Busway Platform Bus Capacity Analysis - QUT ePrintseprints.qut.edu.au/43698/1/Sumeet_Jaiswal_Thesis.pdf · Earlier BRTs were bus shared lane or bus only lane, ... 3.2.1 The size 46

Busway Platform Bus Capacity Analysis

Sumeet Kumar Jaiswal B.E. (Civil), M. Tech (Transportation)

A thesis submitted for the degree of

Doctor of Philosophy

School of Urban Development

Faculty of Built Environment & Engineering

Queensland University of Technology

December 2010

Dedicated to my dear sister Deepti Muley

Sumeet Jaiswal Page i

Keywords

BRT, busway, capacity, lost time, crowd, transit, public transport, interface, dwell

time.

Sumeet Jaiswal Page iii

Abstract

Bus Rapid Transit (BRT), because of its operational flexibility and simplicity, is

rapidly gaining popularity with urban designers and transit planners. Earlier BRTs

were bus shared lane or bus only lane, which share the roadway with general and

other forms of traffic. In recent time, more sophisticated designs of BRT have

emerged, such as busway, which has separate carriageway for buses and provides

very high physical separation of buses from general traffic.

Line capacities of a busway are predominately dependent on bus capacity of its

stations. Despite new developments in BRT designs, the methodology of capacity

analysis is still based on traditional principles of kerbside bus stop on bus only lane

operations. Consequently, the tradition methodology lacks accounting for various

dimensions of busway station operation, such as passenger crowd, passenger

walking and bus lost time along the long busway station platform. This research has

developed a purpose made bus capacity analysis methodology for busway station

analysis. Extensive observations of kerbside bus stops and busway stations in

Brisbane, Australia were made and differences in their operation were studied. A

large scale data collection was conducted using the video recording technique at the

Mater Hill Busway Station on the South East Busway in Brisbane.

This research identified new parameters concerning busway station operation, and

through intricate analysis identified the elements and processes which influence the

bus dwell time at a busway station platform. A new variable, Bus lost time, was

defined and its quantitative descriptions were established. Based on these finding

and analysis, a busway station platform bus capacity methodology was developed,

comprising of new models for busway station lost time, busway station dwell time,

busway station loading area bus capacity, and busway station platform bus capacity.

The new methodology not only accounts for passenger boarding and alighting, but

also covers platform crowd and bus lost time in station platform bus capacity

estimation. The applicability of this methodology was shown through demonstrative


Sumeet Jaiswal Page iv

examples. Additionally, these examples illustrated the significance of the bus lost

time variable in determining station capacities.

Sumeet Jaiswal Page v

Contents

Abstract iii

List of Tables ix

List of Figures xi

Chapter One 1Introduction 1

1.1 General 1

1.2 Background 1

1.3 Research motivation 2

1.4 Research hypothesis 3

1.5 Research aim and objectives 3

1.6 Scope of this research 3

1.7 Relevance of this research 4

1.8 Thesis outline 5

1.9 Publications from this research 7

Two 2Literature Review 9

2.1 Overview 9

2.2 Bus Rapid Transit System 9

2.2.1 BRT defined 9

2.2.2 Busway defined 11

2.3 Bus stop/ station classification 12

2.3.1 Simple stop 12

2.3.2 Enhanced stop 14

2.3.3 Dedicated station 14

2.3.4 Intermodal Terminal or Transit Centre 15

2.4 Role and impact of bus stop/ station 16

2.4.1 Bus dwell time 17

2.5 Busway station 23

2.5.1 Role of busway station 24

2.5.2 Passenger flow at a busway station 24

2.5.3 Platform crowd 26


Sumeet Jaiswal Page vi

2.6 Crowd density and walking speed 28

2.6.1 Pedestrian speed-density-flow relationship 29

2.6.2 TCQSM method 32

2.7 Busway platform bus capacity 36

2.7.1 Development of Bus capacity model 37

2.7.2 Revised capacity model 38

2.7.3 TCQSM methodology of capacity calculation 39

2.8 Gaps in the knowledge 41

Three 3Research Problem Development 45

3.1 Overview 45

3.2 Difference in bus stop and busway station operation 45

3.2.1 The size 46

3.2.2 The demand 46

3.2.3 The passenger boarding process 47

3.3 Problem conceptualisation 48

3.4 Busway operation 49

3.5 Definition of terms 52

3.6 Conclusions 53

Four 4Data Collection and Processing 55

4.1 Overview 55

4.2 State of art in relevant data collection technique 55

4.3 Technique used for data collection in this study 58

4.4 Research methodology 59

4.4.1 Data collection methodology 59

4.4.2 Data extraction methodology 59

4.4.3 Data analysis methodology 60

4.5 Selection of study station 61

4.6 Characteristics of Mater Hill Busway Station 64

4.6.1 Passenger flow at station 66

4.6.2 Bus flow at station 67

4.7 Sequence of data collection 68

4.8 Data processing 71

4.9 Chapter close

71

Sumeet Jaiswal Page vii

Five 5Parameter Analysis and Evaluation 73

5.1 Overview 73

5.2 Measuring platform crowd 73

5.3 Passenger - bus Interface 73

5.3.1 Discussion of passenger - bus interface 78

5.3.2 Time - Space Diagram 79

5.4 Passenger behaviour while waiting 81

5.5 Bus lost time 84

5.6 Passenger - bus interaction 87

5.6.1 Effect of fare collection policy 90

5.7 Chapter close 93

Six 6Modelling Bus Lost Time 95

6.1 Overview 95

6.2 Bus lost time histogram 95

6.3 Probability distribution curve fitting 99

6.3.1 Assessing normality 103

6.3.2 Null hypothesis testing 104

6.4 Log-normal distribution curves for bus lost time 107

6.4.1 Log-normal probability distribution function curve 108

6.4.2 Log-normal cumulative distribution function curve 115

6.4.3 Descriptive characteristics of busway station bus lost time 115


Seven 7Busway Station Dwell Time Model 119

7.1 Overview 119

7.2 Model framework 119

7.3 Busway station bus dwell time model 121

7.4 Example application 122

7.5 Discussion 127


Eight 8Busway Loading Area Bus Capacity Model 129

8.1 Overview 129

8.2 Approach to busway loading area bus capacity model 129

8.2.1 Busway dwell time 130

8.2.2 Dwell time variability 130


Sumeet Jaiswal Page viii

8.2.3 Failure rate 131

8.2.4 Operating margin due to passenger service time variability 131

8.2.5 Lost time variability 132

8.3 Busway loading area bus capacity model 134

8.4 Effective bus capacity of loading area 135

8.5 Busway station platform bus capacity 136

8.6 Example application 136

8.7 Discussion 139


Nine 9Busway Station Efficiency Model 141

9.1 Overview 141

9.2 Loading area blocking 141

9.2.1 Existing approach 146

9.3 Approach to loading area efficiency factor calculation 147

9.4 Loading area efficiency factors for Mater Hill Busway Station 148

9.5 Discussion 149


Ten 10Conclusions 151

10.1 Overview 151

10.2 Summary of this thesis 151

10.3 Contributions of this research 153

10.4 Implications of this research 153

10.5 Conclusions 154

10.6 Recommendations for future work 155

11References 157

Appendix 12

A 13Busway Station Platform Bus Capacity Analysis Worksheet 163

B 14Bus Capacity example application 165

C 15List of Publications 169

Sumeet Jaiswal Page ix

List of Tables

Table 2.1 Pedestrian level of service on walkways 34

Table 2.2 Levels of service for queuing area 34

Table 2.3 Similar type of bus lanes and busway 37

Table 2.4 Failure rates and corresponding ‘z’ values 39

Table 2.5 Efficiency of multiple offline linear loading area at bus stops 41

Table 3.1 Boarding process at a bus stop and at a busway station 48

Table 4.1 Data collection techniques used in past studies 56

Table 4.2 Candidate busway station (outbound) platforms 63

Table 4.3 Fare collection policies at Mater Hill Busway Station 65

Table 4.4 Bus flow rate and passenger demand classification split 69

Table 4.5 Characteristics of analysis time 69

Table 5.1 Duration of passenger – bus interface during off-peak period 74

Table 5.2 Duration of passenger – bus interface during peak period 75

Table 5.3 Passenger boarding and alighting during evening peak period 85

Table 5.4 Bus lost times (LT) during off-peak periods 85

Table 5.5 Bus lost times (LT) during peak period 86

Table 5.6 Descriptive statistics 90

Table 5.7 Fare collection policies and observations at study station 91

Table 5.8 Effect of fare collection policy on boarding time per passenger 92

Table 5.9 Effect of fare collection policy on alighting time per passenger 92

Table 6.1 Descriptive statistics of loading area 1 (Off-peak period) 100



Table 6.4 Descriptive statistics of loading area 1 (Peak period) 101


Sumeet Jaiswal Page x



Table 6.7 Methods for testing normality 103

Table 6.8 Assessing normality for loading area 1 (Off-peak period) 105



Table 6.11 Assessing normality for loading area 1 (Peak period) 106



Table 6.14 Statistical parameters of bus lost time curves 115

Table 6.15 Descriptive characteristics of bus lost times (Peak period) 116

Table 6.16 Descriptive characteristics of bus lost times (Off- peak period) 116

Table 7.1 Example demonstration 123

Table 8.1 Failure rates and corresponding ‘z’ values 131

Table 8.2 Example demonstration 137

Table 9.1 Efficiency factors provided by TCQSM 146

Table 9.2 Occupancy and blocking rates for loading areas at outbound platform of Mater Hill Busway Station (Afternoon peak period)

148

Table 9.3 Number of effective loading areas calculation for bus station platform 149

Table 9.4 Comparison of loading area efficiency results 149

Sumeet Jaiswal Page xi

List of Figures

Figure 1.1 Developed framework for busway station platform bus capacity analysis 4

Figure 1.2 Structure of this thesis 6

Figure 2.1 BRT configurations and their passenger transit facilities 11

Figure 2.2 Lane configuration of a busway and its station (Brisbane, Australia) 12

Figure 2.3 A kerb side simple bus stop 13

Figure 2.4 An enhanced BRT stop 13

Figure 2.5 Mater Hill Busway Station 14

Figure 2.6 Transit centre with sawtooth arrangement of loading areas 15

Figure 2.7 Framework for pedestrian walking behaviour 29

Figure 2.8 Theoretical model of pedestrian flow in single channels 31

Figure 2.9 Empirical relations between travel and density of pedestrians 31

Figure 2.10 Walking speed variations as a function of age 32

Figure 2.11 Pedestrian speed on walkways 32

Figure 2.12 Illustration of walkway level of service 33

Figure 2.13 Illustration of queuing area level of service 35

Figure 2.14 Steps to calculate station bus capacity 36

Figure 2.15 Examples of loading area 41

Figure 2.16 Gap in busway station bus capacity estimation approach 42

Figure 3.1 Concentration of passenger crowding 47

Figure 3.2 Origin and destination of a trip segment at platform 48

Figure 3.3 Different levels of busway operation 49

Figure 3.4 A passenger – bus interface phase at a busway station 50

Figure 3.5 A bus lost time phase at a busway station 50

Figure 4.1 Camera positions at Mater Hill Busway Station (Outbound platform) 59


Sumeet Jaiswal Page xii

Figure 4.2 Processing of data – from collection stage to analysis stage 61

Figure 4.3 Brisbane’s South East Busway route map 62

Figure 4.4 Configuration of Mater Hill Busway Station 65

Figure 4.5 Boarding and alighting at inbound platform of Mater Hill Busway Station 66

Figure 4.6 Boarding and alighting at outbound platform of Mater Hill Busway Station

67

Figure 4.7 Number of buses servicing outbound platform of Mater Hill Busway Station

68

Figure 4.8 Matrix for data mining of passenger demand and bus flow 70

Figure 5.1 Variation in passenger – bus interface 76

Figure 5.2 Variation in passenger – bus interface during off-peak period 78

Figure 5.3 Passenger – bus interface duration and its dependent variables 79

Figure 5.4 Time – space diagram 81

Figure 5.5 Distance to loading areas from the waiting area on the busway platform (Off-peak)

82

Figure 5.6 Effect of loading area on bus lost time 83

Figure 5.7 Variation in bus lost times over platform crowd by loading area 87

Figure 6.1 Off-peak period bus lost time histogram 96

Figure 6.2 Peak period bus lost time histogram 97

Figure 6.3 Bus lost time probability distribution curves (Peak period) 109

Figure 6.4 Bus lost time probability distribution curves (Off-peak period) 110

Figure 6.5 Comparison of peak and off-peak bus lost time probability distribution curves

111

Figure 6.6 Bus lost time cumulative distribution curves (Peak period) 112

Figure 6.7 Bus lost time cumulative distribution curves (Off-peak period) 113

Figure 6.8 Comparison of peak and off-peak bus lost time cumulative distribution curves

114

Figure 7.1 Overview of model form for busway platform dwell time estimation 121

Figure 7.2 Effect of bus lost time on dwell time at loading area 1 (Peak period) 124

Figure 7.3 Effect of bus lost time on dwell time at loading area 1 (Off-peak period) 125

Sumeet Jaiswal Page xiii


Figure 7.5 Effect of bus lost time on dwell time at loading area 2 (Off-peak period) 126


Figure 8.1 Log-normal density curve 133

Figure 8.2 Variation in busway station bus capacity with boarding load per bus 138

Figure 8.3 Effect of bus lost time on busway station bus capacity 138

Figure 8.4 Estimated busway station bus capacities 139

Figure 9.1 Trajectory of bus processing at the Mater Hill Busway Station (Outbound platform)

143

Figure 9.2 Inter-loading area blocking scenarios and associated numbers of effective loading areas.

145


Sumeet Jaiswal Page xiv

Statement of Original Authorship

The work contained in this thesis has not been previously submitted to meet

requirements for an award at this or any other higher education institute. To the best

of my knowledge and belief, the thesis contains no material previously published or

written by another person except where due reference is made.

Signature _____________________

Date _____________________

Sumeet Jaiswal Page xv

Acknowledgements

Completing this PhD has been the most challenging task for me in the journey called

life. In this journey, I own my deepest gratitude to my principal supervisor, Dr.

Jonathan Bunker, and Associate supervisor, Prof. Luis ferreira, for their advice

encouragement, and support.

I would also like to thank my sister and fellow PhD colleague, Miss Deepti Muley, for

her everlasting support in various aspect of my research.

I also greatly appreciate the support, help and expertise received from Queensland

Transport’s Busway Operation Centre, Brisbane and in particular Mr. Jurgen

Pasiezny and Mr. Andrew Haddock.

I am grateful to Mr. Daniel Buntine for his assistance, especially during the data

extraction phases.

I also acknowledge the financial support and assistance provided to me by the

research portfolio and school of urban development.

I am grateful to my colleagues at Bitzios Consulting who always have provided me a

comfortable working environment.

My special thanks go to my parents, relatives and siblings for their efforts and best

wishes which have provided me this unique opportunity. Last, but not least, I would

like to thank my wife Shraddha for her understanding, love and support.

Introduction

Sumeet Jaiswal Page 1

Chapter One

Introduction

1.1 General This chapter establishes the motivation behind this research and defines its aim and

objectives. This is followed by a description of scope and relevance of this research.

This chapter then outlines the structure on this thesis.

1.2 Background Bus Rapid Transit (BRT) is rapidly gaining popularity with urban designers and

transport planners to address the ever increasing needs for fast and assessable, yet

economical and reliable transport. Busway is one form of bus rapid transit (BRT),

and consists of dedicated roadway infrastructure exclusively for use of buses. It is

designed to provide a very high level of physical separation to buses from general

vehicle and other forms of traffic. A key advantage of a busway is that the bus can

serve suburban communities using local and arterial roads and then enter the

busway to run limited stop or line haul. Busways generally have stations located

further apart than on-road bus stops, and in some cases as far apart as a suburban

rail system. Thus, the busway can provide a premium transit service of quality

approaching that of rail. Other reasons for the increasing popularity of busways

include their simplicity to operate the bus service and flexibility to provide more bus

routes and frequencies when the demand arises.

Some of the well known busway networks include Ottawa’s Transitway opened in

1983, Pittsburgh’s Martin Luther King Jr. East Busway opened in 1983, Brisbane’s

South East Busway network opened in 2001, Auckland’s Northern Busway started in

2008, and Adelaide’s O-Bahn Busway, opened in 1986, which is a guided busway

with its unique specially-built track for buses.



1.3 Research motivation The South East Busway in Brisbane, Australia has experienced an exceptional

growth in patronage since its opening in 2001. In the first 6 months of its operation,

the number of passengers grew by 40 percent or by more than 450,000, giving a

daily average patronage of 58,000. Over the first 3.5 years there has been an 88

percent increase in patronage for the busway (Hensher, 2007). In a report published

by the Translink Transit Authority (2009), the South East Busway between Brisbane

Central Business District (CBD) and Eight Mile Plains carried more than 150,000

passengers per day, with sections of busway carrying 18,000 passengers per hour

during the peak. Further to this, there is strong evidence that background patronage

will continue to grow.

In response to this increase, The Translink Transit Authority (TTA) is increasing bus

service frequencies and routes to increase and maintain the system efficiency. Even

so, such high busway patronage means that its stations have crucial roles to perform

for the smooth operation of its busways. Firstly, a station must accumulate

passengers until their desired services arrive. Secondly, a station should facilitate a

smooth process of boarding and alighting of passengers with their desired buses, so

that buses can be accommodated by the station without any non-service related

delays.

However, waiting passengers on the platform can lead to crowding, which can

interfere in passenger boarding vis-à-vis bus dwell time. Therefore, impacts of

platform crowd on bus dwell time needs to be considered in estimating bus

throughput capacity of a station. Currently there is no methodology available to

analyse the effects of platform crowd on the boarding process and bus dwell time.

The established bus capacity analysis methodology for busway (TRB, 2003) is

primarily based on the operational characteristics of a bus stop adjacent to a bus

only lane, which lacks account for the effects of platform crowding. Thus, there is a

need to develop a busway station bus capacity analysis methodology which can

approximate the operation of a busway station.

Introduction


The subject of this thesis is to develop a purpose made methodology for busway

station bus capacity analysis. The hypothesis, aim and objectives of this research

are given in the following sections.

1.4 Research hypothesis Hypothesis:

“Passenger walking and the prevailing crowd at a busway station platform influence

the bus dwell times”

This hypothesis emphasizes that the traditional bus dwell time models cannot be

used for busway analysis because these modes do not account for accumulation of

passengers at the station platform. A detailed discussion on development of this

hypothesis is presented in Chapter 3.

1.5 Research aim and objectives The driving aim of this research is the development of a reliable and robust bus

capacity analysis methodology for a busway station. In order to achieve this aim,

additional objective are defined -

1. Understand operation of a busway station and study the passenger movement

on the platform.

2. Identify and investigate the parameters affecting bus dwell time at a busway

station platform.

3. Develop a robust busway dwell time model.

4. Assess the impact of busway dwell time on busway platform bus capacity.

5. Assess the impact of bus – bus interference on platform bus capacity

1.6 Scope of this research As stated earlier, the motivation behind this research is to study the impact of

platform crowd on bus dwell time at a busway station. The crowding happens only

due to passengers on the platform waiting for arrival of their desired buses, so that

they can board the service. The alighting passengers, on the other hand, quickly

move out of the platform and hence, do not form a part of the standing crowd.

Furthermore, the dwell time of a bus serving only alighting passengers will have



minimal influence on platform crowd. Hence, the data collection for this study was

limited to a busway platform where boarding passengers were dominant.

1.7 Relevance of this research Lack of a dedicated analysis methodology for a busway station could lead to

inaccurate design of the system, such as over estimation of station platform bus

capacity. Incomplete knowledge of factors influencing the busway station operation

can potentially lessen the advantages of a policy improvement, such as smartcard

fare system. This research, by studying a working busway station on Brisbane’s

busway has developed a methodology which accounts for the various dimensions of

busway station operation.

Various dimensions of busway station operation, such as passenger – bus interface

and bus lost time due to passenger walking along the long platform, and bus – bus

interference due to multiple linear loading areas, have been identified and defined.

Based on these dimensions, a framework for busway station bus capacity analysis

methodology has been developed. Figure 1.1 shows the framework of this

methodology. The stepwise procedure for this methodology is given in Appendix A.

Figure 1.1: Developed framework for busway station platform bus capacity analysis

Estimate bus lost time Chapter 6

Estimate busway dwell time Chapter 7

Estimate busway loading area capacity

Chapter 8

Estimate busway loading area efficiency

Chapter 9

Estimate busway station platform capacity

Chapter 8

Introduction


1.8 Thesis outline Figure 1.2 shows the structure of this thesis. This thesis can be divided into three

parts, namely concerning the development of the research problem, setting the

approach towards solutions, and development of solutions.

Development of research problems are discussed in Chapters 1 to 3. This is followed

by the approach used for solving these problems, Chapters 4 and 5. Chapters 6 to 9

develop the solutions for these problems. Chapter 10 syntheses all the analysis

results and findings, and concludes this thesis.

A brief outlines of each chapter is given below –

Chapter 1 (this chapter) establishes the hypothesis, aim and objectives of this

research. It describes the scope and contributions of this work.

Chapter 2 reviews the existing literature relevant to this research and identifies

the gaps in the area of busway analysis.

Chapter 3 develops the research problems and identifies the parameters

influencing busway station platform operation. It sets the directions for

data collection.

Chapter 4 reviews the state of art in data collection and develops a set

methodology concerning collection, extraction and analysis of data for

this research.

Chapter 5 analyses and evaluates the parameters and develops the variables. It

sets the foundation for data analyses.

Chapter 6 establishes the quantitative descriptions of bus lost time and defines

its descriptive characteristics.

Chapter 7 describes development of busway dwell time model

Chapter 8 describes development of busway loading area bus capacity model



Chapter 9 describes development of busway loading area efficiency model

Chapter 10 concludes this thesis and identifies the contribution and innovations of this research. It also provides guidance for further research.

Figure 1.2: Structure of this thesis

Research problem development

Data collection and processing

Parameter analysis and evaluation

Modelling bus lost time

Busway station dwell time model

Busway loading area capacity model

Busway station efficiency model

Conclusions

Introduction Chapter 1

Literature review Chapter 2

Chapter 5

Chapter 3

Chapter 4

Chapter 6

Chapter 7

Chapter 8

Chapter 9

Chapter 10

Development of research problem

Approach for solution

Solutions

Introduction


1.9 Publications from this research This research has lead to publication of one refereed journal paper and six refereed

conference papers. Additionally, one more paper is under review for journal

publications. The complete list of papers is given in Appendix C.

Literature Review


Chapter Two

Literature Review

2.1 Overview This chapter reviews the relevant literature in the field of bus dwell time and busway

station bus capacity estimation. The next section first provides an outline of different

forms of bus rapid transit (BRT) systems. Later section 2.3 outlines the classification

of various bus stops and stations and their impact on the transit system is discussed

in section 2.4.

Section 2.5 discusses in detail the role and impact of a busway station. Since this

research targets the effects of passenger walking and platform crowding at a busway

station upon the bus dwell time, a review of studies related to pedestrian flow and

density is also reviewed in section 2.6.

The past research in the area of busway bus capacity analysis is presented in

section 2.7. The chapter closes with section 2.8 identifying the gaps in existing

knowledge of busway bus capacity analysis.

2.2 Bus Rapid Transit System

2.2.1 BRT defined

The Federal Transit Administration (FTA) of the United States of America defines

BRT as,

“A rapid mode of transportation that can provide the quality of

rail transit and the flexibility of buses.” (Levinson et.al., 2002).

The above definition highlights the operating characteristic of the BRT system - a

bus service which combines suburban door to door service with a high speed line



haul transit of rail transit. A more detailed definition illustrating the design and

implementation of BRT system was given by the TCRP Report 90 -

“BRT is a flexible, rubber-tired form of rapid transit that

combines stations, vehicles, services, running ways, and ITS

elements into an integrated system with a strong identity.”

(TCRP, 2003)

There are many forms of BRT system in use in different parts of the world. Most

common forms are, but not limited to, exclusive bus lanes, and dedicated busways.

Figure 2.1 shows BRT configuration a) with bus lane and b) with busway. An

exclusive bus lane is a traffic lane reserved for bus use only. This is a relatively

cheaper option however it provides limited improvement in transit speed and

reliability. A busway, on the other hand, can be a fully grade separated exclusively

built rightway for buses and can provide greater improvement in transit speed and

reliability.

Photo by: Jaiswal (2009)

a. Exclusive bus lane with bus stop

Literature Review



b. Dedicated busway with busway station

Figure 2.1: BRT configurations and their passenger transit facilities

2.2.2 Busway defined

A BRT system is a service that operates on bus lanes or other transitways in order to

achieve high speed of transit. A busway is a special roadway infrastructure designed

for exclusive use of buses (FTA, 2008). A busway is different to the other BRT

treatments, such as bus lanes and bus priority schemes which are more limited in

their scope. A busway will usually have its own right of way, physically separated

from general traffic. With the greatest level of separation from general traffic and

road intersections, a very high speed bus transit is possible on a busway. For

example, posted speed limits are 90 km/h on certain sections of Brisbane’s South

East busway. Similarly, buses on Canada’s Ottawa transitways have a speed limit of

70 to 90 km/h between stations (Wikipedia, 2009a).

A busway usually has a non-overtaking regime where buses are not able to overtake

one another on the corridor. The necessity of overtaking does not generally arise on

the busway corridor since buses operate at common speed on any given section.

However, busway stations are mostly designed with provision of a passing lane to

facilitate buses to overtake stopped buses. This makes the busway station design



and management highly important for smooth operation of the busway corridor.

Figure 2.2 shows Brisbane busway and its station.


Figure 2.2: Lane configuration of a busway and its station (Brisbane, Australia)

2.3 Bus stop/ station classification There are four types of bus station (FTA, 2004). Depending on their function in the

system they vary in size and amenities. Their brief descriptions are given below. As

noted by the US Federal Transportation Authority, transit stations and their amenities

provide comfort to the passenger and therefore also help in attracting more

patronage.

2.3.1 Simple stop

These stops are simplest of all four types. Such a stop consists of a shelter and

printed information display. Figure 2.3 shows a simple bus stop. These stops

principally cater service to a relatively low level of passenger demand and bus

routes.

Literature Review



Figure 2.3: A kerb side simple bus stop


Figure 2.4: An enhanced BRT stop



2.3.2 Enhanced stop

An enhanced stop is the upgraded version of simple stop with enhanced shelter,

usually designed for the BRT line to distinguish it from other stops. These stops

usually integrate special amenities such as glass wall, water outlet, trash bin and/ or

provision of pay phone. Figure 2.4 shows an enhanced stop.

2.3.3 Dedicated station

Dedicated stations are the high end stations designed specially for a BRT system.

Designs of these stations usually have multiple loading areas for several buses to

stop simultaneously. Their design includes all-weather shelter for passengers,

lighting, level passenger platform boarding and alighting for speedy movement and

high quality passenger information facilities. Amenity wise they may have benches,

water outlet, pay phone and ticket vending machine and food kiosks. Figure 2.5

shows a dedicated busway station on a Brisbane busway corridor.


Figure 2.5: Mater Hill Busway Station

Literature Review


2.3.4 Intermodal terminal or transit centre

An intermodal terminal or transit centre is designed to facilitate the transfer activities

of passengers between different modes, such as rail or ferry and/ or terminus of bus

services. These facilities incorporate a host of amenities such as waiting areas,

benches, water outlets, lighting, pay phones and so on. Usually their platforms have

level boarding and alighting design to facilitate comfortable passenger movement

between bus and platform. They have the provision of more than one loading area,

commonly in sawtooth arrangement to allow smooth and independent movements of

buses in and out of their loading area. Figure 2.6 shows a transit centre provided in a

close proximity to a shopping center.


Figure 2.6: Transit centre with sawtooth arrangement of loading areas



2.4 Role and impact of bus stop/ station Transit stations are gateways to the network. In fact, these stops and stations are the

only points where customers access the transit service. However, scheduling a bus

to observe a stop or station inflates its journey time due to the delay caused by the

stop. The time spent by a bus at a stop to serve its passengers is commonly known

as dwell time. However, in reality the dwell time is not the only component of bus

delay due to stop (Changshan and Murray, 2005). The stop delay can be divided into

four major components, comprising of delay associated with deceleration and

acceleration, delay due to door opening and closing, delay associated with bus

clearance time, and delay due to passenger boarding and alighting.

When a bus is required to observe a stop it needs to decelerate from its cruise

speed. Similarly, after passenger service, the bus needs to accelerate to get back to

its cruising speed. These deceleration and acceleration from/ to cruise speed also

add delay time to the bus journey time. The deceleration and acceleration time delay can be mathematically modelled as follows (Changshan and Murray, 2005;

Wirasinghe et.al., 1981).

0.5 1

1

Equation 2.1

Where,

= Total delay time at stop i associated with bus acceleration and deceleration.

= Bus cruise speed

= Acceleration rate

= Deceleration rate

The door opening and closing time is the time needed to fully open the door from

the closed position and vice-versa. It depends on the mechanical properties of the

bus door. The value for door opening and closing time can be obtained from on-site

observation. In the absence of such data the delay due to door opening and closing

is usually taken as a constant varying between 2 and 5 seconds (TRB, 2003;

Changshan and Murray, 2005)

Literature Review


The Clearance time is the time a bus takes to clear the loading area after

completing the passenger servicing at the stop and making it available for the next

bus. The clearance time can be influenced by the type of loading area (TRB, 2003).

If the loading is on-line, being within the traffic lane, then the clearance time will

equate to the time required by the bus to start up and travel its own length plus the

time for the subsequent bus to pull into the loading area. However, when the loading

area is off-line, or out of the traffic or passing lane, the bus driver needs to find a

suitable gap in the adjacent traffic flow to re-enter into the flow. This re-entry delay to

bus is an additional component of clearance time.

The fourth component of bus stop delay, delay due to passengers boarding and alighting, is the most important of all delays. Unlike the other three delays, this

delay varies from stop to stop and is highly sensitive to passenger demand. The

delay due to passenger boarding and alighting is commonly assumed as a dual

linear function of the number of passengers boarding, and of those alighting.

Collectively, the delay due to the door opening and closing and delay due passenger

boarding and alighting are referred to in the literature as ‘bus dwell time’. The bus

dwell time at a bus stop can be influenced by four main elements – number of

boarding and lighting passengers, fare collection system, vehicle characteristics

such as number of doors and floorplan, and on board crowd levels.

Bus dwell time has been well established in the literature as a significant factor

causing bus bunching and thereby the reliability of transit service (Rajbhandari et.al.,

2003). Maloney and Boyle (1999) observed that the dwell times at stops on surface

roadways constitute about 7 percent of total time for a bus in service when the bus is

running along with the general traffic. The study also pointed out that the time

manoeuvring out of and into traffic, i.e clearance time, constituted around 7 percent.

In another study, Levinson (1983) found that at a CBD stop the dwell time range

from 20 to 60 seconds and for a non – CBD stop the dwell time ranged between 10

to 15 seconds.

2.4.1 Bus dwell time

The Transit Capacity and Quality of Service Manual (TRB, 2003) defines bus dwell

time as the amount of time a bus spends at a stop or at a station to serve boarding



and alighting passengers plus any time required for door opening and closing

operation. At a given stop / station, the dwell time is directly related to passenger

boarding and alighting, fare payment method, vehicle type and size, and in vehicle

circulation. The studies considering these variables for dwell time estimation are

described in the following sub sections.

2.4.1.1 Passenger boarding and alighting

According to Levinson (1983) dwell time for any bus is directly proportional to the

number of passengers it serves. As one of the earliest studies toward the

understanding of effect of passenger boardings and alightings on the bus dwell time,

Levinson’s study found that each boarding or alighting passenger contributes

between 2.6 to 3.0 seconds towards the bus dwell time (Equation 2.2). For a stop

having predominant number of alighting passengers, the study found that each

passenger adds 1.2 to 1.7 seconds towards dwell time (Equation 2.3).

5.0 2.75 Equation 2.2

4.0 1.5 Equation 2.3

Where,

N = Number of boarding and alighting passengers

Guenthner and Shina (1983) found that the number of boarding and alighting

passengers can be best described using a negative binominal function. Their study

highlighted that, even though the total dwell time increases, the service time per

passenger deceases as the number of passengers at a stop increases. A logarithmic

model was developed to estimate the dwell time per passenger (Equation 2.4). This

equation yields a maximum bus dwell time when there are 24 passengers. The dwell

time per passenger at this point is 1.2s per passenger. The authors, therefore,

suggested a dichotomised relationship to estimate dwell time based on number of

passengers, according to Equation 2.4 and Equation 2.5.

5.0 1.2 In N ≤ 23

Equation 2.4

Literature Review


1.2 N ≥ 24 Equation 2.5

Where,

N = Number of boarding and alighting passengers

Following these single variable dwell time models, researchers turned to multi-

variable models to estimate dwell times. The modification was made to improve the

accuracy of dwell time estimation. The modified approach considered the number of

alighting passengers and number of boarding passengers as two independent

variables. Vuchic (2005) noted that the dwell time for a bus, where boarding and

alighting take place via different doors, is the maximum of the boarding time and

alighting time, plus a constant time annotated t0 to reflect station standing,

comprising of lost time at the station due to the opening and closing of door, plus

clearance time. Mathematically,

, Equation 2.6

A modified equation for a system where boarding and alighting from all doors is

permitted was also suggested by Vuchic (2005). The equation naturally pertains to

the busiest door.

Equation 2.7

Where,

st = dwell time at station

ot = station standing time

'b and 'a = number of boarding and alighting passengers respectively

bτ and aτ = respective boarding and alighting per passenger

An identical equation was suggested by the Transit Capacity and Quality of Service

Manual in 1999 (TCRP, 1997) and 2003 (TRB, 2003) (Equation 2.8).



ocbbaad ttPtPt ++= Equation 2.8

Where,

dt = Average dwell times (s)

aP = Alighting passengers per bus through the busiest door (p)

at = Alighting passenger service time (s/p)

bP = Boarding passengers per bus through the busiest door (p)

bt = Boarding passenger service time (s/p)

oct = Door opening and closing times (s)

These multivariate dwell time models imply that alighting and boarding occur in

series, and account only for those alighting through the busiest door, normally

presumed to be the front door, assuming that it is the only door available for

boarding. Any passengers alighting through the rear door are neglected in the

standard model, as their activity occurs in parallel to the front door activity, which is

implied to be time-critical.

To determine the probability of a passenger choosing the front door of the bus to

alight, Zhao and Li (2005) suggested a door choice model for alighting passengers

using the data collected from the Broward County Transit (BCT) in Florida, USA. A

utility based binary choice model was proposed to obtain the probability of choosing

the front door by an alighting passenger. The utility function U was defined as

follows.

( )PMAMTIMEPOINTONBOARDTOTALOFFfU ,,,,= Equation 2.9

Where, )( 1=YP = Probability of passenger choosing front door to alight bus.

TOTALOFF = Total number of alighting passengers at a given stop.

ONBOARD = Total passengers onboard before bus doors were opened at a

given stop.

TIMEPOINT = Dummy variable for a given stop 1 for time point and 0

Literature Review


otherwise.

AM = Dummy variable, 1 for observation during AM and 0 otherwise.

PM = Dummy variable, 1 for observation during PM and 0 otherwise.

The logit model for the utility function was approximated as equation 2.10.

Subsequently, the probability of an alighting passenger using the front door can be

obtained from equation 2.11.

PMAMTIMEPOINTONBOARDTOTALOFFU 6777040980838900213003630 ..... ++−−= Equation 2.10

U

U

eeYP+

==1

1)( Equation 2.11

2.4.1.2 Fare payment method

The complexity in fare collection system lead to increase in the boarding time per

passenger as the number of boarding increases at a stop (Guenthner and Hamat,

1988) as opposed to the decrease as expected (Guenthner and Sinha, 1983). Later,

it was noted that the accuracy of the dwell time model can be improved if variable for

fare medium is also considered (Marshall et.al., 1990). An exponential equation was

developed for dwell time calculation considering fare collection medium and bus

induced delay (Equation 2.13). An average service time per passenger was obtained

as 8 seconds (approximately) under complex fare structure.

8.07 . (R2 = 0.67) Equation 2.12

6.65 . exp 0.39 0.20 (R2 = 0.71) Equation 2.13

Where,

= Average dwell times (s)

= Total number of boarding and Alighting passengers. (p)

= Fare collection medium

= Bus induced delay



In a recent study, Milkovits (2008) showed that with the 100 percent use of a smart

card medium the bus dwell time could possibly be to reduce by 22.8 percent. It was

also found that the advantage of smart (card) media over magnetic strip card can be

1.5 to 2s.

Fare collection systems influence service time because some media of fare

collection require more transaction time than others. The fare collection method

mainly affects the boarding time per passenger (Guenthner and Hamat, 1988) but in

some cases like smartcard, it could affect the alighting time as well. In the absence

of any field data the Transit Capacity and Quality of Service manual (TRB, 2003)

suggests boarding service time per passenger between 2.25s with no on–board

ticket purchase, and 4.3s when exact change is used for ticket purchase.

2.4.1.3 Vehicle type and size

Low floor buses have an advantage of reduced dwell time due to reduced boarding

and alighting time per passenger. This is because all types of passengers find it

easier and quicker to board and alight a low floor bus compared to a conventional

bus having raised floor or steps. This is especially true for elderly people, parents

with baby strollers, small children and disabled persons. A saving of nearly 2

seconds per elderly passenger and around 6 seconds for a parent with child was

observed in a study undertaken at St. Albert Transit, US (Liggett, 1992). Similar

observations were made by Levine and Torng (1994). The data from Ann Arbor

Transportation Authority, US was used where the conversional buses and low floor

buses were providing transit service. The study showed a saving of 0.5s per

nondisabled passenger for low floor buses and hold good for both boarding and

alighting times. Later TCQSM (TRB, 2003) also suggested a reduction of 0.5s in per

passenger boarding time if the bus is low floor.

2.4.1.4 In-vehicle circulation

Standees on-board a bus reduce the speed of passenger boarding and alighting.

These standees block the aisles resulting in difficulties for boarding passenger

movement especially if the standees are standing near the front door. On the other

hand if the standees are standing near the rear exit door, the alighting passenger

may use the front door which in turn increases the boarding time. Zografos and

Literature Review


Levinson (1986) found that the passenger service time increases when the bus

operates beyond its seating capacity. Based on the field study the authors

investigated the bus dwell time and its variations with the size of boarding passenger

group and the number of passengers already on the bus. A linear relationship was

established between passenger service time, each boarding passenger’s rank and

number of passengers on board.

1.56 0.16 0.09 Equation 2.14

Where,

= Service time per boarding passenger (s)

= Rank of boarding passenger

= Number of passengers on board

Most of the studies in the field of bus dwell time have been based on data from

simple bus stops. These studies have been applied to the enhanced bus stop since

these bus stops are basically an improved version of the simple bus stop. However,

busway stations are different than bus stops, especially in terms of the number of

loading areas and therefore the length of station platform. Provision of simultaneous

stopping of buses at a busway station also changes the dynamics of the passenger

service process. At present there appears to be little dedicated study of bus dwell

times at these busway stations.

2.5 Busway station Busway stations can be classified as dedicated stations and can be designed to

facilitate several buses to stop simultaneously. A busway station has linear off-line

loading areas, to allow buses to overtake the stopping bus (FTA, 2004; Rathwell and

Schijns, 2002). A high level of amenities for passenger comfort and convenience is

generally provided. The station can be equipped with map and real time information

system, security and lighting arrangement, benches, pay phone and water outlet.

However, the station design concept can also be driven by local conditions. For

example, the transitway stations in Ottawa, Canada are provided with heated,

enclosed on-platform waiting area due to the harsh winters of the region (Rathwell



and Schijns, 2002). Brisbane, Australia, busway stations design is an open platform

type, with large awnings to provide passenger shelter against sun and rain.

Passenger transfer between platforms can be controlled and is only permitted via

overbridge by stair and elevator or lift.

2.5.1 Role of busway station

The significance of a busway station or, every transit station, substantially increases

when we see transport as a service. A transport service is different from other forms

of services in a sense that it cannot be stored. That is, it must be consumed when

the service is generated or the service must only be generated when the demand

arises. Practically, achieving such equilibrium for mass public transport system is

very hard, if not impossible. Since busway stations are provided with large spacing in

between stations and together with the fact the busway are the corridors with high

passenger demands, a busway station attracts larger volumes of passengers than a

kerb side bus stop. And therefore, in addition to facilitate a smooth interface between

the passenger and bus, another role which the busway stations have to perform is to

accumulate the demand until the service is provided to them.

Busway stations are also crucial from the bus operation standpoint. As mentioned

earlier, stations are the only sections in a busway where buses can overtake

stopping buses. Equally, a busway station can easily interrupt the smooth flow of

buses, especially during the peak period when bus queuing often happens at station

entry.

2.5.2 Passenger flow at a busway station

Although the platform length of a busway station is greater than for a kerbside bus

stop, it creates a difficult situation to organise the passenger flow. The characteristic

of pedestrian flow at the busway station is rather complicated due to the fact that

there is no channalisation opportunity available to the passengers. During the peak

hours, the high but near continuous inward passenger flow at the platform entrance

points suffers a sudden breakdown in flow after entering the platform area. This is

because the outward flow of passenger is not on the basis of ‘as you come, so you

go’. While some passengers might receive their desired bus promptly, some have to

Literature Review


wait for relatively longer for their bus. These waiting passengers lead to the

formation of crowding with pedestrian flow disequilibrium.

The ‘crowd effect’ and unpredictable waiting time for the desired bus can alter the

characteristics of pedestrian flow within the platform area. Although there is little

doubt that the presence of many people in a given platform space causes problems

in terms of reduced walking speed and manoeuvring capacity, it also blocks the

pedestrians’ line of eye sight making the identification of incoming buses difficult.

This increases the reaction time of the pedestrian. Overall it increases the bus dwell

time and delay time.

Passengers are required to take certain decisions and steps after entering the

platform area and before entering the door of their desired vehicle. Passengers in

fact make decisions at three levels, observed Hoogendoorn and Bovy (2004). These

three levels, in the contexts of busways stations, are described below.

Decisions at strategic level: At this level the passenger decides on the activities to

be done after entering the station and the order of their execution. At busway

stations such activities are often limited to seeking information from the information

board and electronic information display unit.

Decision at tactical level: At this level the decision on how to perform the activities

that were decided at the strategic level is taken. Such decisions are basically

dependent upon the prevailing conditions faced by the passenger upon entering the

station platform area. A passenger will not seek information from the information

area should the desired bus being serving the platform. Therefore, it is possible that

a new activity could be decided and some or all activities decided at strategic level

can be discarded at this level. The most important decision at this level is

identification of the probable waiting area and making an assessment upon the most

suitable place. The choice for waiting area depends on the amount and location of

crowding present within the platform area.

Decision at operational level: At this level the pedestrian executes the options

decided at the tactical level. The decisions are mostly related to their walking

behaviour, and route choice to reach at a pre-determined point.



2.5.3 Platform crowd

Passenger crowding at the station platform can be predominantly observed at transit

stations located in or near the Central Business District (CBD) of the city, because of

higher activity. Such activity generates high number of trips from or to other parts of

the city. Crowds can be made up of either a majority of waiting passengers – who

want to board a transit service – or a majority of alighting passengers from buses

that have arrived at the station platform. It is worth noting, at this point, that crowding

because of waiting passengers is likely to have greater impact on the platform bus

capacity than that of crowding because of the alighting passengers, though both

could lead to a reduction in corridor capacity depending on specific circumstances.

Impact on passengers

Firstly, crowding at a platform can make passengers uncomfortable, because

reduced space availability could induce unwanted behavioural changes in them (Hui

and Bateson, 1991) and could also reduce the individual productivity. Fear of

missing the bus or being stranded at the platform may stress certain passengers.

Secondly, reduced air quality at the station because of vehicular pollution could have

lasting effects (Chertok et.al., 2004; Chan et.al., 2002). On the whole, such

conditions make the journey difficult and unpleasant which may lead to passengers

having a lower perception of public transport.

Though not much literature dealing with the effect of passenger crowding at the

platform is available, a study carried out by Cox et.al. (2006) identifies crowding as a

problem across the British rail network. They emphasised that crowding should be

accepted as a possible threat both to the health of passengers and the transit (rail)

industry. Undoubtedly, the aim of providing stations is to facilitate a comfortable

interface from being a pedestrian to a passenger under the given pedestrian density.

However, there is seen to be breakdown in the smooth interface when there is

crowding at platform. The TCQSM (TRB, 2003) suggested that the restricted and

uncomfortable movement caused by the crowd can be the reason for such

breakdown.

Literature Review


Impact on service provider

Cost of transit operation is a major concern for service providers (IAPT, 2007;

Vuchic, 2005). Journey speed, representing travel time, is one of the major attributes

which represents operational effectiveness of a transit network. Under the busway

scenario the delay to the bus at stations is a major factor in reducing its journey

speed. Delays, such as excess dwell times, excess clearance time, time spent in

queue, can be correlated with passenger crowding at the platform area of the station.

Moreover, at each station buses accumulate further delays to their running time

causing a reduction in the available layover time and /or vehicle productivity at the

terminal.

Layover time is the time built into the schedule between arrivals and departures,

used to recover delays and preparations for the return trip. It also includes the driver

rest time. However, when the accumulated delay exceeds the minimum amount of

layover time (the minimum layover time should not be less than the driver rest time),

the service provider is forced to put their reserved vehicle and operator into service

to maintain the schedule.

Equation 2.15 is a mathematical formula suggested by Tyler (2002) to determine the

fleet size required to provide a schedule service.

AhT

F c= Equation 2.15

Where,

= fleet size (vehicles)

= cycle time in minutes

= availability of vehicles

= service headway in minutes

The cycle time in minutes Tc is defined as the time required by a transit vehicle (bus)

for a return journey plus the two layover periods. Keeping travel speed of the bus

constant, which in the case of a busway is quite possible, the journey time increases

and the availability of the bus decreases if the delay at the station(s) increases. This,



according to Equation 2.15, would lead to the requirement of a larger fleet size to

maintain the service. From the economical prospective this is not desirable to the

service provider.

2.6 Crowd density and walking speed Understanding pedestrian walking behaviour is a complex task. Some researchers

have tried to explain this behaviour with the analogy of ants (Nishinari et.al., 2006),

while some have used a lattice-gas model to study the pedestrian walking (Jiang and

Wu, 2005). Nevertheless, the majority of studies in the field of pedestrian flow have

pointed out that pedestrian walking speed reduces with increased density. The

increase in density decreases the human-usable space causing interference to

others, resulting in a decrease in walking speed (Osaragi, 2004). Further, it was

found that the pedestrian walking as a group of three or more tend to be slower than

when walking individually or in couples (Tarawneh, 2001). Beside this, if the

pedestrian walking is in opposite directions or crossing the path of each other, there

could be further reduction in their walking speeds (Smith, 1995).

Pedestrian movement is individualised and dynamic. For a given condition the

walking behaviour of different pedestrian can vary. Moreover, a pedestrian may

behave differently every time he/she faces a situation even if the situation is

constant. This may be because humans have the tendency to learn from their past

experience and make necessary changes in their action in rapid successions.

Pedestrians are capable of changing speed more quickly when gaps arise and can

accelerate to full speed from a standstill or decelerate to avoid collision (Blue and

Adler, 2001). Seyfried et.al. (2005) in their study on pedestrian flow reinvestigated

the fundamental relationship between the density and pedestrian speed. Analysing

the single-file movement of pedestrians they observed a linear relationship between

the speed and the inverse of the density. Pedestrians adapt their speed not only to

the person immediately in front, but to the situation further ahead, the study

concluded. Daamen et.al. (2005) described pedestrian flow in congestion using a

fundamental diagram – the relationship between speed, flow and density. Form their

experimental study, they found that on the boundary of the congested region,

pedestrians may walk in nearly free flow conditions. However, inside the congested

regions the low speeds with high density were observed.

Literature Review


The reduction in speed can be caused by close proximity of other pedestrians.

Pedestrians may like to keep certain distances from others. But keeping such a

distance in an environment of a busway platform is difficult. Often after seeing their

bus arriving at the platform, pedestrians walk hastily without anticipating the possible

movement of others, by squeezing through the crowd. However, not all pedestrians

prefer to do so, usually the older pedestrians want to keep some constant distance

and avoid physical touch with others (Gerin-Lajoie and Richards, 2006).

Pedestrian walking behaviour can be divided into two groups: unconstrained and

constrained flow (Antonini, 2005). The unconstrained walking is characterised by

behaviours independent of the presence of other pedestrians. Passengers under

unconstrained walking tend to walk toward their destination, keeping the original

direction and decide free flow acceleration and deceleration to maintain the desired

speed. On the other hand, in a constrained situation pedestrian walking behaviour is

influenced by the interaction with other pedestrians. In the constrained flow situation

pedestrians have a strong tendency to follow the leader and avoid collision with other

people. This pedestrian walking behaviour framework is reproduced in Figure 2.7.

Source: Antonini and Bierlaire, 2005

Figure 2.7: Framework for pedestrian walking behaviour

2.6.1 Pedestrian speed-density-flow relationship

The fundamental of pedestrian flow theory states that the flow (or volume) is a

product of speed and density and their inter-relationship can be approximated by a

parabolic curve which is similar to motor vehicle flow (Khisty, 1990).

Pedestrian walking behaviour

Keep direction

Constrained Unconstrained

Towards destination

Free flow acc/dec

Collision avoidance

Leader follower



kvq = Equation 2.16

The density is measured in passengers per unit area and hence its numerical value

can be a fraction. To avoid the uses of number of pedestrians as a fraction some

researchers use reciprocal of density, space per passenger, M. In such case

equation 2.16 becomes

Mvq = Equation 2.17

Where,

q = Pedestrian flow (p/m/min) v = Speed (m/min) k = Density (p/m2)

M = Space (m2/p)

The relationships between pedestrian speed, density and flow, which are very much

similar to the three fundamental diagrams of vehicle flow, are shown in Figure 2.8.

Similar to the this speed vs. density curve, Kholshevnikov et.al. (2007) presented a

graph between pedestrian speed and pedestrian density, as shown in Figure 2.9,

after reviewing the different studies for pedestrian egress under various situations,

like building, underground station and experimental setup. The figure demonstrates

that pedestrian walking speed is inversely proportional to density. However it is not

just the crowd density which influences pedestrian walking speed, pedestrian age

also does (Ando et.al., 1988). Work presented by Ando (1988) showed pedestrian

walking speed as a function of age under the free flow condition (Figure 2.10).

Literature Review


Source: Khisty, 1990

Figure 2.8: Theoretical model of pedestrian flow in single channels; (a) Speed versus Density; (b) Volume versus Density; and (c) Speed versus Volume.

Source: Kholshevnikov et.al, 2007

Figure 2.9: Empirical relations between travel and density of pedestrian (door opening). Buildings: different 1; retail buildings 2,3,4; Sport structure 5; Underground station 6,7,8,9; Experimental 10,11,12,13,14.

Spee

d, v

Density, k

A

A/B

Spee

d, v

Flow, q

Flow

, q

Density, k

(a) A2/4B

(b)

(c) A/2B

A2/4B

D (person /m2)

Wal

king

Spe

ed, V

(m/m

in)



Source: Smith, 1995

Figure 2.10: Walking speed variations as a function of age

2.6.2 TCQSM method

TCQSM (TRB, 2003 based on Fruin, 1987) highlighted pedestrian density as the

most significant factor in influencing pedestrian walking speed and represented the

relationship between them graphically as in Figure 2.11.

Source: TRB, 2003

Figure 2.11: Pedestrian speed on walkways

Age

Wal

king

Spe

ed, V

(m/m

in)

Pedestrian space (m2/p)

Wal

king

spe

ed (m

/min

)

Literature Review


The Level of Service (LOS) of transit facilities was divided into six distinct categories,

varying from A to F, based on the pedestrian density. Figure 2.12 illustrates the

characteristics of each LOS and Table 2.1 provides the corresponding threshold

values. LOS A suggests that the pedestrians at a transit facility, such as on

walkways, can freely select their walking speed and would not interfere with each

other. Whereas LOS F represents the situation where walking manoeuvre takes

place in a restrictive environment with frequent and unavoidable contact with others.

LOS A Walking speeds freely selected; conflict with other pedestrian unlikely.

LOS B Walking speeds freely selected; pedestrians respond to presence of others.

LOS C Walking speeds freely selected; passing is possible in unidirectional streams; minor conflict for reverse or cross movement.

LOS D Freedom to select walking speed and pass other is restricted; high probability of conflict for reverse or cross movements.

LOS E Walking speed and passing ability are restricted for all pedestrians; forward moving is possible only by shuffling; reverse or cross movements are possible only with extreme difficulty; volume approach limit of walking capacity.

LOS F Walking speeds are severely restricted; frequent, unavoidable contact with others; reverse or cross movement are virtually impossible; flow is sporadic and unstable.

Source: TRB, 2003 Figure 2.12: Illustration of walkway level of service



Table 2.1: Pedestrian level of service on walkways

Expected Flows and Speeds

LOS Pedestrian Space (m2/p)

Avg. Speed, S (m/min)

Flow per unit width, v (p/m/min) v/c

A ≥ 3.3 79 0-23 0.0-0.3 B 2.3-3.3 76 23-33 0.3-0.4 C 1.4-2.3 73 33-49 0.4-0.6 D 0.9-1.4 69 49-66 0.6-0.8 E 0.5-0.9 46 66-82 0.8-1.0 F < 0.5 < 46 variable variable

Source: TRB, 2003

The above stated level of service(s) is applicable only for walkways. The station

platform does not have exactly the same characteristics as walkways, though

walking activity takes place there. The station platform is an area where passengers

walk to and from the bus (transit vehicle) and where passengers wait (in a group) in

anticipation of their desired bus. In walkways, all pedestrians will be in motion,

whereas, at the platform a large number of passengers will be standing. This

requires commencement of walking from being stood and then forcing their way to

the bus door amongst the other standing passengers, when the passenger sees their

desired bus arriving at a loading area. TCQSM provides a different set of criteria for

determining the level of service for platform (queuing) areas. Figure 2.13 lists the

characteristics of each level of service and Table 2.2 presents the corresponding

threshold values.

Table 2.2: Levels of service for queuing area

LOS

Average Pedestrian Area (m2/p)

Average Inter-Person Spacing (m)

A ≥ 1.2 ≥ 1.2 B 0.9-1.2 1.1-1.2 C 0.7-0.9 0.9-1.1 D 0.3-0.7 0.6-0.9 E 0.2-0.3 < 0.6 F < 0.2 Variable

Source: TRB, 2003

Literature Review


LOS A Standing and free circulation through the queuing area possible without disturbing other within the queue.

LOS B Standing and partially restricted circulation to avoid disturbing others within the queue is possible.

LOS C Standing and restricted circulation through the queuing area by disturbing others is possible; this density is within the range of personal comfort.

LOS D Standing without touching other is impossible; circulation is severely restricted within the queue and forward movement is only possible as group; long-term waiting at this density is discomforting

LOS E Standing in physical contact with other is unavoidable; circulation within the queue is not possible; queuing at this density can only be sustained for a short period without serious discomfort.

LOS F Virtually all persons within the queue are standing in direct physical contact with other; this density is extremely discomforting; no movement is possible within the queue; the potential for pushing and panic exists.

Source: TRB, 2003 Figure 2.13: Illustration of queuing area level of service

While comparing the LOS for walkway and queuing area, it was found that the

walkway LOS used average pedestrian speed as one of the deciding parameter,

while the queuing LOS considered the average inter-person spacing as the deciding

factor. Both the LOS criteria used pedestrian area (space available per pedestrian,

m2/p) as the primary parameter in deciding the LOS; however the threshold values of

pedestrian space for a walkway are higher than the respective values for a queuing

area because walking requires more space to gait.



2.7 Busway platform bus capacity Various BRT line bus capacities have been reported within a range of 200 bus/h to

450 bus/h, translating to a range of 10,000 p/h to 25,000 p/h. However, BRT stations

themselves might not be able to handle such high bus and passenger throughputs,

essentially becoming the line constrictions. Further, when a BRT station serves more

routes, passenger handling on the platform and therefore the passenger/bus

interactions become more complex, which may affect the line capacity. Design of a

Busway (BRT) system, therefore, involves the estimation of station bus throughput

capacity in order to estimate bus line capacity.

It has been known for some time that BRT station bus capacity constricts line

capacity (TRB, 2003, Vuchic 2005). Moreover, as the American Transit Capacity and

Quality of Service Manual (TCQSM) notes, BRT station bus capacity is in turn

governed by capacities and efficiencies of individual loading areas (for more

common station configuration of linear loading areas placed in series). The bus

capacity of individual loading areas, in turn, depends on the bus dwell times at these

loading areas. Figure 2.14 provides the overview of calculation for station bus

capacity using current methodology as suggested by TCQSM (TRB, 2003).

Source: Based on TRB, 2003

Figure 2.14: Steps to calculate station bus capacity

Boarding and alighting load

Boarding and alighting service time

Bus dwell time

Loading area capacity Loading area efficiency

Station capacity

Literature Review


2.7.1 Development of bus capacity model

A comprehensive documentation of methodologies for at-grade busway bus capacity

estimation was presented by Shen et.al. in 1998. The report discussed the principle

behind the adaptation of bus lane capacity methodology from busway capacity

estimation. The different bus lane configuration was co-related with different busway

configurations. Table 2.3 shows different types of bus lanes configurations and

corresponding busway facility. Type 1 bus lanes are similar to a busway with no

overtaking facility, where both cannot pass other buses servicing in the station area.

A type 2 bus lane is similar to a busway with overtaking facility, where buses can

pass other buses servicing at station area. A type 3 bus lane has no similar busways

as busways consist of only two lanes, one in each direction, and buses on busway

are not allowed to use the opposite direction to overtake other buses.

Table 2.3: Similar type of bus lanes and busway

Type Bus lane configuration Similar Busway

1 No use of adjacent lane for buses to pass other buses, right-turn queue and other bus lane obstructions

Busway with no overtaking facility

2 Partial use of adjacent lane for buses to pass other buses, right-turn queue and other bus lane obstructions

Busway with overtaking facility

3 Full use of adjacent lane for buses to pass other buses, right-turn queue and other bus lane obstructions

N/A

Source: Shen et.al, 1998

Based on these similarities it was considered that for a busway the same formula

can be used for calculation of bus lane capacity. The capacity formula of bus lane

with no overtaking facility provided by Highway Capacity Manual (HCM, 1994) is

given in Equation 2.18.

⁄ 3600⁄

Equation 2.18



Where,

= Buses per hour per channel per berth (loading area)

= Bus dwell time at stop (s)

= Clearance time (s)

= Reductive factor to account for variations in dwell times and arrival

= Number of effective berths (loading areas)

= Effective green time per cycle (s)

= Cycle length (s)

2.7.2 Revised capacity model

The initial bus capacity model presented in HCM was revised to better incorporate

dwell time variability of a bus stop (TCRP, 1997). A parameter called coefficient of

variation, , was introduced to account for the fluctuation in bus dwell time. The

fluctuation may arise because of the non-uniform passenger demand for different

buses and different routes. Another parameter Z corresponding to the failure rate

was incorporated with the equation. The failure rate sets how often a bus should

arrive at a stop only to find all loading areas occupied, and is assumed to be

normally distributed. Additional parameters, coefficient of variable and failure rate,

were used in combination of bus dwell time and are termed as operating margin.

This revised capacity model suggests that the (average) dwell time at the loading

area is the fundamental in determining its capacity.

3600 ⁄ ⁄ Equation 2.19

Where,

lB = Loading area bus capacity (bus/h)

3600 = Number of seconds in an hour

Cg = Green time ratio (the ratio of effective green time to total traffic signal cycle length; 1.0 for unsignalised streets and bus facilities like busway without adjacent signal control)

ct = Clearance time (s)

Literature Review



= Standard normal variate corresponding to a desired failure rate

= Coefficient of variation of dwell time

Table 2.4 provide the value of ‘z’ corresponding to desired failure fate.

Table 2.4: Failure rates and corresponding ‘z’ values

Failure rate z

1.0% 2.330

2.5% 1.960

5.0% 1.645

7.5% 1.440

10.0% 1.280

15.0% 1.040

20.0% 0.840

25.0% 0.675

30.0% 0.525

50.0% 0.000 Source: TRB, 2003

This revised equation was subsequently adopted by the TCQSM (TRB, 2003).

Based on this equation the TCQSM, provided a methodology for the capacity

estimation for a Busway station. This methodology involves uses of bus dwell time

model to first estimate the capacity of each individual loading area and then treats

these capacities with efficiencies of respective loading area to obtain final station

platform bus capacity.

2.7.3 TCQSM methodology of capacity calculation

Transit Capacity and Quality of Service Manual (TRB, 2003) is one of the most

comprehensive documents created out of a range of research studies dealing with

transit design and service.



Part 4 of the TCQSM (TRB, 2003), which deals with the bus transit capacity, defined

the dwell time as the average amount of time a bus is stopped at the station platform

to serve passenger movement, including the time required to open and close the

door. The manual suggests three methods for determining the dwell time. One of the

methods is to calculate the dwell time using the mathematical formula as given in

Equation 2.8. The other two methods are field measurement and use default values

respectively. This equation is identical to Vuchic’s equation for dwell time (Equation

2.7). The capacity of individual loading area can be determined using Equation 2.19.

For a platform configuration where more than one Loading Area (LA) is provided, as

in the cases of a busway station platform, the platform bus capacity can be equal to

the summation of the capacities of all the loading areas (Equation 2.20).

Equation 2.20

Equation 2.21

Equation 2.20 provides the theoretical capacity of the platform area. However, where

more than one loading areas are provided, the effective loading area would always

be less than the total integer value of such areas, because of the variation in dwell

time amongst loading areas and the possibility of blocking of front loading area by

preceding loading area/s. Additionally, the effectiveness also depends on the type of

loading area, off-line or on-line (Figure 2.15). The off-line loading area allows buses

to pull out of the line to get on to the loading area, providing overtaking opportunity to

the buses which do not have to stop at the station. It therefore reduces the station

impact on the travel time of certain buses like express bus which does not have the

schedule stop at the station but just have to pass it. On the contrary, at the on-line

loading area, the flow of the traffic gets blocked by the bus serving to the

passengers. Due to this, the number of effective loading areas for an on-line facility

is further reduced. Generally, busway systems have off-line loading facility. The on-

line loading facility can be seen predominantly in the CBD area of the city due to the

scarcity of the road space. For the busway situation where the loading areas are

Literature Review


typically off-line, Table 2.5 provides the variation of efficiency and effective loading

area with increase in the number of physical loading areas. The platform bus

capacity, therefore, is the product of loading area bus capacity and total effective

loading areas of the platform (Equation 2.21)

Table 2.5: Efficiency of multiple offline linear loading area at bus stops

Loading area

number

Efficiency %

Cumulative No. of effective loading

areas

1 100 1.00 2 85 1.85 3 80 2.65 4 65 3.25 5 50 3.75

Source: TRB, 2003

(a) Off-line loading area (b) On-line loading area

Photo by: Jaiswal (2009) Figure 2.15: Examples of loading area

The off-line loading area efficiency factors (Table 2.5) are based on the experience

at the Port Authority of New York and New Jersey’s Midtown Bus terminal (TRB,

2003) published in NCHRP Report 155 (Levinson et.al, 1975).

2.8 Gaps in the knowledge Single most important finding of this literature search is that there is no methodology

specifically developed for busway capacity estimation. The present methodology

given by TCQSM (TRB, 2003) (Section 2.7.3) was designed from the principles of



simple bus stops and bus lanes and is not based on the operation of busway station.

For instance, the present dwell time equation can describe delay caused to a bus at

a simple bus stop but could be less accurate for a busway station where multiple

loading areas are available. The loading area bus capacity relies on the bus dwell

time model, which does not consider appropriate multiple loading operation as well

as influence of platform crowding.

To accumulate the effects of presence of multiple loading areas, the TCQSM (TRB,

2003 based on Levinson et al., 1975) suggested the use of loading area efficiency

factors. However, the suggested efficiency factors are based on the operation of a

bus terminal facility and not based on the operation of busway station facility. The

busway station operation is described in chapter 3.

Based on these finding, Figure 2.16 shows gaps in the present methodology.

Figure 2.16: Gap in busway station bus capacity estimation approach

Specific findings, in relation to busway bus capacity methodology, for this review

include:

The weak link in present methodology is the bus dwell time model, due to its

insensitivity towards passenger walking, platform crowding and multiple

loading areas.

Boarding and alighting load

Boarding and alighting service time

Bus dwell time

Loading area capacity Loading area efficiency

Station capacity

Busway station operation characteristics

Gap in knowledge

Literature Review


The current dwell time models account for passengers only to the point of

their boarding and alighting times. In fact, the boarding and alighting service

themselves have some gap for local conditions.

The efficiency factors based on terminal operation may not reflect the busway

station operation.

General findings for this review include:

Study investigating the effect of busway platform crowding on passenger

service time vis-à-vis on vehicle dwell time has not been found, though

relevant studies related to airport terminals have been published.

Though TCQSM (TRB, 2003) provides the criteria for level of service for

queuing area, it does not relate the transit vehicle dwell time with a level of

service measure.

Though many models for pedestrian flow study were found in the literature,

these models are mainly stand-alone models and are not integrated with

vehicle/ dwell time models.

Based on these finding the research problem was developed, which is discussed in

the next chapter.

Research Problem Development


Chapter Three


3.1 Overview Previous chapter identified the lack of a purpose built methodology for busway dwell

time and capacity estimation. The shortcomings of using the traditional bus stop

dwell time methodology for busway station operation were described. Therefore it is

necessary to first study the operation of a bus stop and a busway station and identify

the differences between them. To start with, preliminary visual observations were

carried out at some of Brisbane’s busway stations including Cultural Centre, South

Bank, Mater Hill and Eight Mile Plains. Similarly, two arterial road bus stops in

Brisbane, at Coorparoo West and Mt. Gravatt Showground, were observed.

The aim of this exercise is to develop the research problem from the gaps

established in Chapter 2 and identify the parameters needed to model bus dwell time

at a busway station. This chapter describes the analysis of preliminary observations.

Section 3.2 presents the difference established here between busway station and

bus stop vis-à-vis passenger boarding process and operation. The research problem

is conceptualised in Section 3.3. Section 3.4 describes the busway operation at four

levels: platform, vehicle, station, and line. The definition of terms identified in this

research is provided in section 3.5 and section 3.6 concludes the chapter.

3.2 Difference in bus stop and busway station operation Although a busway station and a standard kerbside bus stop facilitate passengers

boarding and alighting, there are considerable differences between these facilities.

The most visible difference is in their size. Busway stations are longer and with

multiple loading areas compared to shorter suburban bus stops usually with single or

sometime two loading areas.



3.2.1 The size At a typical bus stop, typically one marked loading area is provided; hence

passengers know the stopping position of the bus entry door with higher certainty

and are generally positioned within half a bus length of the front door. Waiting

passengers, therefore often align themselves accordingly well before the bus comes

to a stop. Hence the passenger walking to bus entry door plays a minimal role in the

bus dwell time at the bus stop. The dwell time of a bus at a bus stop can therefore be

generally been defined as a function of passenger demand and service time per

passenger, separately for alighting and boarding. However, at least three loading

areas are provided at busway stations in Brisbane. This creates uncertainty in

passengers’ minds about the loading area to be occupied by the desired bus and

hence requires passengers walk to the bus entry door after the bus comes to a halt

at the loading area. This therefore makes passenger walking more significant in the

bus passenger servicing process.

3.2.2 The demand Another important factor that differentiates a busway station from a bus stop is the

number of bus routes servicing the station. In Brisbane, a typical bus stop serves

between one and five bus routes. However, for example the Mater Hill Busway

station, being a mainline station, serves over 40 separate routes including a number

of Bus Upgrade Zone (BUZ) high frequency spine services. Since the number of

routes at a typical bus stop is far less, the passenger route groups are far less

diverse. Hence at a bus stop with more uniformity in passengers’ directional

behaviour the passengers are of a common route group. However, at a busway

platform the number of bus routes can be very high. Therefore the passenger route

groups at the busway station can be much higher, leading to crowding. Such

platform crowd density at a busway station platform acts as an obstruction to the

passenger’s walking path (TRB, 2003) and also obstructs the passenger’s line of

sight, resulting in a longer passenger – bus interface. Figure 3.1 compares observed

passenger densities at a bus stop and a busway station.



a) Bus stop b) Busway station platform Source: Jaiswal (2009)

Figure 3.1: Concentration of passenger crowding

3.2.3 The passenger boarding process Observing passenger boarding activities at busway stations and bus stops, the

passenger servicing process can be divided into four distinct phases. The first phase

is the initial reaction of hailing when the passenger first sees their desired bus. The

second phase consists of walking to the bus entry door. The third phase consists of

queuing at the entry door and fourth phase is boarding the bus.

The first and second phase of the boarding process are considered here to comprise

the ‘passenger – bus interface’ stage. The passenger – bus interface starts when

the passenger first sees the desired bus and hails the bus operator and/or starts

walking towards a point by anticipating its stopping location. Similarly, the bus driver,

after seeing the hailing passenger, prepares to stop the bus at the closest available

loading area to the “lead stop” loading area. During this course of action, both the

bus operator and the passenger act independently but anticipate each other. The

third and fourth phases of the boarding process are considered here to comprise the

‘passenger – bus interaction’ stage. However, when there are only one or two

passengers boarding the bus, queuing may not occur and the passenger – bus

interaction stage may consist of only boarding.

On the contrary, at a simple bus stop since the passengers often align themselves

before the bus arrival, as discussed in section 3.2.1, the boarding process essentially

consists of initial hailing, queuing at entry door and boarding with walking component

been minimal or negligible. With respect to these actions, Table 3.1summaries the

difference between busway station and bus stop boarding processes.



i O

D

Table 3.1: Boarding process at a bus stop and at a busway station

Action Bus stop Busway Station 1 Initial reaction to bus / hailing Yes Yes 2 Walking Minimal Can be substantial 3 Queuing Can occur Can occur 4 Boarding Yes Yes

3.3 Problem conceptualisation The difference between boarding a bus at a bus stop and boarding a bus at a

busway station is, therefore, the requirement of a substantial amount of walking on

the busway platform to reach the bus entry door. The nature of this walking can be

illustrated by considering the typical path of passengers at the busway platform from

when they arrive until when they board their desired bus (Figure 3.2).

Where, O Origin D Destination i Information signage Trip route

W Chosen waiting point

Figure 3.2: Origin and destination of a trip segment at platform

Often after entering the platform passengers require to wait for their desired buses to

arrive at the station. These waiting passengers, hence, results into crowding and

affect others walking manoeuvres. Accordingly, the ease in completing walking

segment between waiting (point W) and the bus entry door (point D) depends on the

prevailing passenger density at the platform.

As the passenger density increases it is intuitive to expect that the walking speed

decreases, and therefore the walking time increases. Therefore the time spent by the

passenger to walk from their waiting (point W) to the bus entry door (point D) is likely

to influence the passenger service time for the bus. Higher crowding not only

reduces passengers’ manoeuvrability but also hinders their line of sight to

approaching buses. This increases their reaction time (hailing) to the bus arrival. On

W



the whole, the delay in walking and the time spent in walking, and therefore, the

dwell times, may vary in a manner somewhat proportional to the number of persons

in the crowd who are encountered by the passenger on the way to the bus door.

3.4 Busway operation The sequential impact of the passenger boarding process across the busway

operation can be considered at four levels in a hierarchy as shown in the Figure 3.3.

The figure shows the flow of effects from platform level to network level.

Figure 3.3: Different levels of busway operation

At the platform level: The busway station platform has multiple loading areas,

which necessitates passengers to walk to the bus entry door. As explained before,

the requirement of walking induces an interface of passenger and bus where the

passenger walks with a sense of uncertainty about possible stopping point of bus.

The time required in completing this task could result into Lost Time (LT) for a bus,

which ultimately affects its dwell time. The passenger – bus interface ends when the

passenger enters into the queue at the bus entry door or boards the bus, whichever

occurs first. The part of interface which occurs after the bus has stopped at the

loading area and doors are opened results in bus lost time. Figure 3.4 shows a

LINE Station to station interface Bus Bunching

Bus Clearance / Bus queuing Bus to bus interface STATION

PLATFORM

VEHICLE Passenger – Bus interaction

Passenger – Bus interface Passengers walking

on platform

Boarding & Alighting

Effect Level Process



passenger – bus interface state of a passenger with the desired bus arriving at

loading area 3 (shown in solid circle). Figure 3.5 shows a bus accumulating lost time

because of a passenger – bus interface (shown in dashed circle).


Figure 3.4: A passenger – bus interface phase at a busway station


Figure 3.5: A bus lost time phase at a busway station



At the vehicle level: Lost time, as the term suggests, forces the bus to occupy the

loading area on the platform waiting for the arrival of its first passenger. Hence the

bus accrues the dwell time in its travel time but without any processing of boarding

service. Therefore the Lost Time (LT) is considered here to be incremental to bus

dwell time. The bus dwell time has previously been considered as the summation of

boarding and alighting times for passengers plus door opening and closing time

(TRB, 2003). Therefore the amount of time a bus spends at the busway platform

could be better represented as –

Equation 3.1

Where,

= Bus dwell time

= Bus lost time

= Passenger processing time

= Door opening and closing time

It is worth noting again that the per passenger processing time i.e. boarding and

alighting time per passenger is sensitive to the fare collection system and policy

(TRB, 2003; Milkovits, 2008).

At the station level: With a busway station platform designed to have an

arrangement of off-line linear loading areas, buses stop at the platform to service

passengers one after the other in close proximity. In some cases this may lead to

insufficient space for buses to leave a loading area after boarding and alighting

passengers have been processed and doors closed. This results in an increase in

clearance time for buses under these circumstances. Such increase in clearance

times coupled with increased dwell times reduces station bus capacity, and affects

queuing of incoming buses. Typical busway line design includes an off-line loading

area lane and passing lane in each direction at stations. When bus queues build up

at a station, the through lane upstream will become blocked. Any through buses,

which are not scheduled to service the station, can hence be delayed along with

buses servicing the station. This then influences bus route travel times.



At the line level: Queuing of buses at the station entry makes platform servicing

simultaneous rather than random. When multiple platform loading areas are used by

buses simultaneously with insufficient clearance space, the buses tend to leave the

station in bunches. Hence they will tend to arrive at downstream stations in bunches.

This process is therefore a compounding one, and here is referred to as station –

station interface. It may be further compounded by any signalised intersections either

to control conflicting bus movements at access intersections, as is the case

immediately outwards of Mater Hill station, or to separate conflicting bus and general

traffic movements on at-grade signalised intersections, as is the case on the South

East Busway two stations inwards of Mater Hill station (and therefore upstream of

the platform under study). Intuitively, bus bunching would tend to improve the

efficiency of loading area utilisation at a busway station, hence increasing bus

capacity. Data in the Transit Capacity and Quality of Service Manual (TRB, 2003)

supports this. However, observations from this study suggest that when buses arrive

together on the station platform, the passenger – bus interface is amplified, mainly

because the passengers have less certainty about the bus arrival sequence and the

stopping location of their desired bus. The increased uncertainty can delay

passengers’ reaction to their desired bus. This may lead to an increase in the lost

time component of dwell time, and hence a reduction in bus capacity.

3.5 Definition of terms Three new parameters were identified which explains the distinct operation of

busway and its stations. These parameters are technically defined below –

Passenger – bus interface: The ‘passenger – bus interface’ (IF) is the phase where

the first passenger and the bus driver are involved in a state when they both interact

but perform their respective activities independently.

Passenger – bus interaction: The ‘passenger – bus interaction’ (IA) is defined as

the phase where the first and subsequent passengers and bus driver are involved in

a state when they both interact and perform their respective activities based on each

other’s position or action.



Bus lost time: The ‘lost time’ (LT) is defined as the time lapse between bus stopping

time and the time first passenger puts his/ her foot at the bus floor.

3.6 Conclusions Discovery of the passenger walking component, particularly walking the distance

between the passenger’s original waiting point and bus entry door, has enhanced

knowledge of bus dwell time at a busway station. The literature on pedestrian flow

characteristics clearly established that walking speed reduces under high density

situations. Therefore the hypothesis of this research is –

“Passenger walking and the prevailing crowd at a busway station platform influence

the bus dwell times”.

After establishing the hypothesis the next step was to collect that data at busway

station to test the research hypothesis and model the busway station bus dwell time.

The data related to bus dwell time as well as pedestrian flow behaviour, pedestrian

walking, and pedestrian crowd behaviour were collected at the study busway station.

The details of study station selection and data collection methodology are described

in chapter 4.

Data Collection and Processing


Chapter Four


4.1 Overview This chapter presents the data collection methodology used in this research to

capture the effect of bus lost time on bus dwell time. The chapter starts with a

discussion on different data collection techniques used in past research studies. This

was followed by section 4.3 which provides details of data collection technique used

in this research. Section 4.4 outlines the various methodologies applied for the

purpose of this research. The selection criteria of study station and characteristics of

selected busway station are discussed in section 4.5 and section 4.6 respectively.

Section 4.7 details data collection sequence and section 4.8 discusses about

processing of the collected data.

4.2 State of art in relevant data collection technique This research needed to examine all variables which may explain the effects of

platform crowding, influence of walking to bus entry door, and effects of multiple

loading areas, on delay experienced by buses. Therefore the data required was

multi-dimensional: bus flow and passenger flow, both in time and space. With the

aim of developing an appropriate data collection and processing technique, a review

of techniques used in the past studies related to pedestrian flow and bus stop

operation was performed. Table 4.1 gives the overview of collection techniques used

in past studies.



Table 4.1: Data collection techniques used in past studies

Data Facility Technique used Author

Boarding and alighting

Bus stop Manual count Guenthner and Hamat, 1988

Bus stop Manual count Kim (2004)

Bus route Automatic passenger counter

Rajbandari (2003)

Pedestrian Crowd

Railway station, London, UK

Video recording / image processing

Davies et.al., 1995

Walkway Video recording / image processing

Kilambi et.al, 2008

Walkway Video recording / Neural network

Kong et.al. 1995

Place of worship Video recording / image processing

Johansson et.al., 2008

Pedestrian walking

Walkway Video recording / manual processing

Antonini and Bierlaire, 2005

Road crossing Video recording / manual processing

Antonini and Bierlaire, 2005

Signalised crossway

Time-lapse photography

Lam et.al, 2002

Walkway Video recording / image processing

Teknomo et.al., 2001

Pedestrian flow behaviour

Narrow bottleneck

Lab experiment Seyfried et.al., 2007

Narrow bottleneck

Lab experiment Daamen (2004)

Four-directional crossing flow

Lab experiment Daamen (2004)

The data collection techniques in Table 4.1 are now explained.

Manual count: Manual counting and field studies are the most common approaches

to collect data related to passenger boarding and alighting from transit vehicles

(Guenthner and Sinha, 1983; Levinson, 1983; Kim, 2004). However, costs involved

in manual collection of data, as observed on Milkovits (2008) limits the number of

observations to a handful of stops, operators, and times of day. Moreover, manual



data collection techniques are laborious and may generate errors (Dueker et.al.,

2004).

Automatic counter: An automated data collection system can provide rich data sets

across time of day, stop, route, operator and other significant but rare events such as

lift operation (Dueker et.al., 2004; Rajbandari et.al., 2003). However this technique

can onle be used where transit vehicles are fitted with appropriate sensors such as

automatic passenger counter (APC), automatic fare counter (AFC) and automatic

vehicle locater (AVL). Maikovits (2008) used automatic passenger counter system

installed on Chicago Transit Authority buses to count boardings and alightings to

study bus dwell times.

Time-lapse photography: Rime lapse photography is where a camera, typically in

a fixed position, automatically records a sequence or series of photos with a set time

interval between each image. Individual images may then be analysed in the

laboratory. This technique was first used to investigate the gap acceptance

behaviours of pedestrians (Dipletro and King, 1970). Lam et.al. (2002) used time

lapse technique to study bi-directional pedestrian flow characteristics at a signalized

crosswalk facility in Hong Kong.

Video recording: Past literature shows that vision based techniques such as video

recording is the preferred method of data collection for pedestrian crowd studies. In

this technique a CCTV camera is used to record the events that are later analysed in

the laboratory. Often analysis of video footage was carried out by image processing

algorithms (Davies et.al., 1995) to extract the desired variables. Kilambi et.al. (2008)

presented two methods, one based on heuristic learned during training and another

based on shape models, to estimate the crowd size. Johansson et.al. (2008) used

the video recording technique to study stop-and-go wave and crowd turbulence

phenomena at a high density location.

Experimental: In real life situations pedestrian flow characteristics may be

influenced by factors which are beyond the investigator’s control. To isolate the

effects of such factors on pedestrian walking behaviours, Daamen (2004) carried out

walking experiments in a controlled environment to study uni-directional flow, bi-

directional flow, crossing flow and bottlenecks. The experiment was recorded using



fixed video cameras and later analysed. Seyfried et.al. (2007) used the experimental

study and video recording to test the validity of pedestrian specify flow concept with

respect to bottleneck situation.

4.3 Technique used for data collection in this study The aim of this observation is to gather evidence of how bus dwell times at a busway

platform are influenced due to platform crowding and associated factors. It is

therefore important to record the attributes explaining the approach of the passenger

to the bus entry door. It is equally important to record the passenger flow, in order to

relate it to crowd formation and identify the location of crowding at the platform.

Observations related to the buses, such as service time at the platform, door

opening and closing time, and so on are also required to establish the density –

dwell time relation.

Because of the large numbers of passengers and bus flows at the subject station

platform during the study times, the data collection method needed to be selected

with care. On site manual counting can prove to be very laborious and may be

susceptible to high human error. Literature search on data collection techniques

(Table 4.1) established that video recording technique was most suitable for this kind

of study. Video recording at the station followed by laboratory counting can eliminate

much of the human and machine errors.

The study busway station platform has two CCTV cameras which record platform

activities on a continuous 24 hr basis. These cameras were used for this data

collection effort with the permission and help of TransLink’s Busway Management

Centre (pers. comm. Mr Andrew Haddock and Mr Jurgen Pasiezny). Additionally, a

QUT camera was installed at the front end on the platform to enhance the quality of

data. Figure 4.1 shows the positions of three cameras used in data collection.



Figure 4.1: Camera positions at Mater Hill Busway Station (Outbound platform)

4.4 Research methodology The methodology for this research was divided into three parts: data collection, data

extraction and data analysis. Figure 4.2 illustrates connections between these

methodologies

4.4.1 Data collection methodology

The following methodology was used to collect the data.

1. Identify a suitable busway station platform.

2. Establish camera angles to achieve the best possible view of passenger

movements on the platform.

3. Conduct video recordings.

4.4.2 Data extraction methodology

The purpose of data collection via video recording was to capture all the activities

occurring at the busway station platform area. These recordings were therefore

manually analysed in the laboratory to extract the key attributes explaining the

passenger – bus interface phenomena. The data extraction methodology was

designed to mine the video footage for bus attributes (i.e. bus dwell time, bus

queuing time and bus clearance time) and passenger attributes (i.e walking time,

QUT Camera 3 CCTV 1 CCTV 2



queuing time, etc). For measuring bus attributes, the guidelines from the Transit

Capacity and Quality of Service Manual - Part 4: Bus Transit Capacity (TRB, 2003)

were followed –

1. Record the bus route number and its queue ingress time (if any)

2. Record its queue egress time (if any)

3. Record the loading area number on which it is serves passenger.

4. Record the time at which the bus comes to a complete halt.

5. Record the time of full opening of the bus front door.

6. Count the number of alighting passengers separately from the front and rear

door and number of boarding passengers onto the front door.

7. Record the timing for first and last passenger alighting.

8. Record the timing for first and last passenger boarding.

9. Record the time of full closing of the bus front door.

10. Record the time when bus starts leaving the bay.

Since analysis of the passenger – bus interface has not been done before; the steps

of recording the passenger side data were derived from the concept explained in

section 3.3. The following steps were involved in the measurement –

1. Select a passenger on the platform and tag it (say xi).

2. Record the time of passenger xi’s reaction to desired bus.

3. If passenger xi needed to queue at the bus entry door, due to the boarding

passengers in front, or any passenger alighting from the bus, then record the

queue entry time for that passenger.

4. Record the time when passenger xi boarded the bus.

5. Record the number of passengers on the platform who were crossed or

passed by that passenger xi

6. Record the total number of passenger present on the platform during this

time.

4.4.3 Data analysis methodology

1. Analyse bus dwell times with respect to the loading area.

2. Analyse passenger walking times with respect to the loading area.

3. Analyse passenger – bus interface and its effect on bus dwell time.

4. Develop a bus lost time profile for platform.



Figure 4.2: Processing of data – from collection stage to analysis stage

4.5 Selection of study station For the purpose of this research the busway station platform should meet following

desired criteria:

1. Presence of crowd at platform.

2. Variation in passenger demand thought out the day.

3. Dominance of boarding passengers during at least one of the peak period.

The South East busway in Brisbane, Australia is a16 km corridor with 11 dedicated

stations. Figure 4.3 shows the busway corridor and its stations. Each busway station

has one platform in each inbound (to city) and outbound (from city) direction. The

characteristics of the corridor are such that, during the morning peak, flow of

passengers toward the city is high, contributing to high numbers of passengers

alighting at the inner inbound platforms. This situation reversed during the afternoon

and evening peaks when there are more boarding passengers at the outbound

platforms of the inner stations. Therefore, during the evening peak period, crowding

of boarding passengers can be observed at outbound platforms of the inner busway

stations.

Data collection Data extraction Data analysis On-site In-lab In-lab

Dwell time Bus data

Video recording

Platform Bus Capacity Passenger data

Passenger Walking Time



Figure 4.3: Brisbane’s South East Busway route map

Based on the above listed desired criteria, three busway stations qualified for this

research study. Table 4.2 provides the characteristics of the outbound platform of

these three candidate stations.

Eight Mile Plains Station (16.00 km)

Queen Street Station (0.00 km)

Cultural Centre Station (0.67 km)

Mater Hill Station (2.47 km)

South Bank Station (1.67 km)

Woolloongabba Station (3.17 km) Buranda Station

(4.37 km)

Greenslopes Station (5.97 km)

Holland Park West Station (8.57 km)

Upper Mt Gravatt Station (13.38 km)

Griffith University Station (10.77 km)

Note - Number in bracket shows the distance of the respective station from Queen Street station.



Table 4.2: Candidate busway station (outbound) platforms

Culture Centre South Bank Mater Hill

Photo

Platform length 60m 60m 45m

Number of bus loading area

Four Three Three

Weekday boardings 5000* 2775* 3122*

Weekday alightings 2215** 829** 785**

* Total number of boarding on a weekday Source: TransLink, 2007 ** Total number of alighting on a weekday Photo by: Jaiswal (2009)

Among the candidate stations, the Cultural Centre Busway Station has highest

number of boardings; however, this station also acts as a terminal station for some

buses. Those buses terminating have dwell times due to alighting passengers only.

Since the aim of this research is to investigate the effect of platform crowding on bus

dwell time, and because a terminating bus will have less influence because of

platform crowd, Cultural Centre Busway Station was rejected. The second highest

boarding load was observed at Mater Hill Busway Station. Moreover, Mater Hill

busway station (outbound platform) also has a dominance of boarding passengers

on the platform. The outbound platform of Mater Hill Busway station was therefore

selected for this study.



4.6 Characteristics of Mater Hill Busway Station From Brisbane city, Mater Hill station is the fourth station on the 16 km long South

East busway corridor. Mater Hill Busway station now has three signed and striped

loading areas on its outbound platform as shown in Figure 4.4. Infrequently some

bus operators pull up very close to the dwelling bus ahead of them, thereby creating

a transient fourth loading area. This was much more prevalent prior to the signing

and striping of three loading areas, which occurred midway through this overall

project. This station is located inwards of the confluence of four common-lines; the

mainline South East Busway, the Woolloongabba spur common-line, the Pacific

Motorway access ramps common-line, and the Annerley Road common-line. Hence,

there is a considerable level of passenger demand for this station.

In Brisbane, a passenger can board the bus from the front door only. However, an

alighting passenger can use either the front or rear door. During the annual analysis

periods over the three years of the study, four methods or combinations thereof were

available for passengers to validate their journey. These included two manual

means, being onboard ticket purchase from the bus operator or presentation of a

pre-paid paper ticket. The earlier of two automated means was the use of the 10 trip

saver ticket, a magnetic stripe card dipped on entry into one of two readers located

inside of the front door only. The later of two automated means was the use of the

GoCard smart card, with each bus equipped with four readers; two readers inside

the front door for touch-on and touch-off, and two readers inside the rear door for

touch-off only.

In 2009 TransLink introduced the pre-paid platform policy for the outbound platforms

of three innermost busway stations on the South East busway. Under this policy,

during the busy outbound mid afternoon to evening peak period, all passengers must

have a pre-paid ticket or a GoCard to enter the outbound platform of each of the

Cultural Centre, South Bank, and Mater Hill Busway stations, as no on board ticket

purchasing is permitted. The rationale being to minimise bus dwell times and

therefore improve bus capacity and reduce delays. The details of recordings and fare

policies on the recoding days are given in Table 4.3.



Figure 4.4: Configuration of Mater Hill Busway Station

Table 4.3: Fare collection policies at Mater Hill Busway Station

Month/ Year March 2007 March 2008 April 2009

Fare policy

On board ticket purchase


Pre-paid ticket Pre-paid ticket Pre-paid ticket

10 trip magnetic stripe card into front

door dip readers

10 trip magnetic strip card into front door dip readers

(Phasing out)

GoCard smart card with onboard touch on using readers front door only & touch off using

readers at front and rear doors

(Introduced)

GoCard smart card with onboard touch on using readers front door only

& touch off using readers at front and

rear doors

Pre-paid platform policy

Source: TransLink

Lead Stop L A 2 L A 3

Lead Stop L A 2 L A 3

INBOUND PLATFORM

OUTBOUND PLATFORM

Station entry/ exist

point

Tunnel

Tunnel

Station entry/ exit point Bus direction L A Loading area



4.6.1 Passenger flow at station

Mater Hill Busway station users include school and university students, hospital

employees, patients and visitors, and employees of surrounding businesses. The

peak for inbound platform occurs in morning time and mostly comprise of alighting

passengers (Figure 4.5). The alighting passengers quickly move out of the platform

and hence inbound platform experiences no significant crowding. The outbound

platform has its peak in evening time (Figure 4.6). The outbound platform is more

dominated by boarding passengers. It experiences two non-continuous evening peak

periods (Lee, 2004) with in the peak period. The first afternoon peak occurs following

release of students from two adjacent schools, and the second, evening peak occurs

when hospital, university and other workers commute home.

Source: TransLink, 2007

Figure 4.5: Boarding and alighting at inbound platform of Mater Hill Busway Station

0

100

200

300

400

500

600

5-6

am

6-7

am

7-8

am

8-9

am

9-10

am

10-1

1 am

11-1

2 am

12-1

pm

1-2

pm

2-3

pm

3-4

pm

4-5

pm

5-6

pm

6-7

pm

7-8

pm

8-9

pm

9-10

pm

10-1

1 pm

11-1

2 pm

Num

ber o

f pas

seng

ers

Time of day

Alighting Boarding




Figure 4.6: Boarding and alighting at outbound platform of Mater Hill Busway Station

4.6.2 Bus flow at station

Figure 4.7 shows the scheduled bus flow rate through the study station. This does

not include unscheduled buses such as blank run to depot or terminus (usually in the

counter peak direction) and some express bus services which do not observe the

study station (although much services normally bypass this section of busway).

Buses servicing the study station can be divided in four groups in accordance to their

floorplan –

a. High floor rigid buses

b. Low floor rigid buses

c. High floor articulated buses

d. Low flow articulated buses

It is noted that a greater, but still moderate proportion of buses were articulated

during the later years on the study.

0

100

200

300

400

500

600

700

800

900

5-6

am

6-7

am

7-8

am

8-9

am

9-10

am

10-1

1 am

11-1

2 am

12-1

pm

1-2

pm

2-3

pm

3-4

pm

4-5

pm

5-6

pm

6-7

pm

7-8

pm

8-9

pm

9-10

pm

10-1

1 pm

11-1

2 pm

Num

ber o

f pas

seng

ers

Time of day

Alighting Boarding




Figure 4.7: Number of buses servicing outbound platform of Mater Hill Busway Station

4.7 Sequence of data collection Video footage of the outbound platform of Mater Hill Busway Station was captured at

a similar time of year in three successive years; March 2007, March 2008 and April

2009. This was to ensure that the data collection was free from noise due to school

or university vacation periods, public holidays etc. Each video data set was captured

on a Wednesday, being a typical midweek day. These data collection activities were

carried out with the assistance of the Translink Transit Authority’s Busway

Operations Centre in Brisbane. The passengers on the platform were unaffected by

the video data collection as permanent security cameras were used. These cameras,

mounted on the ceiling of the busway platform awning, record the movements of

passengers on the platform on a 24hr / 7 day basis.

Since it was not feasible to analyse the entire day’s operation, a matrix based

approach was developed to identify the time of day which could produce

homogenous data for analysis. To devolve this matrix, the data collected by

TransLink in November 2007 were examined (Figure 4.6 and Figure 4.7). The bus

0

20

40

60

80

100

120

140

5-6

am

6-7

am

7-8

am

8-9

am

9-10

am

10-1

1 am

11-1

2 am

12-1

pm

1-2

pm

2-3

pm

3-4

pm

4-5

pm

5-6

pm

6-7

pm

7-8

pm

8-9

pm

9-10

pm

10-1

1 pm

11-1

2 pm

Num

ber o

f Bus

es

Time of day



arrival rate and passenger demand were divided in three groups - Low, Medium and

High. The criteria for division are given in Table 4.4.

Table 4.4: Bus flow rate and passenger demand classification split

Group Bus arrival rate Passenger demand

Low < 40 bus/ h < 100 p/h Medium 40 to 80 bus/ h 100 to 200 p/h

High >80 bus/ h > 200 p/h

The derived matrix is given in Figure 4.8. On the horizontal axis, the bus arrival rate

in ascending order (from left to right) and their corresponding hours ending are

plotted. On the vertical axis, the passenger demands for the study station platform

their corresponding hours ending are arranged in ascending order (from top to

bottom). Based on this matrix, three different hours of the day, out of total 18 hours

of day’s operation, were selected for analysis. Each selected hour represents

particular flow characteristics as shown in Table 4.5.

Table 4.5: Characteristics of analysis time

Time of the day Flow characteristics Time period

10:00 am to 11:00 am Medium bus frequency and Medium passenger demand

Morning off-peak

03:00 pm to 04:00 pm High bus frequency and High passenger demand

Evening peak

07:00 pm to 08:00 pm High bus frequency and Medium passenger demand

Evening off-peak

No data set was analysed from the period when passenger demand was low,

because of a lack of statistically significant data. In total the data available for

analysis was comprised of 9 hours (March 2007, March 2008 and April 2009). Note

that the morning peak period (8:00 am to 9:00 am) was not considered for analyses,

because it is negligible for boarding passengers. Instead, the evening off-peak

period (6:00 pm to 7:00 pm) was selected to have diverse set of data for analysis.

The selected hours for data analysis are shown in Figure 4.8 with thick borders.



2 28 29 34 38 42 47 66 74 75 75 75 76 81 83 85 109 115 121

6 7 24 23 22 21 20 8 12 11 13 14 10 15 9 19 16 17 18

0 6 628 24 2439 23 2350 7 758 22 2260 21 21

101 20 20105 10 10136 19 19138 11 11152 13 13159 14 14165 12 12171 8 8180 9 9200 15 15255 18 18325 17 17776 16 16

Observed bus arival rate (bus/hr)

Corresponding hour ending at

Classification splitLow Medium High

Low

Med

ium

H

igh

Cla

ssifi

catio

n sp

lit

Cor

resp

ondi

ng h

our e

ndin

g at

Obs

erve

d pa

ssen

ger d

eman

d (p

/hr)

Note: 6 mean hour ending at 6:00 am, similarly, 17 means hour ending at 5:00 PM

Figure 4.8: Matrix for data mining of passenger demand and bus frequency

Null set

Null set

Null set

Null set



4.8 Data processing For each hour of recording, footage was generated from each of three cameras.

These footages were then processed and synchronised in the laboratory using

Microsoft Movie Maker software. Later, each of the three footages were played

simultaneously on separate computers and manually examined to extract variables

in accordance with the methodologies described in section 4.4.

For each bus servicing at the platform, its first boarding passenger was tracked from

the time they first reacted to the bus until the time they boarded that bus. A

passenger is considered as boarded when no part of his/ her body is out of the bus.

Furthermore, for each one hour time period, 50 passengers randomly selected from

the platform crowd were observed, from the time when they first reacted to their

desired bus until the time when they boarded their bus.

4.9 Chapter close This chapter described the developments of methodologies for data collection,

extraction and analysis. Video recording technique was chosen for the purpose of

data collection, as it provided the opportunity to revisit the footage multiple times and

extract the variable as and when needed.

For this research outbound platform of Mater Hill busway station of Brisbane’s South

East Busway network was selected. The platform has very high amount of boarding

passenger load which leads to platform crowding. This situation is appropriate to

study effect of platform crowd and linear loading area on the bus dwell time and

station operation.

In the next chapter, the data was analysed to evaluate various parameters and

processes identified in chapter 3.

Parameter Analysis and Evaluation


Chapter Five


5.1 Overview This chapter evaluates the two vital parameters of busway station operation. These

two parameters were identified in chapter 3 as: passenger – bus interface, and bus

lost time. The aim of this chapter is to analyse these parameters based on platform

crowding. The outcome of this analysis will be used as inputs for modelling of bus

lost times and bus dwell times at a busway station.

5.2 Measuring platform crowd For this study the crowd at the study platform was measured as the average number

of people waiting on the platform in a 15 minute interval. Crowd density was not used

as a variable because it was not consistent across the platform. For instance, the

crowd density at middle of the platform under low crowd conditions can be same as

that of crowd density of the entire platform area under high crowd conditions.

Therefore the use of density may not properly reflect the total crowd level variation

between periods. A detailed discussion on passenger behaviour and crowd

concentration on the station platform is presented in Section 5.4.

5.3 Passenger – bus interface By definition, the duration of passenger – bus interface is made up of time taken by

passengers to reach the bus entry door from their waiting position having hailed the

bus (Ref: Figure 3.2).

Video recordings were analysed to determine how passenger – bus interface occurs

and to study how platform crowding influences it. For each bus, randomly selected

passengers were manually tracked from the time they first initiated walking after

seeing the desired bus (event 1) to the time when they entered the boarding queue

(event 2). Note that, if they are the first or only passenger, this is just the time when



they boarded the bus. The time stamps of both events were recorded. Therefore,

mathematically the duration of passenger – bus interface can be estimated as –

, , Equation 5.1

Where,

= Duration of passenger – bus interface for passenger, i

, = Time stamp of passenger ‘i' first initiated the walking.

, = Time stamp of passenger ‘i' joined the boarding queue at entry door.

The video recordings from March 2008 were used to analyse the passenger – bus

interface. Table 5.1 provides descriptive statistics of two off-peak periods examined

(10:00 am - 11:00 am and 7:00 pm – 8:00 pm). Table 5.2 provides the descriptive

statistics of peak period (3:00 pm - 3:40 pm). During the off peak period, the

observed number of passengers on the platform in a 15 min interval was between 10

and 27. During the peak period, the observed number of passengers on the platform

in a 15 min interval was between 40 and 67.

Table 5.1: Duration of passenger – bus interface during off-peak period

Duration of Passenger – bus interface

Loading area 1 Loading area 2 Loading area 3

Min 7s 2s 7s

Mean 16s 11s 14s

Max 29s 18s 22s

Std dev 3.2s 3.2s 4.3s

Boarding pax observed 70 27 18

Total boarding pax 165 106 38

Total bus services 57 36 15

Note: Off – peak operation; passengers on platform < 30; total analysis time = 120 minutes



Table 5.2: Duration of passenger – bus interface during peak period

Duration of Passenger – bus interface


Min 1s 3s 6s

Mean 19 s 13 s 11s

Max 31s (51s)^ 25s 38s

Std dev 9s 4.4s 9s

Boarding pax observed 32 22 32

Total boarding pax 149 174 200

Total bus services 12 15 15

Note: Peak operation; # of passenger on platform ≥ 30; total analysis time = 40 minutes

^ In this outlier a passenger was standing at the lead end of the platform and observed the desired

bus stopping at rear loading area (loading area 3). However, before he reached the desired bus at

loading area 3, loading areas 1 and 2 were cleared and the bus driver moved up to loading area 1

causing this passenger to walk back to loading area 1. This resulted in a very high passenger – bus

interface. Such extreme durations were less frequent at the study station and were therefore not

considered further.

Large variations in the duration of passenger – bus interface, both in off-peak period

and peak period, were observed at the platform across all three loading areas. The

passenger – bus interface shows increments in all statistical parameters when the

time period changes from off-peak to peak operation, with the exceptions of

minimum duration of passenger – bus interface for loading area 1 and loading 3.

It was also noted that, during the off-peak period, the minimum duration of

passenger – bus interface was identical for loading areas 1 and 3. However, while

the minimum duration of passenger – bus interface for loading area 1 decreased by

6s, the reduction for loading area 3 was only 1s. In the off-peak the loading area 1

has the higher value of maximum duration of passenger – bus interface compared to

loading area 3. This situation was reversed during the peak period where the

maximum value for duration of passenger – bus interface was observed for loading

area 3.



In order to study the spread of passenger – bus interface, the duration of passenger

– bus interface was averaged over prevailing platform crowding and plotted against

the platform crowd (Figure 5.1).

Figure 5.1: Variation in passenger – bus interface

The Figure 5.1 provides some vital information about the busway station operation

under different prevailing crowd levels. Firstly, a clear distinction between the low

crowd operation and the high crowd operation can be seen in the figure. The

observed duration passenger – bus interface (averaged) formed two well separated

clusters representing off-peak and peak operations.

Secondly, considerable spread in passenger – bus interface is noticeable with

change in crowd level. It can be observed that the spread in the duration of

passenger – bus interface was lesser when the crowd at the platform was relatively

small. As the crowding increased, the passenger – bus interface became more non –

uniform and the spread considerably larger. This means, under increased crowd

conditions, while some passengers have a reduced passenger – bus interface others

have very high interface duration. Low interface duration occurs when the bus stops

close to the passenger’s waiting position, which means passenger walks a shorter

distance. High duration of interface may occur because, as the crowd size increases,

0

5

10

15

20

25

30

35

40

0 10 20 30 40 50 60 70 80Avg

dur

atio

n of

pas

seng

er -

bus

inte

rfac

e (s

)

Passengers on platfrom

LA 1 LA 2 LA 3

Lower spread High spreadHigh crowdingLow crowding



a passenger’s walking speed decreases. The passenger can no longer use the

straight path between the waiting point and the bus entry door, but is forced to

undertake a zig-zag manoeuvre to complete the path. This zig-zag manoeuvring also

increases the distance and thereby increases the walking time.

This analysis (Figure 5.1), however, failed to establish any clear mathematical

relationship between passenger – bus interface duration and platform crowding.

Nevertheless, it clearly illustrated the unevenness of the passenger – bus interface

under the high crowd condition. The analysis suggests that the study busway station

platform operates in two different operational paradigms. During the low crowd and

coinciding low bus flow operation, the passenger – bus interface appears to be

relatively stable and varying with a standard deviation of 4s. On the other hand, the

standard deviation is doubled to 8s, during the high crowd and high bus flow

operation. The duration of passenger – bus interface is more unstable and less

predictable. This instability can cause high variability in bus dwell time. This finding

is important as it confirms that platform crowd affects bus dwell time.

The focus of analysis, therefore, moved to low crowd operation to examine the

correlation between duration of passenger – bus interface and platform crowd, if any.

The plot of average duration of passenger – bus interface and platform crowd is

given in Figure 5.2.



Figure 5.2: Variation in passenger – bus interface during off-peak period

The result shows that duration of passenger – bus interface, generally, increases

weakly with an increase in platform crowding. However, the graph suggests that

there is no strong two dimensional relationship between duration of passenger – bus

interface and platform crowding. The R2 values for all three loading areas shows that

linear regression is not able to explain any correlation.

5.3.1 Discussion of passenger – bus interface The inability to draw a robust mathematical relationship between passenger – bus

interface and platform crowd, and the number of passengers encountered en route,

indicates significant complexity amongst all variables. The duration of passenger –

bus interface depends on the passenger’s walking distance to cover, speed and

straightness of walking path. The straightness of the passenger’s walking path,

indeed, depends on the number of other passengers in the path. However, number

of passenger encounters depends on platform crowd density. Crowd also influences

the passenger’s choice of waiting point. The walking distance for passenger, on its

part, is subjective to its waiting position on the platform relative to the loading area.

Moreover, the relative positions of loading area (used by bus) and waiting point, in

space, decide the passenger’s walking direction with respect to the direction of bus

arrival. The effects of walking direction on passengers bus interface is discussed

R² = 0.0364 R² = 0.1096 R² = 0.0521

0

5

10

15

20

25

0 5 10 15 20 25 30Avg

dur

atio

n of

pas

seng

er -

bus

inte

rfac

e (s

)

Passengers on platform

LA 1 LA 2 LA 3 Linear (LA 1) Linear (LA 2) Linear (LA 3)



latter in Section 5.4. Therefore, the duration of passenger – bus interface is a

function of variables having complex interconnections. Figure 5.3 attempts to present

the connections between these variables in the simplest manner possible.

Figure 5.3: Passenger – bus interface duration and its dependent variables

The figure exemplifies that the passenger – bus interface depends on characteristics

of passenger walking, which in turn depending on the platform crowd and distance of

waiting point from loading area.

In the absence of any prior research on passenger – bus interface and its results on

bus dwell time, it is advantageous here to further explore the passenger – bus

interface duration and its influence on bus dwell time using graphical means.

5.3.2 Time – space diagram Observations made at the study busway station platform indicate that only the

duration of passenger – bus interface of the first passenger impacted bus dwell time.

Because of the simultaneous interfaces of all boarding passengers with the desired

bus, all boarding passengers except the first boarding passenger overlap their

respective passenger – bus interface duration with that of first passenger’s interface.

In addition to this, some part of the first passenger’s interface duration overlaps with

the time taken by the bus to reach the loading area. The above observations can be

better understood using the time-space diagram shown in Figure 5.4. The distance in

space along the line of the platform is represented on the Y axis and the time is

Platform crowd Waiting point Loading area

Walking speed

Number of passenger encountered

Walking direction w.r.t. bus arrival direction

Passenger – bus interface duration

Walking distance



represented on the X axis. The X – X axis, in space, represents the location of the

bus entry door on the loading area. P1, P2, … Pn are the position of the first, second

and nth passenger at time t0 when they first observe their desired bus, B. B’ is the

point in time when bus B reaches a stop at the loading area. P’1, P’

2, … P’n, are the

points in time when the first, second and nth passenger reach the bus entry door. The

lines P1-P’1, P2-P’

2, and Pn-P’n are the idealised trajectories of respective passengers.

The slope of these lines represents the passenger’s walking speed respectively.

Note that these trajectories may not be straight in practice due to the zig-zag path

discussed previously. The B-B’ line is the trajectory of bus and the slope of line

represents bus speed.

Note that the time lapse between B’ and P’1 is identified here as lost time for the bus,

where no real passenger service occurs. This research defines ‘lost time’ as the time

lapse between when the bus comes to rest on its loading area and the time of

boarding of the first passenger. Note that portions on the passenger – bus interface

duration of 2nd … nth passenger occurs during a period termed here as the

passenger – bus interaction period, IA. Therefore, in this case the dwell time for the

bus is equal to sum of the lost time (LT), passenger – bus interaction (IA) and the

door opening and closing time (toc). The variables in Figure 5.4 are described below:

DT = Bus dwell time LT = Bus lost time t0 = Time when passenger(s) first see the desired bus (say at point B in

space) P1, P2,.., Pn = Location of 1st, 2nd,…, nth passenger in the space at time t0. IF1,

IF2,…,IFn represent their respective passenger – bus interface duration.

IA = Duration of passenger – bus interaction to = Door opening time tc = Door closing time



Figure 5.4: Time – space diagram

In the time – space diagram, the loading area position represented by X-X is static.

However, the passenger is not. The passenger may change the waiting position due

to the influence of the level of crowding. This would change the duration of

passenger – bus interface and subsequently the bus lost time. Hence, we analysed

the behaviour of waiting passengers to understand the character of the passenger –

bus interface.

5.4 Passenger behaviour while waiting The bus arrival procedure at the study station, like most stations on Brisbane’s

busway network, follows the ‘lead stop’ operation principle. Under this principle, an

arriving bus should use the lead loading area of the platform. Only when the lead

stop is not available, the bus can use the next available loading area. This means

that, at the study station platform (Figure 5.5), if all the loading areas are empty, the

bus should use loading area 1 to provide service to passenger/s. Only if the front

loading area is occupied or blocked then the next loading area (i.e. loading area 2)

should be used. This created a relatively higher chance that a bus would use the

loading area 1 or 2 over loading area 3 (Table 5.1). This leads to passengers tending

to wait between the middle of lead loading area (loading area 1) and loading area 2

to wait for their desired bus (Figure 5.5). Under the off-peak, low crowd condition,

passengers were observed to wait within this area.

IA

DT

LT

IF1

IF2

IFn

P1 P2

Pn

Time

Spac

e

t0

X X

B

tcB’ P’2 P’

nP’1

to



Figure 5.5: Distance to loading areas from the waiting area on the busway platform (Off-peak)

As the crowd at the platform increases, additional passengers first spread toward the

left side of the shaded area (front area of platform) and later sprawl towards left of

the shaded area (rear area of platform). However, observations at the study station

found that passengers outside of the shaded area moved quickly into the shaded

area whenever a space was created. This behaviour is largely driven by their desire

to minimise the walking to their desired bus, which, due to the ‘lead stop’ operation

principle, has a higher probability of using either the loading area 1 or the loading

area 2 compared to loading area 3. For simplicity, the shaded area and its centroid

i.e. y-y line is used as the reference line.

The distance between the bus door on loading area 2 and the shaded area centroid

y-y is less than the distance between the bus door on loading area 1 and the shaded

area centroid (Figure 5.5). Therefore, in the best case scenario, the lost time for any

bus would be lowest when it uses the loading area 2.

From the reference line i.e. y-y line, the passenger walking to loading area 1 walks

nearly in the direction of approaching bus to loading area 1. In contrast, a passenger

walking to loading area 3 walks in the direction opposite of approaching bus to the

loading area 3. These directional differences between passenger and bus alter their

interface and subsequently the lost time for buses on these loading areas. Walking in

the opposite direction of the approaching bus reduces the overlapping component of

the passenger – bus interface, causing higher lost time. This is because the bus

needs to cover less distance to reach that loading area. In contrast, walking in the

direction of approaching bus slightly increases the overlapping component of the

passenger – bus interface, and could cause a slight reduction in lost time. This

phenomenon of passenger – bus interface is better explained in Figure 5.6, which

illustrates a case where the bus B would experience different lost time, due its first


Front Rear

Direction of bus movement

Where, dy1 > dy2 ; dy1 ≈ dy3

Outbound platform

y

y

dy1

dy2 dy3



passenger P1 standing on reference line y – y, depending on the loading area it uses

to serve passengers.

If the bus ‘B’ uses loading area 2 then it will experience the lowest lost time because

the passenger is standing near to the loading area and hence will not be required to

walk large distance to reach the bus entry door. However, if the bus uses loading

area 3 then it will experience largest lost time because of the passenger is required

to walk large distance and the walking will be in the opposite direction of the bus

arrival. On the other hand, the lost time for the bus B at loading area 1 will be lager

that loading area 2, because of the walking distance increases requirement to walk,

but will be less than that of loading 3 because the passenger’s walking is in the

direction that of bus arrival. The variables in Figure 5.6 are described below:

LT1, LT2, LT3 = Bus lost time at loading area 1, 2, and 3 respectively. P1 = Location of 1st passenger of the bus B.

dy1, dy2, dy3 = Passenger, P1, walking distance to loading area 1, 2, and 3 respectively.

y-y line = Centroid of shaded area (Ref: Figure 5.5)

Figure 5.6: Effect of loading area on bus lost time

P1

L A 1

B

L A 2

L A 3

y - y line

Time

Spa

ce

LT3

LT2

LT1

dy1

dy3

dy2



From the analysis presented in this section and in Section 5.3 we found that the

passenger – bus interface time influences the bus dwell time through a variable

called bus lost time (LT). The analysis illustrated how passenger bus interface

results in bus lost time (Figure 5.4) and explains how loading area is a determinant in

the lost time for a bus (Figure 5.6). In light of these findings, the bus lost time for

individual loading areas is analysed in following section.

5.5 Bus lost time In this research, bus lost time is defined as the time lapse between bus stopping and

the time its first passenger placed his/her foot on the bus floor.

For each bus, the time stamps of its stopping at its loading area (event B1), full

opening of its entry door (event B2) and its first passenger placing his/her foot on the

bus floor (event B3) were recorded from the video recordings. Quite often it was

observed that bus driver opens the entry door simultaneously while stopping at the

loading area, resulting in identical time stamps for event B1 and event B2. The bus

lost time, therefore, from Figure 5.4, can be estimated as -

, , Equation 5.2

Where,

= Bus lost time for bus, j.

= Time stamp of full opening of entry door for bus, j.

, = Time stamp of first passenger placing his/her foot on the floor the bus, j.

If any alighting occurred through the bus entry door during the bus lost time (LT), the

time consumed by the alighting passengers was deducted from the lost time. The

proportion of such observation was relatively very small throughout the analysis

period. This was mainly due to two reasons; first the study station platform had

predominantly boarding passengers during all study periods, and second the

majority of the alighting passengers used rear door of the bus. The number of

boarding and alighting passenger observed during an half hour afternoon peak

period (3:00 pm to 3:30 pm) at the study busway platform are given in Table 5.3.



Table 5.3: Passenger boarding and alighting during evening peak period

Month/ Year Number of buses

Number of boarding

passengers

Number of alighting passengers

Front door Rear door

March 2007 51 348 5 16 March 2008 36 475 4 13 April 2009 62 250 11 42

The video recordings from March 2008 and April 2009 were used to analyse bus lost

time. Table 5.4 provides the descriptive statistics of off-peak periods (10:00 am -

11:00 am and 7:00 pm – 8:00 pm). Table 5.5 provides the descriptive statistics of

peak periods (3:00 pm - 3:40 pm for March 2008; 3:00 pm - 4:00 pm for April 2009).

During the off peak periods, the observed number of passengers on the platform in a

15 min interval was less than 30. During the peak periods, the observed number of

passengers on the platform in a 15 min interval was greater than 40.

Table 5.4: Bus lost times (LT) during off-peak periods

Bus lost time Loading area 1 Loading area 2 Loading area 3

Min 1s <1s 1s

Max 11 s 10s 16 s

Mean 4s 3s 5s


Count 51 30 14

Note: Off – peak operation; # of passenger on platform < 30; total analysis time = 120 minutes



Table 5.5: Bus lost times (LT) during peak period

Bus lost time Loading area 1 Loading area 2 Loading area 3

Min <1s <1s <1s

Max 16s 10s 17s

Mean 4s 3s 5s


# count 61 50 38

Note: Peak operation; # of passenger on platform ≥ 30; total analysis time = 100 minutes

Note that minimum bus lost time, in both the off-peak and peal periods, across all

three loading areas were either 1s or less than 1s. This represents the cases where

the first passenger of the bus happened to be standing right where the bus stopped.

In such cases the first passenger boards the bus almost immediately as soon as it

comes to halt and door opened. Such situations are more common at simple bus

stops where passengers wait for their bus at the signpost for the stop.

Of note, the mean bus lost times for loading areas 1 and 2 were no different between

the off-peak and peak periods. However, the maximum bus lost time for loading area

1 increased in the peak period. On the other hand, loading area 2 experienced a

decrease in bus lost time during the peak period. Loading area 3 however,

experienced an increase in both mean and maximum bus lost times during the peak

periods.

To study the profile of bus lost time between off-peak and peak periods, the mean

bus lost time of each loading area was analysed against platform crowding. Figure

5.7 shows that they have dissimilar changes in their lost time as the crowd level at

the platform increases. For loading area 1 the mean lost time initially increases with

the increase in crowd level. However, with a further increase in crowd level the mean

lost time decreases. The initial increment in the lost time may be due to the crowd

having acted as an obstruction in the path to bus entry door for loading area 1. When

passengers move into the area left of the shaded area (Figure 5.5) this may result in

a decrease in bus lost time for loading area 1.



On the contradictory, the mean lost time for buses using loading area 2 decreases

initially but later increases back. This is reasonable because of the behaviour of

passengers in selecting the area adjacent to loading area 2 (shaded area in Figure

5.5). However, as more passengers tries to fit into this limited space they increases

the obstruction to walking. This results in lost time again increasing back under the

high crowd level.

Whereas, the relatively obstruction free path to loading area 3, during off-peak

period, results in mean lost time for buses on loading area 3 lesser than loading area

1 under the low crowd level condition. However, the mean lost time steadily

increases with platform crowd. This is reasonable for two reasons. First, the

obstructions are likely to increase with crowd level and second, the majority of

passengers must to walk in the direction opposite to the approaching bus.

Figure 5.7: Variation in bus lost times over platform crowd by loading area

The results clearly indicate that bus lost time is not a constant value for the entire

platform. It varies from loading area to loading area and also varies with platform

crowd levels. These observations support the analysis presented in Section 0.

5.6 Passenger – bus interaction This research defines the passenger – bus interaction time as a phase where the

first and subsequent passengers and bus (driver) are involved in a state when they

LA1

LA2

LA3

0

1

2

3

4

5

6

7

8

9

10 20 30 40 50 60 70

Mea

n lo

st ti

me

per B

us (s

)

Number of passengers on platfrom

Off-

peak

per

iod

Pea

k pe

riod

onse

t

Pea

k pe

riod



both interact and perform their respective activities based on each other’s position or

action. The passenger – bus interaction activities are passenger boarding and

passenger alighting. The boarding and alighting activities at a busway station

platform are similar to that at a simple bus stop. The factors influencing the

passenger boarding time and alighting times include fare collection method, bus

floorplan and onboard standees. In the literature, for a simple bus stop, the

passenger boardings and/ or alightings are considered as the part of bus dwell time.

Much literature has been found related to the passenger boarding and alighting at a

simple bus stop. A detail survey of literature on existing bus dwell time model was

presented in Section 2.4.1. For the purpose of this research, the simple bus stop

dwell time model suggest by TCQSM (TRB, 2003), was used (Equation 5.3).

ocbbaad ttPtPt ++= Equation 5.3

Where,







This equation implies that alighting and boarding occur in series, and accounts only

for those alighting through the front door. Any passengers alighting through the rear

door are neglected in the standard model, as their activity occurs in parallel to the

front door activity, which is implied to be time-critical (Vuchic, 2005).

Therefore, the passenger – bus interaction (IA) can be mathematically represented

as –

Equation 5.4



Where,







As discussed in the section 5.3.2, the passenger – bus interaction time is the time

lapse between first passenger boarding and last passenger boarding. A passenger

was considered boarded when no part of his/ her body is out of the bus. For each

bus, time stamps for its first passenger boarding (event PB,1) and last passenger

boarding (event PB,l) were recorded. Similarly, time stamps for its first passenger

alighting from front door (event PAF,1), from rear door (event PAR,1) and last

passenger alighting from front door (event PAF,l) and from rear door (event PAR,l)

were recorded. The number of boardings and alightings was also recorded.

The boarding time per passenger and alighting time passenger at the study station

was estimated as -

Equation 5.5

Equation 5.6

The video recordings from April 2009 were used to estimate the boarding and

alighting times per passenger. The marginal service time was estimated for each

loading area individually. The descriptive statistics of boarding time and alighting

time are provided in Table 5.6.



Table 5.6: Descriptive statistics

Door Minimum Maximum Mean Std. deviation

Loading Area 1 Boarding time per pax Front 2.5s 6.5s 4.2s 2.5s Alighting time per pax Rear 2.6s 8.0s 3.4s 1.4s

Loading Area 2

Boarding time per pax Front 3.0s 7.5s 5.0s 2.3s Alighting time per pax Rear 3.5s 5.0s 4.0s 0.6s

Loading Area 3

Boarding time per pax Front 1.9s 8.0s 4.9s 4.8s Alighting time per pax Rear 2.0s 6.0s 3.6s 1.4s

5.6.1 Effect of fare collection policy During the course of this study, fare collection methods at the study station were

changed. TransLink’s Go Card, a smart card fare collection system, was introduced

between March 2007 and March 2008. Between March 2008 and April 2009 on

board ticket sales were abolished, only during the peak period between 2:30pm and

6:00pm (TransLink Web site). These changes provided a unique opportunity to study

the effects of fare policy changes on marginal service times. However, a detail study

of fare collection policy impact was not originally in the scope of this research. A

short analysis was carried out, findings of which are presented in this section.

The video recordings from March 2007, March 2008 and April 2009 were used to

study the impact of fare collection system on boarding and alighting times per

passenger. The details of recordings and fare policies on the recoding days are

given in Table 5.7.



Table 5.7: Fare collection policies and observations at study station

Month/ Year March 2007 March 2008 April 2009

Time Period Afternoon Peak Afternoon Peak Afternoon Peak

Analysis time 3:00PM to 3:30PM 3:00PM to 3:30PM 3:00PM to 3:30PM

Fare policy



Pre-paid paper ticket Pre-paid paper ticket Pre-paid paper ticket

10 trip magnetic stripe card

10 trip magnetic strip card

(Phasing out)

GoCard smart card (Introduced) GoCard smart card

Pre-paid platform policy

Number of buses 51 36 62

Number of boarding

passengers 348 475 250

Number of alighting

passengers 21 17 53

5.6.1.1 Passenger boarding time Table 5.8 gives the observed boarding time per passenger across all loading areas

of the study station platform for the March 2007, March 2008 (transition period), and

April 2009. Data from the transition period showed an initial increase in average

boarding times for all loading areas. However, later the average boarding time

decreased for all loading areas. The initial increase in service time could be

attributed to the inexperience of users in using GoCard and/or the mixture of the

Magstripe, GoCard and operator as cashier systems in place.

The result highlighted that with removal of on board ticket purchasing, an increased

uniformity in service time per boarding passenger was observed among the three

loading areas on the study station platform. The boarding time was decreased by

approximately 15 percent for loading area 1 and loading area 2, and approximately

40 percent for loading area 3.



Table 5.8: Effect of fare collection policy on boarding time per passenger.

Loading area

March 2007

March 2008

2008 change from 2007

April 2009


1 4.8s 6.8s +42% 4.2s -14%

2 5.9s 6.0s +2% 5.0s -15%

3 8.1s 8.9s +10% 4.9s -40%

5.6.1.2 Passenger alighting time Passengers using the GoCard are required to touch off their cards to a card reader

before alighting the bus to facilitate the accurate fare calculation for their trip. Other

passengers are not required to transact at the time of alighting from the bus. Similar

to touch on while boarding, touch off also requires passengers to place the card less

than 10 cm distance from the card reader and steady for one or more seconds. This

inevitably has led to an increase in alighting time per passenger, over the previous

system used in 2007 where magstripe card holders were not required to transact on

alighting from the bus. Table 5.9 gives the average alighting time per passenger

observed at the study station for the three analysis periods.

Table 5.9: Effect of fare collection policy on alighting time per passenger.

Loading area

March 2007

March 2008


April 2009 2009 change from 2007

1 2.2s 2.0s -9% 3.4s +55%

2 1.9s 2.0s +5% 4.0s +110%

3 2.1s 2.1s 0% 3.6s +71%

The highest increases in alighting time were observed for loading area 2 (110%)

followed by loading area 3 (71%) and loading area 1 (55%). Alighting times were

also found to be affected by the bus type; rigid bus and articulated bus. Passengers

alighting from an articulated bus at the study station showed a tendency towards

using the front door of the bus. As there were still few articulated buses used on the

study platform during this study, no statistically significant results could be produced.

5.6.1.3 Findings This analysis has highlighted that the boarding time per passenger was reduced by a

minimum of 14 percent. This reduction occurred due to various fare collection policy



changes implemented over the analysis period. On the contrary, the alighting time

per passenger increased in excess of 50 percent. The average service time for each

boarding passenger was 4.7s and for each alighting passenger 3.6s. However, more

investigation is required to assess the impact of the smart card system on alighting

passengers, particularly, for a period when alighting passengers are predominant,

such as the morning peak on inbound platforms of inner urban (destination) stations,

for example, the case on the inbound platform at Mater Hill Busway station

5.7 Chapter close This chapter evaluated the new parameters of busway operation that were identified

in chapter 3. Intricate analyses of these parameters were performed, in which

relevant elements and processes related to passenger waiting behaviour and bus

dwell time at the busway station were identified. The analysis highlighted some

relationships between parameters and elements that effect the functioning of the

busway station platform.

In this chapter, the passenger waiting behaviour at the busway station platform was

explained and its influence of passenger – bus interface was studied. By studying

the complex nature of the passenger – bus interface and the relationship with bus

lost time, a sound theoretical framework for busway station bus dwell time was

developed.

In next chapter, the bus lost time model is elaborated. Chapter 7 integrates the bus

lost time variable with existing knowledge base to develop a more comprehensive

busway station bus dwell time model and busway station loading area bus capacity

model. Chapter 8 presents a methodology to estimate busway station loading area

efficiency factors. Chapter 9 propose a comprehensive busway station platform bus

capacity methodology.

Modelling Bus Lost Time


Chapter Six


6.1 Overview Having introduced in preceding chapters the concept of bus lost time and having

analysed it in chapter 5, this chapter discusses the development of stochastic

relationships from the observed bus lost time data set for the study busway station.

The purpose of developing a stochastic model was to generalise the pattern of bus

lost time variations. The generalised bus lost time can be applied to develop an

improved bus dwell time model and bus capacity model for the busway station.

Before estimating the stochastic model for the observed bus lost time it was

necessary to study its frequency density. This analysis was performed using a

histogram technique. Details of analysis are discussed in Section 6.2. Section 6.3

estimates the bus lost time probability distribution function and Section 6.4 fits the

probability distribution curve on observed bus lost time data. As found from the

analysis presented in Section 5.5, the bus lost times depended on the particular

loading area and platform crowd. Further separate probability curves were fitted to

each loading area for each of the peak and off-peak periods. Section 6.5 closes the

chapter and summaries the key contributions to knowledge.

6.2 Bus lost time histogram Figure 6.1 shows the measured bus lost time histogram for the off peak period at the

Mater Hill busway station outbound platform. The values for bus lost time observed

at the station platform range from less than 1s to 16s during this period. The

frequency plot shows that at loading area 1, most of the observed lost times were 1s

to 4s. Similarly at loading area 2, the majority of buses experienced a lost time of 1s

to 3s. However, at loading area 3, the buses experienced a wide range of lost times

with no clustering, with maximum value of lost time observed as 16s.



a) Loading area 1

b) Loading area 2

c) Loading area 3

Figure 6.1: Off-peak period bus lost time histogram

0

2

4

6

8

10

12

14

< 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Cou

nt

Bus lost time

0

2

4

6

8

10

12

< 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Cou

nt

Bus lost time

0123456789

10

<1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Cou

nt

Bus lost time



a) Loading area 1

b) Loading area 2

c) Loading area 3

Figure 6.2: Peak period bus lost time histogram

0

2

4

6

8

10

12

14

<1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Cou

nt

Bus lost time

02468

101214161820

<1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Cou

nt

Bus lost time

012345678

<1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Cou

nt

Bus lost time



Figure 6.2 shows the bus lost time histogram for the peak period at the same

location. The values for bus lost time observed at the station platform range from

less than 1s to 17s. At loading area 1, the bus lost time had widened range

compared to the off-peak period, with the majority of buses experiencing lost time

between 1s and 7s. On the other hand, at loading area 2, most of the buses had

more concentrated lost times between 1s and 3s. Similar to loading area 1, the

majority of buses at loading area 3 experienced the lost time ranging from less than

1s to 7s.

The resulting histograms indicate that the bus lost time, both in off-peak and peak

period, is not normally distributed for any of the loading areas. Graphically we can

interpret that the histograms are positively skewed which suggests that lognormal

distribution or Weibull distribution could be fitted (Devore, 2008).

In probability theory, a log-normal distribution is a probability distribution of a random

variable whose logarithm is normally distributed. A variable might be modelled as

log-normal if it can be thought of as the multiplicative product of many independent

random variables each of which is positive (Wikipedia, 2009b). Since the bus lost

time is indeed a function of complex and dynamic relationships of passenger

walking, bus stopping, and crowd density, the log-normal distribution fitting was

therefore first considered. To avoid syntax error, (log of 0 is infinite), all values of bus

lost time less that 1s was clustered in one group of 0.5s instead of 0s. This

assumption is reasonable because in reality the lost time of 0s for a bus is highly

unlikely due to the fact that passenger takes some time to react to bus stopping and

door opening.

Another most widely used distribution, especially for lifetime distributions in reliability

engineering due to its versatility, is the Weibull distribution. The Weibull distribution

can take on the characteristics of other types of distributions, based on the value of

its shape parameter, ‘β’ (Devore 2008). If the log normal distribution fails to represent

the observed data then Weibull distribution will be applied.

In next section, statistical analyses are presented to fit a probability distribution

curve.



6.3 Probability distribution curve fitting SPPS (originally, Statistical Package for the Social Sciences), was used for the

analysis and curve fitting. Table 6.1 to Table 6.3 provides the descriptive statistics of

off-peak data set of each loading area. Table 6.4 to Table 6.6 provides the

descriptive statistics of peak data set of each loading area.

The four descriptive parameters which are of particular interest are mean, median,

skewness, and kurtosis. These descriptive parameters describe the shape of the

distribution.

The values of mean and median describe the skewness of the data (Hamburg,

1983). If the mean exceeds the median, the distribution is said to have positive

skewness. If the mean is less than the median then the distribution is said to have

negative skewness. From tables note that, for each loading area and period, the

mean of the observed data is greater than the median. This validates the results of

histogram analysis and concludes that the bus lost time at each loading area is

positively skewed distribution.

The value of skewness measures the asymmetry (skewness) of the distribution. For

a perfectly normal distribution the value of skewness is zero, i.e. perfectly symmetric.

For each loading area the value of skewness was found to be greater than 0, as

expected.



Table 6.1: Descriptive statistics of loading area 1 (Off-peak period)

Raw data (X) Log [X]

Statistic Std. Error Statistic Std. Error

Count 50 50

Mean 4.280 0.400 1.245 0.093

95% confidence

interval for mean

Lower bound 3.476 1.058

Upper bound 5.084 1.433

Median 3.000 1.098

Variance 8.002 0.436

Std. deviation 2.828 0.660

Skewness 1.038 0.337 -0.008 0.337

Kurtosis 0.073 0.662 -0.682 0.662




Count 30 30

Mean 2.883 0.363 0.859

95% confidence

interval for mean



Median 2.000 0.691



Skewness 1.911 0.427 -0.136 0.427

Kurtosis 4.823 0.833 0.303 0.833






Count 14 14

Mean 7.500 1.142 1.820 0.192

95% confidence

interval for mean



Median 7.500 2.013

Variance 18.269 0.519


Skewness 0.541 0.597 -1.144 0.597

Kurtosis -0.309 1.154 1.980 1.154

Table 6.4: Descriptive statistics of loading area 1 (peak period)



Count 61 61

Mean 4.066 0.415 1.095 0.107

95% confidence

interval for mean



Median 3.000 1.098

Variance 10.512 0.697


Skewness 1.692 0.306 -0.362 0.306

Kurtosis 3.497 0.604 -0.229 0.604



Table 6.5: Descriptive statistics of loading area 2 (Peak period)



Count 50 50

Mean 2.850 0.244 0.888 0.0823

95% confidence

interval for mean



Median 3.000 1.098



Skewness 2.020 0.337 -0.307 0.337

Kurtosis 6.140 0.662 0.575 0.662

Table 6.6 Descriptive statistics of loading area 3 (Peak period)



Count 38 38

Mean 4.868 0.636 1.259 0.141

95% confidence

interval for mean



Median 4.000 1.386

Variance 15.401 0.758


Skewness 1.515 0.383 -0.440 0.383

Kurtosis 2.223 0.750 -0.097 0.750



Kurtosis is a measure of peakedness of the distribution. Like skewness, the value of

kurtosis for a perfectly normal distribution is 0. A positive kurtosis value means the

distribution is too pointed. A negative value means the distribution is too flat. Higher

kurtosis means more of the variance is due to infrequent extreme deviations, as

opposed to frequent modestly sized deviations (Wikipedia, 2009c).

To fit the lognormal distribution, the natural log of the observed data set was taken

and test for normality was performed. The test of normality of log-data is described in

next section.

6.3.1 Assessing normality There are two ways of testing normality; graphical method and numerical method

(Park, 2008). Table 6.7 provided an overview of these methods. Graphical methods

visualize the distributions of random variables a theoretical distribution (e.g., the

standard normal distribution). These methods are quick and easy to interpret.

Numerical methods evaluate summary statistics such as skewness and kurtosis, or

conduct statistical tests of normality such as Shapiro-Wilk test.

Table 6.7: Methods for testing normality

Graphical Methods Numerical Methods

Descriptive Stem-and-leaf plot, box plot, dot plot, histogram

Skewness, Kurtosis

Theory-driven

P-P plot, Q-Q plot

Shapiro-Wilk test, Shapiro- Francia test, Kolmogorov-Smirnov test (Lillefors test), Anderson-Darling/Cramer-von Mises tests Jarque-Bera test, Skewness-Kurtosis test

Source: Park, 2008

In this research the normality of observed data set was assessed using histogram

plot, and skewness score. The normality of log-data was assessed by testing null

hypothesis about skewness and kurtosis scores.



6.3.2 Null hypothesis testing Skewness measures the vertical displacement of a distribution from a standard bell

shaped normal distribution. Whereas kurtosis measures the horizontal displacement

of a distribution from a standard bell shaped normal distribution. When an observed

data superimposes perfectively on the standard bell shaped normal distribution, its

skewness statistic and kurtosis statistic equals to zero. Therefore value of zero is the

primarily value of interest in assessing the normality of the data (Innes, 2007).

To test the null hypothesis the skewness statistic and Kurtosis statistic score

methods involve construction of 95% confidence interval (z0.025 = 1.96) about a

skewness score (or a kurtosis score). The null hypothesis, in this case, is that the

skewness estimate (or kurtosis estimate) is not significantly different from a value of

zero, i.e. the estimated score is from a standard bell shaped normal distribution. If

the value of ‘zero’ is within the 95% confidence interval than the null hypothesis can

be accepted, otherwise reject the null hypothesis.

The calculations for loading area 1, using off-peak data set, are presented here to

demonstrated the null hypothesis testing method. Using the statistic and standard

error values of skewness, the lower and upper bound of its 95% confidence interval

was determined. The standard z value for 95% confidence is 1.96. The lower bound

of skewness is -0.669 (Statistic plus 1.96 times the std. error) and the upper bound

of skewness is 0.653 (Statistic minus 1.96 times the std. error). Hence the 95%

confidence interval for the skewness ranges from -0.669 to 0.653. Since, zero, the

main value of interest lies with the 95% confidence interval, the null hypothesis is

accepted. This means that natural logarithm of observed data set is normally

distributed and therefore it can be concluded that the off-peak bus lost time data set

for loading area 1 is log-normally distributed.

Similarly, the 95% confidence interval of the kurtosis for off-peak data of loading area

1 ranges from -0.669 to 0.653.

The above steps were repeated for other loading areas. The test of normality for

each loading area during off-peak and peak periods are summarised in Table 6.8 to

Table 6.13. Note that the value of zero for skewness and kurtosis falls within their

respective 95% confidence interval, establishing that natural logarithm of observed



data set is normally distributed. Hence, the bus lost time data set for all loading

areas, in off-peak and peak periods was found to be log-normally distributed.

Table 6.8: Assessing normality for loading area 1 (Off-peak period)



Skewness 1.038 0.337 -0.008 0.337

95% confidence interval for skewness

Lower bound 0.377 -0.669


Kurtosis 0.073 0.662 -0.682 0.662

95% confidence interval for Kurtosis

Lower bound -1.225 -1.980





Skewness 1.911 0.427 -0.136 0.427




Kurtosis 4.823 0.833 0.303 0.833







Skewness 0.541 0.597 -1.144 0.597




Kurtosis -0.309 1.154 1.980 1.154






Table 6.11: Assessing normality for loading area 1 (Peak period)



Skewness 1.692 0.306 -0.362 0.306




Kurtosis 3.497 0.604 -0.229 0.604







Skewness 2.020 0.337 -0.307 0.337




Kurtosis 6.140 0.662 0.575 0.662







Skewness 1.515 0.383 -0.440 0.383




Kurtosis 2.223 0.750 -0.097 0.750






6.4 Log-normal distribution curves for bus lost time This section presents the log-normal probability distribution function (PDF) curve and

cumulative distribution function (CDF) curve of bus lost time for each loading area

under each time period. The standard log-normal PDF is given by Equation 6.1. The

log-normal CDF is given by Equation 6.2.

; ;

1√2

⁄ 0

0 0

Equation 6.1

; ; 0 Equation 6.2

Where,

= An event; in this case bus lost time, whose probability of occurrence is to be calculate

= Mean of In(X)

= Standard deviation of In(X)

X = Variable; in this case observed bus lost time

Note that the parameter µ and σ are not the mean and standard deviation of X but of

In(X). The mean and standard of X can be calculated as

X eµ ⁄ Equation 6.3

X eµ ⁄ 1 Equation 6.4

Std. deviation of X = X Equation 6.5



Where,

= Mean of variable X

= Variance of variable X

6.4.1 Log-normal probability distribution function curve Figure 6.3 provides the bus lost time probability distribution function (PDF) for the

lognormal distribution curves for each loading area for peak time periods. From the

peak period curves, it was noted that all three loading areas have nearly same

probabilities of experiencing a lost time of 4.1s. However loading area 2 has a higher

probability of experiencing lost time of less than 4.1s. Whereas, the loading area 3,

relatively to loading area 1 and 2, has lesser chances of experiencing a lost time less

than 4.1s. On the contrary, the probability of loading area 3 getting a lost time larger

than 4.1s is relatively higher in comparison to loading area 1 & 2.

Figure 6.4 provided the bus lost time probability distribution function (PDF) for the

lognormal distribution curves for loading area 1 and 2 for off-peak time periods. Due

to the insufficient observation during off peak operation of loading area 3 the curve

fitting was not done for it. During off-peak period the loading area 1 and 2 have

identical probability of experiencing a lost time of 2.9s. This value of lost time is less

than that of peak period bus lost time value, which is 4.1s. However, similar to peak

period curves the loading area 2 has higher chances of experiencing a lost time less

than 2.9s in comparison to loading area 1.

The comparison between peak and off peak curves is shown in Figure 6.5. The

figure shows that the off-peak curve for loading area 1 is shifted toward right relative

to the peak period curve. This shift indicates that loading area 1 is more likely to get

higher lost time during low crowd situation. On the contradictory, for loading area 2

the likelihood of a given lost time is fairly same in both periods of operation. In fact,

there are two different movements in probabilities for loading area 2. While the

probability for lost time less than 1.8s and for lost time greater than 5.5s decreased

in peak period, the probability for lost times between 1.8s and 5.5s increased.

These results are consistence with the observations in Figure 5-7 and passenger

behaviours explained in section 5.4.



Figure 6.3: Bus lost time probability distribution curves (Peak period)

Note: LA = loading area

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Prob

abilt

y, f(

x)

Lost time (x)

LA 1 (Peak)

LA 2 (Peak)

LA 3 (Peak)



Figure 6.4: Bus lost time probability distribution curves (Off-peak period)

Note: LA = loading area; OP = Off-peak

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Prob

abilt

y, f(

x)

Lost time (x)

LA 1 (OP)

LA 2 (OP)



Figure 6.5: Comparison of peak and off-peak bus lost time probability distribution curves


0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Prob

abilt

y, f(

x)

Lost time (x)

LA 1 (Peak)

LA 2 (Peak)

LA 3 (Peak)

LA 1 (OP)

LA 2 (OP)



Figure 6.6: Bus lost time cumulative distribution curves (Peak period)

Note: LA = loading area

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Cum

ulat

ive

prob

abilt

y

Bus lost time

LA 1 (Peak)

LA 2 (Peak)

LA 3 (Peak)

85th percentile



Figure 6.7: Bus lost time cumulative distribution curves (Off-peak period)


0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Cum

ulat

ive

prob

abilt

y

Bus lost time

LA 1 (OP)

LA 2 (OP)

85th percentile



Figure 6.8: Comparison of peak and off-peak bus lost time cumulative distribution curves


0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Cum

ulat

ive

prob

abilt

y

Bus lost time

LA 1 (Peak)

LA 2 (Peak)

LA 3 (Peak)

LA 1 (OP)

LA 2 (OP)



The curve parameters µ and σ are given in Table 6.14.

Table 6.14: Statistical parameters of bus lost time curves

Loading area Period µ σ

1 Peak 1.095 0.834 2 Peak 0.888 0.582 3 Peak 1.259 0.870 1 Off-Peak 1.245 0.660 2 Off-Peak 0.859 0.650

6.4.2 Log-normal cumulative distribution function curve Figure 6.6 provides the cumulative distribution function (CDF) of bus lost times for

each loading area for peak time period. The CDF highlights the effect on bus lost

times due to difference in direction of motion of a passenger and the approaching

bus. For a given probability, the lost time for loading area 3 is always higher than that

for loading area 1. Theoretically the CDF for the loading area 1 and 3 should overlap

or near to each other, because the walking distance to both loading area is nearly

same from the shaded area. However, the CDF of loading area 3 is right shifted with

respect to loading area 1. This is because a passenger requires walking in direction

opposite to approaching bus to loading area 3 as against the same directional

motional for the loading area 1. This directional discrepancy resulted in a lesser

slope to the CDF for loading area 3 compared to loading area 1.

The CDF also demonstrates the variations in upper range of bus lost time with

loading area. For example, during the peak period operation, 85 percent of buses

using loading area 1 may have a lost time value less than or equal to 7.1s. On the

other hand, 85 percent of buses using loading area 2 and loading area 3 may have a

lost time value less than or equal to 4.5s and 7.1s respectively.

Figure 6.7 shows the cumulative distribution function (CDF) of bus lost times for

loading area 1 and 2 for off-peak time period. Similar to peak period, in off-period too

the bus lost times at loading 1 are higher compared to loading area 2.



Figure 6.8 compares the bus lost time curves of peak and off-peak period. For

loading area 2 the cumulative probabilities in peak and off-peak are almost identical.

Though for loading area 1 there is increased probability of higher lost time in off-peak

period.

6.4.3 Descriptive characteristics of busway station bus lost time This section presents the descriptive characterises of generalised bus lost time, such

as mean, mode, median and 85th percentile values. In statistics, the mean, mode,

and median are also called as the measures of central tendency which describes the

scores in a distribution. Table 6.15 and Table 6.16 provide central tendency and 85th

percentile values of bus lost time for all three loading areas for peak and off-peak

period respectively.

Table 6.15: Descriptive characteristics of bus lost times (Peak period)

Loading area 1

Loading area 2

Loading area 3

Mode of lost time 1.6 2.0 2.0 Median lost time 3.0 2.4 3.5 Mean lost time 4.2 2.9 5.2 85th percentile lost time 7.1 4.5 8.8

Table 6.16: Descriptive characteristics of bus lost times (off-peak period)

Loading area 1

Loading area 2

Loading area 3

Mode of lost time 2.3 2.0 - Median lost time 3.5 2.4 - Mean lost time 5.3 3.6 - 85th percentile lost time 7.0 4.8 -

6.5 Chapter close This chapter presented the development of probability distribution models for bus

lost times. Statistical analysis of bus lost time established that the busway station

bus lost time is a positively skewed, log-normally distributed variable. A set of five

log-normal distribution curves are developed which approximates the observed data



sets. These probability models can be used to develop a micro-simulation based

model to study busway station operation. Furthermore, understanding of bus lost

time can be applied in improving the methodology of busway station bus capacity

analysis.

Overall, this chapter has three key contributions in evolving the knowledge of

busway station bus lost time. It,

1) Established the quantitative descriptions of the busway station bus lost time.

2) Developed a set of loading area centric and time period specific probability

models.

3) Defined the descriptive characterises of busway station bus lost time.

With bus lost time variable quantified in this chapter, the next chapter discusses the

development of busway station bus dwell time model incorporating the bus lost time.

Busway Station Dwell Time Model


Chapter Seven


7.1 Overview Previous chapters established the bus lost time variable as a significant parameter

influencing bus dwell time at a busway station platform. Along with the traditional

variables of the number of boarding and alighting passengers and respective

marginal time and door opening and closing time, bus lost time is also a determinant

of dwell time of a bus at a busway station.

This chapter describes the development of the busway station bus dwell time model,

incorporating the concept of bus lost time. Section 7.2 describes the framework of

the model. The mathematical model is presented in section 7.3. Section 7.4

discusses the effects of bus lost time on bus dwell time estimation. Section 7.5

summaries the key features of the new model and discusses model implications.

Section 7.6 closes the chapter.

7.2 Model framework Chapter 2 had identified the inadequacies in applying the existing bus stop dwell

time models for busway station analysis. These inadequacies arise due to their lack

of sensitivity towards certain characteristics of the busway station platform operation,

such as the time spent by passengers walking to bus entry door, platform crowding

and multiple loading areas. To achieve a reliable bus dwell time estimation

methodology for a busway station, the busway station bus dwell time model should

account for the characteristics of the busway station in addition to the elements of

passenger service times. Figure 7.1 provides an overview of variables determining

the bus dwell time at a busway station platform.

The number of boarding and alighting passengers and their marginal service

time, during the passenger – bus interaction phase as described in Section 5.6, are



the most traditional and widely used variables in the dwell time estimation. As

discusses in section 5.6, these variables can be modelled using existing bus stop

dwell tide models. For the purpose of this research, the bus stop dwell time model

given by TCQSM, Part 4 (TRB, 2003) was used (Equation 2.5).

The door time is a time consumed by a bus to fully open its doors after the bus has

come to rest and close it back fully after the completion of passenger servicing. The

TCQSM dwell time model also accounts for door time, i.e. door opening and closing

time in its equation (Equation 2.5).

The loading area characteristics perpetuate the requirement of passenger walking

to the bus entry door to board the desired bus. The passenger station demand creates platform crowding, which in turn impacts passenger walking. Analysis

presented in section 5.3 established that passenger walking to the bus entry door

and the crowd encountered en route causes lost time for the bus.

The bus lost time variable results from the complex process between the platform

crowd, passenger walking and the bus itself. As explained in Figure 5.4, the bus lost

time occurs, prior to start of actual boarding, but after the bus has come to rest and

is additive to the traditional variables of the number of boarding and alighting

passengers and door time. The theoretical as well as statistical analyses on bus lost

time, presented in previous chapters, established that the lost time varies across the

platform from loading area to loading area and between peak and off-peak periods.

This implies that the bus dwell time at each loading area varies depending on the

bus lost time at the loading area under the prevailing conditions and is not a uniform

value across the platform length.



Figure 7.1: Overview of model form for busway platform dwell time estimation

7.3 Busway station bus dwell time model Equation 7.1 presents the bus dwell time model for dwell time estimation at a

busway station with multiple linear loading areas. The model is known as BSDT

model, an abbreviation of Busway Station Bus Dwell Time Model.

Equation 7.1

Where,

= Bus dwell time at nth loading area

; = Number of passenger boarding and alighting respectively

; = Service time per boarding and alighting passenger

respectively (s)

= Bus door opening and closing time (s)

= Bus lost time at nth loading area (s)

This model has bus lost time as an additional term compared to the traditional

Equation 2.5. Note that the bus lost time term is loading area specific. Hence this

Station demand (Pax per hour)

Platform crowding

Passenger – bus interface (IF)

Loading area characteristics

Walking distance

Number of boarding and alighting passenger (Pa, Pb)

Bus door passenger processing time (ta, tb)

Bus dwell time (DT)

Door time (toc)

Bus lost time (LT)



bus dwell time model estimates different dwell time values for each specific loading

area.

7.4 Example application To study the impact of the bus lost time variable on bus dwell time estimation, the

BSDT model and the bus dwell time model given by TCQSM (TRB, 2003) were

applied to an example busway station. The latter model does not account for the bus

lost time in its methodology. The difference in estimated dwell time values obtained

from both the models will therefore be due to the influence of the bus lost time. The

effect of bus lost time was analysed at two levels: at the platform level and at the

individual loading area level. For this demonstrative exercise, the peak period

operation of the station was analysed.

The outbound platform of the example busway station has three linear loading areas

with an adjacent passing lane. A boarding load of 7 passengers per bus with

boarding service time of 4s and no alighting load for the front entry door was

considered. The estimation was done with an assumed door opening and closing

time of 2s. The mean bus lost time values obtained from Table 6-15 were used in

BSDT equation. Table 7.1 shows the comparison of the results from two methods.

Unlike TCQSM model which estimated a single dwell time value for all three loading

areas, the BSDT method which considered the bus lost time estimated different

dwell time values for each specific loading area depending on the prevailing

operation condition of the busway station platform. For a constant passenger load,

all three loading areas have different bus dwell times, due to their dissimilar bus lost

times. The lowest bus dwell time was found for loading area 2, whereas the highest

dwell time was found for loading area 3. The example demonstrated the effects of

bus lost time on bus dwell time at platform level.



Table 7.1: Example demonstration

Method Bus dwell time

Loading area 1

Loading area 2

Loading area 3

TCQSM model (without bus lost time)

30s 30s 30s

BSDT model (with bus lost time)

34.2s 32.9s 35.2s

% change +14.0% +9.7% +17.3%

The effects of bus lost time can also be seen at the scale of the individual loading.

The bus lost time is a loading area specific variable and accounts the process

between the bus and the first boarding passenger only. Hence if more passengers

board per bus the proportion of lost time in the total dwell time of bus decreases.

Similarly if the marginal boarding time increases, the share of lost time in the total

dwell time of the bus decreases. To clarify through an example, the bus dwell time

was estimated for loading area 1 with varying boarding load and marginal boarding

time. The influence of bus lost time on bus dwell time gradually decreased with

increase in boarding load. The means that the impact of bus lost time on a bus is

subsidised with increased boarding load.

On the contrary, the influence of lost time increased with a decrease in marginal

boarding time. The inference of this result is that with a decrease in marginal time,

the bus lost time becomes more crucial component in bus dwell time. Figure 7.2

shows the changing influence on bus dwell time at the loading area 1.



Figure 7.2: Effect of bus lost time on dwell time at loading area 1 (Peak period)

Similar kind of trend of bus lost time influence on dwell time can be observed for

other loading areas and for peak and off-peak periods. Figure 7.3 to Figure 7.6

present these graphs. It is worth noting here that the above demonstration example

is designed using mean bus lost time value. The results and shape of these graphs

will change with variation in bus lost time value.

No graph for off-peak period of loading area 3 was plotted because no statistical

analysis was done due to the lack of data.

0

10

20

30

40

50

60

1 2 3 4 5 6 7 8 9 10

Con

tribu

tion

of lo

st ti

me

in d

wel

l tim

e (%

)

Boarding load per bus

tb = 2s

tb = 4s

tb = 5s



Figure 7.3: Effect of bus lost time on dwell time at loading area 1 (Off-peak period)


0

10

20

30

40

50

60

1 2 3 4 5 6 7 8 9 10

Con

tribu

tion

of lo

st ti

me

in d

wel

l tim

e (%

)


0

5

10

15

20

25

30

35

40

45

1 2 3 4 5 6 7 8 9 10

Con

tribu

tion

of lo

st ti

me

in d

wel

l tim

e (%

)

Boarding load

tb = 2s

tb = 4s

tb = 5s

tb = 2s

tb = 4s

tb = 5s



Figure 7.5: Effect of bus lost time on dwell time at loading area 2 (Off-peak period)


0

5

10

15

20

25

30

35

40

45

50

1 2 3 4 5 6 7 8 9 10

Con

tribu

tion

of lo

st ti

me

in d

wel

l tim

e (%

)

Boarding load

0

10

20

30

40

50

60

1 2 3 4 5 6 7 8 9 10

Con

tribu

tion

of lo

st ti

me

in d

wel

l tim

e (%

)

Boarding load

tb = 2s

tb = 4s

tb = 5s

tb = 2s

tb = 4s

tb = 5s



7.5 Discussion The bus dwell time is a crucial factor in the design of bus rapid transit system. Along

with some other factors, it determines journey time and adherence of the bus service

with its schedule. Station bus capacity, which is determinant of line bus capacity,

also depends on the bus dwell times. With the recent increase in emphasis of

providing real time information, accurate estimation of bus dwell times at stations

also becomes more important. The new methodology for dwell time estimation has,

therefore, multiple implications while designing a bus rapid transit system. The key

features of BSDT model are –

1) The new model is sensitive to multiple linear loading area operation.

2) It is also sensitive to the platform crowd and passenger walking to bus entry

door, through the bus lost time variable.

The finding that the impact of bus lost time on bus dwell time varies with boarding

load could potentially help in the development of new tools to optimise service

frequencies to achieve reduced dwell times. Whereas, the discovery of the

relationship between lost time and marginal boarding time relation is vital to derive

full advantages of the improvement in fare collection methods, such as a smart card

system. The benefits in bus dwell from reduction in marginal service time could be

substantially nullified due to bus lost time.

Another important finding is the viability of bus dwell time across the loading area.

This variability will impact the bus capacities of individual loading area and therefore

the combined bus capacity of the platform.

Another important application of the proposed dwell time estimation methodology is

in the area of development of bus arrival algorithms; for example, the real time

information systems and bus priority signal design. The proposed dwell time

methodology, when incorporated with the bus arrival algorithms could greatly

improve the accuracy as the results will be based on the prevailing conditions of the

upstream stations. Additionally, this new methodology can help transit planners in

improving the scheduling of service timetable and this in turn could greatly enhance

the travel time reliability.



7.6 Chapter close This chapter presented a bus dwell time model for a busway station. The new model

is called as BSDT model. The BSDT model considers the bus lost time variable in

the dwell time calculation and therefore reflects more accurately the delay to a bus at

a busway station. Bus dwell time is an important parameter in determining the bus

capacity of a loading area. In the following chapter a refined model for loading area

bus capacity estimation is presented. The new loading area bus capacity model

incorporates the new dwell time model incorporating the bus lost time variable.

Busway Loading Area Bus Capacity Model


Chapter Eight

Busway Loading Area Bus Capacity

8.1 Overview This chapter describes the development of busway loading area bus capacity model.

This new capacity model incorporates the new busway dwell time model, presented

in chapter 7, which in turn incorporated the bus lost time variable, described in

chapter 6.

The next section portrays the approach to model development and discusses the

variables influencing the busway loading area bus capacity. Section 8.3 presents the

busway loading area capacity model. Section 8.4 presents the model for effective

loading area bus capacity and Section 8.5 presents the model for busway station

bus capacity.

Section 8.6 demonstrates calculations for an example case and discusses the

effects of bus lost time on loading area bus capacity. Section 8.7 provides a

discussion on the new model. The chapter closes with section 8.8.

8.2 Approach to busway loading area bus capacity model Loading area bus capacity, by definition, means the number of buses that are able to

provide service at that loading area in a given period of time, say an hour. It

incorporates the time each bus dwells, and time it take to clear the loading area. In

simplistic terms, the loading area bus capacity, B, may be estimated by -

3600 Equation 8.1

A more detailed model for estimation of bus capacity of a loading area depends on

dwell time, clearance time, dwell time variability considering failure rate (TRB, 2003).

Further, this model incorporates, in fairly simplistic terms, the effects of immediately

Model



adjacent signals, which eliminate some of the time available to clear buses. The

Equation 2-19 is reproduced here as Equation 2.19.

3600 ⁄ ⁄ Equation 8.2

Where,

lB = Loading area bus capacity (bus/h)

3600 = Number of seconds in an hour

Cg = Green time ratio (the ratio of effective green time to total traffic signal cycle length; 1.0 for unsignalised streets and bus facilities like busway without adjacent signal control)

ct = Clearance time (s)


= Standard normal variate corresponding to a desired failure rate

= Coefficient of variation of dwell time

The various determinants of busway loading area bus capacity are discussed below.

8.2.1 Busway dwell time Dwell time was earlier defined in chapter 7 as the average amount of time a bus

stopped at a loading area at the busway station to provide the service to its

passengers. This includes the time required for door opening and closing, and the

time lost by the bus due to the delay in arrival of its first passenger.

8.2.2 Dwell time variability The concept of dwell time variability is used by Transit Capacity and Quality of

Service Manual (TRB, 2003) for its kerbside loading area bus capacity model. It

accounts for the consistency (or lack thereof) of dwell times among buses using the

loading area. The dwell time variability in combination with failure rate are used to

provided operating margin for the bus operation.



8.2.3 Failure rate A bus cannot always find the subject loading area empty loading area upon its

arrival. The rate at which such condition occurs is called the failure rate (TCRP,

1997, TRB, 2003).

8.2.4 Operating margin due to passenger service time variability The equation to estimate the operating margin based upon dwell time variability at

kerbside bus stop as given by Transit Capacity and Quality of Service Manual (TRB,

2003) is

Equation 8.3

Where,

= Operating margin for passenger load variability (s)

= Standard normal variable corresponding to a desired failure rate

= Coefficient of variation of bus dwell time


The Equation 8.3 is based on the assumption that the dwell times are normally

distributed. Table 8.1 reproduces the Table 2.4, which provides the value of ‘z’

corresponding to desired failure fate.

Table 8.1: Failure rates and corresponding ‘z’ values

Failure rate z 1.0% 2.330 2.5% 1.960 5.0% 1.645 7.5% 1.440 10.0% 1.280 15.0% 1.040 20.0% 0.840 25.0% 0.675 30.0% 0.525 50.0% 0.000

Source: TRB, 2003



In this research a new bus dwell time model was developed to estimate average

dwell time at busway station which includes a variable for bus lost time. The bus lost

time variable at busway station is loading area specific and depends on the bus and

its first passenger. Unlike kerbside bus stop dwell time, which was assumed as

normally distributed (TRB, 2003), the bus lost time component studied through this

research was found to be log-normally distributed. Therefore, the Equation 8.3 is

modified as Equation 8.4 to make it appropriate for busway station analysis. This

equation is called as passenger service time operating margin. The equation

accounts for the fluctuation in service time due passenger loads between buses and

routes.

, , Equation 8.4

Where,

, = Operating margin for passenger service time variability (s) for nth loading area

= Standard normal variable corresponding to a desired failure rate

, = Coefficient of variation of passenger service time at busway loading area

= Busway dwell time at nth loading area (s), as defined in chapter 7.

= Average bus lost time at nth loading area (s)

In the above equation the bus lost time component of busway dwell time equation

was removed to calculate the operating margin for dwell time variability. The

remaining component i.e. ( ) is identical to that for the standard kerbside

bus dwell time model of TCQSM (TRB, 2003). The coefficient of variation here,

however, is that of the passenger service time. The operating margin due to the bus

lost time variability will be treated separately due to its different statistical properties

from the marginal service times.

8.2.5 Lost time variability Not every bus at a given loading area experiences the same amount of lost time.

The bus lost time fluctuates with the waiting position of its first passenger. Similar to

the concept of operating margin to account the bus dwell time variability bt way of



failure rate (TRB, 2003), this research proposes a variable, lost time operating

margin (toml), to account for bus lost time variability by way of failure rate. The lost

time operating margin is a value by which the bus lost time at a given loading area

will increased for the desired failure rate and is additive to the mean bus lost time.

Figure 8.1: Log-normal density curve

Statistically, the area under the log-normal distribution to the right of point LTi (the

shaded area in Figure 8.1) represents the probability that the lost time for a bus at a

given loading area will be longer than the LTi. Z is the corresponding variates, given

by:

Equation 8.5

Equation 8.5 may be rearranged to yield Equation 8.6

Equation 8.6

Den

sity

Bus lost time, LTLTi

Mea

n lo

st ti

me

Bus lost time operation margin



Note that parameter µ and σ are not the mean and standard deviation of variable,

bus lost time but the log of the variable.

Subsequently, the operating margin for bus lost time can be given by

Equation 8.7

Replacing the mean lost time ( ) with equation 6-3

⁄ Equation 8.8

The generalised model for lost time operation margin is therefore given as:

, ⁄ Equation 8.9

Where,

, = Lost time operating margin for nth loading area (s)

= Standard normal variable corresponding to a desired

failure rate

& = Lost time probability curve parameters for nth loading area

8.3 Busway loading area bus capacity model From indicative logic, the bus capacity of nth loading area of a busway station is

given by:

3600

, , Equation 8.10



Where,

= Bus capacity of nth loading area (bus/hr)

3600 = Number of seconds in 1 hour

= Bus dwell time at nth loading area (s)

, = Operating margin for passenger load variability (s)

, = Operating margin for bus lost time variability at nth loading area (s)


It is noted that this model applies to a loading area away from the influence of

signalised intersection. Equation 8.10 may be enhanced to yield Equation 8.11 when

adjacent signals impact loading area operation. The model is known as BSLC model,

an abbreviation of Busway Station Loading Area Bus Capacity Model.

3600

, , Equation 8.11

Where,

= green time ratio (the ratio of effective green time to total traffic signal cycle length, equals 1.0 for unsignalised streets and bus facilities like busway without adjacent signal control)

8.4 Effective bus capacity of loading area Both Equation 8.10 and Equation 8.11 estimate the maximum bus capacity of the

loading area. However, a linear loading area may potentially reduce the effective bus

capacity of other loading areas because it may block the bus entry to those loading

areas. Any time duration for which a loading was blocked, lessens availability of

effective time in which it can provides any actual service to the passengers.

Therefore the effective loading area bus capacity is, therefore, a product of

maximum bus capacity and its efficiency factor and can be calculated as



, Equation 8.12

Where,

, = Effective bus capacity of nth loading area (Bus/ hr)

= Bus capacity of nth loading area (Bus/ hr)

= Efficiency factor of nth loading area

The discussions and modelling of linear loading areas efficiencies are presented in

chapter 9.

8.5 Busway station platform bus capacity The effective capacity of the busway station platform will be equal to the summation

of effective bus capacities of all variable loading areas, and can be given by

Equation 8.13.

, Equation 8.13

Where,

= Busway station platform bus capacity (Bus/ hr)

, = Effective capacity of nth loading area (Bus/ hr)

= Total number of loading areas available at busway station platform

8.6 Example application The significance of the bus lost time is demonstrated by estimating the station bus

capacity for the same example busway station described in Section 7.4. The details

of the example busway station are redescribed here. The Transit Capacity and

Quality of Service Manual (TRB, 2003) methodology was used for base calculations

and the results were compared with the new methodology considering bus lost time

which is presented earlier in this chapter. For this purpose, a boarding load of 7

passengers per bus with boarding service time of 4s and no alighting load was

considered. The estimation was done with assume failure rate of 7.5%, 60%



coefficient of variation (TRB, 2003), 3 linear loading areas, g/c of 1, 10s clearance

time, and door opening and closing time of 2s. The mean bus lost time obtained from

Table 6-14 and loading area efficiency factors from Table 9.3 were used for revised

capacity design. The loading area efficiency factors for TQCSM method are obtained

from Table 9-1 were used. Table 8.2 shows the comparison of the results from the

two methods. The calculation steps are provided in details in Appendix B.

Table 8.2: Example demonstration

Method Bus dwell time

Station Capacity Loading area

1 Loading area

2 Loading area

3

TCQSM (without bus lost time)

30s 30s 30s 143 bus/hr

BSLC (with bus lost time)

34.2s 32.9s 35.2s 123 bus/hr

% change +14.0% +9.7% +17.3% -13.9%

The above table demonstrated the reduction in station bus capacity value when the

bus lost times were accounted in the busway station bus capacity estimation. For the

given demonstrative example, 13.3 percent reduction in bus capacity was observed.

For the above example, Figure 8.2 shows the variation in busway station bus

capacity station with boarding load per bus, calculated using both methods: TSCQM

and BSLC model. As expected the bus capacity deceases with an increase in

boarding load. This is due to the fact the buses remain at the loading area for longer

with an increase in boarding load. However, the difference in bus capacity estimation

between both methods decreases with an increase in boarding load. This illustrates

the effect of bus lost time on capacity with the increase in boarding load. This finding

is reasonable because, as discussed in past chapters, the lost time for a bus occurs

due to its first boarding passenger only. Other boarding passengers do not contribute

to lost time.



Note: busway station with three linear loading area, no alighting load, tb=4s, failure rate 7.5%, cv=0.6, tc=10s. Figure 8.2: Variation in busway station bus capacity with boarding load per bus

Note: busway station with three linear loading area, no alighting load, failure rate 7.5%, cv=0.6, tc=10s. Figure 8.3: Effect of bus lost time on busway station bus capacity

0

50

100

150

200

250

300

350

400

450

500

1 2 3 4 5 6 7 8 9 10

Bus

way

sta

tion

capa

city

(Bus

/ hr

)


TCQSM

0

5

10

15

20

25

30

35

40

1 2 3 4 5 6 7 8 9 10

Effe

ct o

f bus

lost

tim

e on

sta

tion

capa

city


tb = 2s

tb = 4s

BSLC



Therefore, as the boarding load increases, the share of lost time in the total dwell

time of the bus decreases, resulting in its lesser influence on capacity.

On the contrary, when marginal service time decreases, the influence of lost time on

capacity increases, as seen from Figure 8.3. This is because, with a decrease in

marginal service time, the share of lost time in the total dwell time of the bus

increases.

Figure 8.4 provides busway station bus capacities for various g/C ratios and

boarding loads, assuming no alighting load from entry door. The figure illustrates the

impact of adjacent signal on loading area bus capacity. As seen from the graph, the

station bus capacity decreased with decrease with g/C ratio.

Note: busway station with three linear loading area, no alighting load, tb=4s, failure rate 7.5%, cv=0.6, tc=10s. Figure 8.4: Estimated busway station capacities

8.7 Discussion Determining the reliable and accurate loading area bus capacity is absolutely critical

for design of a bus transit system. Specially, in case of a busway system, where its

stations are the major source of delay to buses, the precise estimation of station bus

capacity becomes more important in determining the line capacity and design period.

0

50

100

150

200

250

300

350

1 2 3 4 5 6 7 8 9 10

Sta

tion

bus

capa

city

Boarding load

g/C = 1

g/C = 0.7

g/C = 0.5



With the bus lost time accounted for the station bus capacity estimation, the new

methodology better approximates the station operation. The key features of the

BSLC model are,

1) It is sensitive to bus lost time and its variability between loading areas.

2) It is sensitive to the platform crowd and passenger walking to the bus entry

door.

The comparison of station capacities, estimated from TCQSM method (TRB, 2003)

and the BSLC model, presented in this chapter, showed that the busway station bus

capacity reduced when bus lost time was considered. This shows that bus lost time

affects the busway station bus capacity and needs to be considered.

With the knowledge of bus lost time characteristics, better operation and

management policies can be developed. For example, the advantages of a reduction

in marginal service time can be undermined because of bus lost time. Any gain in

bus dwell time might be limited until steps towards reducing bus lost time are made.

Furthermore, a busway line can potentially serve as trunk line in a trunk and branch

type arrangement of public transport network. The buses on the trunk line would be

of high frequency and high speed service connected to feeder services, which

branch out deep into the suburban areas. With the improvement in the

understanding of operation of busway and its stations, a better coordinated service

may be created.

All steps involved in busway station platform bus capacity analysis are compiled in a

worksheet provided in Appendix A.

8.8 Chapter close This chapter presented a loading area bus capacity model for busway station. The

new model is called as BSLC model, an abbreviation for Busway Station Loading

Area bus Capacity model. The BSLC model incorporated the new busway dwell time

model (BSDT) and the bus lost time variable.

The next chapter presents the methodology to estimate the loading area efficiency

and effective numbers of loading areas for a busway station platform.

Busway Station Efficiency Model


Chapter Nine


9.1 Overview Under a perfectly ideal situation each loading area of a busway (BRT) station

platform should operate without interfering with the others’ operation. However in a

linear arrangement of loading areas this may not be the case, particularly during the

peak period operation when the bus flow is high. Every loading area can potentially

obstruct the bus entry to the platform (as buses are not permitted to slip into

downstream, unoccupied loading areas) and obstruct exiting from upstream loading

areas. This reduces the number of effective loading areas for the station from the

physical number, as noted by TCQSM (TRB, 2003).

This chapter describes the development of a methodology for estimating busway

loading area efficiency. Section 9.2 first discussed the operation of Mater Hill

Busway station. Later the approach for loading area efficiency calculation was

described in section 9.3.

Section 9.4 estimates the loading efficiencies based on the experience of the Mater

Hill busway station, Brisbane, Australia. Section 9.5 provides a discussion on loading

area efficiency. Section 9.6 closes the chapter.

9.2 Loading area blocking Non-uniform bus dwell time across different loading areas of a busway platform may

lead to blocking of a loading area by its predecessor loading area/s. The Non-

uniformity in bus dwell times may happen due the variation in bus lost time,

illustrated in figure 5-6 and the number of passenger boardings and alightings

between loading areas (TRB, 2003).



The study platform i.e. outbound platform of Mater Hill busway station, Brisbane, has

a total of three loading areas. The station platform reaches its ideal bus capacity

when all the three loading areas are occupied by buses. However, as the buses use

the loading areas on a first come first in basis, the station may reach a non-ideal

capacity when one or both of loading area 1 and 2 are empty but a preceding loading

area is occupied. Loading area 3 cannot experience blocking since it does not have

any predecessor loading area. Figure 9.1 shows bus arrivals and departures

progression observed at the study platform. The ‘red line’ represents buses which

used loading area 1 for passenger servicing. Similarly, the ‘orange line’ and ‘blue

line’ represent buses used loading area 2 and 3 respectively. A dotted line means

bus in queue at station entry, behind the loading area 3. Whereas, a solid line means

the bus is dwelling at a loading area.

The start point of a dotted line represents the entry of the bus in queue. The start of

a solid line represents the arrival of that bus at the loading area. The end point of the

solid line represents the departure of that bus from the loading area after completing

passenger servicing. The length of solid line gives the time spent by the bus at the

loading area, including its dwell time. The length of dotted line gives the time spent

by that bus in the queue. A lack of a dotted line for any bus means that bus it is able

to enter an empty loading area directly upon its arrival at the station platform.

During the non-peak period, bus arrivals at the station platform were random and

most of the buses found an empty loading area at the platform. With the onset of

peak period the flow of bus increased and similarly so their dwell time at the

platform. This caused the loading area to be blocked by buses in its predecessor

loading areas as Figure 9.1 illustrates. Loading area 1 can be seen in this figure

being blocked by either loading area 2, 3, or both. Similarly, loading area 3 blocks

loading areas 2 and 1.



Figure 9.1: Trajectory of bus processing at the Mater Hill Busway Station (Outbound platform)

0 100 200 300 400 500 600 700 800Time (s)

Loading area 1 blocked by loading area 2

Sta

rt of

afte

rnoo

n pe

ak p

erio

d

Loading area 3 blockingloading areas 2 & 1.

Platoon arrivalat platform.

Platoon departurefrom platform. Platoon arrival

at station.



There were five different scenarios of inter-loading area blocking observed at the

study platform as shown in Figure 9.2. A loading area which is empty but blocked is

shown in red shading. A loading area occupied a bus is shown in gray shading. An

empty loading area available for bus is shown in white shading.

Scenario 1: All the three loading areas are available for buses to serve the platform.

The available bus capacity at the platform is 100 percent. The platform is performing

under capacity.

Scenario 2: Loading area 2 is occupied, which results in blockage of loading area 1.

In this scenario, the platform has one third of its capacity blocked. Of the remaining

two thirds of its capacity, one third is in utilization and other third is available for

servicing. The platform is still performing under capacity

Scenario 3: Loading area 2 and 3 are occupied, which results in blockage of loading

area 1. In this scenario, the platform has one third of its capacity blocked and the

remaining two third of its capacity is in utilization. The platform is virtually performing

at full capacity with a loss of one third of its total capacity.

Scenario 4: Loading area 1 and 3 are occupied. However, the bus at loading area 3

is blocking entry to loading 2. In this scenario, like scenario 3, the platform has one

third of its capacity blocked and remaining two third capacity been utilised. The

platform is said to be virtually performing at full capacity with a loss of one third of its

total capacity.

Scenario 5: Loading area 3 is occupied and is blocking loading areas 1 and 2. In

this scenario the platform has two third of its total capacity blocked and only one third

utilised. The platform is said to be virtually performing at full capacity with a loss of

two third its total capacity.

Scenario 6: All the three loading areas are occupied, resulting ideal capacity for the

platform. The platform is performing at full capacity with no loss in capacity.



Scenario Diagram Nel

1

3

2

2

3

2

4

2

5

1

6

3

Note: Nel = Number of effective loading areas Figure 9.2: Inter-loading area blocking scenarios and associated numbers of effective loading areas.

Loading area 1

Outbound platform

Loading area 2 Loading area 3

Loading area 1

Outbound platform


Loading area 1

Outbound platform


Loading area 1

Outbound platform


Loading area 1

Outbound platform


Loading area 1

Outbound platform




These blockings of loading areas result in reduction in their efficiencies causing a

decline in effective number of loading areas. The right most column of Figure 9.2

provides the number of effective loading areas (Nel) available at the platform at the

time under each scenario.

In Figure 9.2, scenarios 2, 3, and 4 have equal numbers of effective loading areas.

However, in scenario 2, the platform is not at its full capacity. There is a spare

capacity of one loading area. Therefore, arrival of one additional bus at the platform

will not result in formation of a queue at platform entry, i.e. behind loading area 3. On

the other hand, in scenario 3 and 4, the platform is at its full capacity, with no

available loading for any further bus. Arrivals of additional buses at the platform

under these scenarios cause these buses to queue behind loading area 3. This

situation could potentially lead to scenario 5, where loading area 3 blocks the entry

to loading areas 1 and 2. This causes the platform to lose its two third of its total

available loading areas. Under a continuous bus arrival situation, this may result in

bus arrivals at station, arrival at platform, and departure from the platform, in

platoons, as highlighted in the Figure 9.1.

9.2.1 Existing approach

The TCQSM (TRB, 2003) noted that each additional loading area provided linearly at

the platform has a reduced contribution to the total effective number of loading

areas. Table 9.1 provides the efficiency factors given by the TCQSM. However,

these suggested efficiency factors are based on the operation of a bus terminal

facility (TRB, 2003; Levinson et.al., 1975) and not based on the operation of busway

station facility.

Table 9.1: Efficiency factors provided by TCQSM

loading area Efficiency %

Cumulative # of Loading Areas

3rd 1 1 2nd 0.85 1.85 1st 0.80 2.65

Source: TRB, 2003



9.3 Approach to loading area efficiency factor calculation Loading area efficiency is a dynamic factor, sensitive to the bus flow rate. It can

change from time to time. The loading area efficiencies can be higher under a low

bus flow condition compared to high bus flow condition. Therefore, an afternoon

peak period was analysed to determine conservative values.

Theoretically, loading area 3 will not be blocked at any time because no bus will

stand in queue if loading area 3 is empty. But a bus on loading area 3 can block the

entry to empty loading area 2. For instance, consider that all loading areas are

occupied by buses and there is a bus queued behind loading area 3. A bus on

loading area 2 moves out of the platform, making this loading area available for the

next bus. But if the bus on loading area 3 is still dwelling at the platform, it blocks the

entry to loading area 2. During data collection, the amount of time that loading area 2

is blocked because of the presence of a bus on its upstream loading area was

recorded as blocked time. The efficiency of loading area 2 is given by Equation 9.1.

, Equation 9.1

Where,

E2 = Efficiency of loading area 2 T3

= Total time that loading area 3 is occupied during time T T2,b = Total time that loading area 2 was empty while a bus occupied loading

area 3 during time T T = analysis period

Similarly a bus on loading area 2 or loading area 3 can block loading entry to empty

loading area 1. The efficiency of loading area 1 of the platform at the station is given

by Equation (4).

, ,

, Equation 9.2



Where,

E1 = Efficiency of loading area 1 T2,3 = Total time that loading area 3 OR loading areas 2 and 3 are occupied

while there is a queue behind loading area 3 during time T T1,b = Total time that loading area 1 was empty while a bus occupied loading

area 2 OR loading area 3 OR both loading areas 2 and 3

The number of effective loading areas (Nel) on the three loading area station platform

may then be calculated using Equation 9.3.

1.0 Equation 9.3

9.4 Loading area efficiency factors for Mater Hill Busway Station Table 9.2 shows the occupancy rate and blocking rate for each loading area during

the half hour analysis time period (T=1800s).

Table 9.2: Occupancy and blocking rates for loading areas at outbound platform of Mater Hill Busway Station (Afternoon peak period)

Loading area Occupied Time Blocked Time

1 790s 372s

2 958s 82s

3 915s 0s

Table 9.3 presents the calculation of the number of effective loading areas on the

outbound platform at Mater Hill Station, during the analysis period, T.



Table 9.3: Number of effective loading areas calculation for bus station platform

Loading Area

Time preceding loading areas

occupied

Time loading area empty while

preceding occupied

Loading Area

Efficiency

Cumulative Loading Areas

3rd NA NA 1.00 1.00

2nd T3 = 814s T2,b = 82s 0.90 1.90

1st T2,3 = 1244s T1,b = 362s 0.71 2.61 Note: Analysis period, T = 1800s

These results indicate that, during peak period, the outbound platform of Mater Hill

busway station has 2.61 effective loading areas out of the available 3 loading areas.

In other words, the platform has a loss of 13 percent of its total available physical

loading areas. The results welcomely resemble the default values from TCQSM

(TRB, 2003) as shown in Table 9.4.

Table 9.4: Comparison of loading area efficiency results

Number of available loading area

Mater Hill Busway station

TCQSM’s Cumulative Loading Areas for Comparison

1 1 1

2 1.9 1.85

3 2.61 2.65

9.5 Discussion The loss in the effective number of loading areas on a busway (BRT) platform results

in a reduction in its bus capacity. This reduction amplifies queuing of buses at the

station platform entry. This, in turn, tends to result in a higher level of platooning.

Platooned arrival means the station is forced to function at its full capacity and with

no opportunity to fall back below capacity. It also means that the last loading area

(Loading area 3, in the case of study station) would have increased usage,

subsequently resulting in a rise in blocking of the downstream loading areas.

The arrival of buses in platoon at a station suggests that buses are carrying the

effects of operation of the upstream station to that station. Similarly, the departure of

buses in platoon from that station shows that buses are carrying forward the effects

of that station. This was referred to in Section 3.4 as station – station interface.



Detailed analyses of platoon arrivals and departures were not in the scope of this

research but warrant further investigation. Further research on bus discipline at

stations, passenger ticking and management, including prepaid ticketing and all-door

usage should also be pursued in future.

9.6 Chapter close This chapter presented a methodology to estimate the efficiencies of linearly placed

loading areas at a busway station platform. The estimated efficiencies were used in

the demonstrative example for station platform bus capacity estimation presented in

section 8.6.

Next chapter concludes this thesis and discusses the future research opportunities in

the area of bus lost time to enhance the models.

Conclusions


Chapter Ten

Conclusions

10.1 Overview This chapter concludes this thesis and summarises the analyses, discussions and

results presented in past chapters. A brief thesis summary is provided in section

10.2, followed by section 10.3 which discusses the contributions of this research to

the existing knowledge and practice. Later, section 10.4 discusses the implications

of the research findings for both theory and practice. Section 10.5 provides the

conclusions of this research work and finally the recommendations for further

research are given in section 10.6.

10.2 Summary of this thesis This thesis presented an intricate analysis of busway station operation and identified

various parameters and processes related to passenger waiting behaviour and bus

dwell time. Chapter 1 established the aim and objectives of this research and the

scope of this thesis. Having set up the research goals, an extensive literature survey

was conducted to study the existing knowledge in the area of busway bus capacity

analysis. The findings of the literature survey were presented in Chapter 2. The

literature search found that no methodology available for busway station bus

capacity analysis which addresses all of the mechanisms considered here to be

important.

Based on the gaps identified, the research problem was formalised in Chapter 3. The

chapter studied the operations of the busway station compared with the traditional

kerbside side bus stop and developed a framework of busway station operation. The

framework identified and elaborated various parameters and processes occurring at

four tiers of busway operation (platform, vehicle, station, and line) and their influence

on each another. The analyses lead to the identification of a previously neglected



process, which this study refers in this thesis as passenger – bus interface, and a

new parameter, termed here as bus lost time.

To study the previously neglected process and parameter, and to identify their

influence on busway station operation, pertinent data was collected. Chapter 4

detailed the development of data collection and processing methodologies for this

research. This chapter presented a matrix based concept for data mining to achieve

homogeneous data for analysis.

Four specific models were then developed in this research, one model each to

estimate bus lost time, busway dwell time, busway loading area bus capacity, and

busway loading area efficiency.

Chapter 6 analysed and modelled bus lost time. Stochastic models were developed

for this previously neglected variable, and its descriptive statistics were established.

With the finding of the bus lost time variable and its roles in operation of a busway

station, a new busway dwell time model was developed in Chapter 7, and a new

busway loading area bus capacity model was developed in Chapter 8. This new

formulation of busway dwell time, in addition to passenger boarding and alighting

load and their marginal service times, also account for bus lost time.

The third model developed in this research was busway loading area bus capacity

model, presented in Chapter 8. The new busway loading area bus capacity model

can better approximate the busway loading area operation for two reasons: firstly, it

incorporates the more accurate busway dwell time model and secondly, it accounts

for variation in bus lost times. A new term, lost time operating margin, was defined,

which allows for variability in bus lost time between buses. Steps for calculating

busway station bus capacity were also presented this chapter.

The total bus capacity of a platform with multiple linear loading areas depends on the

efficiencies of each loading areas. Linear loading areas tend to interfere with the

smooth operation of adjacent loading areas. Such interference causes reductions in

efficiencies. Impacts of one linear loading area on another were studied in Chapter 9.

Based on the experience of the study busway platform, loading area efficiency

Conclusions


models were developed and efficiency factors were estimated for linear loading

areas.

In order to demonstrate the influence the bus lost time has on busway dwell time and

busway bus capacity, demonstrative examples were presented. Through these

examples, the changing patterns of dwell time and bus capacity at a busway platform

were illustrated.

10.3 Contributions of this research From the knowledge prospective, this research identified for the first time the phase

of passenger – bus interface which occurs at a busway station. This insight then lead

to the most important contribution of this research, identification of bus lost time as

an important parameter of influence to bus dwell time. Subsequently, a new bus

dwell time model and loading area bus capacity model were developed for a busway

station.

From the practice perspective, the main contribution of this research is the

development of a more refined tool for estimating busway station bus capacity. In

order to develop this tool, this research has,

i. Developed a framework of busway operation.

ii. Identified for first time the interface between passenger and bus.

iii. Established descriptive characteristics of the busway station bus lost time.

iv. Established the effectiveness of linear loading areas in busway station bus

capacity, under local conditions.

10.4 Implications of this research Though this research used a particular busway station platform as its case study, the

findings of this research may be applied to any station with multiple linear loading

areas. The results from this research have applications both for research and

practice. For research, the knowledge of the passenger – bus interface can be

applied in better understanding passenger movements in the platform area. This can

be used in developing better platform designs to reduce bus lost time.



The concept of bus lost time could be applied to develop a two tier real-time

information system for linear loading area busway station. The first tier of the

information would provide the expected arrival time of a bus at the station. Whereas

the second tier of the information would provide the loading area number for that

bus. Such information should be provided to passengers very close in time to the

actual arrival of the bus, in order to reduce bus lost time.

From a practical application prospective, the new methodology for estimating bus

dwell time at a busway station can help transit planners in improving scheduling and

in turn could greatly enhance the travel time reliability. The new methodology, which

is specifically designed for a linear loading area station can help transit planners to

perform robust capacity analysis of future bus transit system. This tool can naturally

also be applied in evaluating an existing station. Since this methodology of capacity

analysis accounts for the impact of passenger crowd at a busway station, growth in

patronage can be better incorporated in future policies; in particular in the design of

service frequencies.

10.5 Conclusions In conclusion, this thesis found that the traditional approach of bus dwell time

estimation may not be reliable for a busway station. This is because of the

significance of additional complex variables, such as the presence of large crowds,

multiple linear loading areas, multiple bus services, bus queuing and lost times.

These complex variables have been analysed and subsequently, new models for

dwell time estimation and platform bus capacity were presented in this thesis.

A demonstrative example comparing the traditional approach and new approach for

bus dwell time estimation at a busway station showed that the new approach

estimated the varying dwell time values for different loading areas. It showed that

new approach was able to approximate in finer detail the dwell times of a real

busway platform, compared to the single dwell time value for all the loading areas

given by the traditional model.

Conclusions


The comparison of station capacities showed that the bus capacity of a busway

platform reduced with the bus lost time considered. This highlighted the importance

of accounting for bus lost time in deciding the final throughput of the station.

Identifying the variables governing the operation of a busway station is absolutely

necessary to explore a busway system’s full potential. The demonstrative examples

have shown that bus lost time at a platform is one such variable, which manipulates

the operational capacity of a busway station.

10.6 Recommendations for future work This research has provided a comprehensive analysis of a multiple linear loading

area busway station and has provided many insight of busway station functioning.

However, further research is required to test the new approach for other busway

station configurations and modes of operation. Consequently, future research is

required in a number of directions, in order to make the results of this research more

robust and versatile.

This research suggested the use of lognormal probability distribution curves for

estimation of bus lost time for three linear loading areas. It would be interesting to

see the impact on lost time distribution if an additional fourth loading area is added to

the current set of three. The bus lost time estimation approach needs to be refined

with the help of more case studies to make it applicable to a wide range of linear

loading area station operations.

Furthermore, this research has only considered the bus lost time due to its first

boarding passenger. Additional research work in the area of bus lost time is

necessary to study the impact of subsequent passengers. Specifically, the impact of

platform crowd on subsequent passenger walking time and in turn on bus lost time

ought to be studied.

The results of this research are based on a predominantly boarding platform, with

very little passenger alighting activity. The impact of alighting passengers on bus lost

time and passenger walking on the platform was not fully studied. More studies

based on the stations where alighting passengers have a substantial impact on

station operation are recommended. Specifically, predominantly alighting platform



would be located in the inbound direction at inner urban platforms, and the morning

peak period would be expected to be critical.

Additionally, this research also recommends the development of a full scale micro-

simulation model for the busway station to refine and strengthen the Busway station

platform capacity methodology suggested by this research.

References


References

Ando, K., Ota, H. and Oki, T. (1988). Forecasting the flow of people, (in Japanese), RRR, Railway Research Review, 45(8), 8-14.

Antonini, G. and Bierlaire, M. (2005). Capturing interactions in pedestrian walking behavior in a discrete choice framework, Technical Report RO-050921, Operations Research Group ROSO, Institute of Mathematics, Ecole Polytechnique F´ed´erale de Lausanne, CH-1015 Lausanne, Switzerland.

Blue, V. J. and Adler, J. L. (2001). Cellular automata microsimulation for modeling bi-directional pedestrian walkways. Transportation Research Part B, 35, 293-312.

Chan, L.Y., Lau, W.L., Zou, S.C., Cao, Z.X. and Lai, S.C. (2002). Exposure level of carbon monoxide and respirable suspended particulate in public transportation modes while commuting in urban area of Guangzhou, Chain. Atmospheric Environment, 36, 5831-5840.

Changshan, W. and Murray, A. T. (2005). Optimizing public transit quality and system access: the multiple-route, maximal covering/shortest-path problem. Environment and Planning B: Planning and Design, 32, 163-178.

Chertok, M., Voukelators, A., Sheppeard, V. and Rissel, C. (2004). Comparison of air pollution exposure for five commuting modes in Sydney – car, train, bus, bicycle and walking. Health Promotion Journal of Australia, 15(1), 63-67.

Cox, T., Houdmont, J. and Griffiths, A. (2006). Rail passenger crowding, stress, health and safety in Britain. Transportation Research part A, 40, 244-258.

Daamen, W. (2004). Modelling Passenger Flows in Public Transport Facilities. PhD Dissertation, Department Transport & Planning, Delft University of Technology, Netherland.

Daamen, W., Hoogendoorn, S. and Bovy, P. (2005). First-Order Pedestrian Traffic Flow Theory. In Transportation Research Record: Journal of the Transportation Research Board, 1841, 120-127.

Davies, A. C., Yin, J. H. and Velastin, A. (1995). Crowd Monitoring Using Image Processing. IEE Electronic and Communications Engineering Journal, 7(1), 37-47.



Devore, J. L. (2008). Probability and Statistics: For Engineering and the Sciences. Thomson/Brooks/Cole Inc., Belmont, CA.

Dipletro, C. M. and King, L. E. (1970). Pedestrian Gap-Acceptance. In Highway Research Record 308, HRB, National Research Council, Washington, D.C., 1970, pp. 80-91.

Dueker, K., Kimpel, T. and Strathman, J. (2004). Determinants of bus dwell time. Journal of Public Transportation, 7(1), 21-40.

Fruin, John J. (1987) Pedestrian Planning and Design, Revised Edition, Elevator World, Inc., Mobile, AL

FTA (2004). Characteristics of bus rapid transit for decision-making. Department of Transportation, Federal Transit Administration, Report number FTA-VA-26-7222-2004.1. Washington D.C.

FTA (2008). Bus rapid transit. Department of Transportation, Federal Transit Administration, Washington DC. Last accessed November, 2009 at http://www.fta.dot.gov/research_4240.html

Gerin-Lajoies, M. and Richards, C.L. (2006). The circumvention of obstacles during walking in different environmental context: A comparison between older and younger adults. Gait & Posture, 24(3), 364-369.

Guenthner, R. P., and Hamat, K. (1988). Transit Dwell Time Under Complex Fare Structure, Journal of Transportation Engineering, 114(3), 367-379.

Guenthner, R. P., and Sinha, K. C. (1983). Modeling Bus Delays Due to Passenger Boardings and Alightings. In Transportation Research Record: Journal of the Transportation Research Board, 915, 7-13.

Hamburg, M. (1983). Probability and Statistical Analysis for Decision Making. 3rd Ed. Harcord Brace Jovanovich, Inc., New York, NY.

HCM (1994). Highway capacity manual. Special Report, 209, 3rd ed. Transportation Research Board, National Research Council, Washington, D.C.

Hensher, D. A. (2007). Sustainable public transport systems: Moving towards a value for money and network-based approach and away from blind commitment. Transport Policy, 14, 98-102.

Hoogendoorn, S. P. and Bovy, P. H. L. (2004). Pedestrian route-choice and activity scheduling theory and models. Transportation Research Part B, 38, 169-190.

References


Hui, M.K., and Bateson, J.E.G. (1991). Perceived control and the effect of crowding and consumer choice on the service experience. The Journal of Consumer Research, 18(2), 174-184.

IAPT (2007). International Association of Public Transport. Last accessed October, 2007 at http://www.uitp.org/

Innes, P. (2007). Testing and fixing for normality. Factsheet No. 15, University of Queensland Business School, Brisbane, Australia.

Jiang, R. and Wu, Q.S. (2005). The moving behaviour of a large object in the crowds in a narrow channel. Physica A, 364, 457-463.

Johansson, A., Helbing, D., Al-Abideen, H. Z., & Al-Bosta, S. (2008). From crowd dynamics to crowd safety: A video-based analysis. Advances in Complex Systems, Vol 11, pp. 479-527.

Khisty, C.J. (1990). Transportation Engineering – An introduction Prentice-Hall, New Jersey.

Kholshevnikov, V. V., Shields, T.J., Boyce, K.E. and Samoshin, D.A. (in press). Recent development in pedestrian flow theory and research in Russia. Fire Safety Journal. doi: 10.1016 j.firesaf.2007.05.005.

Kilambi, P., Ribnick, E., Joshi, A., Masoud, O. and Papanikolopoulos, N. (2008). Estimating pedestrian counts in groups. Computer Vision and Image Understanding, 110, 43-59.

Kim, Won-Ho (2004). An improved bus signal priority system for networks with nearside bus stop. PhD dissertation, Texas A&M University.

Kong, D., Gray, D. and Tao, H. (2005). Counting pedestrians in crowds using viewpoint invariant training, In Proceeding of British Machine Vision Conference. 5-8 September 2005, Oxford Brookes University, Oxford.

Lam, W., Lee, J. and Cheung, C. (2002). A study of the bi-directional pedestrian flow characteristics at Hong Kong signalized crosswalk facilities. Transportation, 29, 169-192.

Lee, W.C.K. (2004). An investigation into the dwell time at Mater Hill Busway station. Bachelor thesis, Queensland University of Technology, Brisbane.

Levine, J. C. and Torng, G. (1994). Dwell time effects of low-floor bus design. Journal of Transportation Engineering, 120, 914-929.



Levinson, H. S. (1983). Analyzing Transit Travel Time Performance. In Transportation Research Record: Journal of the Transportation Research Board, 915, 1-6.

Levinson, H. S., Adams, C. L. and Hoey, W. F. (1975). NCHRP Report 155: Bus Use of Highways: Planning and Design Guidelines. Transportation Research Board, National Research Council, Washington DC.

Levinson, H. S., Zimmerman, S., Clinger, J. and Rutherford, C. S. (2002). Bus Rapid Transit: An Overview. Journal of Public Transportation, 5(2), 1-30.

Liggett, W.D. (1992). Low-Floor Bus--The St. Albert Experience. paper presented at the CUTTER Fall Conference in Laval, Quebec.

Maloney, M. and Boyle, D. (1999) Components of travel time on the Glendale Beeline bus network. In Transportation Research Record: Journal of the Transportation Research Board, 1666, 23-27.

Marshall, L. F., Levinson, L.C., Lennon, L. C., and Cheng, J. (1990). Bus service times and capacities in Manhattan, In Transportation Research Record: Journal of the Transportation Research Board, 1266, 189-196.

Mikovits, M. N. (2008). Modelling the factors affecting bus stop dwell time: Use of automatic passenger counting, automatic fare counting, and automatic vehicle location data. In Transportation Research Record: Journal of the Transportation Research Board, 2072, 125-130.

Nishinari, K., Sugawara, K., Kazama, T., Schadschneider, A. and Chowdhury, D. (2006). Modelling od self-driven particles: Foraging ants and pedestrians. Physica A, 372, 132-141.

Osaragi, T. (2004). Modeling of pedestrian behaviour and its applications to spatial evaluation. In Proceedings of the Third international Joint Conference on Autonomous Agents and Multiagent Syatems, Washington, USA. IEEE Computer Society, 836-843.

Park, H. M. (2008). Univariate Anlaysis and Normality Test Using SAS, Stata, and SPSS. Working paper. The University Information Technology Service (UITS) Center for Statistical and Mathematical Computing, Indiana University.

Rajbhandari, R., Chien, S. I. and Daniel, J. R. (2003). Estimation of Bus Dwell Times with Automatic Passenger Counter Information. In Transportation Research Record: Journal of the Transportation Research Board, 1841, 120-127.

References


Rathwell, S. and Schijns, S. (2002). Ottawa and Brisbane: Comparing a Mature Busway System with Its State-of-the-Art Progeny. Journal of Public Transportation, 5(2), 163-182.

Seyfried, A., Passon, O., Steffen, B., Boltes, M., Rupprecht, T., and Klingsch, W. (2007). New insight into pedestrian flow through bottlenecks, Transportation Science, 43(3), 395-406.

Seyfried, A., Steffen, B., Klingsch, W. and Boltes, M. (2005). The fundamental diagram of pedestrian movement revisited, Journal of Statistical Mechanics: Theory and Experiment, 10, 1-13.

Shen, L. D., Elbadrawi, H., Zhao, F. and Ospina D. (1998). At Grade Busway Planning Guide. Center for Urban Transportation Research, Florida International University, The State University of Florida at Miami, Miami, FL.

Smith, R. A. (1995). Density, velocity and flow relationships for closely packed crowds. Safety Science, 18, 321-327.

Tarawneh, M. S. (2001). Evaluation of pedestrian speed in Jordan with investigation of some contributing factors. Journal of Safety research, 32(2), 229-236.

TCRP (1997). Transit Capacity and quality of service. TCRP web document 6, Transportation Research Board, Washington, D.C.

TCRP (2003). Bus rapid transit volume 2: Implementation guidelines. TCRP Report 90. Transportation Research Board, Washington, D.C.

Teknomo, K., Yasushi, T. and Hajime, I. (2001). Tracking System to Automate Data Collection of Microscopic Pedestrian Traffic Flow, In Proceeding of The 4th Eastern Asia Society For Transportation Studies, Hanoi, Vietnam, 3(1), 11-25.

TransLink (2007). Unpublished Report (http://www.transport.qld.gov.au/)

TransLink (2009). South East Busway: Rochedale to Springwood. Last accessed November, 2009 at http://www.transport.qld.gov.au/

Translink web site: www.translink.com.au

TRB (2003). Transit Capacity and Quality of service Manual. Transportation Research Board of the National Academics, Washington, D.C. Last accessed February, 2007 at http://www.trb.org/.



Tyler, N. (2002) Accessibility and the bus system: from concepts to practice. Thomas Telford. London.

Vuchic, A. R. (2005). Urban Transit: Operations, Planning, and Economics. John Wiley and Sons, Inc., New Jersey.

Wikipedia (2009a). Ottawa Rapid Transit. Last accessed November, 2009 at http://en.wikipedia.org/wiki/Ottawa_Rapid_Transit

Wikipedia (2009b). Log-normal Distribution. Last accessed November, 2009 at http://en.wikipedia.org/wiki/Log-normal_distribution

Wikipedia (2009c). Kurtosos. Last accessed November, 2009 at http://en.wikipedia.org/wiki/Kurtosis

Wirasinghe, S. C. and Ghoneim, N. S. (1981). Spacing of bus-stops for many to many travel demand. Transportation Science, 15, 210-221.

Zhao, F. and Li, M. (2005) Calibration of highway/ transit speed relationships for improved transit network modelling in FSUTMS. Draft final report, Florida Department of Transportation, Tallahassee, US.

Zografos, K. G., and Levinson, H. S. (1986). Passenger service times for a No-fare bus system, Transportation Research Record 1051, 42-48.

Appendix- A


Appendix – A

BUSWAY STATION PLATFORM BUS CAPACITY ANALYSIS WORKSHEET

Diagram

Facts Platform Outbound Inbound Loading area type = S Linear b Analysis period Peak Off-peak # of loading areas (N) = ________(N ≤ 3)

Input # boarding Passenger Pb = ____ Parameter, µ, for bus lost time (s) (Table 6-14)Marginal boarding time tb = ____ Loading area 1 = _________ # alighting Passenger Loading area 2 = _________

Front door Pa = ____ Loading area 3 = _________ Rear door Pa’ = ____

Marginal alighting time ta = ____ Parameter, σ, for bus lost time (s) (Table 6-14)

Door opening &closing time toc = ____ Loading area 1 = _________ Clearance time tc = ____ Loading area 2 = _________ Failure rate = ____ ~~ Z = ____ Loading area 3 = _________ Green time ratio g/c = ____

Coefficient of variation for passenger service time

Cv,p = ____ Mean lost time (s) (Table 6-15 for peak period & table 6-16 for off-peak period)

Loading area efficiency (Table 9-3) Loading area 1 LT1 = _________

Loading area 1 E1 = ______ Loading area 2 LT2 = _________ Loading area 2 E2 = ______ Loading area 3 LT3 = _________ Loading area 3 E3 = a 1.0 t

Cont…

Space for diagram



Cont…

Calculation steps

1. Busway dwell time

(s) Equation 7-1

2. Operating margin for passenger service time

, , (s) Equation 8-4

3. Operating margin for bus lost time

, ⁄ (s) Equation 8-9

4. Loading area bus capacity

3600

, , (bus/ hr) Equation 8-10

5. Effective bus capacity of loading area , (bus/ hr) Equation 8-12

6. Platform bus capacity

, (bus/ hr) Equation 8-13

Capacity Calculation Loading area LTn DTn tomp,n toml,n Bn En Bef,n

1 ____ ____ ____ ____ ____ ____ ____

2 ____ ____ ____ ____ ____ ____ ____

3 ____ ____ ____ ____ ____ ____ ____

Platform bus capacity (Bs) = _________ (bus/hr)

Appendix- B


Appendix – B

Bus capacity example application

Question: Let say we need to design a virtual busway station say Mater Hill Busway Station with 3 loading areas. For capacity calculation, two methodologies are available – a) Transit Capacity & Quality of Service Manual (TCQSM) 2003 Method b) Busway Station Loading Area Bus Capacity Model (BSLC) 2009 Method (this PhD)

Inputs: 1 Boarding passengers per bus Pb = 7 2 Marginal boarding time tb = 4s 3 Alight passenger per bus Front door Pa = 0 Rear door Pa’ = 0 4 Marginal alighting time ta = 0 5 Door opening and closing time toc = 2s 6 Bus clearance time tc = 8s 7 Failure rate = 7.5% 8 Z value corresponding to the failure rate = 1.44 9 Co-efficient of variation Cv = 60% 10 g/c = 1 11 Effective loading area (TCQSM, 2003) Leff = 2.65

Additional Inputs for BSLC model: 12 Loading area efficiency (Table 9-3) Loading area 1 E1 = 0.71 Loading area 2 E2 = 0.90 Loading area 3 E3 = 1.00 13 Parameter, µ, for bus lost time (Table 6-14) = Loading area 1 µ1 = 1.095 Loading area 2 µ2 = 0.888 Loading area 3 µ3 = 1.259 14 Parameter, σ, for bus lost time (Table 6-14) = Loading area 1 σ1 = 0.834 Loading area 2 σ2 = 0.582 Loading area 3 σ3 = 0.870 15 Mean lost time (Table 6-14) = Loading area 1 LT1 = 4.2 Loading area 2 LT2 = 2.9 Loading area 3 LT3 = 5.2



16 Co-efficient of passenger variation Cv,p = 60%

Solution:

a) By TCQSM, 2003 method

i. Calculate bus dwell time using Equation 2.8 (on page 20)

td = 30s

ii. Calculate loading area capacity using Equation 2.19 (on page 38)

3600

Bl = 54 bus per hour

iii. Calculate station capacity using Equation 2.21 (on page 40)

Bs = Bl * Ne = 54 * 2.65 = 143 bus/hr

b) By BSLC, 2009 method

i. Calculate bus dwell time using revised dwell time equation, Equation 7.1 (on page 121)

Loading area Dwell time For loading area 1 = 34.2s For loading area 2 = 32.9s For loading area 3 = 35.2s

Appendix- B


ii. Calculate operating margin for passenger service time using Equation 8.4 (on page 132)

, ,

Loading area Operating margin for passenger service

For loading area 1 = 25.9s For loading area 2 = 25.9s For loading area 3 = 25.9s

iii. Calculate operating margin for bus lost time using Equation 8.9 (on page 134)

, ⁄

Loading area Operating margin for bus lost time

For loading area 1 = 5.7s For loading area 2 = 2.7s For loading area 3 = 7.2s

iv. Calculate loading area bus capacity using Equation 8.10 (on page 134)

3600

, ,

Loading area Bus capacity For loading area 1 = 47 bus/hr For loading area 2 = 50 bus/hr For loading area 3 = 45 bus/hr

v. Calculate effective loading area bus capacity using Equation 8.12 (on page 136)

,

Loading area Effective bus capacity For loading area 1 = 33 bus/hr For loading area 2 = 45 bus/hr For loading area 3 = 45 bus/hr

vi. Calculate platform bus capacity using Equation 8.13 (on page 136)

,

= 123 bus per hour

Appendix- C


Appendix – C

List of publications

Refereed journal paper publication

• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2010). Influence of platform walking on BRT station bus dwell time estimation: Australian analysis. Journal of Transportation Engineering (ASCE), Vol 136, No 12, pp 1173-1179.

Refereed conference paper publication

• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2010) Modelling Bus Lost Time: An Additional Parameter Influencing Bus Dwell Time and Station Platform Capacity at a BRT Station Platform. In Proceedings 89th Annual Meeting of Transportation Research board, Washington DC, United States.

• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2009) Effects of Fare Collection Policy on Operating Characteristics of a Brisbane Busway Station. In Proceedings 32nd Australasian Transport Research Forum (ATRF), Auckland, New Zealand.

• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2009) Effect of Passenger Crowding at a Busway Station on Dwell Time. In Proceedings 2nd Infrastructure Theme Postgraduate Conference, Queensland University of Technology, Brisbane, Australia.

• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2009) Modelling the Relationship Between Passenger demand and Bus Delays at Busway Station. In Proceedings 88th Annual Meeting of Transportation Research board, Washington DC, United States.

• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2008) Measuring Bus Dwell Time and Platform Crowding at a Busway Station. In Proceedings 31st Australasian Transport Research Forum (ATRF), Gold Coast, Australia.

• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis (2007) Operating Characteristics and Performance of a Busway Transit Station. In Proceedings 30th Australasian Transport Research Forum (ATRF), Melbourne, Australia.



Journal paper (Under review)

• Jaiswal, Sumeet and Bunker, Jonathan M. and Ferreira, Luis. Modelling Bus Lost Time: An Additional Variable Influencing Bus Dwell Time at a BRT Station Platform. Journal of Advance Transportation, Wiley InterScience. (Submitted on 19 July 2010).

Documents

Busway Platform Bus Capacity Analysis - QUT ePrintseprints.qut.edu.au/43698/1/Sumeet_Jaiswal_Thesis.pdf · Earlier BRTs were bus shared lane or bus only lane, ... 3.2.1 The size 46