OVERVIEW OF TRANSPORTATION DEMAND MODELS KSG HUT251/GSD 5302 Transportation Policy and Planning, Gomez-Ibanez

1
OVERVIEW OF TRANSPORTATION DEMAND MODELSKSG
HUT251/GSD 5302 Transportation Policy and
Planning, Gomez-Ibanez

• OUTLINE OF CLASS
• Origins and motivations
• The standard five-step model
• Often called UTPS (Urban Transportation
  Planning System) model
• Passenger Freight
• Urban UTPS
• Intercity
• Subsequent refinements
• Disaggregate models and data
• Simultaneous models
• Stated vs. revealed preference
• Virtual or micro simulation
• Back of the envelope assessment

2
EVOLUTION OF THE MODELS

• Postwar metropolitan growth ? planning for major
  new expressway systems
• Early metropolitan studies
• 1953 Detroit
• 1956 Chicago (CATS)
• 1958 Pittsburgh
• 1962 Federal Highway Aid Act
• 3 Cs Comprehensive, coordinated and continuing
  planning
• 1990 Clean Air Act 1991 Intermodal Surface
  Transportation Efficiency (ISTEA) Act
• Transportation and air quality improvement plans
  must be consistent
• Subsequent refinements
• 1970s Disaggregate models widely adopted
• 1960s and 1980s Simultaneous models limited
  applications
• 1990s Stated preference still controversial
• 1990s-2000s Virtual-micro simulation still
  experimental (TRANSIM program sponsored by DOT,
  EPA, and DOE)

3
COMPLICATIONS OF TRAVEL DEMAND

• P
• Q
• REAL TIME AND SPACE DIMENSION
• Many distinct markets with different Ps and Qs
• SERVICE QUALITY IMPORTANT
• Ps are multidimensional
• SYSTEM INTERDEPENDENCIES
• Cross elasticities are high
• TRANSPORTATION AFFECTS LAND USE
• Long run demand may be significantly different
  from short run demand

4
STEPS IN UTPS MODEL

• LAND USE
• TRIP GENERATION
• TRIP DISTRIBUTION
• MODE SPLIT
• ROUTE ASSIGNMENT

5
TRAFFIC ZONES

6
TRAFFIC ZONES

7
NETWORKS

8
TRIP TABLE (with n zones)

• Oi trips originating in zone i
• Aj trips attracted to zone j
• Tij trips between zones i and j

To 1 To 2 To j To n To all
From 1 T11 T12 T1j T1n O1
From 2 T21 T22 T2j T2n O2

From i Ti1 Ti2 Tij Tin Oi

From n Tn1 Tn2 Tnj Tnn On
From all A1 A2 Aj An
9
TRIP TABLE

• DIFFERENT TRIP TABLES
• BASE AND FORECAST YEARS
• Convention here superscript denotes forecast
  year no superscript denotes base year data
• BY PURPOSE
• Home-based work
• Home-based school
• Home-based shop
• Home-based other
• Non-home based
• BY MODE
• Auto, transit, bike

10
CALIBRATING DATA(BASE YEAR)

• LAND USE INVENTORY BY ZONE
• ORIGIN AND DESTINATION DATA (to build trip table)
• US Census (work trips only often used for up
  date)
• Home interview survey (2 to 5 sample typical)
• Special surveys (taxis, trucks)
• Cordon and screen line counts (cordon around CBD
  screen lines across suburban corridors

11
STEP 1 LAND USE FORECAST

• EARLY AD HOC
• LATER FORMAL MODELS
• Empiric
• Land use in zone f(accessibility of zone,)
• Lowry type
• Distinguish basic (export-oriented) from
  population-serving employment
• Basic employment located exogenously, residences
  of workers and poulation serving employment
  follows
• CURRENT SENARIOS

12
STEP 2 TRIP GENERATION AND ATTRACTION

• (Using land use forecast, forecast Oi and Aj)
• Oi f(residential populationi, auto
  ownershipi, etc.)
• Aj f(square feet of officesj, square feet of
  retail storesj, etc.)

13
STEP 3 TRIP DISTRIBUTION OR ZONAL INTERCHANGE

• (Using Oi and Aj, forecast Tij )
• SIMPLE GROWTH FACTORS
• Tij k Tij
• CORRECTED GROWTH FACTOR
• Tij k (Oi/ Oi) Tij or Tij k (Aj/ Aj)
  Tij
• GRAVITY MODEL
• n
• Tij k Oi (Aj/ Dijb)/ ? (Aj/ Dijb)
• j1
• Where Dij is the impedance between zones i and
  j and k and b are empirically determined from the
  base year data

14
STEP 4 MODAL SPLIT

• (Split Tij into transit, highway, etc.)
• TRIP END MODELS
• Transits share of Tij F(incomei, densityi,
  etc.)
• DIVERSION CURVES
• 100
• Percent
• using
• transit
• 0
• 0.5 1.0 1.5
• Ratio of transit time or cost to auto time or
  cost
• DISAGREGATE MODELS

15
STEP 5 ROUTE ASSIGNMENT

• AD HOC
• MINIMUM PATH
• Linear programming
• CAPACITY CONSTRAINED MINIMUM PATH

16
COMMON CRITICISMS OF UTPS(and responses)

• STRUCTURE OF MODEL UNREALISTIC
• LAND USE AND TRANSPORT USUALLY ASSUMED
  INDEPENDENT (may be true in some cases)
• TRAVEL DECISIONS ARE SIMULTANEOUS NOT SEQUENTIAL
  (simultaneous modeling hard)
• TRANSPORT OMITTED FROM SOME STEPS (only from trip
  generation and attraction)
• TRANSPORT CHOICES DONT FEED BACK ON PERFORMANCE
  OF TRANSPORT SYSTEM (usually iterate model until
  inputs and outputs consistent)
• MODELS ARE EXPENSIVE TO CALIBRATE (for big
  decisions worthwhile for small decisions can
  often use only one or two steps of model)
• NO PEAK HOUR MODEL (time-of-day models in
  infancy)

17
USES OF UTPS-LIKE MODELS TODAY
PASSENGER FREIGHT
URBAN UTPS common for major investments Parts of UTPS used for smaller projects (esp. mode split and route assignment) No models
INTERCITY UTPS-like models used occasionally for major investments Mode split models common Carrier share models common UTPS-like models used only rarely (mainly developing countries) Mode split models common
18
REFINEMENTSDISAGGREGATE DATA AND MODELS

• Idea Calibrate models with data on individual
  travelers rather than on zonal aggregates
• Advantages
• Uses data more efficiently
• (avoids loss in variation that comes from
  aggregating individual data by zones)
• Coefficients less likely to be biased
• Estimated with logit or probit instead of
  ordinary regression (dependent variable is
  discrete)
• 1.0 x x x x x
• Probability
• of picking
• transit
• 0.0 x x x x x x
• xobservation relative convenience of auto
  vs. transit

19
REFINEMENTSDISAGGREGATE DATA AND MODELS

• Typical logit specification
• Pm eUm / ? eUi
• All modes i
• Where Pm probability person will pick
  mode m
• Um measure of utility of mode m
• e base of the natural log
• Example with two modes auto and bus
• Pauto eUauto / (eUauto eUbus )
• Pbus eUbus / (eUauto eUbus )
• Utility of a mode is assumed to be linear
  function of variables measuring
• Performance of the modes (travel time and cost)
• Socio economic characteristics of the travelers,
  and
• Dummy variables for each mode

20
REFINEMENTSDISAGGREGATE DATA AND MODELS

• Example mode to work in SF (Essays, p. 20)
• Four modes drive alone, carpool, walk to bus,
  drive to bus
• U -0.0412 (travel cost in cents / travelers
  wage rate)
• -0.0201 (in vehicle time in minutes)
• -0.0531 (out-of-vehicle time in minutes)
• -0.89 (dummy for drive alone)
• -2.15 (dummy for carpool)
• -0.89 (dummy for walk to bus)
• Derivation of value of travel time (useful as
  check on model reasonableness and for project
  evaluation)
• Value of time (coefficient for
  time)/(coefficient for cost)
• (lost utility/min)/(lost utility/)
  /min.
• SF example above
• In-vehicle time (-0.0201)/(-0.0412/wage) 0.49
  wage rate
• Out-of-vehicle time (-0.0531)/(-0.0412/wage)
  1.29 wage

21
REFINEMENTSSIMULTANEUOS MODELS

• Idea Eliminate sequential structure
• 1960s Direct demand models (with aggregated
  data)
• Tijpm Trips from i to j by purpose p and mode m
• Tijpm f(characteristics of zones i and j,
  service i to j, etc.)
• 1980s Nested logit models (with disaggregated
  data)
• Example vacation destination and mode choice
  model in U.S. (Essays, p. 22)
• DEST 1 DEST 2 DEST 3 DEST 4
• AUTO AIR RAIL BUS
  AUTO AIR RAIL BUS
• Difficulties
• Relatively data intensive
• Many choices and independent variables, so need
  many observations and much information per
  observation
• Results sometimes very sensitive to specification

22
REFINEMENTSSTATED PREFERENCE

• Distinction
• REVEALED PREFERENCE revealed by actual behavior
• STATED PREFERENCE revealed by survey
• Motivation New modes of travel (example
  high-speed rail in the United States)
• Difficulties Do respondents
• Understand choice?
• Take choice seriously?
• Have incentives to misrepresent preferences?
• (Same issues as in debate among environmental
  analysts over contingent valuation)

23
REFINEMENTSVIRTUAL OR MICRO SIMULATION

• Idea Model individual travelers and activities
  to give more spatial and temporal detail and
  (hopefully) more accuracy
• POPULATION LIKE LAND USE FORECAST
• SYNTHESIZER
• ACTIVITY LIKE TRIP GENERATION AND ATTRACTION
  PLUS
• GENERATOR TRIP DISTRIBUTION
• ROUTE INNOVATIVE IN THAT HANDLES TRIP CHAINS
  AND
• PLANNER INTERMODAL BETTER SOLVED BY
  MINIMIZING
• GENERALIZED COST
• TRAFFIC
• SIMULATOR THE STEP THAT WAS THE INSPIRATION
• EMMISSIONS
• ESTIMATOR

24
TIPS FOR BACK OF THE ENVELOPE ASSESSMENTS

• FIND THE RELEVANT TARGET
• Easier to assess whether target is too high or
  too low
• Obvious choices proponents forecast or
  breakeven traffic
• COMPARE WITH CURRENT TRAFFIC AND TREND
• How much more do you have to get?
• CONSIDER ALTERNATIVE SOURCES
• Usual (1) Normal growth, (2) induced traffic
  (stimulate market), (3) other modes, (4) other
  carriers
• SEGMENT MARKET
• Usual by O D, purpose (passenger), commodity
  (freight), season or time of day
• ASSESS QUALITY AS WELL AS PRICE
• Usual travel time, frequency, reliability, etc.
• COMPARE WITH SIMILAR MARKETS