Industrial, and Service
• 2000 GDT Street
• Networks
• 2006 Pre-Coded
• Transit Networks

10
Basic Steps for Applications

• Base Scenario
• Develop Transit System Network
• Enter other input data (interliners, transfer
  centers, )
• Update provided socio-demographic to base year
• Run base scenario
• Base Validation
• Enter observed ridership data (system, route,
  etc.)
• Run validation
• Develop Alternative Scenarios
• Run Alternatives
• Compare
• Ridership, Performance Measures

11
User Needs and Capabilities

• FDOT/planners wanted a software tool for
  short-term transit service planning
• Desired something designed with transit in mind
  i.e. walk access sensitive
• Serve as FDOT provided ridership estimation
  technique for TDPs
• User friendly
• Non-GIS expert compatible
• Brings consistency and embeds knowledge in model
  data and coefficients

12
Lessons Learned Understanding Users

• Capabilities of agency planning staff
• Knowledge of service planning
• Knowledge of GIS
• Knowledge of data bases
• Resources available locally
• Time to learn and use model
• It is not clear that agency
• staff are sufficiently
• experienced to integrate
• model results and expert
• judgement

13
FDOT and Project Team Response

• Data bases embedded in model for Florida
• Training offered
• Web site discussion group
• Technical support built into model
  development/maintenance efforts
• Some ongoing model enhancements programmed
  (updates for ARCGIS versions, fixes to identified
  problems, etc.)

14
Lessons Learned Why Change?

• Why plan future service when we dont have money
  to operate it?
• I heard the model didn't work in city xxxx
  therefore we dont want to use it.
• We dont have dedicated staff, any GIS
  experience, money to travel to training etc.
• I heard you were changing the model. (rumor has
  it Microsoft keeps changing their software)

15
Lessons Learned Staff vs Consultants

• Consultants offer strong methods and breadth of
  experience.
• Consultant approach fails to fully integrate the
  modeling capability within staff which enables
  its use for other applications.
• The expertise and context knowledge to do quality
  service planning is often lacking in small or
  fast growth areas.

16
User Expectations for Accuracy and Precision

• The challenge of stop level data is the
  temptation to value the model based on the
  accuracy at the stop level.
• Transit agencies like stop level accuracy yet
  often dont have stop levels counts to compare
  forecasts to.
• Transit trip Attractions are difficult to
  appropriately value with current
  socio-demographic and employment data bases.

17
Model Ownership and Governance

• Who should own/support the software and
  documentation?
• Florida has been very supportive
• Other entities have helped
• Lack of permanent funding commitment or mechanism
  for shared funding
• Highly dependent on a single consultant/person
• Flexibility, exclusivity
• Transparency intellectual property

18
Amortizing Model Costs over Multiple Uses

• Justifying investment by using tool for more
  purposes
• Accessibility / equity analysis
• Data inventory tool
• Operating cost
• model/tool
• Land use development
• impact assessment tool

19
Future Development Priorities

• Client Driven
• Future pop/employment scenarios tools
• Corridor focus feature boundary treatments, run
  time improvements
• Stronger treatment of special generators
• Calibration for additional modes
• Lrt
• Brt
• Metro
• People mover
• Commuter rail

20
Future Development Priorities

• Researcher Driven
• Variable Special Generator Variable (units of
  trip ends?)
• Input data precision improvements

21
Future Development Priorities

• Researcher Driven
• Park and Ride Model Overlay
• Walk access ? Roadway access
• Enhance trip production/attraction data for
  non-home end
• Data and knowledge of trips at home end well
  understood with modest range of variation
• Knowledge of work or non-home end highly variable
  and with far less data available

22
The model said I took the bus here.
Youve either got dirty data or you are in the
error term.
23
TBEST Version 3.2 - coming soon

• Latest developments
• Online Map Access
• Expanded Socio-economic Growth Capabilities
• Sector/Sub-area scenarios
• Expanded Model Capacity
• Loaded Network Output
• Improved Map Draw Speed
• Many other improvements

24
TBEST Website

• User Manuals/Guides
• Download Software
• Publications/Reports
• User Forum / Ask Questions
• Request Support
• TBEST Website - http//www.tbest.org/

25
(No Transcript)