Webinar: Modelling mode and route choices on public transport systems

Modelling Mode and Route Choices on

Public Transport Systems

Sebastián Raveau Pontificia Universidad Católica de Chile

BRT Centre of Excellence Webinar

December 5, 2013




with the collaboration of:

Juan Carlos Muñoz Pontificia Universidad Católica de Chile

Juan de Dios Ortúzar Pontificia Universidad Católica de Chile

Louis de Grange Universidad Diego Portales

Zhan Guo New York University

Nigel H.M. Wilson Massachusetts Institute of Technology

Carlo Giacomo Prato Technical University of Denmark

Origin Destination

The trip begins by heading

in the opposite direction…

It’ is better to use the Yellow Line,

but 9 out of 10 use the Red Line!

Attribute Red Line Yellow Line

Transfers 1 1

Time 23:40 23:43

Density 5 pax/m2 3 pax/m2

First leg 90 % 50 %

How do we change these

travelers’ decision?

Understanding travellers is essential in Transportation Planning

and Design.

Identify and quantify the factors that affect the public transport

users’ behaviour.

Explore differences across modes, in multi-modal public

transport networks.

Compare the preferences of public transport users in different

systems and contexts.

Study’s objectives

Contents

Route Choice

Background

Study Case 1

Metro Networks

Study Case 2

Multimodal Network

Results &

Analysis

Extensions &

Applications

Conclusions

Traditional route choice models usually consider just tangible

variables related to the level of service.

travel time

fare

number of transfers

These models are sometimes refined including socio-economic

variables of the travellers.

Route choice modelling

Route Choice

Background

However, this approach ignores other relevant elements that

influence route choice as:

comfort and safety

transfers accessibility

network topology

aesthetics

These variables are subjective and hard to quantify.

Route choice modelling

Route Choice

Background

Pathfinding Criteria

Route Choice

Background

Pathfinding Criteria

Route Choice

Background

Some people follow different criteria when deciding how to get

from one point to another:

the fastest way

the cheapest way

avoid walking

avoid transferring

But most consider many factors at the same time, depending on

their preferences and information!

Route Choice

Background Pathfinding Criteria

Route Choice

Background Pathfinding Criteria

Santiago London

Survey date 2008 1998-2005

Length 78 Km 324 Km

Lines 5 11

Stations 85 255

Transfer stations 7 72

Daily trips 2,300,000 3,400,000

Survey size 28,961 16,300

Study Case 1

Metro Networks Analyzing travellers decisions on Metro Networks

What do people take into account?

In-vehicle time

Waiting time

Walking time (when transferring)

Number of transfers

Transfer stations layout

ascending

at level

descending

travel time

components

Study Case 1

Metro Networks


In-vehicle time

Waiting time


Number of transfers


Transfer stations infrastructure

assisted or

semi-assisted or and

non-assisted

travel time

components

Study Case 1

Metro Networks


In-vehicle time

Waiting time


Number of transfers



Mean occupancy

Possibility of not boarding

initial occupancy ≥ 75% in London

initial occupancy ≥ 85% in Santiago

travel time

components

transfer

experience

Study Case 1

Metro Networks


In-vehicle time

Waiting time


Number of transfers



Mean occupancy


Possibility of getting a seat

initial occupancy ≤ 25% in London

initial occupancy ≤ 15% in Santiago

travel time

components

transfer

experience

Study Case 1

Metro Networks


In-vehicle time

Waiting time


Number of transfers



Mean occupancy



Route distance

Number of stations

Angular cost

comfort and

crowding

travel time

components

transfer

experience

2d sin

Study Case 1

Metro Networks


Origin

Destination 1d

2d

3d

1 21 2

2 2d sin d sin

T1

T2

Angular Cost =

Study Case 1

Metro Networks

2

1


In-vehicle time

Waiting time


Number of transfers



Mean occupancy



Route distance

Number of stations

Angular cost

Reasonable route

comfort and

crowding

travel time

components

Transfer

experience

topological

variables

Easy to obtain!

Easy to obtain!

Easy to obtain!

Study Case 1

Metro Networks

Defined based on the schematic maps

Schematic map’s effect

Study Case 1

Metro Networks

We want to understand the impact of the Metro network

schematic map on the users’ behaviour

Schematic map’s effect

Study Case 1

Metro Networks

Set of alternative routes

A key element when dealing with probabilistic route choice

models is the definition of the alternatives for the OD pairs of

interest

Santiago

generated based on the actual choices

→ 2 to 4 alternative routes

London

generated based on a labeling approach

→ 2 to 6 alternative routes

Study Case 1

Metro Networks

C-Logit Model

for Route Choice

Estimation results

Attribute London Underground Santiago Metro

Travel Time - 0.188 - 16.02 - 0.095 - 19.57

Waiting Time - 0.311 - 7.39 - 0.139 - 5.07

Walking Time - 0.216 - 6.14 - 0.155 - 8.23

Number of Transfers - 1.240 - 4.37 - 0.632 - 4.06

Ascending Transfers - 0.138 - 2.57 - 0.323 - 2.73

Even Transfers 0.513 3.53 n. a. (2) n. a.

Descending Transfers 0.000 (1) n. a. 0.000 (1) n. a.

Assisted Transfers 0.000 (1) n. a. 0.000 (1) n. a.

Semi-Assisted Transfers - 0.328 - 6.83 n. a. (2) n. a.

Non-Assisted Transfers - 0.541 - 6.79 - 0.262 - 6.23

Mean Occupancy - 2.911 - 3.48 - 1.018 - 5.60

Getting a Seat 0.098 2.08 0.092 3.41

Not Boarding - 0.430 - 6.06 - 0.380 - 2.97

Angular Cost - 0.065 - 5.87 - 0.024 - 5.48

Map Distance - 0.358 - 5.76 - 0.274 - 5.69

Number of Stations - 0.316 - 5.52 - 0.147 - 3.10

Turning Back - 0.725 - 8.12 - 0.141 - 9.76

Turning Away - 0.968 - 8.00 - 0.226 - 7.11

Commonality Factor - 0.146 - 3.92 - 0.548 - 3.33

Adjusted r 2 0.566 0.382

Study Case 1

Metro Networks

Attribute London Santiago

1 min waiting 1.65 min in-vehicle 1.46 min in-vehicle

1 min walking 1.15 min in-vehicle 1.62 min in-vehicle

1 (basic) transfer 6.60 min in-vehicle 6.63 min in-vehicle

1 % of occupancy 0.16 min in-vehicle 0.11 min in-vehicle

Seating 0.52 min in-vehicle 0.97 min in-vehicle

Not boarding 2.29 min in-vehicle 3.99 min in-vehicle

1 station 1.68 min in-vehicle 1.54 min in-vehicle

Turning back 3.86 min in-vehicle 1.48 min in-vehicle

Turning away 5.15 min in-vehicle 2.37 min in-vehicle

Marginal rates of substitution

Study Case 1

Metro Networks

Transfer valuations in London

Transfer Type Getting

a seat Intermediate

Not

boarding

Ascending

Assisted 06.81 min 07.33 min 09.62 min

Semi-assisted 08.56 min 09.07 min 11.36 min

Non-assisted 09.69 min 10.21 min 12.49 min

At level 03.35 min 03.87 min 06.15 min

Descending

Assisted 06.08 min 06.60 min 08.88 min

Semi-assisted 07.82 min 08.34 min 10.63 min



Study Case 1

Metro Networks

Transfer valuations in Santiago

range in London 3.35 to 12.49 min

range in Santiago 5.67 to 16.76 min


Transfer Type Getting

a seat Intermediate

Not

boarding

Ascending Assisted 09.05 min 10.02 min 14.01 min


Descending Assisted 05.67 min 06.63 min 10.62 min


Study Case 1

Metro Networks

Transantiago - Santiago, Chile

Study Case 2

Multimodal Network

34 communes

7 million people

700 sq Km

10 million daily trips

55% in public modes


Study Case 2

Multimodal Network

10 zones

feeder bus lines

trunk bus lines

express bus lines

Metro


Study Case 2

Multimodal Network

30,000 daily trips

(7am to 12 pm)

1% of all the city trips

1,892 respondents

access to all modes

The objective is to expand the behavioural models obtained

form Metro, to the entire public transport system.

Some new explanatory variables are:

fare

distinguish travel time by mode

distinguish transfers by modes involved

variability of in-vehicle and waiting times

When travelling in frequency-based networks, the travellers

might follow different route choice strategies.

Analyzing travellers decisions on Transantiago

Study Case 2

Multimodal Network

Choosing a itinerary

Choosing an hyper-path → considering common lines

Route choice strategies

Study Case 2

Multimodal Network

We found that 66.6% of the travellers that could choose their

routes considering common lines, didn’t do so...

One might argue that considering common lines is a personal

characteristic, rather than the behaviour of everyone.

We propose modelling two types of individuals:

Those who consider common lines

Those who don’t consider common lines

Route choice strategies

Study Case 2

Multimodal Network

Logit probability of considering common lines

Attribute Parameter t-Value

Income – More than 1,000€/month - 0.940 3.22

Income – 500€/month to 1,000€/month - 0.327 3.45

Income – Less than 500€/month - 0.000 base

Frequency - Al least once a week - 1.322 4.98

Frequency - Al least once a month - 0.766 3.71

Frequency – Rarely/Never - 0.000 base

Age – Less than 30 years old - 0.399 2.90

Age – More than 30 years old - 0.000 base

Constant - 2.051 - 5.76

Log-Likelihood - 800.66

r 2 0.525

Study Case 2

Multimodal Network

Mode/route choice results

Study Case 2

Multimodal Network

Consider

Common Lines

Do Not Consider

Common Lines

Variable Parameter t-value Parameter t-value

Fare (CLP) - 0.041 - 2.32 - 0.050 - 2.45

In-vehicle time (min) - 0.625 - 2.17 - 0.477 - 2.39

Waiting time (min) - 1.601 - 4.37 - 1.217 - 3.78

Walking time (min) - 1.856 - 2.11 - 1.353 - 2.43

Bus-bus transfer - 2.822 - 2.98 - 2.139 - 2.23

Bus-Metro transfer - 2.201 - 2.32 - 1.849 - 2.63

Metro-Metro transfer - 1.939 - 2.33 - 1.673 - 2.09

Travelling seated 1.886 2.88 1.652 2.33

Not boarding - 1.890 - 1.97 - 1.533 - 2.04

Log-Likelihood - 1,512

r 2 0.487


Study Case 2

Multimodal Network

Variable Consider

Common Lines

Do Not Consider

Common Lines

In-vehicle time (min) € 1.35 per hour € 0.88 per hour

Waiting time (min) € 3.51 per hour € 2.25 per hour

Walking time (min) € 4.06 per hour € 2.50 per hour

Bus-bus transfer € 0.11 per transfer € 0.07 per transfer

Bus-Metro transfer € 0.08 per transfer € 0.06 per transfer

Metro-Metro transfer € 0.07 per transfer € 0.05 per transfer

Travelling seated € 0.07 per leg € 0.05 per transfer

Not boarding € 0.07 per vehicle € 0.05 per transfer

Those who consider common lines are more sensitive to the

different attributes.

Extensions &

Applications Using the model for policy

Change in the Santiago Metro Map

Apply the model to different cities and systems

Extensions &

Applications Some extensions to this work

Extensions &


Map design optimization

Application to journey planner

Extensions &


Public transport users take into account a wide variety of

attributes when choosing routes.

The modelling effort should be on what we can explain, rather

than in what we can’t explain.

Network’s topology, and specially the way it’s presented to users

on a daily basis, is relevant.

Different individuals follow different strategies when choosing

routes.

Conclusions

What did we learn today?

Don’t forget that we are dealing

with individuals, whose behaviour is

hard to understand and model

Conclusions

What did we learn today?




BRT Centre of Excellence Webinar

December 5, 2013