How Does Your Model Measure Up

1
How Does Your Model Measure Up

• Presented at
• TRB National Transportation Planning Applications
  Conference by Phil Shapiro
• Frank Spielberg
• VHB

May, 2007
2
Disclaimer

• The presentation that follows is based on a
  survey conducted for the TRB Committee to
  determine the State-of-the-Practice of MPO
  Travel Forecasting Procedures in 2005. The Final
  Report of the TRB Committee has not been
  published. Therefore, minor details of the data
  are subject to change but the overall findings
  are valid.

3
Study Objectives

• gather, organize, describe and interpret
  information on the current state of the
  practice....
• Present cogent picture
• Summarize data in ways that aid TRB panel judge
  correct characterization

4
Survey Purpose

• Obtain travel forecast procedures from broad
  sample of MPOs
• Sufficient detail to permit assessment
  categorization of methods
• Survey Instrument Design
• Respond to 21 committee questions
• Additional questions identified by study team

5
Survey Process

• Web-based survey designed by VHB
• Pre-tested by five MPOs
• Distributed in late June to 381 MPOs
• TRB (Jon Williams), AMPO, NARC and AASHTO
  encouraged responses

6
Surveys Sent and Responses Received(by MPO size)
7
MPOs Providing Responses
8
General Findings
9
Model Characteristics

• Majority of MPOs use 4-step process
• Few MPOs use tour-based methods
• Many MPOs omit mode choice
• Some MPOs do no travel forecasting

10
Trip Generation

• Unit of travel
• Total Person Trips for midsize and large MPOs
• Vehicle trips Total Person Trips evenly
  split for small MPOs
• Trip generation model
• Cross-classification for trip productions
• Regression analysis for trip attractions

11
Trip Distribution

• Gravity Model dominant methodology
• Distributes person trips
• Impedance
• Mostly based on travel time over highway network
• Significant portion of large MPOs use function
  combining highway and transit times or other
  factors
• Less than 1/2 of reporting MPOs apply some type
  of adjustment factors
• K factors
• Time penalties
• Some do not use K factors because they do not
  validate model results

12
Mode Choice

• Home-based work mode choice model
• Used by 94 of large MPOs
• 54 of mid-size MPOs
• 21 of small MPOs
• Functional form overwhelmingly multinomial or
  nested logit

13
Assignment
14
Feedback of highway and transit times
15
Top Ten Best Features as Reported by MPOs

• Multipath Transit Assignment

Actually eleven there was a tie for tenth
place
16
What method is used for transit assignment?

• Assignment of transit trips is typically a single
  path method based on minimum time, weighted or
  unweighted, or impedance

17
Top Ten Best Features as Reported by MPOs

• Multipath Transit Assignment
• Detailed Operations Outputs
• Accurate Land Use Data

18
How are demographic forecasts allocated to
sub-areas or zones?

• Allocation of population, households and
  employment forecasts to TAZs within a region is
  completed by negotiation or based on Master Plan
  Zoning in over half of reporting MPOs.

19
Top Ten Best Features as Reported by MPOs

• Multipath Transit Assignment
• Detailed Operations Outputs
• Accurate Land Use Data
• Intersection Impedances
• GIS Base
• Multiplicity of Transit Purposes

20
What trip types are represented in the
trip generation models?

• Overall 60 of MPOs report using four person trip
  categories. These are HB Work, HB shop/retail,
  NHB and one other purpose.
• Only 30 of large MPOs use only these purposes.
  Large MPOs, on average, use between 5 and 6 trip
  purposes.
• Cross classification models are the dominant type
  of trip production model

21
Top Ten Best Features as Reported by MPOs

• Multipath Transit Assignment
• Detailed Operations Outputs
• Accurate Land Use Data
• Intersection Impedances
• GIS Base
• Multiplicity of Transit Purposes
• Mode/Mode Choice Model

22
What is the form of the mode choice model?
Large MPOs

• 72 of large MPOs use nested logit models.

23
Top Ten Best Features as Reported by MPOs

• Multipath Transit Assignment
• Detailed Operations Outputs
• Accurate Land Use Data
• Intersection Impedances
• GIS Base
• Multiplicity of Transit Purposes
• Mode/Mode Choice Model
• Standardized Model
• Zone/Network Details

24
How many zones are modeled?
25
How completely do networks represent
transportation facilities?

• Most large MPOs model 100
• Freeways
• Major arterial
• Minor arterials
• Most Large MPOs do not model local roads

26
Top Ten Best Features as Reported by MPOs

• Multipath Transit Assignment
• Detailed Operations Outputs
• Accurate Land Use Data
• Intersection Impedances
• GIS Base
• Multiplicity of Transit Purposes
• Mode/Mode Choice Model
• Standardized Model
• Zone/Network Details
• Well calibrated and validated

27
When was the model last validated?

• Large MPOs
• Over 50 validated models since 2004
• Around 80 validated models since 2001

28
From the last model validation, what is the
percent root mean square error comparing traffic
counts with highway link volumes for all links
with counts and for each functional or volume
class?

• RMS error for all link classes tends to range
  from 20 to 40 with a 50th percentile value of
  about 30.

29
From the last model validation, what is the
percent root mean square error comparing traffic
counts with highway link volumes for all links
with counts and for each functional or volume
class?

• For freeways, RMS ranges from 0 to 30 and the
  50th percentile RMS error is 19.

30
From the last model validation, what is the
percent root mean square error comparing traffic
counts with highway link volumes for all links
with counts and for each functional or volume
class?

• As with all link classes, RMS error for major
  arterials tends to range from 20 to 40 with a
  50th percentile value of about 30.

31
Top Ten Best Features as Reported by MPOs

• Multipath Transit Assignment
• Detailed Operations Outputs
• Accurate Land Use Data
• Intersection Impedances
• GIS Base
• Multiplicity of Transit Purposes
• Mode/Mode Choice Model
• Standardized Model
• Zone/Network Details
• Well calibrated and validated
• Ease of use/Flexibility

32
Top Ten Worst Features as Reported by MPOs

• Lack of tour-based or activity model

33
Are you working toward any activity or
tour-based approaches to replace the existing
trip distribution model?

• Large MPOs are more likely to be working toward
  replacing the existing trip distribution model
  with an activity or tour-based method than small
  or medium-sized MPOs.

34
Top Ten Worst Features as Reported by MPOs

• Lack of tour based activity model
• No mode choice model
• Lack of commercial vehicle/truck/freight modeling

35
How is freight movement modeled?

• About 20 of the agencies reporting used a
  synthetic trip table, about 25 use a factoring
  procedure.
• The remaining 55 report use of an Other
  method, generally a gravity model.

36
Top Ten Worst Features as Reported by MPOs

• Lack of tour based activity model
• No mode choice model
• Lack of commercial vehicle/truck/freight modeling
• Lack of toll/HOT lane modeling
• Lack of recent quality calibration
• Lack of recent travel survey data

37
When was the last home interview survey conducted?

• Since 1991, 93 percent of large MPOs, compared
  with 80 of all MPOs, conducted a household
  travel survey.
• 22 of large MPOs conducted a household travel
  survey between 2001 and 2005, compared with 32
  percent of all MPOs.
• 2/3 of large MPOs have conducted a survey since
  1996.

38
Top Ten Worst Features as Reported by MPOs

• Lack of tour based activity model
• No mode choice model
• Lack of commercial vehicle/truck/freight modeling
• Lack of toll/HOT lane modeling
• Lack of recent quality calibration
• Lack of recent travel survey data
• Lack of time of day modeling
• Trip generation
• Lack of detail/quality of mode choice on model

39
What transit modes are modeled? Large MPOs

• Drive Alone, 2 person auto, and local bus are the
  most frequently included modes.
• Rail appears in about a third of the models, and
  where rail is used it is typically in a nested
  structure with walk-to-rail and drive-to-rail
  treated separately.
• Over 1/3 or the large MPOs claim to treat Walk as
  a separate mode and 26 list Bike as a mode used
  in their mode choice model.
• About one-third of the large MPOs report having a
  separate mode-of-access model.

40
Top Ten Worst Features as Reported by MPOs

• Lack of tour based activity model
• No mode choice model
• Lack of commercial vehicle/truck/freight modeling
• Lack of toll/HOT lane modeling
• Lack of recent quality calibration
• Lack of recent travel survey data
• Lack of time of day modeling
• Trip generation
• Lack of detail/quality of mode choice on model
• Land use forecasting

41
How are demographic forecasts allocated to
sub-areas or zones?

• Allocation of population, households and
  employment forecasts to TAZs within a region is
  completed by negotiation or based on Master Plan
  Zoning in over half of reporting MPOs.

42
The End