Traditional/Aggregate Travel

Demand Modeling

Prof. K. V. Krishna Rao

IIT Bombay
Organisation of the Presentation
• Categories of Demand Models
• Overview of Urban Transport Modelling System
• Urban Transportation Planning Process
– Pre-analysis Phase
– Technical Analysis Phase
– Post-analysis Phase
• Good Modelling Practices
• Base Year Demand Model
• Forecasting
• Behavioural Travel Model
Categories of Demand Models
 Sequential Versus
Simultaneous Models
• Sequential model
– The travel demand is modeled in sequential steps of trip
generation, trip distribution, modal split and assignment
• Simultaneous model
– If two or more steps in the sequential approach are combined
then it results in a simultaneous model.
– Simultaneous demand models are also called as direct demand
models
– Examples
• Combined modal split and distribution model
• Intercity travel demand model
THE URBAN TRANSPORTATION MODEL SYSTEM
INPUTS
•TRANSPORTATION SYSTEM CHARACTERISTICS
•LAND USE – ACTIVITY SYTEM CHARACTERISTICS

URBAN TRANSPORTATION MODEL SYSTEM (UTMS)

TRIP GENERATION
(How many trips?)

TRIP DISTRIBUTION
(Where do they go?)

MODE CHOICE
(By what mode?)

TRAFFIC ASSIGNMENT
(By what route?)

OUTPUTS
TRAFFIC FLOWS ON NETWORK LINKS
•Quantity (Volume)
•Quality (Speed)
 Cross-section model versus
Temporal model
• Cross-section model
– model that uses data on dependent and independent
variables collected at one point in time for several
spatial units (e.g., TAZs)
– The traditional four stage urban travel demand model
is a cross-sectional model
• Temporal model
– Model that uses panel data (collected at different
points of time) on dependent and independent
variables for a single spatial unit (e.g., airport, city,
etc)
 Trend model versus
Econometric Model
• Trend Model
– Causal variable is only time
– All growth factor models are trend models
– e.g., linear, exponential, logistic trend models
• Econometric Models
– The econometric variables (e.g., GDP, employment,
car ownership, etc.,) that cause the changes in
demand are used as independent variables
– Traditional four stage model is an econometric model
 Aggregate versus
Disaggregate Models
• Aggregate Models
– The demand model that uses summaries of data is an
aggregate model
– The traditional four stage urban travel demand model
is an aggregate travel demand model as it uses zonal
summaries or aggregate data
• Disaggregate Model
– The demand model that uses the data on individual
decision making unit as it is and explains the
behaviour of the decision making unit when
confronted with alternatives is a disaggregate model
 Top-down model versus
Bottom-up Model
• Top-down model
– Top down models are also known as market share
models. A single model is developed for a larger
spatial unit (state) and the demand for the smaller
units (regions, individual airports, etc.) with in the
larger unit is worked out by market share methods
• Bottom-up model
– Separate models are developed for each smaller unit,
and when aggregated across all smaller units the
demand for the bigger unit comes out
Overview of UTMS
 UTMS
• Metropolitan Planning Organisation (MPO) divides the urban
region into 100 - 2000 Traffic Analysis zones (TAZs)
• MPO develops networks that represent all major roads and
public transport lines, generally termed as strategic network
• Performs household travel survey, cordon line and screen line
surveys, speed surveys, etc.
• Collects data relating to demography, land use and activity
system
• Establishes travel pattern in terms of trips made by purpose,
time, mode, route and destination.
• This data is then used to estimate several submodels, viz.,
trip generation, distribution, modal split and assignment
representing the travel pattern.
 UTMS
• The models would include explanatory variables relevant to all
major policy issues, such as transit fares, road tolls, land use
policies, etc.
• The model set is then validated by comparing the base year
volumes obtained by the calibrated model with actual traffic
volume counts in certain parts of the region
• This validation is done also for public transport ridership,
regional modal shares and number of trips
• After the model is validated, it is run for those future years of
interest, based on the amount and location of projected
population and employment
• Generally, models are run on a 20-year horizon for facility
planning and are also run on intermediate years for the
requirement of transit ridership estimation or emission
deadlines, etc.
URBAN TRANSPORTATION PLANNING PROCESS

PREANALYSIS PHASE
• Problem/Issue Identification
• Formulation of Goals and Objectives
• Data Collection
• Generation of Alternatives

TECHNICAL ALYSIS PHASE

• Land Use –Activity System Model
• UTMS
• Impact Prediction Models

POSTALYSIS PHASE
• Evaluation of Alternatives
• Decision Making
• Implementation
• Monitoring
 Pre-analysis Phase
• Define problem broadly
– Examples
• Maximise public transport accessibility
• Achieve reasonable level of service on roads

• Identify broad objectives

– Examples
• Public transport area coverage >95%
• Congestion inside transit vehicles < 5 standees/m2
• LoS of Arterial Road network ≥ D
 Overcrowding in Suburban Trains

10 April 2019 15
 Generation of Alternatives
• Generate alternatives by judgment
• Narrow them down to manageable number by
DELPHI technique
• “No Action Alternative” needs to be considered
for comparative evaluation
• Generate enough number of
– transit investment schemes
– road network investment schemes
– land use options
 PROJECTS IN YEAR: 2011

Passenger Water Transport

JVLR

Metro – Phase I: Andheri – Ghatkopar

- Versova SC Link Road

Bandra – Worli Sea Link Sewri – Worli Fast

Corridor

MTHL
Western Freeway Sea Link Eastern Freeway
PROJECTS IN YEAR: 2021

Passenger Water Transport JVLR

Metro Phase II:Colaba – Mahim - Charkop /

Mankhurd
SC Link Road
Metro – Phase I: Andheri – Ghatkopar
- Versova

Sewri – Worli Fast

Bandra – Worli Sea Link Corridor

MTHL
MTHL-RAIL
Western Freeway Sea Link
Eastern Freeway
 Metro – Phase II: Ghatkopar – Mulund &
Charkop - Dahisar
PROJECTS IN YEAR: 2031

Passenger Water Transport JVLR

Metro – Phase I: Andheri – Ghatkopar

- Versova SC Link Road
Metro Phase II:Colaba – Mahim - Charkop /
Mankhurd

Sewri – Worli Fast

Bandra – Worli Sea Link
Corridor

MTHL/RAIL
Western Freeway Sea Link
Eastern Freeway
 Metro –– Phase
Metro Phase II: Ghatkopar –– Mulund
II: Ghatkopar Mulund &
&
Charkop
Charkop -- Dahisar
Dahisar
PROJECTS IN YEAR: 2041

Metro
Phase III

Passenger Water Transport JVLR

Metro – Phase I: Andheri – Ghatkopar

- Versova SC Link Road
Metro Phase II:Colaba – Mahim - Charkop /
Mankhurd
Sewri – Worli Fast
Bandra – Worli Sea Link Corridor

MTHL/RAIL
Western Freeway Sea Link
Eastern Freeway
 Plan Evaluation Criteria
• Measures of congestion
– LoS (A-F) on each network link
• Travel Delay
– Person hours
• Measures of pollution
– Number of tons of each pollutant produced per day
• Economic welfare (equity measure)
– Net benefit to travelers in terms of reduction in
generalised cost of travel
– Net benefits to travelers by income class
 Data Collection
• Data gathering and data cleaning is part of an ongoing process
• This exercise is fundamental to accurate travel modeling
• Household travel survey need to be performed every decade,
coordinated with the national census
• Usual traffic surveys such as cordon line and screen line counts and
O-D surveys need to be performed more frequently (three times in a
decade)
• Panel surveys on the same household can be used to model
household and firm location behaviour
• Survey of employment is needed to supplement national economic
census
• Demographic data need to be collected from census.
• Land use data need to be collected from development plan sheets
supplemented with satellite imagery
• Survey firms to determine goods movement by commodity type
 Procedure for
Eliminate bias in HH data and
compute expansion factors Generating
Base Year O-D
Expanded partial O-D Matrices
matrices

Outer cordon O-D

O-D matrices with all trips O-D surveys at terminals
Workplace based surveys

Comparison of trips from O-D

matrices with screen line counts
Adjustment of matrices

Load matrices on to the network and Network data

compare the assigned and observed Screen line data
link flows. Validate matrices. Select Cordon line data
appropriate assignment technique Bus route network

Validated O-D matrices for

base year
 Technical Analysis Phase
• Activity Forecasts
– Population and employment forecasts are taken from
the planning body (for example MMRDA projects
population using ratio method)
– Other important variables in travel models like
household income and size need to be forecast
– Allocate the households and employment to the traffic
analysis zones
– A land use model would be more appropriate in
allocating the future activities to zones
 Technical Analysis Phase
• Car ownership model
• UTMS
– Trip generation
– Trip distribution
– Modal split
– Assignment/ route choice
• Impact Prediction Models
– Travel impacts
– Air pollution impacts
– Noise pollution impacts
– Ecological impacts
– Social impacts
 TRIP PRODUCTION TRAVEL DEMAND ESTIMATION
ZONE TRIP TRIP DISTRIBUTION

1 47 TO ZONE

2 66 1 2 3
3 110
1 10 18 19
FROM
ZONE TRIP 2 30 32 4 MODE SPLIT
ZONE

1 45 3 5 40 65 MODE 15
I
2 90 MODE 25
3 88 ROUTE A II
5
ROUTE B 17
TRIP ATTRACTION
ROUTE C 3
TRIP ASSIGNMENT
 Post-analysis Phase
• Plan Evaluation
– Done using economic, equity and
environmental measures
– Reduction in generalized cost categorized by
at least three income groups represents social
equity
– Environmental measures consider pollution
impacts, noise impacts, ecological impacts,
social impacts
 Good Modeling Practice
• Time Representation
– Peak and off-peak periods
• Data Gathering
– Household travel survey every decade with tours
– Vehicle speed surveys
– Data for urban model
• Activity Forecasts
– GIS land use model or economic urban model
 Good Modeling Practice
• Car Ownership
– Discrete choice model, dependent on land use, parking costs,
and accessibility by mode
• Trip Generation
– Walk and bicycle modes
– More trip purposes
– Dependent on car ownership
– Three or more time periods
• Trip Distribution
– Full Model Equilibration
– Composite costs used (all modes, all costs)
– All-day trip tours represented
 Good Modeling Practice
• Mode Choice
– Discrete choice models used
– Land use variables in transit, walk, and bike models
• Goods Movement
– Fixed trip tables
• Assignment
– Capacity-restrained
– Cleaned-up link capacities
– Speeds calibrated
– Three or more time periods
 BASE YEAR TRAVEL DEMAND PROCESS
Validated base year Trip productions
O-D matrices Trip attractions

Planning Varibels
Trip Generation Modelling  Population
for internal trips  Residential workers
 Trip production equations  Vehicle ownership
 Trip attraction equations  Employment

 Assignment of PT trips on to the public

transport network Road network data
PT network data
 Assignment of highway trips on to the
road network taking PT flows as preloads

PT time/cost skims Calibration of a disaggregate

Highway time/cost skims mode choice model

Calibration of Gravity trip distribution

model for internal trips
ASSIGNMENT OF PUBLIC TRANSPORT TRIPS
• Assignment based on Generalized Time

 In-vehicle travel time

 Waiting time
 Transfer time
 Walk time
 Fare
 Discomfort

• GT = IVTT + WTFAC * WT + TRFAC * TR +

WKFAC * WKT + FARE / VOT + DCFAC * DC
• Path building based on GT
• Assignment by Logistic choice function based on GT
Attribute No. of Values Units
Levels
Waiting 3 levels 2, 5, 8 Minutes
time
1 level - Minutes
Travel time Computed
Travel time
Rs. 0.75, Rs.1.0, Rs.1.25
Travel 4 levels and Rs. 1.50 per km Rs. Per
Cost trip

1 and 2 On a scale
Discomfort 2 levels of 1-5
 Existing Mode Metro

Waiting Stated Waiting 3 Levels

Time Time
Travel Time Stated Travel Time Computed

Travel Cost Stated Travel Cost 4 Levels

Discomfort Stated Discomfort 2 Level

Choice Scale

Definitely Probably Can’t Say Probably Definitely

Existing Existing Metro Metro
1 2 3 4 5
 eVMetro
Pr( Metro / EM )  VMetro VEM
e e
VMetro =  WTMetro +  TTMetro +  TCMetro +  DCMetro + CONST
VEM =  WTEM +  TTEM +  TCEM +  DCEM
Pr (Metro/EM) = probability of shifting to Metro
VMetro = deterministic component of utility of Metro mode
VEM = Utility of Existing Mode
WT = waiting time
TT = travel time
TC = travel cost
DC = discomfort
, , ,  =parameters to be estimated using SP data
CONST = constant that explains the unobserved effects
Mode  (WT)  (TT)  (TC)  (DC) CONST 2-statistic

TW -0.0625 -0.0518 -0.085 -1.144 - 0.69

CAR - -0.0183 -0.0139 -0.0151 - 0.32

BUS -0.6255 -0.5569 -1.665 -3.124 - 0.90

II Rail -0.3952 -0.3346 -1.572 -2.365 - 0.96

I Rail -0.3308 -0.2756 -0.8267 -2.291 - 0.543

Mode WT TT DC
(Rs/hr) (Rs/hr) (Rs/hr per
unit shift in DC)
TW 44.00 36.50 27.00

CAR - 79.00 65.00

BUS 22.50 20.00 1.90

II Rail 15.00 13.00 1.50

I Rail 24.00 20.00 2.80

 FORECASTING

Projection of Planning Variables

using Land-use / demographic
models for the future year

Apply trip-end equations and

obtain future year trip-ends of
internal trips

Apply calibrated gravity Previous cost/time

model and obtain O-D matrix skims for initial run
for internal trips

Apply mode choice model and

obtain PT, car and two-wheeler O-D
matrices of passenger internal trips
 Obtain truck matrix and mode-wise
external O-D matrices by Furness
•Matrix of daily PT (bus+rail+ taxi+walk) method using growth factors
passenger trips
•AM peak and PM peak matrices of car, Regional peak hour to daily
two-wheeler and truck trips in PCU flow ratios
Passenger - PCU conversion
Assignment of PT passenger trips on to the factors
public transport network
Assignment of peak-hour PCU trips on Road network data and PT
road network taking peak-hour PT & truck network data for the scenario
PCU flows as preloads under consideration

No Link costs
stable?

Yes
Final Link flows
PT Loadings (Bus, MRTS, Rail, Taxi, walk)
MRTS Boardings and Alightings
Final PT and Highway Cost/Time skims
Behavioural Travel Modelling
 Contents
• Definition of Behavioural Model
• Factors Influencing Choice
• Utility Theory
• Variable Specification
• Maximum Likelihood Estimation
• Goodness-of-fit Statistics & Variable Selection Process
• Demand Elasticities, Subjective Value of Time and
Prediction Success Tables
• Drawbacks of MNL Model
• Nested Logit Model
• Example of a revealed preference mode choice model
from home interview survey data
Definition of a Behavioural Model

A behavioural model is one that explains the

decisions of an individual when confronted with
several alternatives.
Travel as a Choice Process
Factors Influencing the Choice of Mode
• Socioeconomic Characteristics of Trip Maker
- Car Availability and/or ownership
- Possession of driving license
- Household Structure
- Income
- Residential Density
• Characteristics of Journey
- Trip purpose
- Time of day of travel
• Characteristics of Transport System
- Travel time
- Waiting Time
- Travel cost
- Comfort & Convenience
- Reliability & regularity
- Protection & Security
20
METRO 15

20

METRO.
Contents of a Course on Behavioural Travel
Modelling
• Introduction to behavioral travel modeling
• Individual choice theory: binary choice models, multinomial and
multi-dimensional choice models, issues in model specification,
methods and statistics of model estimation with emphasis on
maximum-likelihood estimation, aggregation and forecasting with
discrete choice models, validation and transferability aspects,
ordered multinomial models, nested logit models
• Survey design and analysis: travel surveys and their role in transport
planning, survey methods, precision and accuracy in travel surveys,
sample design, sampling procedures, survey format, pilot surveys,
survey administration, collection of stated and revealed preference
data, survey data processing.
• Advanced concepts: accommodating unobserved population
heterogeneity in choice behavior, mixed logit models, joint stated
preference and revealed preference modeling, and longitudinal
choice analysis
• Discrete choice models for integrated land use and transport
modeling, review of state-of-the-art and future directions.
 Texts/References
• Ortuzar, J. D. and Willumsen, L.G., Modelling Transport, John Wiley
& Sons, New York, 1996.
• Domencich, T.A. and McFadden, D., Urban Travel Demand: A
Behavioral Analysis, North-Holland, 1975.
• Ben-Akiva, M. and Lerman, S, Discrete Choice Analysis: Theory and
Application to Travel Demand, MIT Press, 1985.
• Oppenheim, N., Urban Travel Demand Modeling: From Individual
Choices to General Equilibrium, John Wiley, 1995.
• Borsch Supan Axel , Econometric analysis of discrete choice,
Springer-Verlag, Berlin, 1987.
• Richardson, Ampt, and Meyburg, Survey Methods for Transport
Planning, Eucalyptus Press, 1995.
• Selected papers from journals such as Transportation Research,
Transportation Science, and Transportation Research Record.
• Class Notes