1. Using Cube for Public Transport Planning
An Overview
Andreas Köglmaier
Regional Director
As some of our clients use Cube mainly for the modelling of road traffic and are less familiar with the aspects of public transport modelling I thought it is a good idea to provide an overview of what needs to be considered when using Cube for public transport planning processes. Obviously this presentation has not the intention to provide a comprehensive step by step guide to develop a pt model. It rather should be first step to get involved in PT modelling for new users and provide other aspects to consider for existing PT modellers.

2. Content Reasons for modelling public transport
Public transport data availability
Considerations regarding network representation
Representing passenger behaviour (mode choice and route choice)
Public transport model development in Cube
Specific considerations (fare, park and ride, crowding)
Examples of Cube public transport models
Why should we model PT? What data do we need to have? How can we represent the network (Supply)?
How can we represent the demand?

3. Reasons for building a public transport model
Forecast public transport demand and revenue
Analyse PT projects (economic analysis)
Optimise the routing of PT lines
Testing different ticket systems
Revenue allocation for integrated ticket systems with multi operators
Estimate future run times
Before we build a model we should always consider what is the purpose of the model

4. Public transport data availability
System data (Supply)
Network data (line routing, stopping pattern, timetable)
Observed vehicle speeds (delays)
Reliability (delays, cancelations, accessibility of vehicle)
Quality of vehicles and station infrastructure (comfort)
Ticket and tariff System

5. Public transport data availability
Passenger data (Demand)
Boarder and alighter information
Vehicle loadings (crowding information)
Passenger origin and destination information
Demographic Information (age group, car ownership, …)
Usage of different ticket types
Mode choice and route choice behaviour (stated and revealed preference surveys)

6. Representing the public transport system
Physical infrastructure (roads, tracks, stations)
Stick network or shape network
Station single node or detailed modelling of interchanges
Speeds taken from network or from service definition
WalkTime = ~1 min
Streets
Bus stop node
Rail Platform
Rail

7. Representing the public transport system
Service definition and routing
Decision on which services to include (special school runs)
Consider simplification of network
Coding of services as one or two way lines (one way streets)
Organisation of the system (operators, modes, fares)
Understand which parameters influence route and mode choice
Level of representation depends on set up of demand model

8. Cube represents complexity of PT Systems
The route, where it stops (boarding and alighting: both, only boarding, only alighting), and its variations (sub-routes) by period of the day. Details such as boarding delays at selected stops..
The type of vehicle providing the service (bus, light rail, heavy rail, ferry…etc.) and its capacity (seated and crush). (capacity restraint is provided in PT)
Information about multiple PT operators and their operating/fare policies
Full representation of circular routes
Uses function of roadway speed, set travel times via time points, used fixed time.
Highway network
Transit line
Cube provides the possibility to have this representation:
It can do a everything therefore it provides the following files:
Highway network
Transit line
System data

9. Representing the passenger behaviour
Understanding route and mode choice behaviour
Choosing between bus and rail (mode or route choice?)
Considerations for mode choice model
Group passengers in homogenous groups with similar route choice behaviour
Consider ticket choice model
Considerations for route choice
Value of different trip components
Each user-class may have unique factors which determine:
Which routes are reasonable…
How much walking?
How many transfers?
What are the access characteristics?
What is the maximum cost?
What is the maximum travel time?
Which routes are desirable…
What is the perceived time for walking, transferring, waiting, riding (by mode)?
Are there rebates or special fare systems?
What is the sensitivity to crowding/capacity restraint?
What is the value of time?
What is the willingness to pay?
AUTO
DRIVE ALONE
SHARED RIDE
SHARED RIDE 2
SHARED RIDE 2+
TRANSIT
WALK
BUS
LIGHT RAIL
COMMUTER RAIL
CIRCULATOR
DRIVE
CIRCULATOR`

10. Representing the passenger behaviour
Deviding the trip in its components

11. Representing the passenger behaviour
Home
Rail Platform
Bus Stop
Transfer
Transit path
Mode Choice, Route Choice and Wait curve
Auto path
Toll
Work
Parking

12. PT Methodology
Compiles Data and Simplifies Network (Produces NET file)
Enumerates “Acceptable” Discrete Routes for every O-D (RTE file)
Finds least cost route
Enumerates all other routes within defined limits
Evaluates Routes and Performs Analysis:
Decisions on access and assignment to individual lines through a series of logit choice models.
Route evaluation through a hierarchical logit choice model.

13. Route Enumeration
Enumeration finds a traveler's reasonable public transport routes from origin to destination
Identifies full discrete routes
The route should move progressively from the origin to the destination
Travelers tend to select journeys that are simpler – that are direct or involve few interchanges
Travelers are unwilling to walk very long distances
These principals are used to constrain the potentially huge computational task of identifying all reasonable routes
The process can be considered analogous to a traveler using a map to reject routes which are very long relative to more direct alternatives
Creates a dataset of the ‘reasonable’ routes between each origin and destination by user class

14. Route Evaluation
Evaluation ‘qualitatively judges’ the routes calculated in the route enumeration stage of PT
The elements that can be used in this process
Limit on the number of transfers
The difference (actual & percentage) difference between the minimum cost route and the evaluated route
Limit on non-transit cost (walk/drive access)
Limit on waiting and transfer times
Limit on In-vehicle costs
Specified by user class – or – market segment
Provides attractive and reasonable routes along with their probability of being used and the costs of each of the routes.
Calculates the % probability of using each path using choice models at each decision point

15. Report and Visualize PT Results & Inputs
Reports
Graphics
Record Processing
Select Link Analysis

16. Fare System Representation
All sorts of complex fare systems can be modeled by user class
FREE – No cost incurred
FLAT – One fixed cost per use
DISTANCE – Possible boarding cost + unit cost per distance or cost lookup table
FROMTO –Fare zone matrix based on the start/end zones
COUNT –Counts number of fare zones crossed, sum number of zones crossed
ACCUMULATE – Each fare zone has a fare and when crossed adds to cost

17. Representing Park and Ride
Input Car & PT cost matrices
Define Catchment Area, City zone & Station zone
Calculate Car cost: Catchment to Station & opposite
Calculate PT cost: Station to City & opposite
Combine to calculate Catchment to City via station cost & opposite
Consider cost penalty for parking time.
Where there is more than one station in a catchment, the process is repeated for each station to determine the station with the minimum cost for each zone in the catchment.
Output a Park and Ride cost matrix

18. Representing Park and Ride
Rail Station
Rail
Bus Stop
PNR Lot
Rail Platform
Escalator link
PNR lot–stop access connector
Buses
Streets
Driveway link 18

19. Representing crowding
Vehicle capacity required as input
Understand changes in comfort
Seating capacity and crush capacity
Define crowding curve
Penalties
In-vehicle penalty
Boarding penalty
Iterative process: Loaded demand from an iteration is used to update the following for the next iteration
Link travel times
The probability of boarding a line at a particular stop
Link travel time adjustment
The link travel time adjustment models passenger perceptions that travel time is more onerous when they have to stand (rather than sit), or when the vehicle is crowded. The adjustment is represented by a “crowding factor,” which is multiplied by in-vehicle time to give the perceived value of “crowded in-vehicle time.”
Wait time adjustment
In the context of a transit leg from a boarding point to an alighting point with several lines operating, demand (without crowding) is allocated using the service frequency model (SFM), or service frequency & cost model (SFCM). The wait time adjustment redistributes the initial SFM or SFCM loadings whenever any line does not have the available capacity to take its allocated demand. This excess demand is reallocated to other lines which have spare unused capacity, and incurs additional wait time.
The additional wait time (due to not being able to take the first service) may make this route less attractive, and so cause diversion of demand to other enumerated routes for the origin-destination pair.

20. Multimodal model Spain
4. Exemplos internacionais de utilização do CUBE – Europa
Multimodal model Spain
Passengers: Cars, Railways, Airplanes
Freights: Highways and Railways
MAIN FEATURES AND PECULIARITIES
A tool enabling to test nationwide and/or regional transport policies, as well as specific network/services scenarios.
Multimodal model:
Passengers: road (car-bus)-rail (conventional-high speed) and air.
Freight: road-rail (loading only).
2 ½ step model: modal split + assignment + pre-load-calculation
Highly aggregated data: 390 zones. Short-mid distance trips matrices (<50 km) are not included in the modelling.

21. De Lijn Public Transport Model (North Belgium)
Autonomous Public Transport Company in Flanders ( habitants)
508 million travelers in 2008
More than busses – 360 trams
50% service by external operators
De Lijn = decentralized organization
Headquarters: support, coordination & strategy
5 operational regional entities and over 60 workplaces

22. Master Plan Budapest, Hungary
4. Exemplos internacionais de utilização do CUBE – Europa
Master Plan Budapest, Hungary
Support establishment of project selection and project prioratization

23. Multimodal Model Lisboa, Portugal
demand and revenue forecast multimodal model, for a 30 year period, with the following characteristics:
Traffic counts and passenger counts in all transit modes;
Surveys in both public and private systems;
Reveled preference
Stated preference
Multimodal model with separated assignment models for public and private transport;
Peak hour
Off peak hour.
Complex fare system
Specific modeling for park & ride users
Modal split model.

24. Calcutta Light Rail Model Calcutta, India
4. Exemplos internacionais de utilização do CUBE – Ásia
Calcutta Light Rail Model Calcutta, India
Capital of the Indian state of West Bengal.
The city of Kolkata has 4.5 million residents, and the metropolitan area, including suburbs, has a population of approximately 15.7 million, making it the third most populous metropolitan area in India and the 13th most populous urban area in the world

25. Thailand Multimodal Model
4. Exemplos internacionais de utilização do CUBE – Ásia
Thailand Multimodal Model
Urban Master Plans
Bangkok
Khisanulok
Modelos urbanos

26. Multimodal Model Hong Kong
Developed a series of hierarchical models:
A strategic model representing interaction with mainland China, called the Cross Boundary Model (CBM)
A conventional 4-stage model for HK Special Administrative Region
Detailed models for PT and rail assignment
Detailed modeling of highway route choice
Local area models including microsimulation (Dynasim)

27. Transport Model Beijing, China
Key Statistics:
10m population
Over 20m daily trips
Rapidly growing car availability
Transport networks:
Metro with plans for extensive expansion
Expressways
BRT
Extensive Bus Network
Some 30% of trips in Beijing are made by bicycle: Cube allows them to be modeled as a separate mode with an appropriate pcu factor.
Dedicated bicycle lanes are modeled in Cube as a separate link class which is banned to other vehicle types in the assignment; bicycles also have a separate speed-flow curve
As some bicycle lanes are only separated by a road markings (rather than physical segregation), a special mechanism has been set up in Cube to check the vol/cap ratio of these links and reflect overspill to the main carriageway

28. Multimodal Model Jakarta
Key Statistics:
8.5m Jakarta
22m total in Greater Jakarta (Jabodetabek)
More than 30m daily trips
Large public transport market – but nearly all on buses
Motorcycle and cars
Highly congested road network
Transport networks:
Toll Road network on key arterials and around CBD
Heavily used other roads
Cube Analyst (Matrix estimation) used to refine matrices by multi-user classes
Highways assignment model – using multi-user class equilibrium assignment in HIGHWAY with capacity constrained congestion modeling
Tolls represented by fixed point toll or distance based component (depending on toll road)

29. Multimodal Model Hanoi, Vietnam
Key Statistics:
3m Population
6m daily trips
- Around 75% of trips made by motorcycle
10% public transport; grown from 2% only 5 years ago
Average road speeds <20kph
Transport networks:
Bus network – modernized and expanded in past 5 years
Bus Rapid Transit and Rail Lines under planning

30. Multimodal Model - Melbourne
Train tram bus
Two user classes PT walk access and PT park and Ride
Regional Rail Services and use of airport shuttle bus handled by submodels
4 time periods; clustered run to bring down run time to less than 24 hours on 4 core machine
Taking account crowding increases run time significantly

31. Highspeed Rail model, California USA
Evaluate HSR alternatives
Statewide
Into and out of the San Francisco Bay Area
Produce performance and evaluation measures
Ridership and revenues, user benefits
Time and cost savings for new riders
Impacts on other modes
Use existing models to build high speed rail networks
Develop Logit mode choice models from new data
Perform Assignment to look at ridership
Use Cube PT (Public Transport) Module to:
Code transit route networks
Access and Egress to trains
Park & Ride and Pedestrian / Bike Catchment Area
Define Fares and penalties
Model Service Scenarios

32. Thoughts on ideal modelling program
Four inter-related elements…
Data collection program
Model design/structure
Validation/testing procedures
Training
…where each element is developed with the other three elements in mind
Data collection program captures sufficient information for validation/testing
Validation/testing procedures correspond with the model design/structure
Training is sufficient to allow for proper execution or maintenance of the model design/structure
Etc. 32

33. Thank you! Andreas Köglmaier Regional Director