Modal Split

Vehicular
Mode Choice
Occupancy
Models
Levels

Mode wise
Person Trip Vehicular Trip
person trip
Interchange Interchange
interchange
Matrix matrices
matrices
 CHOICE???
• More than one option / alternatives
• Evaluation of options based on set of criteria / attributes /
characteristics of present options and / or option likely to
be added in future
• Dependant on utility / disutility / generalised cost of
individual option
• Utility / Disutility / Generalised cost is weighted sum of
different attributes
• Higher the utility >>> More is the satisfaction >>> Higher
probability of choice
Travel as a Choice Process
 MODE CHOICE

Mode choice model : Trip-maker’s behaviour regarding the

selection of travel modes.

Aggregate mode choice models - Based on zonal information

Disaggregate mode choice models - Based on
individual/household data.
 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)
- Trip purpose
- Time of day of travel
• Characteristics of Transport System (Travel mode)
- Travel time
- Waiting Time
- Travel cost
- Comfort & Convenience
- Reliability & regularity
- Protection & Security
20
METRO 15

20

METRO.
CLASSICAL MODAL SPLIT MODELS

 Trip-end Modal Split Models : Modal split before the trip-distribution phase.

● Transport patronage is relatively insensitive to the service characteristics of the

transport modes and determined principally by the socio-economic
characteristics of the trip makers.
● Used for medium and smaller sized cities.

 Trip-interchange Modal Split Models : Modal split after the trip distribution
phase.

● Most of the models that have been developed, incorporated measures of the
relative service characteristics of competing modes, as well as measures of the
socio-economic characteristics of the trip makers.
● Used for larger urban area.
 LAND USE & SOCIO-ECONOMIC CHARACTERISTICS

TRIP GENERATION EQUATION

MODAL SPLIT MODEL TRIP DISTRIBUTION MODEL

TRIP ENDS BY MODE ORIGIN – DESTINATION VOLUMES

TRIP DISTRIBUTION MODEL MODAL SPLIT MODEL

O-D VOLUMES BY MODE O-D VOLUMES BY MODE

Trip End Modal Split Models Trip Interchange Modal Split Models

Differences Between Modal Split Model Types

 Types of Mode Choice Models
• Diversion Curves & Diversion Surfaces
Empirical

• Discrete Choice Models

Utility (Disutility / GC) Maximisation (Minimisation) approach
DISCRETE CHOICE MODELS

 Models focus on individual behaviour rather than zonally

aggregated modal choice behaviour.

 Discrete choice models postulate that the probability of

individuals choosing a given option is a function of their
socio-economic characteristics and the relative attractiveness
of the option.

 The decision making of the individuals is to be modeled by

the use of probabilities of choice.
0  Pij  1 for all i & j
ni
 Pij = 1 for all i
j=1
where, Pij is the probability of individual i choosing the alternative j,
ni defines the entire set of available alternatives for individual i
 The basis for estimating the probabilities of choice for each
alternative lies in evaluation of utilities of alternatives (Luce
and Raiffa, 1957).

Uij = ƒ(Xj,Si)
Where, Uij = Utility of alternative j perceive by an individual
i
Xj = Attributes of alternative j
Si = Attributes of individual I

 Types of Utility Functions

 Mode (Choice) specific utility function
 Attribute specific (Choice abstract) utility function
 Discrete Choice Models
 Discriminant Analysis

 Probit Analysis

 Logit Analysis
Binary Logit (BL): Choice between 2 modes
Multi Nomial Logit (MNL): Choice between more than 2
modes
Hierarchal (Nested) Logit (HL/NL): Choice between cluster
of modes
LOGIT ANALYSIS

 Logit analysis assumes that the probability of the occurrence of an event

varies with respect to function F(x) as a sigmoid curve called logistic curve.

● BINARY LOGIT MODEL:

The Binary Logit Model can be written as,

exp Gi(Xjk)
Pik =
1 + exp Gi(Xjk)

Form Of Binary Logit Curve

where Pik is the probability of individual i choosing the alternative k

Xjk are the relative measures of the attributes of alternative k against
alternative j,
Gi(Xjk) is the generalized cost function
● MULTINOMIAL LOGIT (MNL) MODEL

exp Ui(Xk)
Pik =
 expUi(Xk)
j

T
Ui (Xk) = a0i +  aitXjt
t=1

where , T is the number of characteristics in the common utility of alternative j

ait are the coefficient of the characteristics X jt
● NESTED LOGIT MODEL

CHOICE

PUBLIC TRANSPORT PRIVATE TRANSPORT

BUS TRANSIST I. P. T MT NMT

T B
TWO WHEELER CAR BICYCLE WALKING
 exp (Up)
P(P) =
exp (Up) + exp (Upr)

exp (Upr)
P(Pr) =
exp (Up) + exp (Upr)
By moving to the lower public transportation level, the conditional probabilities of
choosing the Bus (b) or the I.P.T (ipt), given the decision to travel by public transport,
become

exp (Ub)
P(b/p) =
exp (Ub) + exp (Uipt)

exp (Uipt)
P(ipt/p) =
exp (Ub) + exp (Uipt)
To calculate unconditional probabilities of choosing Bus or IPT, use following
equations

P(b) = P(b/p) X P(p)

P(ipt) = P(ipt/p) X P(p)

The public transport utility expression is expressed as

Up = ap + …… + an * Xn +  * Logsum

Where, Logsum = ln { exp (Ub) + exp (Uipt) }

Mode Choice Models: Case Study
 NORTH SECTOR
POPULATION-3,26,074
WEST SECTOR AREA-19.03 SQ.KM.
POPULATION-2,50,410 DENSITY-171.35 PPHA
AREA-19.63 SQ.KM.
DENSITY-127.59 PPHA 2
EAST SECTOR
POPULATION-6,08,233
6 AREA-14.75 SQ.KM.
DENSITY-412.50 PPHA

CENTRAL SECTOR 3
POPULATION-4,13,598
AREA-8.89 SQ.KM.
DENSITY-465.24 PPHA
1

SOUTH-WEST SECTOR
POPULATION-2,31,789
AREA-19.10 SQ.KM.
DENSITY-121.36 PPHA SOUTH SECTOR
POPULATION-6,03,689
5 AREA-31.50 SQ.KM.
DENSITY-191.66 PPHA

LEGEND :

SMC BOUNDRY TITLE : SURAT CITY STUDY SECTORS

SECTOR BOUNDRY
SCALE :
RIVER

MAP NO: 4.1

CITY TRAVEL BEHAVIOUR ANALYSIS
Modal Split Analysis

Income group IPT Bus Car 2W Bicycle Walk Total %

LIG 26.36 1.63 0.00 37.01 18.36 16.64 100.00

LMIG 17.40 3.08 0.72 42.13 21.66 15.02 100.00

MMIG 15.80 1.99 4.28 49.74 17.26 10.92 100.00

HMIG 11.10 0.22 11.04 61.58 9.51 6.54 100.00

HIG 17.94 0.30 24.46 49.02 5.02 3.27 100.00

CITY LEVEL 16.92 1.59 6.22 48.76 15.18 11.35 100.00

General Trip Rate Observations

Compulsory Voluntary
Total
(daily) trips (non daily) trips
Income group

tphw tpcd tphw tpcd tphw tpcd

LIG 20.6 1.08 2.9 0.13 23.5 1.21

LMIG 39.5 1.63 4.8 0.17 44.3 1.8

MMIG 41.4 1.60 5.1 0.17 46.5 1.77

HMIG 40.6 1.57 5.8 0.19 46.4 1.76

HIG 36.7 1.30 6.1 0.18 42.8 1.48

CITY LEVEL 35.8 1.44 5.0 0.17 40.8 1.61

tphw: trip per household per week

tpcd: trip per capita per day
Mode Wise % Weekly Work Trips Per Household

Income group IPT Bus Car 2W Bicycle Walk Total %

LIG 9.14 4.89 - 54.81 27.90 3.26 100.0

LMIG 8.07 3.63 1.63 77.83 4.37 4.47 100.0

MMIG 4.75 2.05 7.51 80.62 0.97 4.10 100.0

HMIG 2.68 0.67 13.42 81.88 - 1.34 100.0

HIG 2.74 0.90 31.01 64.45 - 0.91 100.0

CITY LEVEL 5.23 2.27 11.21 72.95 5.59 2.76 100.0

Income Group Wise % Weekly HH Trips By Diff Mode

Income group IPT Bus Car 2W Bicycle Walk

LIG 28.39 35.06 - 12.21 81.09 19.24

LMIG 30.18 31.27 2.84 20.86 15.28 31.69

MMIG 18.97 18.88 13.97 23.06 3.63 31.08

HMIG 12.00 6.91 27.97 26.23 - 11.38

HIG 10.45 7.88 55.22 17.64 - 6.61

Total 100.00 100.00 100.00 100.00 100.00 100.00

 Trip Length Distribution
Average Modal Trip Length Range

Modal trip length range Average Trip Length

Income group
(km) (km)

LIG 2-4 5.81

LMIG 2-4 5.96

MMIG 4-6 6.1

HMIG 4-6 6.28

HIG 4-6 5.64

CITY LEVEL 5.95

Trip Length Distribution

Cumulative Trip Length Distribution

Trip Length Distribution

Cumulative Trip Length Distribution

Trip Length Distribution

Cumulative Trip Length Distribution

INFLUENCING FACTORS

• Travel time

• Travel cost

• Comfort

• Convenience

• Higher ownership of two-wheelers.

• The work purpose trip mode choice by individuals across all the income groups is
found to be highly captive to personalized mode of two-wheelers.
 CALIBRATION OF UTILITY FUNCTIONS
• A utility function measures the degree of satisfaction derived by an individual from
the travel mode choices.

• The magnitude of utility depends on the characteristics of each choice as well as on

the socio-economic characteristics of the individual making choice.

• The mode specific utility functions are calibrated separately for each existing modes
incorporating most influential travel attributes of the respective modes.

• However it becomes impossible to estimate the utility associated with new mode as
the necessary base year data required for the calibration of its utility function would
not be available.
• This problem is resolved by developing the choice-abstract (attribute specific) utility
function.

• This approach is based on the hypothesis that while making choice people perceive
goods and services indirectly in terms of their attributes, each of which is weighed
identically across choices (Lancaster, 1966).
 The Structure Of Utility Functions
 The general form of the utility function calibrated in this study is,

Uki = a0 + a1X1k + a2X2k + a3X3k + a4X4k

Where, Uki = Utility of kth mode perceived by ith income group

a0 = Travel utility constant

a1, a2, a3, a4 = Coefficient of Travel Attributes

X1k = Travel time in min. by kth mode

X2k = Travel cost in Rs. by kth mode

X3k = Travel comfort by kth mode

X4k = Travel convenience by kth mode

 • The utility of a mode is measured in terms of importance of a travel attribute and
individual satisfaction level modified by mode bias factor().

Ukq = Ijq  (Sjq)

j
Where, Ukq = Utility value for kth mode perceived by qth individual,
I jq = Importance for jth attribute of kth mode for qth individual,
S jq = Satisfaction for jth attribute of kth mode for qth individual,

Inputs Of Utility Functions

• Importance & Satisfaction for travel attributes of time, cost, comfort & convenience.

• Travel time and travel cost per trip are found in minutes and Rs.

• The comfort of travel by a mode is measured as driving ease for private transport
and seating comfort for public transport linguistically
 Rating Scale Of Comfort
Rating Driving Ease Seating Comfort
scales (For Personalized Transport Mode) (For Public Transport)
4 Excellent Seated comfortably
3 Good Seated
2 O.K. Standing comfortably
1 Bad Crush load condition

• The convenience is measured in terms of access distance, waiting time, no. of

interchanges in case of public transport(IPT & Bus).
Rating Scale Of Convenience For Public Transport

No. of Access Time + Waiting Time (min.)

interchange
0-5 5-10 10-15 ≥15

0 25 20 15 10
1 20 15 10 5
2 15 10 5 5
 Rating Scale Of Convenience For Personalized Transport

INCOME GROUP
MODE
LIG LMMIG HMIG HIG
Bicycle 10 10 - -
Two-wheeler 100 100 100 100
Car - 60 80 80
 START

DEFINING THE VARIABLES

INTRODUCE MODE BIAS

FCTOR ()

BUILDING MLR MODEL ORIGIN SOFTWARE

CHECKING THE
PARAMETER
ACCEPTABILITY

VARIATION IN
INPUTS

STATISTICAL ANALYSIS

END

Fig. 7.1 Flow Chart For The Calibration Of Utility Function

INCOME BASED MODE CHOICE MODELS (InMoCeM)

• The multinomial logit models are proposed.

• The general formulation of the model is:

exp(Uik)
Pik = n
 exp(Uik)
k=1

Where Pik = Share of kth mode for ith income group

Uik = Utility of kth mode for ith income group

• The convenience of a mode in case of LIG, LMMIG & HMIG is found to be sensitive
to the relative vehicle ownership in addition to traffic condition, parking facilities
particularly for the choice of two-wheeler as a travel mode.
 INCOME GROUP WISE UTILITY FUNCTIONS

For LIG,
U = 0.00515 + 0.00611X1 – 0.0207X2 + 0.00986X3 + 0.00739X4
(R2=0.88)
For LMMIG,
U = -1.28895 + 0.000999X1 – 1.01622X2 + 0.01179X3 + 0.02052X4 (R2=0.87)

For HMIG,
U = -0.25028 + 0.01884X1 – 0.04592X2 - 0.03967X3 + 0.01029X4
(R2=0.83)
For HIG,
U = -2.84443 - 0.000887X1 + 0.00588X2 + 0.05811X3 + 0.03533X4 (R2=0.97)
 Mode Share Computation for LIG

X1 X2 Pk Observed
MODE X3 X4 U* exp(U) Mode
(Min.) (Rs.) (%) share (%)

IPT 20 5 1.5 10 0.11254 1.119117 15 9.45

BUS 20 5 0.5 10 0.10268 1.108137 15 5.05

2W 16 5 4 178 1.35427 3.873932 52 56.65

BICYCLE 35 0 2 10 0.31262 1.367002 18 28.85

Total 7.468188 100 100

*U = 0.00515 + 0.00611X1 – 0.0207X2 + 0.00986X3 + 0.00739X4

Mode Share Computation for LMMIG

X2 Observe
X1 Pk d mode
MODE (Rs. X3 X4 U* exp(U)
share
(Min.) (%)
) (%)
0.34900
IPT 20 5 1 10 -1.05266 3.5 6.7
8
BUS 25 5 1 10 -1.04771 0.35074 3.6 3.0
0.97079
CAR 18 15 2 60 -0.02964 10 4.8
5
BICYCL 0.35440
35 0 1 10 -1.03731 3.6 2.8
E 7
7.82168
2W 16 3.5 4 160 2.0569 79.3 82.7
5
9.84663
Total 100 100
5
*U = -1.28895 + 0.000999X1 – 1.01622X2 + 0.01179X3 + 0.02052X4
 Mode Share Computation for HMIG

X1 X2 Pk Observed
MODE X3 X4 U* exp(U) mode
(Min.) (Rs.) (%) share (%)

IPT 20 5 1 10 -0.03985 0.96093 10 2.7

BUS 25 5 1 10 -0.05435 1.05585 10 0.7

CAR 18 15 4 80 0.06456 1.06669 10 13.6

2W 16 5 3 220 1.96635 7.14455 70 83

Total 10.22803 100 100

*U = -0.25028 + 0.01884X1 – 0.04592X2 - 0.03967X3 + 0.01029X4

InMoCeM - HIG

Mode Share Computation for HIG

X1 X2 Pk Observed mode
MODE X3 X4 U* exp(U)
(Min.) (Rs.) (%) share (%)

IPT 20 12 1 10 -2.3803 0.0925 2 2.8

2W 16 3.5 3 100 0.869 2.3844 63 65.63

CAR 25 13 4 80 0.2682 1.3077 35 31.57

Total 3.7846 100 100

*U = -2.84443 - 0.0008876X1 + 0.00588X2 + 0.05811X3 + 0.03533X4

Comparison Of Predicated And Observed Mode Share
…………….Later……………….