TCRP Report 13

T
R A N S I T
O O P E R A T I V E
E S E A R C H
SPONSORED BY
The Federal Transit Administration
TCRP Report 13
Rail Transit Capacity
Transportation Research Board

National Research Council
R O G R A M
TCRP OVERSIGHT AND PROJECT

SELECTION COMMITTEE
CHAIR
ROD J. DIRIDON
Intl Institute for Surface Transportation
Policy Study
MEMBERS
SHARON D. BANKS
AC Transit
LEE BARNES
Barwood, Inc.
GERALD L. BLAIR
Indiana County Transit Authority
MICHAEL BOLTON
Capital Metro
SHIRLEY A. DELIBERO
New Jersey Transit Corporation
SANDRA DRAGGOO
CATA
LOUIS J. GAMBACCINI
SEPTA
DELON HAMPTON
Delon Hampton & Associates
RICHARD R. KELLY
Port Authority Trans-Hudson Corp.
ALAN F. KIEPPER
New York City Transit Authority
EDWARD N. KRAVITZ
The Flxible Corporation
PAUL LARROUSSE
Madison Metro Transit System
ROBERT G. LINGWOOD
BC Transit
GORDON J. LINTON
FTA
WILLIAM W. MILLAR
Port Authority of Allegheny County
MIKE MOBEY
Isabella County Transportation Comm.
DON S. MONROE
Pierce Transit
PATRICIA S. NETTLESHIP
The Nettleship Group, Inc.
ROBERT E. PAASWELL
The City College of New York
JAMES P. REICHERT
Reichert Management Services
LAWRENCE G. REUTER
WMATA
MICHAEL S. TOWNES
Peninsula Transportation Dist. Comm.
FRANK J. WILSON
New Jersey DOT
EDWARD WYTKIND
AFL-CIO
EX OFFICIO MEMBERS
JACK R. GILSTRAP
APTA
RODNEY E. SLATER
FHWA
FRANCIS B. FRANCOIS
AASHTO
ROBERT E. SKINNER, JR.
TRB
TDC EXECUTIVE DIRECTOR
FRANK J. CIHAK
APTA
SECRETARY
ROBERT J. REILLY
TRB
TRANSPORTATION RESEARCH BOARD EXECUTIVE COMMITTEE 1996
OFFICERS
Chair: James W. VAN Loben Sels, Director, California Department of Transportation
Vice Chair: David N. Wormley, Dean of Engineering, Pennsylvania State University
Executive Director: Robert E. Skinner, Jr., Transportation Research Board
MEMBERS
EDWARD H. ARNOLD, Chair and President, Arnold Industries, Lebanon, PA
SHARON D. BANKS, General Manager, AC Transit, Oakland, CA
BRIAN J. L. BERRY, Lloyd Viel Berkner Regental Professor & Chair, Bruton Center for Development Studies,
The University of Texas at Dallas
LILLIAN C. BORRONE, Director, Port Commerce, The Port Authority of New York and New Jersey (Past
Chair, 1995)
DWIGHT M. BOWER, Director, Idaho Department of Transportation
JOHN E. BREEN, The Nasser I. Al-Rashid Chair in Civil Engineering, The University of Texas at Austin
WILLIAM F. BUNDY, Director, Rhode Island Department of Transportation
DAVID BURWELL, President, Rails-to-Trails Conservancy, Washington, DC
E. DEAN CARLSON, Secretary, Kansas Department of Transportation
RAY W. CLOUGH, Nishkian Professor of Structural Engineering, Emeritus, University of California, Berkeley
JAMES C. DELONG, Director of Aviation, Denver International Airport, Denver, CO
JAMES N. DENN, Commissioner, Minnesota Department of Transportation
DENNIS J. FITZGERALD, Executive Director, Capital District Transportation Authority, Albany, NY
DAVID R. GOODE, Chair, President & CEO, Norfolk Southern Corporation
DELON HAMPTON, Chair & CEO, Delon Hampton & Associates, Washington, DC
LESTER A. HOEL, Hamilton Professor, Civil Engineering, University of Virginia
JAMES L. LAMMIE, Director, Parsons Brinckerhoff, Inc., New York, NY
ROBERT E. MARTINEZ, Secretary, Virginia Department of Transportation
CHARLES P. OLEARY, JR., Commissioner, New Hampshire Department of Transportation
CRAIG E. PHILIP, President, Ingram Barge Co., Nashville, TN
WAYNE SHACKELFORD, Commissioner, Georgia Deparment of Transportation
LESLIE STERMAN, Executive Director, East-West Gateway Coordinating Council, St. Louis, MO
JOSEPH M. SUSSMAN, JR East Professor, Civil and Environmental Engineering, MIT
Officials
MARTIN WACHS, Director, Institute of Transportation Studies, University of California, Los Angeles
EX OFFICIO MEMBERS
MIKE ACOTT, President, National Asphalt Pavement Association
ROY A. ALLEN, Vice President, Research and Test Department, Association of American Railroads
ANDREW H. CARD, JR., President and CEO, American Automobile Manufacturers Association
THOMAS J. DONOHUE, President and CEO, American Trucking Associations
FRANCIS B. FRANCOIS, Executive Director, American Association of State Highway and Transportation
DAVID GARDINER, Administrator, U.S. Environmental Protection Agency
JACK R. GILSTRAP, Executive Vice President, American Public Transit Association
ALBERT J. HERBERGER, Maritime Administrator, U.S. Department of Transportation
DAVID R. HINSON, Federal Aviation Administrator, U.S. Department of Transportation
T. R. LAKSHMANAN, Director, Bureau of Transportation Statistics, U.S. Department of Transportation
GORDON J. LINTON, Federal Transit Administrator, U.S. Department of Transportation
RICARDO MARTINEZ, National Highway Traffic Safety Administrator, U.S. Department of Transportation
JOLENE M. MOLITORIS, Federal Railroad Administrator, U.S. Department of Transportation
DHARMENDRA K. SHARMA, Research and Special Programs Administrator, U.S. Department of Transportation
RODNEY E. SLATER, Federal Highway Administrator, U.S. Department of Transportation
ARTHUR E. WILLIAMS, Chief of Engineers and Commander, U.S. Army Corps of Engineers
TRANSIT COOPERATIVE RESEARCH PROGRAM

Transportation Research Board Executive Committee Subcommittee for TCRP
JAMES W. VAN LOBEN SELS, California Department of Transportation (Chair)
DENNIS J. FITZGERALD, Capitol Dist. Transportation Authority, Albany, NY
LILLIAN C. BORRONE, The Port Authority of New York and New Jersey
LESTER A. HOEL, University of Virginia
GORDON J. LINTON, U.S. Department of Transportation
ROBERT E. SKINNER, JR., Transportation Research Board
DAVID N. WORMLEY, Pennsylvania State University
R A N S I T
O O P E R A T I V E
E S E A R C H
Report 13
Rail Transit Capacity
TOM PARKINSON
Transport Consulting Limited
Vancouver B.C. Canada
with
IAN FISHER
University of British Columbia
Subject Area
Public Transit
Research Sponsored by the Federal Transit Administration in

Cooperation with the Transit Development Corporation
T R AN S P O R T AT I O N R E S E AR C H B O AR D
NATIONAL RESEARCH COUNCIL
NATIONAL ACADEMY PRESS

Washington, D.C. 1996
R O G R A M

The nations growth and the need to meet mobility,
environmental, and energy objectives place demands on public
transit systems. Current systems, some of which are old and in need
of upgrading, must expand service area, increase service frequency,
and improve efficiency to serve these demands. Research is
necessary to solve operating problems, to adapt appropriate new
technologies from other industries, and to introduce innovations into
the transit industry. The Transit Cooperative Research Program
(TCRP) serves as one of the principal means by which the transit
industry can develop innovative near-term solutions to meet
demands placed on it.
The need for TCRP was originally identified in TRB
ssful Special
National
Report 213Research for Public Transit: New Directions,
published by
in 1987 and based on a study sponsored by the Urban
ities
executed
Mass Transportation Administrationnow the Federal Transit
Administration (FTA). A report by the American Public Transit
Association (APTA), Transportation 2000, also recognized the need
for local, problem-solving research. TCRP, modeled after the
longstanding and succe
Research Program, undertakes research and other technical activ
in response to the needs of transit service providers. The scope of
TCRP includes a variety of transit research fields including
planning, service configuration, equipment, facilities, operations,
human resources, maintenance, policy, and administrative practices.
TCRP was established under FTA sponsorship in July 1992.
Proposed by the U.S. Department of Transportation, TCRP was
authorized as part of the Intermodal Surface Transportation
ittee.
Efficiency Act of 1991 (ISTEA). On May 13, 1992, a memorandum
agreement outlining TCRP operating procedures was
the three cooperating organizations: FTA, the National Academy of
Sciences, acting through the Transportation Research Board (TRB),
ittee
hout
defines
levelsCorporation,
and
and the
Transitfunding
Development
Inc. (TDC), a nonprofit
educational and research organization established by APTA. TDC is
responsible for forming the independent governing board, designated
as the TCRP Oversight and Project Selection (TOPS) Comm
Research problem statements for TCRP are solicited periodically
but may be submitted to TRB by anyone at any time. It is the
responsibility of the TOPS Committee to formulate the research
program by identifying the highest priority projects. As part of the
evaluation, the TOPS Comm
expected products.
Once selected, each project is assigned to an expert panel,
pact
if products
appointed
by the fail
Transportation Research Board. The panels prepare
titioners.
project statements (requests for proposals), select contractors, and
provide technical guidance and counsel throughout the life of the
project. The process for developing research problem statements and
ities
selecting research agencies has been used by TRB in managing
cooperative research programs since 1962. As in other TRB
activities, TCRP project panels serve voluntarily wit
compensation.
Because research cannot have the desired im
to reach the intended audience, special emphasis is placed on
disseminating TCRP results to the intended endusers of the research:
transit agencies, service providers, and suppliers. TRB provides a
series of research reports, syntheses of transit practice, and other
supporting material developed by TCRP research. APTA will
arrange for workshops, training aids, field visits, and other activ
to ensure that results are implemented by urban and rural transit
industry prac
The TCRP provides a forum where transit agencies can
cooperatively address common operational problems. The TCRP
results support and complement other ongoing transit research and
training programs.
TCRP REPORT 13
Project A-8 FY 93
ISSN 1073-4872
ISBN 0-309-05718-3
Library of Congress Catalog Card No. 96-60739
Price $39.00
Cooperative Highway
NOTICE
The project that is the subject of this report was a part of the Transit Cooperative
Research Program conducted by the Transportation Research Board with the
approval of the Governing Board of the National Research Council. Such approval
reflects the Governing Boards judgment that the project concerned is appropriate
with respect to both the purposes and resources of the National Research Council.
The members of the technical advisory panel selected to monitor this project
and to review this report were chosen for recognized scholarly competence and
with due consideration for the balance of disciplines appropriate to the project. The
opinions and conclusions expressed or implied are those of the research agency that
performed the research, and while they have been accepted as appropriate by the
technical panel, they are not necessarily those of the Transportation Research
Board, the Transit Development Corporation, the National Research Council, or the
Federal Transit Administration of the U.S. Department of Transportation.
Each report is reviewed and accepted for publication by the technical panel
according to procedures established and monitored by the Transportation Research
Board Executive Committee and the Governing Board of the National Research
Council.
Special Notice
The Transportation Research Board, the Transit Development
Corporation, the National Research Council, and the Federal Transit
Administration (sponsor of the Transit Cooperative Research Program)
do not endorse products or manufacturers. Trade or manufacturers
names appear herein solely because they are considered essential to the
clarity and completeness of the project reporting.
Published reports of the

are available from:
Transportation Research Board
National Research Council
2101 Constitution Avenue, N.W.
Printed in the United States of America
FOREWORD
By Staff
Transportation Research
Board
This report will be of interest to transportation and rail-transit planners,

designers, and operators responsible for determining the passenger-carrying capacity
of rail lines for rapid rail transit, light rail transit, commuter rail, and automated
guideway transit. The report provides a comprehensive description of the factors that
determine rail transit capacity and easy-to-use procedures for estimating practical
achievable rail transit capacity under a variety of conditions, calibrated with
extensive, current, North American field data. The procedures are provided in two
forms: a simple method of estimation in which rail capacity for typical or average
conditions can be read from a graph based on train length and type of signal system
and a more comprehensive method that allows for user control over additional
variables. To assist in the more comprehensive method, a computer spreadsheet was
developed in this project and is available free of charge on disk or through the
Internet World Wide Web from the American Public Transit Association (APTA). A
description of the spreadsheet and information on how to obtain it is provided in the
Summary at the beginning of this report. Examples of applications for the rail transit
capacity information found in this report include analyzing project planning and
operations for new starts and extensions; evaluating transit line performance;
establishing and updating service standards; assessing the capacities of new signaling
and control technologies; and, estimating changes in system capacity and operations
for environmental impact assessments and land-use variations.
In the past several decades, many developments have taken place that directly
affect North American rail transit performance, vehicles, operations, and system
technologies. Such developments include the extension and modernization of rail
rapid transit and commuter rail systems; the introduction of proof-of-payment fare
systems; the requirements of the Americans with Disabilities Act (ADA); and the
construction of new light rail transit, automated guideway transit, heavy rail transit,
and commuter rail systems. Consequently, data and procedures related to estimating
rail transit capacity need updating to take into account these recent developments.
Rail-transit capacity information available in TRB Special Report 209, Highway
Capacity Manual, is based on operating experiences from the 1970s and the early
1980s. While providing broad guidelines and general approaches to determining rail
transit capacity, it does not fully reflect current experience.
There has been a need to identify and document the factors affecting rail transit
capacity and collect data on current values of the factors in order to update and
expand the range of applications for this information taking into account vehicles,
station designs, fare policies, train control technologies, and operating practices that
better reflect actual North American rail transit experience. There also has been a
need for information and procedures for estimating rail transit capacity, which
includes both the number of people and the number of vehicles past a point per unit
of time, and relates to stations, routes, junctions, and other controlling transit system
features.
Under TCRP Project A-8, research was undertaken by Transport Consulting

Limited to (1) obtain current information on rail transit capacity, including a) factors
affecting capacity; b) current values for parameters affecting capacity under a range
of operating conditions; and c) current values for maximum passenger and vehicle
capacities achieved under various operating practices and loading standards and (2)
provide appropriate methodologies for estimating the capacity of future rail transit
systems and modifications to existing systems. The scope included investigation,
evaluation, and documentation of current North American experience in rail transit
capacity for light rail transit, rapid rail transit, commuter rail, and automated
guideway transit.
To accomplish this effort, the researchers conducted a comprehensive survey of
existing literature on rail transit capacity experience and capacity analysis
methodologies. In addition, a survey of 63 rail transit operators in the United States,
Canada, and Mexico was performed to determine actual line-by-line capacity and
capacity constraints of each system. Extensive field surveys were also conducted to
determine passenger boarding rates and dwell times for different rail transit modes,
platform heights, and fare collection methods. Quantitative analyses then produced
easy-to-use procedures for estimating achievable rail transit capacity. Thus, the
report is a valuable resource for transportation and rail transit planners, designers,
and operators.
CONTENTS
INSIDE THE REPORT................................................... ix
PROBLEM STATEMENT ............................................. ix
RESEARCH OBJECTIVES ........................................... ix
RESEARCH APPROACH.............................................. ix
FOOTNOTES AND REFERENCES ...............................x
SUMMARY ...................................................................... xi
S1 INTRODUCTION .................................................... xi
S2 CAPACITY............................................................... xi
S3 GROUPING ............................................................. xii
S4 TRAIN CONTROL.................................................. xii
S5 STATION DWELLS .............................................. xiii
S6 LOADING LEVELS............................................... xiv
S7 OPERATING ISSUES............................................ xiv
S8 CAPACITY DETERMINATION.............................xv
S9 THE RESULTS.........................................................xv
S10 COMPARISONS .................................................. xvi
S11 INCREASING CAPACITY................................. xvii
S12 ECONOMIC ISSUES .......................................... xvii
S13 CONCLUSIONS................................................. xviii
COMPUTER DISK ..................................................... xviii
USER GUIDE................................................................. xix
THE REPORT.............................................................. xix
THE SPREADSHEET ................................................. xix
1. RAIL TRANSIT IN NORTH AMERICA .................. 1
1.1 INTRODUCTION..................................................... 1
1.2 LIGHT RAIL TRANSIT........................................... 1
1.2.1 Introduction......................................................... 1
1.2.2 Status .................................................................. 1
1.2.3 Ridership............................................................. 2
1.3 RAIL RAPID TRANSIT........................................... 2
1.3.1 Introduction......................................................... 2
1.3.2 Status .................................................................. 3
1.3.3 Ridership............................................................. 3
1.4 COMMUTER RAIL.................................................. 5
1.4.1 Introduction......................................................... 5
1.4.2 Status .................................................................. 5
1.4.3 Ridership............................................................. 6
1.5 AUTOMATED GUIDEWAY TRANSIT ................. 6
1.5.1 Introduction......................................................... 6
1.5.2 Status .................................................................. 7
1.5.3 Ridership............................................................. 7
2. CAPACITY BASICS .................................................... 8
2.1 INTRODUCTION..................................................... 8
2.2 TERMINOLOGY...................................................... 8
2.2.1 Definitions .......................................................... 8
2.2.2 Footnotes and References ......................................8

2.3 GROUPING.....................................................................8
2.4 THE BASICS...................................................................8
2.5 DESIGN VERSUS ACHIEVABLE CAPACITY .........10
2.5.1 Service Headway....................................................11
2.5.2 Station DwellsPractical Issues............................12
2.6 LINE CAPACITY .........................................................12
2.6.1 Train Control Throughput ......................................13
2.6.2 Commuter Rail Throughput ...................................13
2.6.3 Station Dwells ........................................................13
2.7 TRAIN/CAR CAPACITY.............................................14
2.7.1 Introduction ............................................................14
2.7.2 Car Capacity...........................................................14
2.7.3 Design-Specific Capacity .......................................14
2.7.4 Car Dimensions ......................................................15
2.7.5 Car Capacity Calculation Alternatives ...................15
2.7.6 Train Length Alternative ........................................15
2.7.7 Train Capacity ........................................................15
2.8 STATION CONSTRAINTS..........................................16
3. TRAIN CONTROL AND SIGNALING..........................17
3.1 INTRODUCTION .........................................................17
3.2 FIXED-BLOCK SYSTEMS..........................................17
3.2.1 Cab Signaling .........................................................19
3.3 MOVING-BLOCK SIGNALING SYSTEMS...............19
3.4 AUTOMATIC TRAIN OPERATION...........................22
3.5 AUTOMATIC TRAIN SUPERVISION .......................23
3.6 FIXED-BLOCK THROUGHPUT.................................24
3.6.1 Station Close-In Time.............................................24
3.6.2 Computer Simulation..............................................25
3.6.3 Calculating Line Headway......................................26
3.6.4 Calculating Station Headway..................................28
3.7 SENSITIVITY...............................................................30
3.7.1 Grades.....................................................................30
3.7.2 Line Voltage ...........................................................31
3.7.3 Acceleration............................................................31
3.7.4 Braking ...................................................................32
3.7.5 Train Length ...........................................................32
3.8 MOVING-BLOCK THROUGHPUT ............................32
3.8.1 Fixed Safety Distance .............................................32
3.9 TURN-BACK THROUGHPUT ....................................34
3.10 JUNCTION THROUGHPUT......................................35
3.11 SUMMARY.................................................................36
4. STATION DWELLS.........................................................38
4.1 INTRODUCTION .........................................................38
4.2 LITERATURE REVIEW ..............................................38
4.3 DWELL CONSTITUENTS...........................................39
4.4 DOORWAY FLOW TIMES .........................................41
4.4.1 Flow Time Hypotheses...........................................41
4.4.2 Flow Time Results .................................................42
4.4.3 Effect of Door Width on Passenger Flow Times....43
4.5 ANALYZING FLOW TIMES.......................................43
4.5.1 Data Transformation...............................................44
vi
4.5.2 Comparisons ....................................................... 45
4.5.3 Prediction of Door Movement Time Using
Boarding and Alighting....................................... 46
4.5.4 Prediction of Doorway Movement Time Using
Number Boarding and Alighting Plus the Number
Standing .............................................................. 48
4.5.5 Prediction of Dwell Time from Doorway
Movement Time.................................................. 49
4.5.6 Estimating the Controlling Dwell....................... 49
4.6 SUMMARY ............................................................... 50
5. PASSENGER LOADING LEVELS............................. 51
5.1 INTRODUCTION...................................................... 51
5.2 STANDARDS............................................................ 51
5.3 SPACE REQUIREMENTS........................................ 53
5.4 VEHICLE CAPACITY.............................................. 55
5.4.1 Commuter Rail .................................................... 55
5.4.2 Existing Systems ................................................. 56
5.4.3 Vehicle Specific Calculations ............................. 57
5.4.4 Results Of The Calculation ................................. 58
5.5 LENGTH.................................................................... 59
5.6 LOADING DIVERSITY............................................ 60
6. OPERATING ISSUES................................................... 65
6.1 INTRODUCTION...................................................... 65
6.2 TRAIN PERFORMANCE ......................................... 65
6.3 OPERATING VARIATIONS .................................... 66
6.4 OPERATING MARGINS .......................................... 67
6.5 ESTIMATING MARGINS ........................................ 70
6.6 OPERATING WITHOUT MARGINS ...................... 70
6.7 SKIP-STOP OPERATION......................................... 71
6.8 PASSENGER-ACTUATED DOORS........................ 72
6.9 OTHER STATION CONSTRAINTS ........................ 72
6.10 IMPACT OF AMERICANS WITH DISABILITIES
ACT (ADA).............................................................. 73
7. GRADE SEPARATED RAIL CAPACITY
DETERMINATION..................................................... 76
7.1 INTRODUCTION..................................................... 76
7.2 THE WEAKEST LINK............................................. 76
7.3 GROWTH AND ACHIEVABLE CAPACITY ........ 77
7.4 SIMPLE PROCEDURE ............................................ 77
7.5 COMPLETE PROCEDURE ..................................... 78
7.5.1 Determining the Maximum Load Point Station ..79
7.5.2 Determining the Control Systems Minimum Train
Separation ........................................................... 79
7.5.3 Determining the Dwell Time.............................. 81
7.5.4 Selecting an Operating Margin........................... 84
7.5.5 Selecting a Passenger Loading Level ................. 84
7.5.6 Determining an Appropriate Loading Diversity
Factor .................................................................. 85
7.5.7 Putting it All Together........................................ 86
8. LIGHT RAIL CAPACITY DETERMINATION........ 87
8.1 INTRODUCTION ..................................................... 87
8.1.1 Selecting the Weakest Link................................ 87
8.1.2 Other Capacity Issues......................................... 87
8.2 SINGLE TRACK ...................................................... 87
8.2.1 Calculating Single-Track Headway Restrictions .......88
8.3 SIGNALED SECTIONS................................................. 89

8.4 ON-STREET OPERATION ........................................... 89
8.4.1 Empirical Approach................................................. 90
8.4.2 Practical Issues......................................................... 90
8.4.3 Determining On-Street Capacity.............................. 91
8.5 PRIVATE RIGHT-OF-WAY WITH GRADE
CROSSINGS .................................................................. 91
8.5.1 Pre-emption.............................................................. 92
8.5.2 Grade Crossings and Station Dwell Times .............. 92
8.6 TRAIN LENGTH AND STATION LIMITATIONS ..... 93
8.6.1 Street Block Length ................................................. 93
8.6.2 Station Limitations................................................... 93
8.7 WHEELCHAIR ACCESSIBILITY EFFECTS .............. 93
8.7.1 Introduction.............................................................. 93
8.7.2 High Platforms......................................................... 94
8.7.3 Low Platform Methods ............................................ 94
8.7.4 Low-Floor Cars........................................................ 95
8.8 CAPACITY DETERMINATION SUMMARY ............. 95
9. COMMUTER RAIL CAPACITY
DETERMINATION......................................................... 97
9.1 INTRODUCTION ......................................................... 97
9.2 TRAIN THROUGHPUT ............................................... 97
9.2.1 Station Constraints.................................................. 98
9.2.2 Station Dwells......................................................... 98
9.3 TRAIN CAPACITY ...................................................... 99
10. AUTOMATED GUIDEWAY TRANSIT CAPACITY
DETERMINATION .......................................................... 101
10.1 INTRODUCTION ....................................................... 101
10.2 TRAIN CONTROL SEPARATION............................ 101
10.3 PASSENGER FLOW TIMES AND DWELLS........... 102
10.4 LOADING LEVELS ................................................... 102
10.5 OFF-LINE STATIONS ............................................... 102
11. FUTURE RESEARCH ................................................... 103
11.1 INTRODUCTION ....................................................... 103
11.2 SERVICE RELIABILITY........................................... 103
11.3 STATION DWELLS ................................................... 104
BIBLIOGRAPHY & REFERENCES ................................. 106
GLOSSARY........................................................................... 109
A1. APPENDIX ONE REVIEW OF NORTH AMERICAN
RAIL TRANSIT CAPACITY ANALYSIS
METHODOLOGIES ............................................................ 119
A1.1 INTRODUCTION ...................................................... 119
A1.2 LITERATURE SUMMARIES ................................... 119
A1.3 REVIEW SUMMARY ............................................... 152
A1.3.1 Basics and Cautions............................................. 152
A1.3.2 Influencing Factors.............................................. 152
A1.3.3 Grouping.............................................................. 153
A1.3.4 Light Rail Specifics ............................................. 153
A1.3.5 Station Constraints............................................... 153
A1.3.6 Conclusions ......................................................... 153
A2. APPENDIX TWO RAIL TRANSIT SURVEY ....... 154
A2.1 INTRODUCTION ...................................................... 154
vii
A2.1.1 Purpose of Survey.......................................... 154
A2.1.2 Survey Methodology ..................................... 154
A2.2 RIDERSHIP INFORMATION ............................. 158
A2.2.1 Collection and Availability of Ridership
Information ............................................................... 158
A2.3 CAPACITY AND POLICIES............................... 158
A2.3.1 Loading Standards ......................................... 158
A2.3.2 Train Length .................................................. 158
A2.3.3 Pass-Ups ........................................................ 159
A2.3.4 Event Ridership ............................................. 159
A2.3.5 Ridership/Capacity Ratio............................... 159
A2.4 HEADWAY LIMITATIONS ............................... 159
A2.4.1 Signaling........................................................ 159
A2.4.2 Turnbacks ...................................................... 161
A2.4.3 Junctions........................................................ 161
A2.4.4 Station Approach........................................... 161
A2.4.5 Single Track .................................................. 161
A2.4.6 Station Dwells ............................................... 162
A2.4.7 Other Headway Constraints........................... 162
A2.5 STATION LIMITATIONS ........................................ 162

A2.5.1 Full Parking Lots ................................................ 162
A2.5.2 Ticketing Line-Ups ............................................. 162
A2.5.3 Congested Platforms ........................................... 162
A2.5.4 Other Congestion ................................................ 163
A2.5.5 No TransfersTransfer Cost.............................. 163
A2.5.6 Safety and Security ............................................. 163
A2.5.7 Long WalksPoor Access ................................. 163
A2.5.8 Other Station Constraints.................................... 163
A2.6 DWELL TIMES ......................................................... 163
A2.6.1 Fare Collection.................................................... 163
A2.6.2 Platform Height................................................... 164
A2.6.3 Wheelchair Effects.............................................. 164
A2.7 SCHEDULE ADHERENCE ...................................... 164
A2.8 COMMENTS AND RESULTS.................................. 164
A3. APPENDIX THREE DATA TABULATIONS ............ 165
A3.1 INTRODUCTION...................................................... 165
A3.2 NOTES ON THE TABLES........................................ 165
viii
COOPERATIVE RESEARCH PROGRAMS STAFF

ROBERT J. REILLY, Director, Cooperative Research Programs
STEPHEN J. ANDRLE, Manager, Transit Cooperative Research Program
CHRISTOPHER W. JENKS, Senior Program Officer
EILEEN P. DELANEY, Editor
KAMI CABRAL, Assistant Editor
HILARY M. FREER, Assistant Editor
PROJECT PANEL A-8

JEROME M. LUTIN, New Jersey Transit (Chair)
WOLFGANG S. HOMBURGER, University of California, Berkeley
BARTHOLOMEW M. KANE, Transportation & Distribution Associates, Dallas, TX
GEORGE KRAMBLES, Oak Park, IL
HERBERT S. LEVINSON, New Haven, CT
S. DAVID PHRANER, Port Authority of New York and New Jersey
FRANKLIN L. SPIELBERG, SG Associates, Annandale, VA
DAVID M. WEISS, MTA New York City Transit
RON JENSEN-FISHER, FTA Liaison Representative
JEFFREY MORA, FTA Liaison Representative
RICHARD A. CUNARD, TRB Liaison Representative
ACKNOWLEDGMENTS
This research was performed under TCRP Project A-8 by
Transport Consulting Limited of Vancouver, British Columbia,
Canada.
The report was written by Tom Parkinson, president of TCL,
the principal investigator, and Ian Fisher of the University of
British Columbia, the research associate with contributions from
the associate investigator Ian Graham, operations planner for the
British Columbia Rapid Transit Company and Dr. Jonathan
Berkowitz the projects statistical advisor. Murray McMillan of

the Vancouver Sun assisted in editing the report.
The research team acknowledge with thanks the time and effort
of numerous people in rail transit operating agencies who
provided information for the rail transit survey and assisted the
team in the field data collection. TCRP Program Officer
Christopher Jenks deserves many thanks for his help and support
throughout this project.
ix
evaluating transit line performance,

establishing and updating service standards,
studying environmental impacts,
assessing the capacities of new signaling and control
technologies,
estimating changes in system capacity and operations over
time, and
assessing capacity impacts in land-development studies
where transit is expected to provide a significant role in site
access.
Inside the Report

This report has three main sections. This introductory section,
paginated with roman numerals, contains the Problem Statement,
Research Objectives and Research Approach of the project,
followed by the Summary and a User Guide.
In the main section, the first two chapters, Rail Transit In North
America and Capacity Basics, describe the industry and capacity
issues. The following four chapters: Train Control and Signaling,
Station Dwells, Passenger Loading Levels, and Operating Issues
develop the methodology. These are followed by chapters seven
through ten, which present capacity calculation methods for the
four rail transit groups, respectively: Grade Separated Rail, Light
Rail, Commuter Rail and AGT. The final chapters present
recommendations and suggestions for Future Research followed
by a Bibliography and Glossary.
In the third and final section, three appendices summarize the
Literature Reviewed and the Data Survey, and Tabulate the Data
used in the project. In particular Table A 3.3 provides a detailed
listing of all North American individual transit routes and
ridership.
Research Objectives
The objectives of this research have been to obtain current
information on rail transit capacity and to provide appropriate
methodologies for estimating the capacity of future rail transit
systems and of modifications to existing systems, taking into
account generally accepted theory and observed operating
practices.
Effort has been divided among the four rail modes:
Light Rail Transit (LRT)
Rail Rapid Transit (Heavy Rail) (RT)
Commuter Rail (Regional Rail) (CR)
Automated Guideway Transit (AGT)
Problem Statement
In the past several decades, many developments have taken place
that directly affect North American rail transit performance,
vehicles, operations, and systems technologies. These
developments include the extension and modernization of rail
rapid transit and commuter rail systems, the introduction of the
proof of payment fare collection system, the requirements of the
Americans with Disabilities Act (ADA), and the construction of
new light rail, automated guideway transit (AGT), rail rapid
transit, and commuter rail systems. Consequently, data and
procedures related to estimating rail transit capacity need
updating to take these developments into account.
Rail transit capacity information available in the 1985 Highway
Capacity Manual is based on operating experiences from the prior
two decades. While providing broad guidelines and general
approaches to determining rail transit capacity, it does not fully
reflect current experience.
There is a need to identify and document the factors affecting
rail transit capacity and collect data on current values of these
factors in order to update and expand the range of applications for
this information. The research must take into account vehicles,
station designs, fare policies, train control technologies, and
operating practices that better reflect North American rail transit
experience. There is also a need for information and procedures
for estimating transit capacity. Rail transit capacity, as defined for
this project, includes both the number of people and the number
of vehicles past a point per unit of time, and it relates to stations,
routes, junctions, and other controlling transit system features.
Examples of applications for new rail transit capacity
information include the following:
project planning and operations analysis for new starts and
extensions,
Research Approach
The study has taken a structured and methodical approach that
makes maximum use of previous work and existing data. The
North American rail transit industry monitors ridership carefully,
usually as part of the Federal Transit Administration (FTA)
(UMTA) Section 151 reporting. Annual summary reports are also
prepared by American Public Transit Association (APTA),
Canadian Urban Transit Association (CUTA), and individual rail
operators. Less frequently published reports summarize rail
equipment rosters with quantities, dimensions and other
information.
These data have been augmented by direct contacts with each
agency to determine peak-point ridership, theoretical and actual
minimum headways, limitations on headways, individual car
loadings, locations and frequencies of pass-ups, and other relevant
factors.
The initial data collection was used as an input into an analytic
framework containing the above capacity influencing factors with
particular emphasis on achieving accurate real-life calibration for
each factor.
Additional data needs were identifiedconcentrating on
systems with heavily used rail lines. The only accurate way to
determine the true maximum capacity of a car is when there are
pass-ups. That is when passengers wait for the next train on a
routine day-by-day basis. There are only an estimated six
locations in the United States and Canada where pass-ups occur
on rapid transit, all were visited.
1
FTAFederal Transit Administration. Section 15 of the Urban Mass

Transportation Act of 1964, as amended. Uniform System of Accounts and
Records and Reporting System.
x
Based on the analytic framework and data collected,
quantitative analysis was carried out and calibrated, with
formulae and constants determined to provide a
comprehensive method for determining rail transit capacity
over a wide range of variants for each of the four rail modes.
A practical method of using the data and determining capacity
has been developed in two categories. The first category is a
simple method containing basic parameters with constants for
major variables that reflect typical or average conditions. The
second category is more complete, adding further variants,
including capacity adjustments for grade and line voltage.
To assist in using the results of this research, a computer disk
has been prepared containing spreadsheets into which system

variables can be inserted. (See Summary for availability.)
Footnotes and References

To avoid duplication, references are shown as (R23) and refer to the
bibliography in Chapter Twelve and the literature review item of
the same number in Appendix One. Footnotes are shown by an
italicized superscript number8 referenced to the bottom of each
page.
xi
Summary
S1 INTRODUCTION
Rail transit systems in North America carry 5 billion passengers
each year. Fifty-three agencies operate 207 routes of the four rail
transit modes with a total length of 8,200 km (5,100 mi),
providing 29 billion passenger-kilometers of service annually.
Two systems dominate. The largest operator, Sistema de
Transporte Colectiva (STC) in Mexico City, has recently
overtaken MTA New York City Transit in ridership. STC carries
1,436 million passengers annually, 29% of the continents total.
MTA-NYCT carries 1,326 million passengers annually, 27% of
the continents total, 50% of the United States total. Adding all
New York City area rail operators makes the New York area the
continents largest user of rail transit with 1,585 million
passengers annually, 32% of the continents total, 59% of the
United States total. Together the rail transit systems in the New
York area and in Mexico City account for 61% of all unlinked rail
passenger trips in North America. Summary data is shown in
Tables S.1 and S.2.
Rail transit plays a vital role in five metropolitan areas carrying
over 50% of all work trips and, in three regions, over 80% of all
central business district (CBD)-oriented work trips. Rail transit
plays an important but lesser role in another six regions. Other rail
transit systems carry a smaller proportion of all regional trips but
fill other functionsdefining corridors, encouraging densification
and positive land-use development, reducing congestion and
providing reliable, economic and environmentally responsible
capacity in overloaded corridors.
S2 CAPACITY
This study has concentrated on the achievable capacity of the four
rail transit modes: rail rapid transit, light rail, commuter rail and
automated guideway transit.
Table S.1 North American rail ridership by mode
Table S.2 Transit ridership summary (million)
Achievable Capacity
The maximum number of passengers that can be
carried in an hour in one direction on a single
track allowing for the diversity of demand.
The basics of rail transit capacity are very simplethe product of

how many trains can be operated in the peak hour and by the
number of passengers that will fit on those trains. However, as
many contributors to this field have pointed out, some of the
factors in this seemingly simple calculation vary widely, none
more so than the density of loading. Leroy Demery(R22) states this
succinctly in reference to new rail transit lines in the USA:
... long before crowding levels...... reached New York

levels, prospective passengers would choose to travel by
a different route, by a different mode, at a different time,
or not at all........outside the largest, most congested
urban areas, the level of crowding that transit
passengers appear willing to tolerate falls well short of
theoretical design or maximum vehicle capacity...
Determining how many passengers will fit on a train is a policy

issue subject to significant economic constraints. The actual
levels in North America vary by a factor of six to one from
Mexico Citys Line 2 to most commuter rail systems where
universal policies provide a seat for all longer distance
passengers. The range on rail rapid transit in the United States is
less at approximately three to one. The project has reduced this
range further with recommended loading ranges for rail rapid
transit and light rail of two to one.
The other largest variable in the determination of achievable
capacity is the operating margin. An operating margin must be
added to the minimum train separation time plus maximum
station dwell to arrive at the closest practical train headwayand
so maximum throughput. Although rail transit is noted for reliable
and regular operation, minor delays are routine and an operating
marginand the associated end-of-line schedule recovery time
are essential to prevent delays from compounding. Service
designed so that routine irregularities do not spread from one train
to another is desirable and is said to be operating with a
noninterference headway.
The range of operating margins on close headway rail rapid
transit in North America exceeds four to one. After analyzing this
range, the project recommends a range of 15 to 25 secjust less
than two to one.
At the maximum load point station it is possible to calculate the
minimum train separation possible with a given train control
system with some precision, and the portion of station dwell
xii
Figure S.1 Basic capacity calculation (all line capacity components in seconds)
related to passenger flow with reasonable accuracy. It is,

however, a classic case of statistical spurious accuracy to pursue
these definable elements with too much rigor when other factors
vary so widely. The well-stated caution from Richard Soberman,
one of the earlier workers in this field, should always be kept in
mind:
The capacity of transit service is at best an elusive
figure because of the large number of qualifications
that must be attached to any measure of capacity that
is adopted.
S4 TRAIN CONTROL
The three major designs of train control system offer progressive
increases in capacity. By far the most common constraint is the
close-in movement at the maximum load point station.
Occasionally another heavy-use station with mixed flow may
require longer dwells and become the constraint. The minimum
headway can be readily calculated with the only uncertainty
being the safety separation factor. Logical safety separation
factors were developed for each generic type of train control and
showed close correlation to field experience. A summary of the
results is shown in Figure S.2 and Table S.3.
S3 GROUPING
For the purpose of capacity analysis and determination, the four
modes of rail transit in this study can be grouped into specific
categories based on the type of alignment and rolling stock.
The first category is fully segregated, signaled, double-track
right-of-way, operated by electrically propelled multiple-unit
trains. This is the largest category encompassing all rail rapid
transit, all non-institutional automated guideway transit,1 several
light rail sectionsfor example, the Market Street subway in San
Francisco, and several commuter rail lines on the East Coast.
This category represents 94% of all rail transit ridership on the
continent.
The second category is light rail without fully segregated
tracks, divided into on-street operations and private right-of-way
with grade crossings. The third category is commuter rail other
than services included in category one. In each of these
categories the basic capacity analysis is determined by the flow
chart shown in Figure S.1.
Occasionally the throughput bottleneck is not the maximum
load point station but a junction, a heavy-use station with an
entry speed restriction or a turn-back movement. Generally these
constraints can be avoided by good design and should not be
accepted on new systems.
1
The Morgantown Automated Guideway Transit system, with off-line

stations, is not classed as a public operation by APTA, but is included as a
transit operation in this report.
Figure S.2 Moving block headways with 45 sec dwell and 25sec operating margin compared with conventional fixed
block systems
xiii
Table S.3 Headway result summary in seconds with 200m
(660-ft) trains (8-10 cars) VSD = variable safety distance
The minimum train separation is based on systems designed

for the greatest throughput with typical equipment performance.
Many systems are not designed for this maximum throughput but
use a more economical train control system with lower
capabilities. In this case the design capabilities of the train
control system must be obtained and used in the achievable
capacity calculation.
The headway calculations can make allowances for grades into
and out of stations and reductions in line voltage. Adjustments
for speed restrictions on the approach to the maximum load point
station are also accommodated with a distance-speed chart that
permits a manual adjustment to the approach speed. Where
available, or on systems with unusual circumstances, the use of a
comprehensive suite of simulation programs is recommended.
The components of a typical rail rapid transit system with full
length trains, a 45-sec station dwell and the recommended
midrange operating margin are shown in Figure S.3.
Figure S.3 Headway components for cab-control signaling

that compose the typical North American minimum headway
of 120 sec
S5 STATION DWELLS
As Figure S.3 shows, the station dwells are the largest
component of the minimum headway, and they are also a partly
controllable item. One disconcerting result of the field survey,
which concentrated on lines at or close to capacity, is the
relatively small proportion of dwell time productively used for
passenger flowshown in Figure S.4. This is discussed as a
potential area for future research in Chapter Eleven.
Although it was not possible to equate flow times with door
width, statistical analysis produced a good fit between passenger
volumes and dwells for all level loading situations, independent
of mode and system. This result avoided having separate
equations for a variety of situations.
The majority of the field data collection involved doorway
flow time. The results are summarized in Figure S.5. The most
surprising result was the consistently faster loading rate up light
rail steps compared to alighting down the steps.
A special survey of passenger flows at special events a
football game and a rock concertdisproved the theory that
flows would be faster. In the limited sample observed they were
slightly slower than in normal peak periods. This can be
attributed to the many riders to special events not accustomed to
transit use.
On the few light rail systems with on-board fare collection,
boarding time was 31% slower. The exact-fare collection process
Figure S.4 Toronto Transit Commission King Station S/B

dwell time components: am peak period (part) (flow time
averages 31% of total dwell)
added one second per passenger on average. Light rail with

lowlevel loadingwith steps on the car as distinct from lowfloor carsproduced times per passenger that averaged exactly
double those for level loading, an additional 2.05 sec per
passenger.
Flow ratesand the resultant dwell timesfor light rail with
on-board fare collection or low-level loading were not used in
xiv
Figure S.5 Summary of rail transit doorway average flow

times
Figure S.7 linear passenger loading heavy rail cars
Table S.4 Linear load summarypassengers per meter
Figure S.6 Linear passenger loading of articulated LRVs.
the calculation of maximum achievable capacity. On-board fare

collection through a single door is not possible at significant
passenger volumes. All North American light rail systems with
on-board fares use station fare collection at busy trunk stations.
Maximum achievable capacity with steps is an oxymoron. The
busiest light rail trunk, San Franciscos Market Street subway,
uses cars equipped with folding steps to provide level loading.
The other heavy trunk light rail line, in Boston, also operates at
less than half the maximum achievable capacity of three-car
articulated light rail trains operating close to the minimum
headwayprimarily because of the level of demand but also, in
part, because of longer dwells caused by the low-level loading.
S6 LOADING LEVELS
A comprehensive survey of theoretical and actual car capacity
resulted in a detailed methodology to select seating arrangements
and standing densities that produce car and train loading levels.
The recommended result to base loading on the linear length of a
car or train is summarized in Figures S.6 and S.7 and Table S.4.
Three levels of loading diversity were reviewed. The diversity

of loading within a car and between cars of a train was
incorporated in the recommended linear loading levels. The more
important diversity between the peak-within-the-peak and the
full peak hour is shown in Table S.5. The recommended loading
diversity factors based on actual North American experience are
0.80rail rapid transit
0.75light rail
0.60commuter rail
S7 OPERATING ISSUES
The field survey, plus data provided by several operators,
showed a surprising amount of headway irregularities. An index
was developedthe coefficient of variation of headwaysbut
no relationship could be found between this and headway, dwell
or train control separation. The potential savings from controlling
dwell were demonstrated by a few operators who combined close
headways with brisk operation. This topic is suggested as an area
for future research in Chapter Eleven.
A wide range of data was compiled to determine actual operating
margins. A selection is shown in Figure S.8. The recom-
xv
Table S.5 Diversity of peak hour and peak 15 min
between A and B trains could extend dwells slightly. Passengeractuated doors, a common light rail feature, have no effect at
systems close to capacity as at heavy volumes train operators
control the doorsdisabling the passenger actuation.
The Americans with Disabilities Act (ADA), timing
wheelchair boarding and alighting movements, and agency plans
to meet ADA requirements were reviewed. This led to the
conclusion that ADA would probably have no negative
consequences on maximum achievable capacity but possibly
positive ones as better visual but audio messaging could reduce
doorway delays from passengers who are uncertain what train to
board or alight from. All heavy volume rail transit will adopt
level loading where wheelchair movements can be as fast as
those of other passengerssometimes faster.
S8 CAPACITY
DETERMINATION
Service is only one train per hour and is not included in the average.
Capacity determination was broken down into the four modes

and into simple and complete methods. Over 90% of North
American rail transit fits into the main category of Chapter
Seven, Grade Separated Rail Capacity Determination, and in
reality any rail transit system intending to offer the maximum
achievable capacity will be in this category.
The simple methodology uses two charts that provide a modest
range for rail transit with typical parameters. The charts (Figures
S.11 and S.12) offer variants for heavy rail and light rail with
either cab-control or moving-block signaling systems.
The complete method takes the user through a series of steps
that require some judgment. The first call is to determine the
weakest link in the capacity chain, then calculate or pick a dwell
timethree methods are given. Other calls include the operating
margin and the passenger loading level.
Three subsequent chapters deal with the specifics of light rail,
commuter rail and automated guideway transit. Equations to
determine the headway constraints of light rail single-track
sections are developed. The results for selected parameters are
shown in Figure S.9. Commuter rail is unique in that train
capacity is the total number of seats in the train less an allowance
of 5-10%. Commuter rail throughput outside the main
category of electric multiple-unit operation on dedicated tracks
cannot be calculated but must be obtained from the capabilities
of the specific signaling system, or more commonly from the
number of trains contracted with the owning railroad.
Figure S.8 Headway components of selected North American

rail rapid transit systems (in seconds)
S9 THE RESULTS
mended range to be applied in capacity determination is 15 to 25

sec.
Other operating issues were reviewed. Skip-stop operation and
passenger-actuated doors were found not to influence maximum
achievable capacity. Skip-stop operation still requires all trains to
stop at the maximum load point station. Passenger transfers
Figure S.10 shows the capability of various train control systems

with trains of different length. Figure S.11 shows the dwell time
and achievable capacity relative to hourly, directional platform
volumes at the maximum load point station. Figure S.10
contributes to the main results shown in Figure S.11 and Figure
S.12. These latter two figures together constitute the simple
method of capacity determination based on the assumptions of
Table S.6.
xvi
Table S.6 Simple method performance assumptions
Figure S.9 Light rail travel time over single-track section.

(with a speed limit of 55 km/h and various numbers of stations
train length 56 m, dwell time 20 sec, operating margin 20 sec,
other data as per Table 8.2.)
Figure S.11 Achievable capacity with a multiple-command

cab-control signaling system and peak-hour average loading
of two passengers per square meter for one track of a grade
separated rail transit line
Figure S.10 Minimum train separation versus length
S10 COMPARISONS
The highest capacity double-track rail rapid transit is believed to
be the Yamanote line in Tokyo reaching 100,000 passengers per
peak-hour direction. Hong Kongs busiest line carries 75,000 and
some European lines reach 60,000. In past eras high ridership
was sustained on rail rapid transit and light rail or streetcar lines
in several North American cities. This is no longer the case.
In North America, Mexico Citys Line 2 with 75,000
passengers per peak-hour direction is the heaviest. In the United
States and Canada, no lines exceed 50,000. NYCTs two-track

trunk combining lines E and F (Queens Blvd. Express) carries
49,800 while the busiest four-track trunk is the Lexington
Avenue line used by the 4, 5 and 6 services with 63,200
passengers per peakhour direction.
In theory a four-track line could carry double the capacity of
two tracks if the services were independent. However, where
local and express services are inter-worked, the New York ratio
of up to 50% additional capacity is modest and for maximum
capacity determination four tracks of local and express service
can be considered capable of carrying 180% of the passengers
per peak hour on two tracks.
xvii
Table S.7 Peak-hour ridership summary 1993
rolling stock, staff and operating funds. There are few urban
corridors in North America where demand requires this
maximum achievable capacity.
S11 INCREASING CAPACITY

Figure S.12 Achievable capacity with a moving-block
signaling system and peak-hour average loading of two
passengers per square meter for one track of a grade
separated rail transit line Caution: With the exception of San
Franciscos Muni metro, signaled grade separated light rail lines
are rarely provided with the minimum headway capabilities
represented by the capacity ranges in Figure S.11 and Figure
S.12.
Outside New York and Mexico City the heaviest rail rapid
transit lines are Torontos Yonge subway with 26,900 passengers
per peak-hour direction, Montreals Orange line with 24,400,
followed by WMATA with 15,300 and BART with 14,900.
With the exception of New York and Mexico City, none of the
existing rail rapid transit trunks are close to the maximum
achievable capacity range with conventional train control of
34,000 to 40,000 as shown in Figure S.11.
The story with light rail is similar. The busiest trunks appear to
be Bostons Green Line subway with the Massachusetts Bay
Transportation Authority (MBTA) giving a rough estimate of
10,000 passengers per peak-hour direction. San Franciscos
Market Street subway is estimated to be carrying 7,000 to 8,000,
with the third busiest trunk in Philadelphia handling 4,100 in the
peak hour. These usage figures are well below the maximum
achievable capacity range for light rail of 19,000 to 21,000 from
Figure S.12.
The heaviest commuter rail ridership is on the LIRR into
Manhattan with 41,500 passengers per peak-hour direction,
followed by Metro North into Grand Central with 36,000 and the
C&NW in Chicago with 22,300all multiple-rack trunks which
exceed all but the four busiest rail rapid transit lines on the
continent, three of which are in Mexico City.
All line and trunk ridership data are tabulated in Appendix
Three (A3) and summarized in Table S.7.
The achievable capacity data developed in this report are a
measure of the supply of service given an adequate supply of
Where higher capacity is required there are the obvious steps of

running longer trains and increasing loading levels. However, the
commonly operated rail rapid transit train length of 180 m (600
ft) is regarded as close to a practical maximum, and increasing
loading levels is contrary to the need to make rail transit more
attractive with higher quality service.
The two most appropriate ways to increase achievable capacity
are through advanced train control systems and shorter station
dwells. Processor-based train control systems have now gained
acceptance and will become standard in the future. They offer a
20 to 30% increase in throughput and the possibility, through
sophisticated automatic train supervision components, of better
service regulation. They also make more efficient operation
possible. Driverless operation has accumulated 10 years of safe
experience in Vancouver and Miami and 30 years on some
automated guideway transit systems. Acceptance elsewhere is
slow but the advantages are considerable, not only in operating
economies but in the ability to operate shorter trains more
frequently throughout the service day a feature highly
appreciated by users and a contributor to ridership growth.
Potentially some of these economies can be translated into less
crowded conditions for future generations of passengers.
Capacity can be maximized by avoiding junctions near heavy
stations and ensuring that terminal and turn-back locations do not
have constraintsproviding multiple platforms when necessary.
Inefficient use of station dwell time is common on several
North American systems. Improvements not only have the
potential to increase capacity in the order of 5 to 20%with the
existing number of carsbut also to reduce costs, reduce travel
times and attract more passengers.
This is an area suggested for future research in the next
chapter. While much of the dwell time relates to operating
practices, improvements in signage, platform markings and
interior car design can all contribute to shorter dwells.
S12 ECONOMIC ISSUES

This project has not dealt with economic issues where
limitations in the size of the car fleet or the operating budget
xviii
restrict the number of trains operated. While this is one possible
topic for future research, it is relatively straightforward to
estimate the capacity given a set number of trains.
The throughput in trains per hour can be estimated by
determining the round-trip time plus layover time and any
terminal operating margin in minutes and dividing this into 60.
The result is then multiplied in turn by the number of trains for
throughput in trains by hour. Multiplying again by the passenger
loading on a train (see Chapter Five, Passenger Loading Levels,
or Figures S.6 and S.7) gives a maximum hourly capacity.
Multiplying this again by the loading diversity factor, 0.6, is
recommended for commuter rail with an increase to 0.9 possible,
by 0.8 for rail rapid transit, and by 0.75 for light rail to produce
an achievable capacity in passengers per peak-hour direction per
track.
S13 CONCLUSIONS
The study has achieved its goals of surveying the North
American rail transit industry and providing a complete range of
information to determine the maximum achievable capacity of
each mode.
The principal methodology can be found on an easy-to-use but
comprehensive computer spreadsheet. Although few new rail
transit lines will be concerned with the upper range of achievable
capacity, the methods are applicable to existing systems and
allow an examination of the impact of many variables on
capacity.
This approach is particularly valuable in analyzing the impact
of single high-use stations. The changes in capacityand so the
cost to provide that capacitycan be compared by examining
alternates such as double-faced platforms or spreading the load
between two closely spaced stations.
The results of this project show maximum achievable
capacities, based on reasonable loading levels, that are more
conservative than earlier work in this field. As demands for
improved standards grow, loading levels will likely decrease and
the achievable capacity shown in this study will not only be
appropriate but may have to be further reduced.
Computer
Disk
A 1.44 MB, 3.5" IBM-formatted high-density disk is available on
request, containing spreadsheet and database files from the
project. The spreadsheet files are designed to allow users to input
basic system parameters from which the maximum achievable
capacity will be calculated and presented as a single estimate in
passengers per peak-hour direction. Suggested default parameters
are provided for all entry areas.
Apple Macintosh users with compatible programs should be
able to read and use some of these files using their Apple File
Exchange program. Transport Consulting Limited regrets that it
cannot provide the disk or files in formats other than those
described below.
THE DISK IS NOT REQUIRED TO CALCULATE

CAPACITY. BOTH THE SIMPLE AND MORE
COMPREHENSIVE METHODS DOCUMENTED IN THIS
REPORT CAN BE CARRIED OUT USING EITHER
MANUAL OR COMPUTER TECHNIQUES.3
The disk contains the following capacity calculation files which
are also available to download from the Internet at APTAs
dissemination
site
on
the
World
Wide
Web:
http://www.apta.com/tcrp
All project spreadsheet work has been carried out in Microsoft

Excel 5.0 for Windows. The generic Lotus 1-2-3, and Quattro
Pro files are suitable for either the DOS or Windows version of
these programs. However they do not contain the charts,
equations, color and user-friendly formatting of the Excel
version, nor the component that estimates dwell from hourly
station passenger volumes. This latter process, described in
Chapter Four, Station Dwells, would not translate to a generic
version. Use of the Excel version is recommended whenever
possible.
USING THE SPREADSHEETS Instructions,

together with a printout of sections of the capacity spreadsheet
are contained in the next sectionUser Guide.
ADDITIONAL DATA FILES
The projects
database file is included as TCRPA-8.MDB, and a selection of
the field data collection as a spreadsheet, A8DATASS.EXE.
TCRPA-8.MDB is in Microsoft Access(TM) 2.0 format. Note
that this format cannot be read by Access version 1.0 or 1.1. The
file A8DATASS.EXE, when executed, expands to the
spreadsheet field data file A8DATASS.XLS in Microsoft Excel
5.0 format. TCL regrets that disk space prevents including other
formats. Both files require their respective programs running
under Microsoft Windows(TM) and should be possible to import
into other database or spreadsheet programs.
CAUTION Reasonable care has been taken in obtaining

and transcribing data. However the data is from various sources
and for different years1992 through 1995. The accuracy of the
originating agencys data cannot be verified. In particular
ridership data may only be accurate within 10%. The capacity
calculation spreadsheets are intended to assist in the estimation
of capacity under a variety of normal conditions. Not all
variables or system specific conditions can be accounted for.
Consequently Transport Consulting Limited can provide no
assurance or warrantee of the suitability or accuracy of these
3
The process that estimates dwell from hourly station passenger volumes
calculations has compound logarithmic functions and should only be
attempted by experienced spreadsheet users.
xix
programs for any specific purpose. The disks by request have
been checked to be free from common known viruses. No such
assurances can be given for copies of the programs obtained
from other sources.
LIMITATION of LIABILITY
In no event will
Transport
Consulting Limited,
the
Federal
Transit
Administration, the Transit Cooperative Research Program, the
Transportation Research Board, or the National Research
Council be liable for direct, indirect, special, incidental or
consequential damages arising out of the use or inability to use
these computer files and their documentation, even if advised of
the possibility of such damages.
ORDERING The disk is available on request to

American Public Transit Association
c/o TCRP Dissemination
1201 New York Ave., N.W.
FAX (202) 898-4019
e-mail: tcrapta.com
Include name and mailing address on request.
Internet
The spreadsheets can be downloaded from APTAs TCRP
Dissemination site on the World Wide Web.
http://www.apta.com/tcrp
CORRECTIONS Transport Consulting Limited would

appreciate notification of any errors or problems with the disk
and will make reasonable attempts to prepare a corrected version.
e-mail [email protected].
The contractor regrets that it otherwise cannot enter into
correspondence regarding use of, or problems with, the programs
on the disk, or the conversion for use in other programs or with
other operating systems.
The spreadsheet files will operate reasonably on any IBM
compatible computer with a 386 or higher CPU running
Windows and 4MB of RAM. Microsoft Access 2 requires a
minimum of 6MB of RAM to run reasonably. When expanded,
the total files require less than 3 MB of hard disk space.
User Guide
THE REPORT
The basics of rail transit capacity are straightforward. The hourly
throughput of trains is determined, multiplied by the number of
passengers per train, then adjusted by a loading diversity factor
that compensates for the fact that trains are not evenly loaded
over a peak hour.
However there are many nuances to these basics that can
become complex resulting in this report having several sections
with complicated mathematics. For ease of use, capacity
calculation methods are divided into two: a simple method and a

complete method. Spreadsheets are available on request to
perform the math for the complete method. This user guide
provides assistance in obtaining an understanding of rail transit
capacity and performing either the simple or complete
calculations.
STARTING OUT
The preceding summary, this user guide, and the first two
chaptersChapter One, Rail Transit In North America and
Chapter Two, Capacity Basicsshould be read by all users.
Readers wanting to use the simple method of capacity estimation
can use the preceding summary section or jump to the beginning
of the appropriate application chapter. Chapter Seven, Grade
Separated Rail Capacity Determination covers the majority of
North American rail transit fully segregated, signaled, double
track right-of-way, operated by electrically propelled multipleunit trains; Chapter Eight, Light Rail Capacity Determination for
light rail; Chapter Nine, Commuter Rail Capacity Determination
for commuter rail and Chapter Ten, AGT Capacity
Determination for automated guideway transit.
More details of capacity nuances and methodology
development can be consulted as needed in Chapter Three, Train
Control and Signaling; Chapter Four, Station Dwells; Chapter
Five, Passenger Loading Levels and Chapter Six, Operating
Issues. To avoid the details on train control systems and the more
complex mathematics, start Chapter Three at section 3.6.4 and in
Chapter Five omit section 5.5.
These last two chapters are also of value to the general reader
as they deal with factors that can greatly effect capacity. Loading
levels can make a greater than three to one difference between
policies that provide a seat for most passengers to ones that allow
high levels of standing. Operations and reliability go hand in
hand and there can be almost a 50% difference in capacity
between a system incorporating a substantial operating margin to
achieve good reliability and one where the need for capacity
reduces the operating margin almost to nothing.
THE SPREADSHEET
Whether you can use the spreadsheet or not, this section provides
a step-by-step guide to capacity calculation and should be read
by all users. This section is abstracted from the Excel version of
the spreadsheet but, like the generic version of the spreadsheet,
necessarily omits the user-friendly color coding and the
embedded charts and equations, instead referring to specific
sections of the report. If you can run Excel do so and omit this
section. The Excel spreadsheet is self-explanatory. It is based on
TCRP Report A-8 and is applicable to all grade separated electric
multiple-unit rail transit with level loading.
CAUTION This capacity calculation spreadsheet is intended to
assist in the estimation of rail transit capacity under a variety of
normal conditions. Not all variables or system specific conditions
can be accounted for. Consequently Transport Con Consulting
Ltd can provide no assurance or warrantee of the
xx
suitability or accuracy of these programs for any specific
purpose.
LIMITATION of LIABILITY In no event will Transport
Consulting Ltd., the Federal Transit Administration, the Transit
Cooperative Research Program, the Transportation Research
Board or the National Research Council be liable for direct,
indirect, special, incidental or consequential damages arising out
of the use or inability to use these computer files and
documentation, even if advised of the possibility of such
damages.
THE SPREADSHEET IS NOT INTENDED TO STAND
ALONE AND SHOULD BE USED ONLY IN
CONJUNCTION WITH THE REPORT AND THE
EXAMPLES AND EXPLANATIONS THEREIN
CONVERSION Do not import the Excel 5.0 file into another

spreadsheet. Certain functions do not translate. Instead use the
generic version of the spreadsheet RAILCAP.WK1 specifically
converted for DOS or windows versions of Lotus 1-2-3, Quattro
Pro, or other spreadsheets. When opening the file always check
to ensure correct values are obtained by comparing the results in
the default column with the adjacent entry column. Excel users
must install the solver add-in.
SIMPLE ACHIEVABLE CAPACITY ESTIMATION The

report contains simple methods to estimate achievable capacity
of rail transit that does not require use of the spreadsheet. Refer
to Figures S.11 and S.12 in the report, also reproduced on line
390 of the Excel spreadsheet. This is the preferred method rather
than using this spreadsheet with default values. It provides faster
results and a reasonable range of values with less chance of error.
COMPLETE METHOD OF CAPACITY ESTIMATION

Achievable capacity is the maximum number of passengers that
can be carried in an hour, in one direction, on a single rail transit
track, allowing for the diversity of demand. There is no precise
value. The density of passengers on a carthe loading level
can vary from system to system by up to a factor of three.
Similarly an allowance for irregularities, the operating margin,
can range widely depending on prioritiesmaximum capacity or
the most reliable operation. Values for the loading level and
operating margin are inputs into this methodology. The default
values can be used but reference to the report is recommended to
select an appropriate value for each specific system.
The best method to estimate capacity is with a complete
system simulation involving models of the signaling system,
power supply system and train performance. The following
methodology involves simplifications and approximations.
Correctly applied with reasonable input values, it should
estimate capacity within 10%. Incorrectly used it can
produce erroneous values.
ALWAYS CHECK THE RESULTS WITH THE RANGES

IN THE REPORT, AND FIGURES S.11 AND S.12, TO
ENSURE THEY ARE REASONABLE. IF IN DOUBT USE
THE RANGES FROM THE REPORT.
step 1
DETERMINING THE WEAK LINK
Rail transit capacity is set by the weakest link or bottle-neck on a

system. This may be at a flat junction or at the terminal turnback.
Such constraints should not be tolerated on a new system. Where
they may exist on an existing system, Chapter Seven of the
report shows methods to calculate such headway restrictions and
in turn, the achievable capacity. By far the most common bottleneck is the time for one train to replace another at the busiest
maximum load pointstation.
On light rail systems a possible weak link is any single-track
section over 400 to 600 m long. A separate spreadsheet
LRSINGLE.XLS or WK1 contains the equation to calculate the
headway restrictions due to single track. Light rail may also be
limited by on-street operation or by grade crossings, as discussed
in Chapter Eight. However the most common limitation is that of
any signaled section. The methodology of Chapter Three (step 2)
can be used for light rail when the signaling is designed for
maximum throughput. Otherwise, the design headway of the
system should be used.
If you are sure that the weak link is the time for one train to
replace another at the busiest station, then proceed to the next
step that is applicable to rail rapid transit (heavy rail), light rail
with segregated right-of-way signaled for maximum throughput,
all automated guideway transit with on-line stations and
commuter rail with electric multiple-unit equipment using rapid
transit type signaling. For other capacity determination refer to
the report.
step 2
CALCULATING SIGNALING SYSTEM

THROUGHPUT AT THE PEAK-POINT
STATION
The minimum train separation includes any safety distances or

times, the time to brake into a station and to accelerate out until
the platform is clear for the next train to enter. Refer to Equation
3-15, the station minimum headway formula, for conventional
signaling.
The spreadsheet applies this equation for conventional three
aspect, cab control and moving-block signaling. Insert your
system and train values in the blue column4 or use the defaults
(red column). The results are shown in the yellow cells.
where H(s) = Station minimum train separation without dwell

or operating margin, and
va= Optimum approach speed to maximum load point station
4
The spreadsheet BLUE for values is shown as a light tone, RED, default
values as a dark tone, YELLOW for results as a heavy border.
xxi
Spreadsheet (part) RAILCAP.XLS showing default data
This table lists mean dwells at the maximum load point station of
several systems. Your choice should be from 30 to 60 sec. The
high value would be for a rail rapid transit system with heavy
mixed flows, the lower value for uni-directional flows under
optimal conditions. A default of 45 sec is recommended where a
specific value is not self evident.
2) Use the methodology of Chapter Four to estimate a dwell
based on the hourly flow, by direction, at the maximum load
point station. This methodology is calculated in the Excel
spreadsheet but omitted from the generic spreadsheet.
step 4
If your system is not designed for minimum train separation
insert the value of H(s) obtained from a simulation or
specification in the above results box and transfer to Step 7.
NOW check that there are no speed restrictions on the maximum
load point station approach that would prohibit a train operating
at the optimal approach speed va in the above results boxes.
Refer to Figure 3.5 which shows the distance a train would be
from the station platform stopping point at the respective speeds.
If there are no speed restrictions (due to curves or switches or
safety speed limits) then proceed to the next step.
IF there are speed restrictions within this distance then manually
type in the restriction in the respective result boxes above in
kilometers per hour. The station minimum train separation in the
cell above will automatically increase from the calculated level.
step 3
ESTIMATING OR CALCULATING
THE DWELL TIME
Refer to Chapter Four, Station Dwells, for a detailed discussion.

Dwells cannot be determined precisely. You have two choices.
1) Select a reasonable value from the table below.
Peak-period dwells for heavily used systems
OPERATING MARGIN SELECTION
Refer to Chapter Six, Operating Issues, for a detailed discussion.

An operating margin is essential for regular running. If the
minimum headway consisted only of the minimum train control
separation plus the maximum dwell, any minor incident, delay or
extended dwell would result in interference between trains.
The more operating margin that is allowed then the lower the line
capacity and the greater the probability of even performance.
Determining an operating margin requires a balancing act
between these two desires. The table below (Table 4.17 in the
report) offers guidance based on the projects field survey. For
maximum capacity, a range of 15 to 25 sec is recommended. A
default value of 20 sec is used in the spreadsheet. If your
priorities are to avoid irregular running at the expense of
maximum passenger capacity then a higher operating margin
could be appropriate.
Alternately from this table you can select a controlling dwell
consisting of the mean dwell plus two standard deviations and
omit or minimize any operating margin. One approach is to use
the higher of this or dwell plus operating margin.
Controlling dwell examples (seconds)
xxii
step 5
SELECTING THE LOADING LEVEL
Refer to Chapter Five, Passenger Loading Levels, for a detailed

discussion. Levels vary widely across North America from the
loaded conditions on certain New York trunks and on Mexico
City meter lines to the more relaxed levels that provide almost a
seat for every passenger. In fact, a seat for every passenger is the
common standard on all commuter rail lines.
There are two approaches. 1) Select a loading level, in
passengers per meter of car or train length, from the heavy rail
figure below (Figure 7.3 in the report), 7.0 passengers per meter
of train length is recommended, or from the figure for articulated
light rail below (Figure 7.4 in the report), 6.0 passengers per
meter of train length is recommended.
2) Calculate the capacity of a specific car by entering the

dimensions, the type of seating and the standing density in
Equation 5-2. This calculation is contained in the spreadsheets.
step 6
SELECTING THE LOADING

DIVERSITY FACTOR
Refer to Chapter Five, Passenger Loading Levels, and Chapter

Seven, Grade Separated Rail Capacity Determination, for
detailed discussion. The next step is to select a loading diversity
factor based on the rail mode and the type of system. Consult the
table below (Table 5.14) for actual diversity factors of various
systems. Unless there is sufficient similarity with an existing
operation to use a specific figure, the recommended loading
diversity factors are 0.80 for heavy rail, 0.75 for light rail and
0.60 for commuter rail operated by electric multiple-unit trains.
Diversity of peak hour and peak 15 minutes5
Linear passenger loading of heavy rail cars
This peak hour diversity factor is the same as the peak-hour factor (phf) in
the Highway Capacity Manual(R47)
6
Linear passenger loading of articulated light rail cars
xxiii
step
THE FINAL STEP CALCULATING THE

ACHIEVABLE CAPACITY OF A RAIL
RAPID TRANSIT SYSTEM ON
SEGREGATED TRACK WITH TRAINS
OPERATING AT THE CLOSEST SPACING
PERMITTED BY THE SIGNALING
In this final step, the results of the preceding steps are brought
together and multiplied to produce the estimated achievable
capacity of the system.
Total headway is the sum of the signaling minimum headway
plus dwell time and operating margin or dwell time plus two
standard deviations. Dividing this sum into 3600 produces the
number of trains per hour, which must then be multiplied by the
passengers per meter, the train length and the loading diversity
factor to produced the achievable capacity in passengers per
peak-hour direction per track.
ALWAYS CHECKTHAT THE FINAL RESULT IS

REALISTIC BY REFERRING TO THE FOLLOWING
FIGURE. IF THE RESULTS ARE ABOVE THESE
LEVELS THEN YOU HAVE EITHER SELECTED
UNREALISTIC INPUT DATA OR MADE AN ERROR. IF
IN DOUBT USE THE DEFAULT VALUES FROM
FIGURES S.11 AND S.12.
Passengers per peak hour per direction
Data from preceding steps (default values shown)
Typical maximum passenger capacities of grade-separated

rail transitexlcuding all-seated commuter rail.
CAUTION Light rail signaling is rarely designed for minimum
headway. No light rail line in the United States and Canada
carries more than 10,000 passengers per peak-hour direction.
1. Rail Transit in North America

1.1 INTRODUCTION
Rail transit plays a significant role in moving people in North
American cities. In U.S. urbanized areas exceeding 200,000 in
population, 35% of all transit trips in 1993 took place on one of
the four rail modes with rail rapid transit alone accounting for
28% of these trips.
The four rail modes consist of Automated Guideway Transit
(AGT), Commuter Rail (CR), Light Rail Transit (LRT) and
Rail Rapid Transit (RT), often called Heavy Rail. Table 1.1 and
Figures 1.1 and 1.2 give a condensed look at some of the key
North American statistics for each mode.
Table 1.1 Comparison of key modal statistics
Figure 1.1 Rail transit annual passenger trips by mode

(billions, Fiscal Year 1993)
1.2 LIGHT RAIL TRANSIT

1.2.1 INTRODUCTION
Light rail transit (LRT) started as a modification of streetcar
operation to allow higher speeds by separating it from street
traffic. LRT is characterized by its versatility of operation as it
can operate separated from other traffic below grade, at-grade,
on an elevated structure, or together with road vehicles on the
surface. Service can be operated with single-car or multiple-car
trains. Electric traction power is taken from an overhead wire,
thus eliminating the restrictions imposed by having a live thirdrail at ground level. (An exception is Southeastern
Pennsylvania Transportation Authoritys [SEPTA] gradeseparated Norristown high-speed line which uses third-rail
current collection.) This flexibility helps to keep construction
costs moderate and explains the popularity this mode has
experienced since 1978 when the first of 14 new North
American light rail transit systems was opened in Edmonton,
Alberta.
These newer light rail transit systems have adopted a much
higher level of segregation from other traffic than earlier
systems enjoyed. Boston opened a downtown streetcar subway
in 1897 with Philadelphia and Newark following later. New
Jersey Transits (NJT) Newark City Subway, opened in 1935,
also benefits from extensive surface private right-of-way.
Segregation from motor traffic permits higher speeds, greater
schedule reliability and improved safety. Modern signal preemption and progression methods have also made on-street
operation faster and more reliable.
Passenger loading can be accomplished at street level with
steps on the cars, or at car floor level with high-level platforms.
The lines in Calgary, Edmonton, Los Angeles and St. Louis
operate entirely with high-platform access. The San Francisco
Municipal Railway uses moveable steps on its cars to allow
them to use both high-platform stations and simple street stops.
Pittsburgh takes a different approach and has two sets of doors
on its light rail vehicles, one for high platforms the other for
low-level loading. Most other systems use low-level loading
with steps. Low-floor cars, already popular in Europe, have
been ordered for Portland and Boston to provide floor-level
loading without the need for steps or high platforms.
Wheelchair access also benefits because lifts are not required
with low-floor cars.
1.2.2 STATUS
Figure 1.2 Rail transit annual passenger-kilometers by

mode (billions, Fiscal Year 1993)
There are currently 23 light rail transit systems in operation in

North America (Table 1.2). This total includes the traditional
streetcar lines in Toronto and New Orleans. Lines that are
primarily operated for heritage and tourist purposes, such as
those in Memphis and Seattle, are not included in this study.
2
The recent popularity of light rail transit is apparent in that
12 of the surveyed light rail systems have opened since 1980.
Older streetcar systems in Boston and Philadelphia survived
the widespread replacement of streetcars with buses following
the two world wars thanks to city center tunnels that gave them
rapid access to downtown. San Franciscos streetcars benefited
from two tunnels that provide strategic routes under major hills
in that city. Pittsburghs streetcars survived for similar reasons.
These older systems have been modernized with new cars, and,
in the cases of Pittsburgh and San Francisco, with tunnels
penetrating the downtowns of their respective cities.
Toronto is the last city to operate a largely conventional
streetcar network. Torontos streetcars must share most their
routes with vehicular traffic, a condition which leads to
relatively low speed service. Many of the other older streetcar
systems with light rail characteristics must also operate with
general traffic on substantial portions of their routes. Such is
the case in San Francisco and Philadelphia where tunnels
bypass downtown traffic congestion and surface in outlying
areas.
1.2.3 RIDERSHIP
Ridership information collected by light rail transit systems is
not as comprehensive as for the other modes with many
systems only reporting the total number of passengers carried
on an average weekday. Peak-hour and peak-15-min flows
were obtained for a number of systems but this important data
Table 1.2 North American light rail transit systems
Figure 1.3 Weekday ridership for the 15 busiest North

American LRT lines (thousands, Fiscal 1993)
was not available for some of the major light rail transit
systems. As a result, average weekday ridership for major
routes is shown in Figure 1.3 with the available peak flows
shown in Figure 1.4. Data for the TTCs traditional streetcar
lines are not included but may be found in Appendix (A3). Few
light rail lines operate near capacity, with the exception of the
trunk portions of San Franciscos Muni Metro and Bostons
Green Line.
It is worth noting that the first and fourth busiest light rail
transit lines in North America, Calgary Transits South (201)
and Northeast (202) lines, operate mostly at-grade; downtown
operation is on a transit mall shared with buses.
1.3 RAIL RAPID TRANSIT

1.3.1 INTRODUCTION
1
2
Historic, conventional street car line.

Conventional streetcar network with little segregation of tracks.
Rail rapid transit (heavy rail) is by far the predominant urban

rail travel mode in North America. Systems are listed in Table
3
Table 1.3 North American Rail rapid transit systems
Figure 1.4 Peak-hour and peak-15-min directional flows for

light rail transit trunks (passengers per hour per direction,
Fiscal 1993)3
1.3. Figures 1.1 and 1.2 illustrate the lead rail rapid transit has
over the other rail modes in both annual passenger trips and
annual passenger kilometers. Rail rapid transit is characterized
by fully grade-separated rights-of-way, high-level platforms
and high-performance, electric multiple-unit (EMU) cars.
The expeditious handling of passengers is enabled through
the use of long trains of up to 11 cars running a frequent
service. Loading and unloading of passengers at stations is
rapid due to level access and multiple double-stream doors.
Power is generally collected from a third-rail but can also be
received from overhead wires as in Cleveland, Bostons Blue
Line and Chicagos SkokieSwift.4 Third-rail power collection,
frequent service and high operating speeds generally necessitate
the use of grade-separated pedestrian and vehicular crossings.
Grade crossings are an exceptional feature on third rail systems
in Chicago and New York.
San Francisco and Oakland is wide to allow the high overall

speed required to compete with the automobile. The Canadian
and Mexican systems are exceptions. Despite being of
relatively recent construction, they have loading and station
spacing standards similar to older lines in the United States. BC
Transits SkyTrain is included in the rail rapid transit category
rather than light rail or automated guideway categories. It most
closely resembles rail rapid transit system in operating practices
and right-of-way characteristics.
The costs of constructing fully grade-separated rights-of-way
(subway or elevated) for rail rapid transit have limited new
systems in recent years although extensions are being planned
or built in several cities.
1.3.3 RIDERSHIP
Two of the 18 rail rapid transit systems operating in North
America, the Sistema de Transporte Colectiva in Mexico City
and MTA - New York City Transit, carry two-thirds of all riders
using this mode. Figure 1.5 shows the dominance of these two
1.3.2 STATUS
A distinction can be made between the generally older systems
where high passenger densities are routine and stations are
spaced closely together, and newer systems that tend to place a
higher value on passenger comfort and operating speed.
BART in the San Francisco Bay area is a prime example of
the latter category with fast trains where most of the passengers
have upholstered seats. BART station spacing outside downtown
3
15-minute data not available for most light rail lines. MBTA Green line
trunk data estimated by MBTA staff.
Skokie Swift has light rail characteristics. The CTA defines it as rail rapid.
Figure 1.5 Concentration of rail rapid transit ridership

(billions of annual riders, 1993 data)
Figure 1.6 MTA-NYCT subway tracks in Midtown

Manhattan7
regions relative to the rest of the continent.5 Rail rapid transits

efficiency in moving large volumes of passengers in densely
populated areas is evident in these, the two largest metropolitan
areas in North America. Rail rapid transit plays a key role in
enabling such concentrated settlements to exist. In 1992, 50.9%
of business day travel into the Lower Manhattan hub was by
rail rapid transit. In the 7 - 10 am time period this share
increases to 62.2%.6
The 794-km route New York subway system is one of the
largest and most complex in the world. This extensive subway
system carries almost twice as many riders as does the local bus
system. Most lines are triple or quadruple tracked to allow the
operation of express services. A large number of junctions
permit trains to be operated on a variety of combinations of line
segments to provide an extensive network of service. Figure 1.6
shows the complexity of subway tracks in Midtown Manhattan.
Figure 1.7 illustrates the peak-hour and peak-15-min passenger flow rates for the 15 busiest rail rapid transit trunk lines in
North America outside Mexico City.8 The STC in Mexico City
is not included because passenger crowding up to 6 passengers
per m 2 is beyond what is acceptable elsewhere in North
5
The New York data used in the chart also includes the relatively small
contributions of the Port Authority Trans-Hudson (PATH) and the MTA Staten Island Railway.
6
New York Metropolitan Transportation Council, Hub-Bound Travel 1992,
December 1993.
7
From New York Railway Map, courtesy John Yonge, 1993 Quail Map
Company, 31 Lincoln Road, Exeter, England
Figure 1.7 Peak-hour and peak-15-min flows for the busiest

15 North American rail rapid transit trunks8
America. For comparison, the peak hourly flow on the STCs

busiest line (Line 2) is 75,300 with nine car trains every 115
sec. The graph uses trunks rather than routes in order to group
those services sharing tracks together. All the trunks listed are
double tracked and have at least one station used by all routes
serving the trunk.
When four track lines in New York are taken into
consideration, the maximum load is a combination of the
Lexington Avenue Express and Local at 63,200 passengers per
peak-hour direction with almost comparable volumes on the
combined Queens Boulevard lines at Queens Plaza. Detailed
rail rapid transit ridership data can be found in the tables of
Appendix Three.
8
Peak 15-min flow data were not available for all lines for which peak-hour
data were available.
1.4 COMMUTER RAIL

1.4.1 INTRODUCTION
Commuter rail is generally a long distance transit mode using
trackage that is a part of the general railroad system, some of
which is used exclusively for passengers. Track may be owned
by the transit system or access may be by agreement with a
freight railroad. Similarly train operation may be by the transit
agency, the track owner or a third-party contractor.
Service is heavily oriented toward the peak commuting
hours, particularly on the smaller systems. All-day service is
operated on many of the mainlines of the larger commuter rail
systems and the term regional rail is more appropriate in these
cases.
Commuter rail scheduling is often tailored to the peak travel
demand rather than operating a consistent service throughout
the peak period. Where track arrangements and signaling
permit, operations can be complex with the use of local trains,
limited stop expresses and zoned expresses. Zoned expresses
are commonly used on busy lines with many stations where
express trains serve a group of stations then run nonstop to the
major destination station(s).
Diesel and electric power are both used for traction on
commuter rail lines. Electric traction is capital intensive but
permits faster acceleration while reducing noise and air
pollution. It is used mainly on busy routes, particularly where
stops are spaced closely together or where long tunnels are
encountered. Both power sources can be used for locomotive or
multiple-unit operation. All cars in a multiple-unit train can be
powered or some can be unpowered trailer cars, which must be
operated in combination with powered cars. Electric multipleunit (EMU) cars are used extensively in the New York,
Philadelphia and Chicago regions with the entire SEPTA
regional rail system in Philadelphia being electrified. SEPTA
and GO Transit (Toronto) are the only systems with lines
routed through the center city. There are currently no diesel
multiple-unit commuter trains in North America although this
will change once commuter rail service begins in Dallas.
Locomotive-hauled commuter trains are standard for diesel
operation and are becoming more common on electrified lines
as a way to avoid the high costs of multiple-unit cars. New
Jersey Transit and SEPTA have both purchased electric
locomotives as an economical alternative to buying multipleunit cars. Other systems place a high value on the flexibility of
multiple-unit cars in varying train length. The STCUM in
Montral is replacing a mixed fleet of multiple-units and
electric locomotives with a standard new multiple-unit design.
Commuter rail train length can be tailored to demand with
cars added and removed as ridership dictates. This is
particularly easy with multiple-unit equipment and can result in
trains of anywhere from 2 to 12 cars in length. Where train
length is constant all day, unneeded cars can be closed to
passengers to reduce staffing needs and the risk of equipment
damage.
Commuter rail is unique among the transit modes in that a
high priority is placed on passenger comfort as journeys are
long and the main source of competition is the automobile. All
lines operate with the goal of a seat for every passenger except
for the busy inner portions of routes where many lines funnel
together and a frequent service is provided. Such is the case for
the 20-min journey on the Long Island Rail Road (LIRR)

between Jamaica and Penn Stations. Service between these
points is very frequent (trains on this four-track corridor operate
as close as 1 min apart in the peak hours) as trains from
multiple branches converge at Jamaica to continue to
Manhattan.
Commuter rail cars are generally designed with the
maximum number of seats possible, although this tradition is
changing somewhat where wheelchairs and bicycles must be
accommodated. A number of common approaches are taken to
achieve maximum seating over the car length. The simplest is
the use of 2+3 seating where five seats are placed in each row
as opposed to the usual four. This can be done quite easily in
wide railroad-type cars and brings the number of seats per car
to around 120. It is not especially popular with passengers. 2+3
seating is used by many operators including the LIRR and the
MBTA in Boston. However, 2+3 seating places a constraint on
aisle width, which may be problematical with increasing
demands for wheelchair movement.
The other main approach to increasing car capacity is to add
additional seating levels to the car, subject to any height
restrictions, such as tunnels and underpasses, on the rail lines.
The gallery type car is one example and adds an upper seating
level to the car with an open well to the lower level. The well
serves to permit ticket collection and inspection from the lower
level but does limit the upper level to single seats on each side.
Gallery cars can typically seat 150 to 160 passengers and are
used most extensively by Chicagos Metra. A more recent
development is the so-called bi-level car,9 which has upper and
lower levels over the center of the car with an intermediate
level at each end over the trucks. Torontos GO Transit
popularized this design with relatively spacious seating for 160.
It is now also being used by Metrolink in Los Angeles, the
Coaster in San Diego and BC Transits West Coast Express.
Passenger access to commuter rail trains can be from
platforms at floor level or ground level with the former
commonly used on busy lines or at major stations to speed
passenger movements. Standard railway type traps in the
stepwells allow cars to use both types of platform but require
the train crew to raise and lower the trap door above the steps.
The EMU cars used by the Northern Indiana Commuter
Transportation District on the South Shore line out of Chicago
and some New Jersey Transit cars employ an extra set of doors
at the center of the car that are used exclusively at high
platform stations, while the end doors are fitted with traps in the
conventional manner for use at high-and low-platform stations.
This arrangement is also being used on the new EMU cars
being delivered to the STCUM for use on Montreals Mount
Royal tunnel line.
1.4.2 STATUS
Commuter rail services operate in 13 North American
metropolitan regions. These include the recently started Coaster
service between San Diego and Oceanside, California. There
has been rapid growth in this mode as a result of the availability
of government funding and the relatively low capital costs of
the mode.
9
Less commonly known as tri-level cars as there are technically three floor
levels.
6
Table 1.4 North American commuter rail systems
Figure 1.8 Annual ridership for the 10 busiest North

American commuter rail systems (millions, Fiscal Year 1993)
Dallass DART is expected to start commuter rail service in fall

1996.
Extensions and expansions are planned on other systems to
enlarge the service area and provide additional parking for
patrons. With many commuter rail lines serving low-density
suburban areas, the provision of adequate customer parking is a
key to maximizing ridership. To meet this need, some agencies,
such as Metra, are building stations whose primary purpose is
to allow parking capacity to be expanded at low cost in
relatively undeveloped areas. (See Table 1.4.)
Figure 1.9 Peak-hour and peak-15-min flows for the busiest

15 commuter rail trunks10 (Fiscal Year 1993)
plans should mean continued ridership growth for MBTA
service in the future. Figure 1.9 shows the hourly and 15-minpeak riderships for the 15 busiest commuter rail lines in North
America. Although the New York area is dominant in total
commuter rail ridership, it is interesting that 10 of the 15
busiest individual routes are outside the New York area.
1.5 AUTOMATED
GUIDEWAY TRANSIT
1.4.3 RIDERSHIP
1.5.1 INTRODUCTION
Ridership is highly concentrated New York(3) and

Chicago(1) metropolitan systems are the four busiest on the
continent, as shown in Figure 1.8. GO Transit in Toronto, one
of the first of the new generation of commuter rail systems,
ranks fifth. Bostons MBTA has had ridership double over the
last decade thanks to extensive capital investment. Expansion
Automated guideway transit (AGT) is the newest of the rail

transit modes and has played a relatively minor role in North
10
Ridership data for SEPTA is from Regional Rail Ridership Census, 199394, copyright Southeastern Pennsylvania Transportation Authority, July
1994.
7
Table 1.5 North American automated guideway transit
systems (surveyed systems)
American transit. As the name suggests, the operation of these

systems is completely automated with personnel limited to a
supervisory role. Inherent in the definition of this mode is the
need for guideways to be fully separated from other traffic.
Cars are generally small and service frequentthe name people
mover is often applied to these systems, which can take on the
role of horizontal elevators.
mandated by the Federal Transit Administration (FTA).

Ridership on many systems is also likely affected by seasonal
patterns and less pronounced peaking than occurs on transit
systems. Regardless of these qualifications, the total daily
ridership on the 37 nontransit systems amounts to almost
670,000 compared to just over 40,000 on the three public AGT
lines.
Table 1.6 Daily ridership for North American automated

guideway transit systems (source: Transit Pulse11 and
database, various years, 1992-1994)
1.5.2 STATUS
Automated guideway transit systems operate in regular transit
service in three U.S. cities plus the AGT system at the West
Virginia University campus in Morgantown, WV. This 5-km
line features off-line stations that enable close headways, down
to 15 sec, and permit cars to by-pass intermediate stations. The
cars are small, accommodating only 23 passengers, and are
operated singly. On-demand service is possible at off-peak
hours.
The transit operations surveyed (Table 1.5) include the
Detroit People Mover, Miami MetroMover and the VAL line in
Jacksonville, FL. The latter line, at less than a kilometer in
length, is to be replaced with a more extensive automated
monorail. The Detroit line has remained unchanged from
opening in 1987 while the Miami MetroMover added two
extensions in 1994.
The vast majority of AGT systems are, however, not
operated by transit systems. Many lines serve institutions (such
as the Morgantown line), airports and recreational facilities.
The ridership table in the following section shows the
dominance of these nontransit systems.
1.5.3 RIDERSHIP
Given the small number of transit agencies operating AGT, the
amount of transit ridership data is limited. Even among the
transit agencies, ridership data collection is limited to all-day
ridership counts. Data from West Virginia University in
Morgantown show their line carries 16,000 riders per day with
a peak one-way hourly flow of 2,800.
Daily ridership data are shown in Table 1.6. Caution should
be exercised with many of these figures as the non-transit
systems are not required to provide the reporting accuracy
11
Transit Pulse, PO Box 249, Fields Corner Station, Boston, MA
02122.
2. Capacity Basics
2.1 INTRODUCTION
Capacity is an important measure of a rail transit systems
passenger-handling capability. It is determined to ensure that a
line is built, expanded or re-equipped with adequate facilities to
handle the peak-hour passenger demands both in the near and
long term, comfortably and safely. Other applications for
capacity information are as follows:
project planning and operations analysis for new starts and
extensions,
evaluating transit line performance,
establishing and updating service standards,
studying environmental impacts,
assessing the capacities of new signaling and control
technologies,
estimating changes in system capacity and operations over
time, and
assessing capacity impacts in land-development studies
where transit is expected to provide a significant role in
site access.
This chapter defines capacity and develops an initial
framework to analyze and determine the capacity of rail transit
modes in North America.
2.2 TERMINOLOGY
2.2.1 DEFINITIONS
The North American rail transit industry is inconsistent in its use
of terminology. Numerous reviewed reports use the same term
to mean different things. Several reports develop their own
definitions.
Chapter 13 provides a project glossary derived from the TRB
and APTA transit glossaries. These definitions are used
consistently throughout the report. Where reference must be
made to an alternative definition, the variation is clearly noted in
the text or via an accompanying footnote.
Note that headway and capacity are inversely related and this
can be a source of confusion. The minimum or closest headway
delivers the maximum capacity.
italicized superscript number8 referenced to the bottom of each

page.
2.3 GROUPING
Following the extensive literature review and data collection, for
the purpose of capacity analysis, the four modes of rail transit in
this study have been grouped into categories based on
alignment, equipment, train control and operating practices.
The first category is fully segregated, signaled, double-track
right-of-way, operated by electrically propelled multiple-unit
trains. This is the largest category encompassing all rail rapid
transit1, all noninstitutional automated guideway transit2, several
light rail sectionsfor example, the Market Street subway in
San Francisco, and several commuter rail lines on the east coast.
This category is termed Grade Separated Rail.
The second category is light rail without fully segregated
tracks, divided into on-street operations and right-of-way with
grade crossings. Streetcar only operations (Toronto and New
Orleans) is a sub-set of the on-street section.
The third category is commuter rail other than services in
category one.
The fourth category is automated guideway transit (AGT).
Although most AGT is a sub-set of the main category, Grade
Separated Rail with very short trains, the use of off-line
stationson certain systemsis unique to this mode and
requires separate examination. Off-line stations can also
increase the capacity of more conventional rail transit as
discussed in Chapter Six, Operating Issues.
Each of these categories is provided with its own chapter with
the procedures for determining capacity.
2.4 THE BASICS

Professor Richard Soberman
Handbook(R19) states:
in
the
Canadian
Transit
The capacity of transit service is at best an elusive

1
2.2.2 FOOTNOTES AND REFERENCES

To avoid duplication, references are shown as (R23) and refer to
the Bibliography of Chapter 12 and the literature review item of
the same number in Appendix One. Footnotes are shown by an
Chapter 7 Grade Separated Rail Capacity Determination

Chapter 8 Light Rail Capacity Determination
Chapter 9 Commuter Rail Capacity Determination
Chapter 10 AGT Capacity Determination
The minor exceptions where there are grade crossings on rail rapid transit
(CTA) will be discounted. Routes with more than two tracks will be
discussed relative to express, local and skip-stop service. However, it is
not intended to otherwise develop unique capacity calculations for
multiple track routes.
The Morgantown automated guideway transit, the only North American
example of AGT with off-line stations, is not classed as a public operation
by APTA.
9
figure because of the large number of qualifications
that must be attached to any measure of capacity that is
adopted.
Most of the capacity calculations in the literature add
constants, multipliers, reductive factors or other methods to
correlate theory with practice.
In this study emphasis has been placed on reducing the
number of qualifications and quantifying, describing and
explaining adjustments between theory and practice in
determining rail transit capacity.
The literature is in general agreement on a definition of rail
transit capacity as:
whether those spaces going by each hour will be usedthey

would be fully used only if passengers uniformly filled the trains
throughout the peak hour. This does not occur and a more
practical definitionsometimes referred to as practical
capacityis required. Achievable capacity takes into account
that demand fluctuates over the peak hour and that not all
trainsor all cars of a trainare equally and uniformly full of
passengers.
Achievable Capacity
carried in an hour in one direction on a single
track allowing for the diversity of demand.

carried in an hour, in one direction on a single track.
Several papers add refinement to compensate for diversity of
loading within the maximum peak hour. This compensated
definition was referred to in some cases as the practical
maximum rail transit capacity. Other definitions added
qualifiers such as: sustainable over a peak hour without
impedance (to other trains) or the less restrictive without
unrecoverable delays to trains.
This study is oriented to practical results and it would be
logical to include peak-hour diversity in the definition of
maximum capacity. In North America the diversity factor of
total peak-hour capacity to peak-within-the peak capacity ranges
from 0.70 to 0.95. The latter high factor, relates only to a few
lines in New York and Mexico City. Most rail transit fits into
the range of 0.75 to 0.90.
However, in practice it is correct, if somewhat misleading, to
quote a maximum hourly capacity of 60,000 passengers, or
passenger spaces, per peak-hour direction when, as passengers
do not arrive evenly over the peak hour to fill this capacity, the
actual number of passengers carried in one hour is 45,000.
This introduces the issue of supply and demand. This study
determines supplythe number of passenger spaces per peak
hour per track that is providednot the number of passengers
actually carried. Although demand is not within the scope of the
study, a secondary issue has been added to examine demand
with particular respect to station constraintsinadequate
platform size, number of exits, ticketing throughput and parking
limitationsdiscussed in Chapter Six, Operating Issues.
To avoid any confusion between supply and demand, and to
avoid confusion with other work, the study uses two definitions
of capacity.
Unless otherwise stated, reference in the study to passenger

capacity means the achievable capacity of a single line.
Reference to single track is necessary as most rail rapid trunk
routes in New York3 have three or four tracks while the Broad
Street subway in Philadelphia and the North Side L in Chicago
have four tracks. The capacity of four-track lines is not a simple
multiple of two single tracks and varies widely with operating
practicesthe merging and dividing of local and express
services and train holding at stations for local-express transfers.
The result is that, given adequate demand, four tracks can
theoretically increase capacity by 80% over a double-track
linealthough 50% is more typical. A third express track does
not necessarily increase capacity at all when restricted to the
same close-in limitations at stations with two platform faces.
Design capacity has two factors, line capacity and train
capacity, and can be expressed as shown in Figure 2.1. In turn
the achievable capacity can be expressed as shown in Figure 2.2.
The basic capacity expression can be expanded as shown in
Figure 2.3. This expression of Figure 2.4 determines the number
of trains per hour and is the inverse of the closest or minimum
headway. The relevant minimum train separation in seconds is
the minimum time to approach and leave a station, i.e., the time
from when a train starts to leave a station until the following
train can berth at that station. This is referred to as the close-in
time.
Figure 2.1 Basic design capacity expression
Design Capacity
The maximum number of passenger spaces
past a single point in an hour, in one
direction on a single track.
Design capacity is similar to, or the same as, maximum capacity,
theoretical capacity or theoretical maximum capacity
expressions used in other work. It makes no allowance for
Figure 2.2 Basic achievable capacity expression

3
All New York four-track trunks merge into double-track sections, tunnels
or bridges, crossing the Harlem and East Rivers.
10
margin or a margin can be added separately to the denominator
of the expression. Chapter Four, Station Dwells, develops the
methodology and analysis of dwells. Chapter Six, Operating
Issues, discusses and develops operating margins.
The expression of Figure 2.4 determines train throughput at
the controlling stationusually the maximum load point station.
In rare cases speed restrictions or heavy mixed passenger flows
may dictate that other than the maximum load point station
controls the closest achievable and repeatable headway.
From the above expressions the framework can be expanded
to include other variables. Figure 2.5 outlines the project.
The next section in this chapter discusses the relationship
between design and achievable capacity, followed by sections
expanding and explaining the components of the project flow
chart.
Figure 2.3 Expanded design capacity expression
Figure 2.4 Line capacity expression (train throughput/hour)
In determining this minimum headway, the train separation is

based on line clear close-in, with successive green signals
governing the following train. Such a headway is termed
noninterference. The minimum line headway is determined by
the critical line condition, usually the close-in at the maximum
load point station.
The entire stretch of a line between junctions and turnbacks,
where train density is physically constant, is governed by this
one critical close-in. In a small number of cases the critical
governor of headway is the terminal maneuver. In the Rail
Transit Survey nine4 out of 58 responding systems cited turn
backs as a constrainttwo light rail, five rail rapid transit and
two commuter rail operators. In comparison, 34 operators cited
train control limitations as a capacity constraint.
Junctions are not usually headway constraints. In the projects
Rail Transit Survey, only four out of 58 responding systems
cited junctions as a constrainttwo commuter rail and two
heavy rail operators. This reflects the good design of the busiest
systems in the survey where potential junction constraints are
minimized by grade separation. Chapter 3, Train Control and
Signaling, develops analytic methods for determining the closein time at stations, or headway limitations at junctions and
turnbacks, for a variety of train control systems.
The other factor in the expression controlling dwell is based
on actual station dwell time adjusted to a controlling value over
the peak hour. The controlling dwell may contain an operating
4
A closer examination of turnback constraints shows that many are due to

operating practicesnot physical constraints.
2.5 DESIGN VERSUS

ACHIEVABLE CAPACITY
The objective of this project is to provide guidelines and meth
ods that can be used for real-world evaluation of rail transit
capacities. As such it is appropriate to consider the difference
between design and achievable capacity.
Design capacity, in passengers per hour per direction (pphpd),
is often calculated using the following factors:
number of seats per car,
number of standees per car (= standing area standee
density),
number of cars per train, and
train headway (minimum headway determined by a
combination of the signaling system, station dwell, and
terminus constraints).
Such an approach, however, does not incorporate many realworld factors that may reduce the actual number of regular
riders that the system can or could sustain.
Standing densities are not as absolute as the typical four
passengers per square meter implies; people will crowd in
more tightly in some situations than in others.
It is rarely possible to equalize loading densities perfectly
in a multi-car train; some car positions invariably carry
more passengers on average than others.
Many factors can reduce train performance (propulsion
faults or differences, door problems, operator variation),
which may not only increase the sustainable average
headway, but will increase the variation in headway, and
consequently the passenger load waiting for that train.
Minimum headway, by definition, leaves no margin for
schedule recovery from even minor delays, leaving the
system susceptible to more variation in service.
Passenger demand is usually distributed unevenly within
the peak; there may be predictable waves of demand,
corresponding to specific work start and finish times.
Since passengers are essentially a perishable commodity
(i.e., may not tolerate being forced to wait for later
11
Figure 2.5 Project outlineanalytic framework flowchart (Circled numbers denote the relevant report chapter)
departures), the capacity rate requirement for the peak 10
to 15 min may have to be significantly higher than the
average for the peak 1 or 2 hr.
There is day-to-day fluctuation in demand. Some may be
associated with the day of the week (peaks have become
lighter on Mondays and Fridays as more people move into
shorter or flexible work weeks), seasonally (lighter in the
summer and at Christmas time), weather and special
events. Beyond those identifiable factors, which may be at
least partially anticipated, are essentially unpredictable,
random variations in demand.
Passengers are resilient to a degree, and will tolerate overcrowding or delay on occasion. This is an important safety
valve that permits at capacity systems to accommodate
special events or recover from service delays, with perhaps
less difficulty than would be predicted.
Achievable capacity is the product of the design (maximum)
capacity and a series of reality factors, most of which
downrate the ideal. These factors are not absolutes, since they
reflect human perception and behavior, as well as site-specific
differences (expectations, cultural attitudes and the
transportation alternatives). This study has endeavored to derive
these factors from observation and understanding of existing
North American rail rapid transit operations and combine them
into a single diversity factor. Chapter Five, Passenger Loading
Levels, details existing diversity factors and recommends factors
for new systems.
2.5.1 SERVICE HEADWAY

Design (minimum) train operating headway is a function of
signaling system type and characteristics, including block
lengths and separation;
operating speed at station approaches and exits or other

bottlenecks such as junctions;
train length; and
station dwells.
A review and comparison of signaling and train control systems
is included in Chapter Three, Train Control and Signaling.
Table 2.1 presents minimum headway constraints under current
conditions on 53 of the systems surveyed. (Six operators stated
there were no constraints, three did not respond.) These stated
constraints are not necessarily absolute; many systems are not
operating at or close to capacity and have therefore not
exercised all of the relatively easy improvements that could be
made within their existing plant and technology. In particular
several of the turn-back constraints relate more to operating
practices than physical limitations.
Achievable headway must account for additional factors that can
affect the separation of individual trains:
Operator performance: Differences among operators can
Table 2.1 Headway constraints by mode
12
have a significant effect, depending on the number of
variables under direct operator control:
delay in initiating station departure (even if signaled by
an automatic dispatching system);
acceleration and deceleration rates (especially the latter
for manual positioning of trains at station stops);
maximum speed (particularly where an automatic
emergency brake may be imposed for overspeed); and
train separation (anticipation of signals, or following
distance in purely manual operation).
Vehicle performance: Primarily the performance of
propulsion; weak trains can impose a constraint on the
entire line.
External interference: A shared operating environment
(street-running, grade crossings, lift or swing bridges) can
impose delays that affect headways, both in a predictable
pattern (e.g., average street congestion, traffic light
timing) as well as randomly (grade crossing incidents,
exceptional traffic congestion due to traffic incidents
elsewhere, bridge operation).
Schedule recovery: Systems that attempt to operate at the
absolute minimum headway have no margin for schedule
recovery in the event of a delay. When operating at the
short headways implied in most high-volume situations,
delays of even a couple of minutes will have some effect
on passenger loading. If there is no allowance for the
above variations, then the gap, and delays to all following
trips, will be perpetuated until the end of the peak period.
Schedule recovery (over and above any labor contractual
requirements for operator layover) is essentially a
judgment call, based on probabilities and consequences of
delays, but ultimately determined by assessment of the
passenger market.
The methodology for determining service headway with most of
the above variables is developed in Chapter Three, Train
Control and Signaling. Operating margins and schedule
recovery allowances are developed in Chapter Six, Operating
Issues.
Propulsion and door interlocking: delay before the train

stops, or after the doors close.
Door operation: actual opening and closing time, plus
door warning time and any other fixed system constraints
on door operation.
Passenger volume: average number of passengers
boarding and alighting. In unconstrained, uni-directional
situations, passengers can board or alight at a rate of better
than 2 sec per passenger per single-stream doorway width.
Passenger crowding: Efficiency of pedestrian movement
is very sensitive to crowding; in the densities that are of
concern to systems that are near capacity, movement is
reduced to a slow shuffle as passengers vie for space either
in the car or on the platform. The rate is further slowed
when there is a mix of boarding and alighting.
Number, width and spacing of vehicle doors.
Platform circulation: If platforms are too narrow, or exit
paths limited, congestion on the platform can cause delays
in unloading a train; this can affect the overall station
dwell.
Single/dual platform loading/unloading: Door operation
on a single side of a train is the norm; however, some
systems configure busy stations with platforms on both
sides of a train, to allow either for segregation (off-loading
one side; loading on the other), or to split the combined
passenger movement.
High or low level platform loading/unloading.
The methodology for determining station dwells is developed in
Chapter Four, Station Dwells.
2.6 LINE CAPACITY

Line capacity is the maximum number of trains that can be
operated over a line in a peak hour. As shown in Figure 2.6, there
are two principal factors in determining line capacity which
are almost equal in weight. First is the capability or throughput
2.5.2 STATION DWELLSPRACTICAL

ISSUES
Station dwells affect the overall round-trip time, and thus can
affect the productivity of a given fleet if multiple trips are being
made. (This is of virtually no consequence for trippers,
including many commuter rail operations, which make only one
trip in each peak period.) Mid-route station dwells also affect
the inservice speed, and thus the service attractiveness. Roundtrip time and fleet size issues are not necessarily related to
maximum capacity, and are therefore not directly addressed by
this study.
However, station dwells do become a factor in capacity when
they combine with minimum operating headway to create a
constraining headway bottleneck in the system. Typically this is
a concern on fully segregated systems that are operating long
trains on close headways; busy stations, especially major
passenger interchanges, can produce block occupancy times that
limit the entire system.
Station dwells are governed by the following:
Figure 2.6 Line capacity flowchart (Not all wheelchair

options apply to commuter rail)
13
of the train control system, adjusted for various constraints,
principally those at terminals and at any junctions or single track
sections. Second is the dwell time at stations.
Both factors can be further subdivided into the three
categories based on alignment, equipment, train control and
operating practices. In turn, light rail and commuter rail lines
that are not in the principal segregated double-track category,
must be divided by high- or low-level loading and by the
method of handling wheelchairs.
shown in Figure 2.7 with the wide variation in dwells and

individual train headways.
The three constituents of dwell in this example are shown in
Figure 2.8, using NYCT Grand Central data from Figure 2.7.
The three main components of dwell are
Passenger flow time,
Door open time after flow ceases, and
Waiting to depart time after doors close.
These components vary widely from system to system. One
example, with a high ratio of dwell time used for passenger
2.6.1 TRAIN CONTROL THROUGHPUT

The number of trains per hour that is theoretically possible is
dependent on the different signaling systems including
conventional
block
signaling,
cab
signaling,
and
communication- or transmission-based signaling systems with
moving blocks. Chapter Three, Train Control and Signaling,
describes different signaling systems and develops empirical
methods to estimate their throughput. More precise throughput
determination requires the use of computer simulations.
2.6.2 COMMUTER RAIL THROUGHPUT

Certain line capacity issues are specific to commuter rail
operation. Commuter rail signaling generally must
accommodate trains of different lengths and speeds, and
contract operations may set limits on the number of trains per
hour.
Earlier in this chapter, commuter rail was divided into two
classes: those lines that emulate rapid transit with electric
multiple-unit operation on dedicated tracks (mainly in the New
York City area) and all others. Both classes need special
treatment for line capacity as they use railroad type signaling or
train control, different operating practices, and trains with
widely varying length and performance.
Figure 2.7 Dwell component of headway
2.6.3 STATION DWELLS

Station dwells and train control system minimum separation are
the two major factors in determining line capacity. In many
circumstances dwells are the dominant factor. The third factor in
headway is any operational allowance or margin. In some cases
this margin can be added to the dwell time to create a
controlling dwell time. An example of this is the dwell
component of headways at one of the small number of rail
transit lines in North America that are at capacitylines 4 and 5
at Grand Central Station in New York.
The average dwell is 64 seconds39% of the average head
way of 165 sec. The minimum train separation at this location is
55 sec. The residual of headway minus dwell and train
separation is 46 sec. This can be regarded as a surrogate for the
operating margin. The need for a suitable margin is clearly
Figure 2.8 Average headway components in seconds
14
Figure 2.10 Train capacity flow chart
Figure 2.9 Station dwell components in seconds

NYCT Grand Central February 8, 1995 (NOTE some dwell
times may have been extended due to local and express trains
waiting for each other)
flow, is shown in Figure 2.9 The importance of station dwells is
clear from these three figures. The methodology to determine
dwell times is contained in Chapter Four, Station Dwells, and
their associated operating margin in Chapter Six, Operating
Issues.
diversity factor to compensate for uneven car loadings over

multiple-car trains (see Figure 2.10). Car capacity is often
quoted at the crush loading level. This is inappropriate because
such loading levels are rarely, if ever, achieved in practice.
Rather, crush loading is a worst case level for which a cars
structure, propulsion and braking systems are designed.
Typically the North American crush level is based on 6
passengers per square meter (6/m2) (1.8 sq ft per passenger),
after making allowance for seated passengers and space lost to
cabs and any equipment cabinets or stepwells. In reality, the
typical maximum standing loads in North America range
between four and five passengers per m2 (2.2 to 2.7 sq ft per
passenger) while the average over all systems through the peak
period is only two passengers per square meter (5.4 sq ft per
passenger).
The only true means of measuring achievable car capacity is
on those systems where pass-ups occur. That is where
passengers wait for the next train rather than crowd onto the one
in their station. Avoiding pass-ups is the goal of any transit
system, so these are rare, but where they do occur, they provide
hard data on achievable car capacity.
Determining full car capacity and pass-up capacity is
discussed in the next sections relative to interior arrangements,
type of system, old or new, and time of peak loading.
2.7.2 CAR CAPACITY

Commuter Rail Dwells Dwells on many commuter rail lines
are set by schedule or policy and can be relatively independent
of passenger flows; consequently, they have a lesser effect on
capacity than occurs on other modes. In these cases, the lower
commuter rail deceleration and acceleration rates become more
significant, particularly on busy lines such as Chicagos Aurora
service where a wide range of express services is offered. The
exceptions where dwell times are more significant are the highvolume, high-platform operations using electric multiple-unit
operation on dedicated tracks. These lines, which are mostly in
the New York City area, are included in the Grade Separated
Rail category described in section of this chapter.
2.7 TRAIN/CAR CAPACITY

2.7.1 INTRODUCTION
Train capacity is the product of passengers per car and the
number of cars, adjusted to achievable capacity case using a
There are two approaches to the calculation and evaluation of

car capacitydesign-specific and dimensional average
(generic).
2.7.3 DESIGN-SPECIFIC CAPACITY

If a specific car design has already been chosen, capacity
calculation is relatively straightforward, as follows:
Number of seats: Assume each seat occupied by one
passenger.
Standing area: Usable floor area (m2 or ft2), excluding an
envelope of space for knees and feet of seated passengers,
particularly in longitudinal (side-facing) seats.
Standing density: A generally accepted average for shortdistance sustainable peak loading is 4 passengers per
square meter (2.6 sq ft per passenger), this may be reduced
for longer distance trips, or where service policy or local
conditions dictate otherwise.
15
Standing efficiency: A factor that is used explicitly to
increase or decrease the expected standing density, based
on characteristics of the standing space.
Wheelchair adjustment: With more and more rail systems
becoming wheelchair accessible, and with an increasing
number of wheelchair users being integrated into the
regular transit system, a small adjustment may be required
for wheelchair users. Typically a wheelchair occupies 1.2
to 1.5 m2, or the equivalent of two to six standing
passengers. The wheelchair adjustment factor is the
average number of wheelchairs per car, times two to six.
Typically wheelchairs represent such a small component
of ridership that their overall effect on system capacity is
negligible.
Baggage adjustment: Similar to wheelchairs, some
adjustment may be required if significant numbers of other
large objects (bicycles, suitcases, etc.) are carried on
board. On most systems the overall effect is negligible, but
it could be a factor in lines that serve airports or
recreational areas.5
2.7.4 CAR DIMENSIONS

If a specific car design has not been chosen, a generic car can
be developed for capacity calculation. This approach avoids
biases that may result from a somewhat arbitrary selection of
existing transit systems. For example, a Portland LRT car with
relatively generous seating and a New York MTA subway car
designed primarily for standees may both be representative of
their respective modes, but they do not indicate the range of
possibilities for each.
The factors that control car capacity are as follows:
Car length: Nominal length from center of couplers
allows for calculation of multi-car train lengths.
Car width: Car width at seat back height, typically 0.8
above the floor, is often 0.10 to 0.15 m wider than at
floor/platform level), recognizing that passengers hips and
shoulders are wider than the space required for head and
feet. Car width is usually described for exterior dimensions
and can be converted to interior width by assuming a
sidewall thickness of 0.05 to 0.10 m.
Nonpassenger space: Out of the nominal rectangular
envelope of the car, nonpassenger space must be deducted
for drivers cabs (which can be omitted in a fully
automated system), equipment lockers and bulkheads (if
any), and the endwalls of the car (including a typical 300
mm distance to end of the coupler).
Seat density: Seating density can range from a low of 1.5
pass/m2, typical for commuter rail or long-distance
suburban rapid transit, to a high of over 2.0 pass/m2 on
some heavy rapid transit lines that have put a premium on
overall seating capacity. This is a service quality policy
that is independent of other operating attributes.
Seating ratio: As with seat density, the percentage of
passengers to be seated is a site-specific design and policy
decision.
5
Adjustments similar to those for wheelchairs and baggage can also be

made for systems that allocate space for bicycles or strollers. Such space
usage will be dealt with in narrative form.
Standing density: Car floor space not occupied by

seating, or designated for wheelchair, baggage or bicycle
storage, can accommodate the typical 4 passengers per m2,
or may range widely (from 1.5 to7 passengers per m2 in
North America).
Long-established systems in large, older cities (New York,
Philadelphia, Chicago, Toronto, etc.) sustain higher car loadings
because people are used to it and because of limitations on the
alternativeshigh levels of traffic congestion, long driving
times and high parking fees. Newer systems offer more space
per passenger to be more attractive and competitive with
alternative travel options.
2.7.5 CAR CAPACITY CALCULATION

ALTERNATIVES
Three aspects of car capacity discussed aboveseat density,
seating ratio and standing densityare policy issues. Policy
decisions on service levels and interior design can make a three
to one difference between the capacities of two systems with the
same given train length and the same minimum train control
headway.
This suggests that for many capacity calculations, detailed
determination of seating and standing space may be
unnecessary, or, for new systems where vehicles have not been
specified, not possible. There are two possible simplified
methods for determining car capacity: the gross area alternative
and the train length alternative. Both methods can still have a
range of capacities as determined by the policies of a specific
system.
2.7.6 TRAIN LENGTH ALTERNATIVE

This alternative offers the simplest method of establishing
capacity based on policy decisions of seating type and quantity,
and standing density. This method is developed in Chapter Five,
Passenger Loading Levels.
2.7.7 TRAIN CAPACITY

Design train capacity is simply the product of car capacity and
the number of cars per train. The latter in turn will be
constrained largely by site-specific factors:
platform length (especially on existing systems)
on-street constraints (street-running light rail).
Achievable capacity is affected by systematic variation in
loading within the traintrain loading diversity. This can be
significantly influenced by station design. The factor is closest
to 1.0 if the majority of station entrances distribute passengers
effectively along the length of the platform, or if biases in some
locations are offset in others. In peak conditions, passengers will
learn to spread out, but this process is rarely perfect, and pass-up
conditions or excessive crowding will occur on some cars, while
others are less heavily loaded. Existing loading diversities
16
are tabulated in Chapter Five, Passenger Loading Levels, and
levels are recommended for use in calculating achievable
capacity.
2.8 STATION CONSTRAINTS

In rare cases station capacity constraints can reduce achievable
capacity by limiting the flow of passengers to the platform and
trains. Although this study is concerned with supply rather than
demand, a section of Chapter Six, Operating Issues, discusses
the following factors:
Station capacityincluding occupancy limits imposed by
the NFPA6 130 fire codes.
National Fire Prevention Association
Platform flow restrictions due to the number and width of

exit and entry passageways and vertical circulation
components.
Parking space inadequacies at park and ride stations.
Fare collection system capacityfare collection
arrangements are normally developed to match passenger
demand, including the use of manual collection for special
high demand events (football games, parades etc.) Only in
unusual circumstances will fare collection restrictions limit
capacity. One such circumstance is those few light rail
systems that collect fares (at some or all stops) as
passengers board. On-board fare collection on commuter
rail services is not regarded as a capacity issue although it
can be an operating problem on crowded trains.
17
3. Train Control and Signaling

Automatic train protection (ATP)
Automatic train control2 (ATC or ATO)
Automatic train supervision (ATS)
3.1 INTRODUCTION
Signaling has been a feature of urban rail transit from the
earliest days. Its function is to safely separate trains from each
other. This includes both a separation between following trains
and the protection of specific paths through junctions and crossovers. The facilities that create and protect these paths or routes
are known as interlockings.
Additional functions have been added to basic signaling,
starting, again from a very early date, with automatic train stops.
These apply the brakes should a train run through a stop signal.
Speed control can also be added, usually to protect approaches
to junctions (turnouts), sharp curves between stations and
approaches to terminal stations where tracks end at a solid wall.
Automatic trains stops are in universal use. Speed control is a
more recent and less common application, often introduced in
conjunction with automatic train control or to meet specific
safety concerns.
Rail transit signaling is a very conservative field maintaining
high levels of safety based on brick-wall stops and fail-safe
principles. A brick-wall stop means that the signaling separation
protects a train even if it were to stop dead, an unlikely though
possible event should a train derail and strike a structure. This
protection allows for a) the following trains failure to observe a
stop signal, b) driver and equipment reaction time, and c) some
impairment in the braking rate.
Fail-safe design principles ensure that failure of singleand
often multiplecomponents should never allow an unsafe
event. Traditionally in North America this involves the use of
heavy railroad style relays that open by gravity and have
nonwelding carbon contacts. Compact, spring opening,
European-style relays or solid state (electronic or computer
controlled) interlockings are now being accepted. Here
equivalent safety is provided by additional logic, duplicate
contacts or multiple polling processors.
The rigor with which fail-safe principles have been applied to
rail transit has resulted in an exceptional safety record.
However, the safety principles do not protect against all
possibilitiesfor example, a derailed train could interfere with
the safe passage of a train on an adjacent parallel track. Nor do
they protect against all possible human errors whether caused by
a signal maintainer, dispatcher or train driver. An increasing
inability to control the human elementresponsible for threequarters of rail transit accidents or incidents1has resulted in
new train control systems using technology or automation to
reduce or remove the possibility of human error.
Train control, or more properly automatic train control, adds
further features to basic signaling. Automatic train control is an
ill-defined term but usually encompasses three levels:
1
PARKINSON, TOM, Safety Issues Associated with the Implementation of

ATCS-Type Systems, Transportation Development Centre, Transport
Canada, August 1989.
Automatic train protection is the basic separation of trains and

protection at interlockings. In other words, the signaling system
as described above.
Automatic train control adds speed control and often
automatic train operation. This can extend to automatically
driven trains but more commonly includes a driver, operator or
attendant who controls the train doors and observes the track
ahead.
Automatic train supervision attempts to regulate train service.
It can be an integral feature of automatic train control or an
addon system. The capabilities of automatic train supervision
vary widely from little more than a system that reports the
location of trains to a central control office, to an intelligent
system that automatically adjusts the performance and stop
times of trains to maintain either a timetable or an even headway
spacing.
Automatic train protection and automatic train control
maintain the fail-safe principles of signaling and are referred to
as vital or safety critical systems. Automatic train supervision
cannot override the safety features of these two systems, and so
it is not a vital system.
This chapter describes and compares the separation
capabilities of various train control systems used on or being
developed for rail transit. It is applicable to the main rail transit
grouping of electrically propelled, multiple-unit, gradeseparated systems. Specific details of train control for commuter
rail and light rail modes are contained in the chapters dealing
with these modes.
These descriptions cannot include all the complexities and
nuances of train control and signaling but are limited to their
effect on capacity. More details can be found in the references
and in the bibliography. All urban rail transit train control
systems are based on dividing the track into blocks and ensuring
that trains are separated by a suitable and safe number of blocks.
Train control systems are then broken down into fixed-block
and moving-block signaling systems.
3.2 FIXED-BLOCK SYSTEMS

In a fixed-block system, trains are detected by the wheels and
axles of a train shorting a low-voltage current inserted into the
rails. The rails are electrically divided into blocks. Originally
this required a rail to be cut and an insulating joint inserted.
Only one rail is so divided. The other rail remains continuous to
handle the traction power return.
2
Sometimes termed automatic train operation to avoid confusion with the

overall term automatic train control.
18
By moving from direct current to alternating current circuits,3
the blocks can be divided by an inductive shunt4 connected
across the rails, avoiding the need for insulated joints. These are
called jointless track circuits and both rails are then available for
traction power return. A track circuit can be any reasonable
length. Each circuit is expensive so lines use the minimum
required for appropriate headways. Circuits will be short where
trains must be close together, for example in a station approach,
and can be longer between stations where trains operate at speed.
The signaling system knows the position of a train only by the
relatively coarse measure of block occupancy. It does not know
the position of the train within the block; it may have only a
fraction of the train, front or rear, within the block. At block
boundaries, the train will occupy two blocks simultaneously for
a short time.
In the simplest two-aspect block system, the signals display
only stop (red) or go (green). A minimum of two empty blocks
must separate trains, and these blocks must be long enough for
the braking distance plus a safety distance. The safety distance
can include several components, including sighting distances,
driver and equipment reaction times, and an allowance for
partial brake failure, i.e. a lower braking rate.
Automatic train stops have long been a feature of rail transit
(almost from the turn of the century). These prevent a train
running through a red signal by automatically applying the
emergency brakes should the driver ignore a signal. Called a trip
stop, the system consists of a short mechanical arm beside the
outer running rail that is pneumatically or electrically raised
when the adjacent signal shows a stop aspect. If a train runs
through this signal, the raised arm strikes and actuates a trip
cock on the train that evacuates the main air brake pipe. Full
emergency braking is then applied along the length of the train.
To reset the trip cock the driver must usually climb down to
track side and manually close the air valve.5
A two-aspect signaling system does not provide the capacity
normally required on busy rail transit linesthose with trains an
3
Alternating current track circuits use different frequencies, combinations of

frequencies or modulated frequencies. In all cases care must be taken to
avoid interference from on-board vehicle equipment. Modern high power
chopper and VVVF (variable voltage, variable frequency) three phase ac
motor control equipment can emit considerable levels of EMI (electro
magnetic interference). The systems engineering to coordinate and avoid
such interference is difficult and complex and is beyond the scope of this
report.
4
In essence, the shunt shorts the small alternating current track circuits
while presenting a low resistance to the high direct currents.
5
Resetting the trip cock is understandably an unpopular task and consumes
time. Consequently drivers may approach a trip cock cautiously at less
than the optimal speed, particularly when closely following another train.
In this case they expect the signal aspect to change as they approach but
cannot be certain. Automatically driven trains will typically operate closer
to the optimal speeds and braking rates and so can increase throughput.
There are times when it is operationally desirable to operate through a
stop signal and its associated automatic train stop, particularly when the
train ahead is delayed in a station and following trains wish to close up to
expedite their subsequent entry to the station. The process is commonly
called key by from an arrangement where the driver must lean out of the
cab and insert a key in an adjacent electrical switch. However, the most
common arrangement no longer involves a key, merely a slow movement
of the train into the next block, which lowers the trip stop before it is
struck by the train. The train must then proceed on visual rules toward the
train ahead. In recent years an increase in the number of incidents caused
by this useful, time saving, but not fail-safe, procedure has caused several
systems to prohibit or restrict its use.
hour or better. Increased capacity can be obtained from multiple

aspects where intermediate signals advise the driver of the
condition of the signal ahead, so allowing a speed reduction
before approaching a stop signal. Block lengths can be reduced
relative to the lower speed, providing increased capacity.
The increased number of blocks, and their associated relay
controls and color-light signals, is expensive. There is a
diminishing capacity return from increasing the number of
blocks and aspects as shown in Figure 3.1. This figure also
shows that there is an optimal speed to maximize capacity.
Between stations the line capacity is greatest with maximum
running speeds of between 40 km/h (25 mph) with three aspects
to 55 km/h (34 mph) with 10 aspects. At the station entry
invariably the critical point for maximum throughputoptimal
approach speeds are from 25 km/h (15 mph) to 35 km/h (22
mph).
In North America, the most common block signaling
arrangement uses three aspects. In Europe and Japan, a small
number of systems extend to four or five aspects.
Optimizing a fixed-block system is a fine art, with respect
both to block lengths and to boundaries. Block lengths are also
influenced by grades because a trains braking distance
increases on a down grade and vice-versa. Grades down into a
station and curves or special work with significant speed
restrictions, below the optimal levels given above, will reduce
throughput and so reduce capacity. Fortuitously, one useful
design feature of below-grade systems is a gravity-assisted
profile. Here the stations are higher than the general level of the
running tunnel. Trains use gravity to reduce their braking
requirements in the station approach and to assist them
accelerating away from the stations. This not only reduces
energy consumption, equipment wear and tear and tunnel
heating, but also reduces station costs because they are closer to
the surface, allowing escalators and elevators to be shorter.
More important to this study, it increases train throughput
altogether a good thing.
Requiring a train operator to control a trains speed and
commence braking according to multiple aspect color-light
signaling requires considerable precision to maximize
throughput. Coupled with the expense of increasing the number
of aspects an improvement has been developed over the past
three decadescab signaling.
Figure 3.1 Throughput versus number of signal aspects(R26)
19
3.2.1 CAB SIGNALING

Cab signaling uses a.c. track circuits such that a code is inserted
into each circuit and detected by an antenna on each train. The
code specifies the maximum allowable speed for the block
occupied and may be termed the reference or authorized speed.
This speed is displayed in the drivers cabtypically on a dual
concentric speedometer, or a bar graph where the authorized
speed and actual speed can be seen together.
The authorized speed can change while a train is in a block as
the train ahead proceeds. Compared to color-light signals, the
driver can more easily adjust train speed close to the optimum
and has less concern about overrunning a trip stop. Problems
with signal visibility on curves and in inclement weather are
reduced or eliminated.
Cab signaling avoids much of the high capital and
maintenance costs of multiple-aspect color-light signals,
although it is prudent and usual to leave signals at interlockings
and occasionally on the final approach to and exit from each
station. In some situations, dwarf color-light signals can be used.
In this way trains or maintenance vehicles that are not equipped
with cab signalingor trains with defective cab signalingcan
continue to operate, albeit at reduced throughput.
Reducing the number of color-light signals makes it
economically feasible to increase the number of aspects and it is
typical, although not universal, to have the equivalent of five
aspects on a cab-signaling system. A typical selection of
reference speeds would be 80, 70, 50, 35 and 0 km/h (50, 43,
31, 22 and 0 mph).
Signal engineers may argue over the merits of block-signaling
and cab-signaling equipment from various manufacturers
particularly with respect to capital and maintenance costs,
modular designs, plug versus hard-wired connections and the
computer simulation available from each maker to optimize
system design. However, for a given specification, the
throughput capabilities vary little provided thatthe signaling is
optimized as to block length, boundary positioning and, when
applicable, the selection of reference speeds. Consequently a
listing or description of different systems is not relevant to
capacity determination.
3.3 MOVING-BLOCK
SIGNALING SYSTEMS
Moving-block signaling systems are also called transmissionbased or communication-based signaling systemspotentially
misleading because cab signaling is also transmission based.
A moving-block signaling system can be likened to a fixedblock system with very small blocks and a large number of
aspects. Several analytic approaches to moving-block systems
use this analogy. However a moving-block signaling system has
neither blocks nor aspects. The system is based on a continuous
or frequent calculation of the clear (safe) distance ahead of each
train and then relaying the appropriate speed, braking or
acceleration rate to each train.
This requires a continuous or frequent two-way
communication with each train, and a precise knowledge of a
trains location, speed and length; and fixed details of the line
curves, grades, interlockings and stations. These may be

contained in a table that allows changes to be made without the
normal full rigor required for changes to safety-critical software.
Temporary changes can be easily made to add speed restrictions
or close off a section of track for maintenance work.
Based on this information, a computer can calculate the next
stopping point of each trainoften referred to as the target
pointand command the train to brake, accelerate or coast
accordingly. The target point will be based on the normal
braking distance for that train plus a safety distance.
Safety Distance Braking distance is a readily determined or

calculated figure for any system. The safety distance is less
tangible because it includes a calculated component adjusted by
agency policy. In certain systems this distance is fixed;
however, the maximum throughput is obtained by varying the
safety distance with speed and locationand, where different
types of equipment are operated, by equipment type.
In theory, the safety distance is the maximum distance a train
can travel after it has failed to act on a brake command before
automatic override (or overspeed) systems implement
emergency braking. Factors in this calculation include
system reaction time;

brake actuation time;
speed;
train load (mass)including any ice and snow load;
grade;
maximum tail winds (if applicable);
emergency braking rate;
normal braking rate;
train to track adhesion; and
an allowance for partial failure of the braking system.
The safety distance is frequently referred to as the worst-case

braking distance, but this terminology is misleading. The truly
worst case would be a total braking failure. Worst case implies
reasonable failure situations, and total brake failure is not
regarded as a realistic scenario on modern rail transit equipment
that has multiple braking systems. A typical interpretation of the
safety distance assumes that the braking system is three-quarters
effective.
Train Position and Communication Without track circuits to

determine block occupancy, a moving-block signaling system
must have an independent method to accurately locate the
position of the front of a train, then use look-up tables to
calculate its end position from the length associated with that
particular trains identification. The first moving-block systems,
developed in Germany, France and the United States, all used
the same principlea wire laid alongside or between the
running rails periodically transposed from side-to-side, the
zigzag or Grecian square arrangement. The wire also serves to
transmit signals to and from antennas on the train.
The wayside wires are arranged in loops so that each train
entering a loop has a precise position. Within the loop, the control
system counts the number of transpositions traversed, each a
20
fixed distance apart m (82 ft) is typical although much shorter
distances have been used. Between the transpositions, distance
is measured with a tachometer.6
The resultant positioning accuracy can be in the order of
centimeters and with frequent braking rate feedback can result
in station stop accuracy within 20 cm (8 in.) or better.
The use of exposed wayside wires is abhorred by
maintenance-of-way engineers, and recent developments
portend changes to existing systems and for the many movingblock signaling systems now under development. Inert
transponders can be located periodically along the track. These
require neither power nor communication wiring. They are
interrogated by a radio signal from each train and return a
discrete location code. Positioning between transponders again
relies on the use of a tachometer. Moving-block signaling
systems already have significantly lower costs for wayside
equipment than do fixed-block systems, and this arrangement
further reduces this cost as well as the occupancy time required
to install or retrofit the equipmentan often critical factor in
resignaling existing systems.
Removing the positioning and communicating wire from the
wayside requires an alternate communication system. This can
most economically be provided by a radio system using overthe-air transmission, wayside radiating cables, intermittent
beacons or a combination thereof.
As with any radio system, interruption or interference with
communications can occur and must be accommodated. After
the central control computer has determined any control action,
it will transmit instructions to a specific train using the
identification number of the trains communication system. It is
clearly vital that these instructions are received by and only by
the train they were determined for.
There are numerous protocols and/or procedures that provide
a high level of security on communication systems. The data
transmission can contain both destination codes and error codes.
A transmission can be received and repeated back to the source
to verify both correct reception and correct destination, a similar
process to radio train order dispatching. If a train does not
receive a correctly coded confirmation or command within a set
time, the emergency brakes will be automatically applied. The
distance a train may travel in this time intervaltypically less
than 3 secis a factor in the safety distance.
Data Processing The computers that calculate and control a

moving-block signaling system can be located on each train, at a
central control office, dispersed along the wayside or a
combination of these. The most common arrangement is a
combination of on-board and central control office locations.
The first moving-block signaling systems used mainframe
computers with a complex interconnection system that provided
high levels of reliability. There is now a move toward the use of
much less expensive and space-consuming personal computers
(PCs).
6
Tachometer accuracy is helped by the ability for continual on-the-fly

calibrations as the distance between each transposition is fixed and
known. This fully compensates for wheel wear but not for slip or slide.
Errors so caused, while small, can be minimized by the use of current
sophisticated slip-slide control or, where feasible, placing the tachometer
on an unmotored axle.
PCs and their local area networks (LANs) have been regarded
as less robust than mainframe systems, and as suspect for use in
safety-critical applications. The first major application occurred
in Vancouver in 1994 when, after 10 years of mainframe
operation, the entire SkyTrain train control system was changed
to operating on PCs with Intel 486 CPUs. Reliability has
increased in the subsequent 15 months of operation. However, it
is not possible to attribute this improvement solely to the new
hardware because new software was also required by the change
in operating systems. The proprietary computers and software
on each train were not changed.
Safety Issues Safety on rail transit is a relative matter. It

encompasses all aspects of design, maintenance and operations.
In fixed-block signaling, electrical interlockings, switch and
signal setting are controlled by relay logic. A rigorous discipline
has been built around this long established technology which the
use of processor-based controls is now infiltrating.
A moving-block signaling system is inherently processor
controlled. Processor-based train control systems intrinsically
cannot meet the fail-safe conventions of traditional signaling.
Computers, microprocessors and solid-state components have
multiple failure opportunities and cannot be analyzed and tested
in the same way as conventional equipment.
Instead, an equivalent level of safety is provided on the basis
of statistical failure modes of the equipment. Failure analysis is
not an exact science. Although not all failure modes can be
determined, the statistical probability of an unsafe event7 can be
predicted.
Determining failure probability is part of a safety assurance
plana systematic and integrated series of performance,
verification, audit, and review activities, including operations,
maintenance and management activities that are implemented to
assure safe and satisfactory performance. The plan can cover a
specific area, such as software, or can encompass the entire
system, where software would be but one aspect. Such a plan
will usually include a fault tree analysis.
The typical goal in designing processor-based systems is a
mean time between unsafe failures of 109 hours, or some
114,000 years.8 After due allowance for statistical errors and the
incorporation of a large safety margin, this is deemed to be
equivalent to or better than the so-called fail-safe conventional
equipment.
The possibility of even a low incidence of unsafe failure may
give cause for concern and the acceptance of processor-based
signaling, particularly moving-block systems, has been slow.
However the safety of conventional rail transit signaling is not
as absolute as is often made out. Minor maintenance errors can
cause unsafe events. An estimated three-quarters of rail transit
accidents are attributed to human error.9
Two methods are used to achieve the high levels of safety on
processor-based control systems. One is based on redundancy,
where two or more computers operate with the same software.
The output of both or the output of at least two out of three
7
An unsafe event may be referred to as a wrong-side failure.

PARKINSON, TOM, Safety Issues Associated with the Implementation of
ATCS-Type Systems, Transportation Development Centre, Transport
Canada August 1989.
9
Ibid.
8
21
must coincide before a comparator circuit transmits a command.
Thereafter, the safety consequences of the output can be
considered in a conventional fashion. This method is a
hardware-intensive solution.
The other method is based on diversity. Two sets of software,
created and verified by independent teams, are run on the same
or separate computers. Again their output must agree before any
commands are executed. This is a software-intensive solution.
Because software development can account for over half the
cost of a moving-block signaling system, and with hardware
costs decliningparticularly with the use of PCsthe
hardware-intensive approach to redundancy is invariably the
most economic. However, the relative cost of software
development, testing, commissioning and safety assessment is
expected to drop with the introduction of modular code
blockssafety critical portions of software that remain
unchanged from system to system.
In some regards, software-based systems, once fully tested
and commissioned, are less prone to unsafe errors created during
equipment installation and maintenance. However there are
three major remaining areas of concern.
1. Revisions to software may be required from time to time
and can escape the full rigor of a safety assurance plan.
2. Removing track circuits also removes broken rail
detection. While no specific data for rail transit have been
found, the Southern Pacific Railroad found that fewer than
2 percent10 of broken rails were detected in advance by
track circuitsit appears that most breaks occur from the
stress of a train passing. Nevertheless, some moving-block
signaling systems have long track circuits added to detect
broken rails.
3. Removing track circuits also eliminates the detection of
any and all vehicles whose wheels and axles short across
the rails. A major hazard exists if maintenance vehicles, or
a train with a defective train control system, enter into or
remain in an area where automatically controlled trains are
run. This requires a rigorous application of operating rules
and requires the defect correction and reentry into the
control system or removal of an automatic train protection
failed train, before service can resume in the occupied
area.
This potential hazard can be reduced by adding axle
counters at various locations. These count entry and exit
into a specified track section. In conjunction with
appropriate software, they will prevent an automated train
from following an unequipped train at an unsafe distance.
However, an unequipped train is not so protected but
depends on the driver obeying rules, whether using line-ofsight operation, or depending on any remaining wayside
signals.
Hybrid Systems There are times when an urban rail transit

system shares tracks with other services, such as long distance
trains, whose equipment is impractical or uneconomic to equip
with the moving-block signaling system. Use of axle counters
for the safety of unequipped rolling stock substantially reduces
10
capacity. To avoid this reduction while still obtaining the close

headway of the moving-block system for the urban or short
distance trains requires a hybrid design.
The SACEM system developed by Matra is employed in Paris
and Mexico City11 The SACEM combines a fixed-block system
with a transmission based system. Conventional blocks are
subdivided into smaller increments that permit those trains,
equipped with a continuous communication system, to operate
on closer headways. Unequipped trains continue to be protected
by the basic block system. As equipped trains operate through
some signals displaying red an additional aspect must be added
to such signalsindicating that the signal is not applicable to
that specific train.
SACEM has a throughput capability between fixed-block and
moving-block signaling systems that depends on the mix of
equipped and unequipped trains. The manufacturer claims an
increase in capacity up to 25%, which is comparable to the
general 30% increase of moving-block over fixed-block
signaling systemsall else being equal. The two equipped rail
transit lines in Mexico City do not have any unequipped long
distance trains with their longer braking distances and so should
obtain the maximum capacity improvement.
While classed as a hybrid system, SACEM does not use
moving-blocks and is really an overlay system. Shorter blocks
applicable to certain trains onlyare overlaid onto a
conventional fixed-block system.
Moving-block signaling systems have been installed by the SEL
(Standard Electrik Lorenz) of Stuttgart, Germany, and its
Canadian subsidiary SEL Canada. Both are now part of the
Alcatel group, a French consortium.
The Alcatel SelTracTM system has evolved through five
generations over two decades. There are some 20 worldwide
installations of which five are in North America: Vancouver,
Toronto, Detroit, San Francisco and Orlando (Disneyworld
monorail).
The SelTrac system uses an inductive loop to both
communicate with trains and, through the loop transpositions, to
determine positioning. Processing power is centralized with the
on-board computers limited to processing signals and
controlling the vehicle subsystems. The use of Intel x86
processors to control critical train movements was introduced in
1994. Transponder positioning has been developed to reduce
hardware costs and improve failure management. In addition,
SelTrac includes an integrated automatic train supervision
subsystem.
The second manufacturer with a system in service is also
French. Service started on Line D of the Lyon metro in 1992
using Matra Transports MaggalyTM system. The Maggaly
system uses inductive transmission with positioning
transponders and places the bulk of the processing power onboard. Line data are stored on-board with the wayside
equipment limited to system management and providing the
location of a leading train to its immediate follower.
The advantages of moving-block signaling systems are

considerable. Beyond the capacity increase of interest to this
study, the concept offers the potential for lower capital and
maintenance costs, flexibility, comprehensive system management capabilities and inherent bi-directional operation. The
11
Ibid.
Line A and Line 8.
22
slow acceptance of processor based train control systems may
explain why most conventional train control suppliers have
stayed away from this concept until the recent selection of
moving-block systems by London Transport and New York City
Transit, together with several smaller systems. This selection is
not necessarily based on the capacity increases but as much on
the economics and relative ease of installing the system on top
of a conventional signaling system on existing lines that must
remain in operation throughout the conversion, modernization or
replacement.
Subsequent to the London and New York decisions, many
manufacturers have announced the development of movingblock signaling systems.
General Railway Signal is developing its ATLASTM system.
This is a modular based concept that allows various forms of
vehicle location and communication systems. A feature is a vital
stored database and low requirements for the vehicle-wayside
data communication flow.
Union Switch & Signal is developing its MicroBlokTM
which shares some similarity with Matras SACEM, overlaying
virtual software based blocks on a conventional fixed block
system. With radio based communications and vital logic
distributed on the wayside, the system uses some concepts
developed for the Los Angeles Green Line which entered
service in August 1995.
AEG Transportation Systems FlexiblokTM shares some
features with MicroBlok and SACEM. It is a radio-based system
designed for both standalone use and for incrementally adding
capacity and features to traditional train control systems.
Operational and safety responsibilities are distributed through
the system, which incorporates nonproprietary interfaces
conforming to Open System Interconnect protocol standards.12
AEGs US division, previously Westinghouse Electric
Transportation Systems, is developing a transmission-based
train control system tailored to the North American market.
Harmon Industries UltraBlockTM system is radio based
with transponder positioning technology. Line profile
information is stored on-board. Vital processing is distributed
along the wayside.
Siemens Transportation Systems is developing a movingblock system based on its Dortmund University people mover,
an under-hanging cabin system that has been in service since
1984.
CMW (Odebretch Group, Brazil) is supplying a radio-based
overlay system to the So Paulo metro with distributed
processing. The system is claimed to reduce headways from 90
to 66 sec. As section 4.7 of this chapter shows, such close
headways are only possible with tightly controlled station dwells
which are rarely achievable at heavy volume stations.
Morrison Knudsen (with Hughes and BART) is developing a
moving-block signaling system based
on
military
communication technology. The system uses beacon-based,
ranging spread spectrum, radio communications which are less
susceptible to interference and can tolerate the failure or loss of
one or more beacons.
12
The proprietary nature of many moving-block signaling systems is a

concern to potential customers who are then captive to a particular
supplier. Traditional train control systems in theory allow many
components from different manufacturers to be mixed and matched.
However, particularly with the introduction of solid state interlockings,
this is not always feasible.
NOTE: The above discussion represents the best information

available to the researchers at the time this report was written.
Other suppliers may exist and omissions were inadvertent. This
discussion is not intended to endorse specific products or
manufacturers.
All moving-block systems that base train separation
on a continually adjusted distance to the next stop or
train ahead (plus a safety distance) should have
substantially similar train throughput capabilities.
Capacity for a generic moving-block signaling system
is developed in section 3.8 of this chapter, based on
information from existing systems (Alcatel and
Matra).
Those systems under development (above) that
succeed in the market can reasonably be expected to
have comparable capacities. However, there is
insufficient information to confirm this.
3.4 AUTOMATIC TRAIN

OPERATION
Automatic acceleration has long been a feature of rail transit. A
driver no longer has to cautiously advance the control handle
from notch to notch to avoid pulling too much current and so
tripping the line breaker. Rather, relays, and more recently
micro-processors, control the rate of acceleration smoothly from
the initial start to maximum speed.
Cab signaling and moving-block signaling systems transfer
speed commands to the train and it was a modest step to link
these to the automatic acceleration features, and comparable
controlled braking, to create full automatic train operation
(ATO). The first North America application occurred in 1962 on
NYCTAs Times Square Shuttle, followed in 1967 by
Montreals Expo Express, then, in short order by PATCOs
Lindenwold line and San Franciscos BART. Most new rail
transit systems have incorporated ATO since this innovative
period.
The drivers or attendants role is not necessarily limited to
closing the doors, pressing a train start button and observing the
line ahead. Drivers are usually trained in, and rolling stock is
provided with, manual operating capabilities. PATCO pioneered
the concept of having drivers take over manual control from
time to time to retain familiarity with operations. Manual
driving under cab controls, limited color-light signaling or radio
dispatching is routine, if infrequent, on many ATO-equipped
systems when there is a train control failure or to provide
signaling maintenance time.
Dispensing entirely with a driver or attendant is controversial.
In 1965 the driverless Transit Expressway was first operated in a
controlled environment in Pittsburgh. This Automated Guideway
Transit (AGT) system, and similar designs, have gained widespread acceptance in nontransit usage as driverless people
movers in airports, amusement parks and institutional settings.
Morgantowns AGT was the first public transit operation to gain
acceptance for driverless operation when it opened in 1968. After
a long gap Miamis downtown people mover opened in 1985
with the Detroit People Mover and the full-scale urban rail transit
23
SkyTrain system in Vancouver starting the following year.
Driverless public transport is now well established in these cities
but no subsequent operations have chosen to follow, despite
their record of safety, reliability and lower operating costs.
Fundamental concerns with driverless automatic train operation
clearly remain.
Automatic train operation, with or without attendants or
drivers, allows a train to more closely follow the optimum speed
envelope and commence braking for the final station approach
at the last possible moment. This reduces station to station travel
times, and more important from the point of capacity, it
minimizes the critical station close-in timethe time from when
one train starts to leave a station until the following train is
berthed in that station.
In the literature Klopotov(R32) makes claims of capacity
improvements of up to 15% with ATO. Bardaji(R10) claims a 5%
capacity increase with automatic regulation. Other reports allude
to increases without specific figures. None of the reports
substantiate any claims. Attempts to quantify time
improvements between manual and automatic driving for this
study were unsuccessful. Any differences were overshadowed
by other variations between systems.
Intuitively there should be an improvement in the order of 5
to 10% in the station approach time. As this time represents
approximately 40% of station headway, the increase in capacity
should be from 2 to 4%.
The calculations used to determine the minimum station
headway assume optimal driving but insert a time for a drivers
sighting and reaction timein addition to the equipment
reaction time. The calculations in this report compensate for
ATO by removing the reaction times associated with manual
driving.
3.5 AUTOMATIC TRAIN

SUPERVISION
Automatic Train Supervision (ATS) encompasses a wide variety
of options. It is generally not a safety-critical aspect of the train
control system and may not need the rigor of design and testing
to its hardware and software that characterizes other areas of
train control. At its simplest it does little more than display the
location of trains on a mimic board or video screen in the central
control or dispatchers office.
One step up in sophistication provides an indication of ontime performance with varying degrees of lateness designated
for each train, possibly grouped by a color code or with a digital
display of the time a train is behind schedule. In either case
corrective action is in the hands of the variously named
controller, dispatcher or trainmaster.
Urban rail transit in North America is generally run to a
timetable. Those systems in Europe that consistently operate at
the closest headways (down to 90 sec) generally use headway
regulation that attempts to ensure even spacing of trains rather
than adhere strictly to a timetable. Although it appears that
keeping even headways reliably provides more capacity, this is
an issue of tradition, operating rules and safety13 that is beyond

the scope of this study.
In more advanced systems where there is ATO, computer
algorithms are used to attempt to automatically correct lateness.
These are rare in North America and are generally associated
with the newer moving-block signaling systems.
Corrective action can include eliminating coasting, increasing
line speed, moving to higher rates of acceleration and braking
and adjusting dwell timesusually only where these are
preprogrammed. Such corrective action supposes that the system
does not normally work flat out.
The Vancouver system is an example of unusually
comprehensive ATS strategies. Here trains have a normal
maximum line speed of 80 km/h (50 mph) which ATS can
increase to 90 km/h as a catch up measurewhere civil speed
restrictions so permit. Similarly acceleration and braking can be
adjusted upwards1414 or downwards by 10%.
In normal operation trains use less than their full performance
which reduces energy consumption and maintenance, and leaves
a small leeway for on-time corrective action. Together, these
strategies can pick up 2 to 3 min in an hour.
Correcting greater degrees of lateness or irregularity generally
involves manual intervention using short turn strategies or
removing slow-performing or defective trains from service.15
This is difficult to implement in the peak period and common
practice is to let the service run as best it can and wait to make
corrections to the timetable until after the peak period.
A further level of ATS strategies is possiblepredictive
control. Although discussed as a possibility, this level is not
known to be used in North America. In predictive control a
computer looks ahead to possible conflicts, for example a merge
of two branches at a junction. The computer can then adjust
terminal departures, dwell times and train performance to ensure
that trains merge evenly without holds, or are appropriately
spaced to optimize turn-arounds at any common terminal.
The nonvital ATS system can also be the host for other
features such as on-board system diagnostics and the control of
station and on-board information through visual and audio
messagesincluding those required by ADA.
Summary ATS has the potential to improve service regularity

and so help maximize capacity. However, the strategies to correct
irregular service on rail transit are limited unless there is close
integration with ATO and the possibilities of adjusting train
performance and station dwells. Without such strategies, ATS
allows dispatchers to see problems but remain unable to address
them until the peak period is over. In Chapter Six, Operating
13
Certain Russian systems that maintain remarkably even 90-sec headways

require drivers to close doors and depart even if passenger flow is
incomplete.
14
A trains performance is limited by motor heating characteristics.
Corrective actions that increase performance also increase heating.
Depending on ambient temperature this can only be carried out for a
limited period before the trains diagnostic equipment will detect overheating and either cut one or more motors out or force a drop to a lower
performance rate.
15
One North American system is known to use a skip-stop strategy for
seriously late trains, that is running through a station where the train
would normally stop. Akin to the bus corrective strategy of set downs
only, no pick-ups, this is both unusual and can be difficult for passengers
to accept.
24
Issues, an operational allowance to compensate for irregular
operation is developed. A sophisticated ATS system in
conjunction with a range of feasible corrective actions can
reduce the desired amount of operating margin time.
3.6 FIXED-BLOCK
THROUGHPUT
Determining the throughput of any rail transit train control
system relies on the repetitive nature of rail transit operation. In
normal operation trains follow each other at regular intervals
traveling at the same speed over the same section of track.
All modern trains have very comparable performance. All
low-performance equipment in North America is believed to
have been retired. Should a line operate with equipment with
different performance and/or trains of different length, then the
maximum throughput rates developed in this section should be
based on the longest train of the lowest performing rolling stock.
Trains operating on an open line with signaling protection but
without station stops have a high throughput. This throughput is
defined as line or way capacity. This capacity will be calculated
later in this section although it has little relevance to achievable
capacity except for systems with off-line stations. Only
Automated Guideway Transit, or some very high capacity lines
in Japan, can support off-line stations.
Stations are the principal limitation on the maximum train
throughputand
hence
maximum
capacityalthough
limitations may also be due to turn-back and junction
constraints. The project survey of operating agencies indicated
that the station close-in plus dwell time was the capacity
limitation in 79% of cases, turnback constraints in 15%, and
junctions in 5% of cases. Further inquiry found that several
turnback and junction constraints were self-imposed due to
operating practices and that stations were by far the dominant
limitation on throughput.
In a well-designed and operated system, junction or turnback
constrictions or bottlenecks should not occur. A flat junction can
theoretically handle trains with a consolidated headway
approaching 2 min. However, delays may occur and systems
designed for such close headways will invariably incorporate
grade-separated (flying) junctions. Moving-block signaling
systems provide even greater throughput at flat junctions as
discussed in section 3.10.
A two-track terminal station with either a forward or rear
scissors cross-over can also support headways below 2 min
unless the cross-overs are long, spaced away from the terminal
platform, or heavy passenger movements or operating practices
when the train crew changes ends (reverses the train) result in
long dwells. The latter two problems can be resolved by
multiple-platform terminal stations, such as PATHs Manhattan
and Hoboken terminals and Mexico Citys Indios Verdes
station, or by establishing set-back procedures for train crews. 16
16
Set back procedures require the train crew or operator to leave the train at
a terminal and walk to the end of the platform where they board the next
entering train which can be immediately checked and made ready for
departure. On a system with typical close headways of two minutes this
requires an extra crew every 30 trains and increases crewing costs by
some 3%less if only needed in peak periods. The practice is unpopular
with staff as they must carry their possessions with them and cannot enjoy
settling into a single location for the duration of their shift.
In this chapter the limitations on headway will be calculated

for all three possible bottlenecks: station stops, junctions and
turnbacks.
Nine reports in the literature survey provide detailed methods
to calculate the throughput of fixed-block rail transit signaling
systems:
AUER, J.H., Rail-Transit People-Mover Headway
Comparison(R9)
BARWELL, F. T., Automation and Control in
Transport(R11)
BERGMANN, DIETRICH R., Generalized Expressions
for the Minimum Time Interval between Consecutive
Arrivals at an Idealized Railway Station(R13)
DELAWARE RIVER PORT AUTHORITY, 90 Seconds
Headway Feasibility Study, Lindenwold Line(R21)
GILL, D.C., and GOODMAN C.J., Computer-based
optimisation techniques for mass transit railway signalling
design(R26)
JANELLE, A., POLIS, M.P., Interactive Hybrid Computer
Design of a Signaling System for a Metro Network(R31)
LANG, A SCHEFFER, and SOBERMAN, RICHARD M.,
Urban Rail Transit Its Economics and Technology(R39)
VUCHIC, VUKAN R., Urban Public Transportation
Systems and Technology(R71)
WEISS, DAVID M., and FIALKOFF, DAVID R.,
Analytic Approach to Railway Signal Block Design(R73)
All the reports deal with station stops as the principal limitations
on capacity and use Newtons equations of motion to calculate
the minimum train separation, adding a variety of nuances to
accommodate safety distances, jerk limitations, braking system
and drivers reaction times plus any operating allowance or
recovery margin. In the following section a classical approach is
examined, followed by a recommended practical approach
derived from the work of Auer(R09) in combination with
information from several other authors. Then an examination is
made of the sensitivity of the results to several system variables.
3.6.1 STATION CLOSE-IN TIME

The time between a train pulling out of a station and the next
train enteringreferred to as close-inis the main constraining
factor on rail transit lines. This time is primarily a function of
the train control system, train length, approach speed and
vehicle performance. Close-in time, when added to the dwell
time and an operating margin, determines the minimum possible
headway achievable without regular schedule adherence
impactsreferred to as the noninterference headway.
When interference occurs, trains may be held at approaches to
stations and interlockings. This requires the train to start from
stop and so increases the close-in time, or time to traverse and
clear an interlocking, reducing the throughput. With throughput
decreased and headways becoming erratic, the number of
passengers accumulated at a specific station will increase and so
increase the dwell time. This is a classic example of the maxim
that when things go wrong they get worse.
25
The minimum headway is composed of three components:
the safe separation (close-in time),
the dwell time in the station, and
an operating margin.
The block length must be greater than or equal to the service

stopping distance.18
Equation 3-2
Station dwells are discussed in Chapter Four, Station Dwells,

recovery margins are discussed in Chapter Six, Operating
Issues.
where
3.6.2 COMPUTER SIMULATION

The best method to determine the close-in time is from the
specifications of the system being considered17, from existing
experience of operating at or close to capacity or from a
simulation. It is common in designing and specifying new rail
transit systems, or modernizing existing systems, to run a
variety of computer simulation models. These models are used
to determine running times, to optimize the design of track
work, of signaling systems and of the power supply system.
Where the results of these models are available they can provide
an accurate indication of the critical headway limitation
whether a station close-in maneuver, at a junction or at a
turnback.
Such models can be calibrated to produce accurate results. In
particular, many simulation models will adjust train
performance for voltage fluctuations in the power supplya
variant that cannot be otherwise be easily calculated. However
caution should be exercised in using the output from
simulations. Simulations can be subject to poor design, poor
execution or erroneous data entry. In particular, increments of
analysis are important. The model will calculate the voltage,
performance, movement and position of the front and rear of
each train in small increments of time, and occasionally in
increments of distance or speed. Such increments should
approach one tenth of a second to produce accurate close-in
times.
Simulation programs are also often proprietary to a specific
consultant or train control, traction substation or vehicle
supplier. They require considerable detailed site and equipment
data. As such, they may not be practical or available for
determining achievable capacity, making it necessary to
calculate the throughput of the particular train control system by
more general methods.
If the minimum headway is not available from the system
designers or from a simulation, then straightforward methods
are available to calculate the time. Here train separation is based
on a line clear basissuccessive green signals governing the
following train. The minimum line headway is determined by
the critical line condition, such as the close-in at the maximum
load point station plus an operating margin. The entire stretch of
line between junctions and turnbacks, where train density is
physically constant, is controlled by this one critical time.
The classical expression for the minimum headway of the
typical rail transit three-aspect block-signal system is
The train control design engineers will be aiming to minimize the close-in
time and information from this source, particularly if the result of an
accurate simulation, is invariably the most accurate way to determine
practical capacity.
=
=
=
=
=
=
=
d
M
=
=
headway in seconds
block length approaching station (m)
station dwell time in seconds
service stopping distance (m)
length of the longest train (m)
maximum approach speed (m/s)
average acceleration rate through the
station platform clear-out (m/s2)
braking rate (m/s2)
headway adjustment combining operational
tolerance and dwell time variance
(constant)
Although the headway adjustment factor, M, can encompass a

variety of items, it is difficult to encompass all the variables that
can affect headway. These include
any distance between the front of the train and the start of
the station exit block,19 particularly if the train is not
berthed at the end of the platform;
control system reaction time;
on manually driven trains, the train operator sighting and
reaction time;
the brake system reaction time;20
an allowance for jerk limitation;21
speed restrictions on station approaches and exits whether
due to speed control for special work or curves; and
grades approaching and leaving a station.
In addition, the length of the approach block and the approach
speed are not readily obtainable quantities. Consequently this
traditional method is not recommended and an alternate
approach will be developed, based, in part, on the work of Auer.
This uses more readily available data accommodating many of
the above variables. This approach encompasses both manually
and automatically driven trains, multiple command cab controls,
and, by decreasing block length, a moving-block system.
Even so, it should be borne in mind that not all variables can
be included, and assumptions and approximations are still
needed. This approach, while more comprehensive than many in
the literature, is not as good as using information from signaling
18
19
20
Equation 3-1
17
H(t)
BL
Dw
SD
L
vap
a
21
On close headway systems block lengths may be less than the service
stopping distance. New York has approach blocks down to 60m (200)
and lengths as short as 15m (50) occur on some systemsparticularly
automated guideway transit systems.
This allows for blocks that do not start at the end of the platformat the
headwallor shorter trains that are berthed away from the headwall.
Older equipment may have air brakes applied by releasing air from a
brake control pipe running the length of the train (train-lined). There is a
considerable delay as this command passes down the train and brakes are
applied sequentially on cars. Newer equipment uses electrical commands
to control the air, hydraulic or electric brakes on each car and response is
more rapid.
Limitations applied to the start and end of braking and the start of
acceleration to limit the rate of change of accelerationcommonly, if
somewhat erroneously called jerk.
26
engineers, based on actual block positions, or from a
comprehensive and well-calibrated simulation.
Chapter Six, Operating Issues, and added into the headway

calculation by mode in Chapters Seven through Ten.
Substituting for Smin and removing Som produces
3.6.3 CALCULATING LINE HEADWAY

On a level, tangent (straight) section of track with no
disturbances the line headway H(l) is given by:
Equation 3-3
where
H(l)
Smin
L
vl
=
=
=
=
line headway in seconds

minimum train separation in meters
length of the longest train in meters
line speed in m/s22
The minimum train separation corresponds to the sum of the

operating margin and safe separation distance shown in Figure
3.2. It can therefore be further subdivided: (all in meters)
Smin = Ssbd + Std + Som
where
Smin
Ssbd
Std
Som
=
=
=
=
Equation 3-5
There are several components in the safe braking time. The
largest is the time to brake to a stop, using the service brake. A
constant K is added to assume less than full braking efficiency
or reduced adhesion75% of the normal braking is an
appropriate factor. There is also the distance covered during
driver sighting and reaction time on manually driven trains, and
on automatically driven trains brake equipment reaction time
and a safety allowance for control failure. This overspeed
allowance assumes a worst case situation whereby the failure
occurs as the braking command is issued with the train in full
acceleration mode. This is often termed runaway propulsion.
The train continues to accelerate for a period of time tos until a
speed governor detects the overspeed and applies the brakes.25
Equation 3-4
mininimum train separation distance

safe braking distance
train detection uncertainty distance
operating margin distance23
The safe braking distance is based on the rail transit assumption

of brick-wall stops using a degraded service braking rate.24 The
train detection uncertainty reflects either the block length or the
distance covered in the polling time increments of a
movingblock signaling system. The operating margin distance is
the distance covered in this time allowance. This will be omitted
from further consideration in this section. It is developed in
Equation 3-6
where
Sbd
Sbd
K
Sbr
Sos
=
=
=
=
safe breaking distance in meters

service braking distance in meters
braking safety factor
train operator sighting and reaction distance
and/or braking system reaction distance in
meters
= overspeed travel distance in meters
The distance to a full stop from speed Vl at the constant

service braking, deceleration or retardation rate is given by:
Equation 3-7
where
ds = service deceleration rate in m/s2
To be rigorous, the safe braking distance should also take into

account grades, train loadpassenger quantities and any snow
and ice load and, in open line sections, any tail wind. These add
complexities beyond the scope of this study and, except for
downgrades, contribute a very minor increment to the result.
Consequently they have been omitted. The effect of grades will
be examined in the sensitivity analysis at the end of this section.
Modern rail transit equipment uses a combination of friction
and electrical braking,26 in combination with slip-slide controls,
to maintain an even braking rate. An allowance can be added for
the jerk limiting features that taper the braking rate at the
beginning and end of the brake application.
25
26
Figure 3.2 Distance-time plot of two consecutive trains

(acceleration and braking curves omitted for clarity)
22
Can be worked in feet with speed in feet per second. 10 mph=14.67 ft/sec,
10 km/h = 2.78 m/s
23
Auer used the term service control buffer distance.
24
Some workers use the emergency braking rate. As this is highly variable
depending on location, equipment, and wheel to rail adhesion, it is not
recommended.
As the braking so applied is usually at the emergency rate, a case can be

made that this component may be discounted or reduced.
Electrical braking is both dynamicwith recovered energy burned by
resistors on each car, or regenerative braking with recovered energy fed
back into the linehere it feeds the hotel load of the braking train,
adjacent trains, is fed back to the power utility via bi-directional
substations or is burned by resistors in the substation. The latter two
modes are rare. Regenerative braking was common in the early days of
electric traction. It then fell out of use when the low cost of electricity
failed to justify the additional equipment costs and maintenance. With
increased energy costs and the ease of accommodating regeneration on
modern electronic power conversion units, regeneration is now becoming
a standard feature. Regeneration is sometimes termed recuperation.
27
The distance an automatically operated train moves until the
overspeed governor operates can be expressed as
Equation 3-8
where
Sos
ts
al
vl
=
=
=
=
between trains.27 The value of B for moving-block signaling

systems can be equal to or less than unity and is developed in
the next section.
Accepting these approximations and substituting Equations 310 and 3-11 in Equation 3-9 produces
overspeed distance
time for overspeed governor to operate
line acceleration rate in m/s2at vl
line speed
Substituting Equations 3-6, 3-7, and 3-8 in Equation 3-5 and

adding a jerk limiting allowance produces
Equation 3-12
where
H(l)
L
vl
K
=
=
=
=
tos
tjl
tbr
al
ds
=
=
=
=
=
Equation 3-9
where
tbr = train operator sighting and reaction time

and/or braking system reaction time
tjl = jerk limiting time allowance
Service acceleration is said to be following the motor curve as it

reduces from the initial controlled rate to zero at the top,
maximum, or balancing speed of the equipment. The
acceleration rate at a specific speed may not be readily available
and an approximation is appropriate for this itema small
component of the total line headway time. On equipment with a
balancing speed of 80 km/h, the initial acceleration is
maintained until speeds reach 10-20 km/h then tapers off,
approximately linearly until speeds of 50-60 km/h, then
approximately exponentially until it is zero. At line speeds
appropriate to this analysis the line acceleration rate can be
assumed to be approximate to the inverse of speed so that for
intermediate speeds
Equation 3-10
where
vl
vmax
al
as
= line speed in m/s

= maximum train speed in m/s
= line acceleration rate in m/s2
= initial service acceleration rate in m/s2
The train detection uncertainty distance is not readily available

but can be approximated as either the block length(s)again not
easily obtainedor the braking distance plus some leeway as a
surrogate for block lengths on a system designed for maximum
throughput. This quantity is particularly useful as a simple
method to adjust for the differences between the traditional
three-aspect signaling system, cab controls with multiple aspects
(command speeds) and moving-block signaling systems.
line headway in seconds

length of the longest train in meters
line speed in m/s
braking safety factorworst case service
braking is K% of specified normal rate
typically 75%
separation safety factorequivalent to the
number of braking distances (surrogate for
blocks) that separate trains
overspeed governor operating time28 (s)
time lost to braking jerk limitation (s)
operator & brake system reaction time (s)
line acceleration rate in m/s2
service deceleration rate in m/s2
North American rail transit traction equipment tends to have

very similar performance derived from the work of the
Presidents Conference Committee (PCC) in the mid 1930s. The
chief engineer, Hirschfeld,29 placed subjects on a moving
platform and determined the acceleration rate at which they lost
their balance or became uncomfortable. A wide variety of
subjects were tested including people who were pregnant,
inebriated or holding packages. From this pioneering work, the
PCC streetcar evolved and with it rates of acceleration and
deceleration (and associated jerk30) that have become industry
standards. The recommended maximum rate is 3.0 mphps (1.3
m/s2) for both acceleration and deceleration.
Attempts have been made to increase these rates, specifically
on the rubber tired metros in Montreal and Mexico City, but
subsequently these were reduced close to the industry standard.
Except for locomotive hauled commuter rail, almost all rail
transit in North America operates with these rates. The main
difference in equipment performance is the maximum speed.
Most urban rail systems with closer station spacing have a
maximum speed of 50-60 mph (80-95 km/h), light rail typically
has a maximum speed of 50 mph (80 km/h),31 while streetcars
have a maximum in the range of 40-50 mph (65-80 km/h). The
few suburban type rail rapid transit systems have a higher
maximum of 70-80 mph (110-130 km/h)BART in San
Francisco and PATCO in Philadelphia are the principal
examples.
Equation 3-11
27
where B is a constant representing the increments or percentage

of the braking distanceor number of blocksthat must
separate trains according to the type of train control system. A
B-value of 1.2 is recommended for multiple command cab
controls. A value of 2.4 is appropriate for three-aspect signaling
systems where there is always a minimum of two clear blocks
On existing systems the results can be calibrated to actual performance by

adjusting the value of B.
tos+ tjl+ tbr may be simpified by treating as a single valuetypically 5 sec
for systems with ATO, slightly longer with manual driving.
29
HIRSCHFELD, C.F., Bulletins Nos. 1-5, Electric Railway Presidents
Conference Committee (PCC), New York, 1931-1933.
30
jerkrate of change of acceleration.
31
SEPTAs Norristown line is a higher speed exception.
28
28
The higher gearing rates required for these higher speeds
result in either a reduced initial acceleration rate or, more
typically, an acceleration rate that more rapidly reduces (follows
the motor curve) as speed increases.
Braking rates are invariably uniform. Emergency braking
rates vary widely and are significantly higher and more
sustainable on equipment fitted with magnetic track brakesall
streetcars, most light rail and the urban rail transit systems in
Chicago and Vancouver.
This relative uniformity of rates allows a typical solution of
Equation 3.11 using the following data for a cab control system
with electrically controlled braking and a train of the maximum
length in North American rail transit.
The results of applying typical rail transit data to Equation 3-9
are shown in Figure 3.3 using the data values of Table 3.1.
Table 3.2 Breakdown of line headway time components
Table 3.1 Data values for line headway

33
32
The 3-sec figure is conservative. For automatically driven trains, a time of

1 sec is appropriate and can drop as low as 0.2 sec on AGT systems. The
higher figure is useful on cab control systems. When the overspeed
detection occurs, and alarm is sounded in the cab to allow the driver to
apply service braking and so cancel the automatic application of
emergency brakesavoiding wheel flats and passenger discomfort or loss
of balance. The delay time is then based on typical manual reaction times
of 2 to 3 sec. With entirely manual operation this term becomes a
surrogate for driver sighting and reaction time. Values of 2 to 5 sec have
been quoted in the literature. 3 sec is an appropriate value.
Overspeed time is applicable to automatically driven trains.
These are somewhat theoretical, showing headways down to

31.5 seconds120 trains per hour. There is a clear minimum at
50 km/h (31 mph). Obviously restricting train line speed to so
low a value would be uneconomic, requiring a larger number of
cars to meet a given demandwhich would, in any event,
diminish because of the slow travel times deterring passengers.
The equation and results will be applied in Chapter 10 for
automated guideway transit with off-line stations and will be
used as a basis for determining realistic headways with station
stops.
To this end it is useful to examine the value of the
components in the line headway, shown in Table 3.2 with all
figures in seconds. Columns one through five in this table
represent, consecutively, the first five terms of Equation 3-12.
The time to travel the length of train and the factored braking
time predominate. No value has been assigned to the brake
system reaction time. The time associated with the runaway
acceleration is small. Equation 3-12, adjusted to compensate for
grades and line voltage variations, is included in the spreadsheet
on the computer diskette. For manual calculations, the equation
can be simplified to:
Equation 3-13
where the constant 4 is approximately the rounded up sum of
columns 3, 4 and 5 plus a small allowance for brake reaction
time. This should be increased to 7 for manually driven systems
to add the train operator sighting and reaction time.
The next step is to accommodate station stops. Reference to
the literature will show numerous ways to calculate the station
headway. This approach is based on adapting the line headway
equation.
3.6.4 CALCULATING STATION HEADWAY

Figure 3.3 Line headway versus speed
Station headway, the time for one train to replace another at the
maximum load point station, is by far the most common capacity
29
limitation. Having derived an expression for line headway that
uses readily available information with as few approximations
as possible, it is possible to adapt this to station headway by
changing line speed to approach speed and solving for this
speed,
adding a component for the time a train takes to clear the
platform,
adding the station dwell, and
adding an operating margin.
The time for a train to clear the platform is
Equation 3-14
minimum headway in Table 3.4 show a distinct optimum

approach speed for fixed-block systems. Moving-block
signaling systems, which adjust their separation according to
speed, are discussed in the next section. The values are
calculated in Table 3.5 with different values of dwell and
operating margin times. Speeds are rounded to the nearest km/h
or mph reflecting the approximations used in their derivation.
As Figure 3.4 deals with maximum length trains, running at
minimum headways, at the longest dwell35 station, dwell times
of 30 sec may not be possible and the lower values of H(s) are
unlikely. The above calculations do not take into account any
speed restriction in the station approach. Reference to Figure 3.4
shows a rapid fall off in throughput as the approach speed
decreases. Speed restrictions may be due to curves, special
work, or speed controls approaching a terminal station. The
Figure 3.5 shows the speed of a braking train against
Adding Equation 3-14 to 3-12 plus components for dwell and an

operating margin produces the station headway
Table 3.3 Data values for station headway
Equation 3-15
where
H(s) = station headway in seconds

= length of the longest train in meters
L
= distance from front of stopped train to start
D
of station exit block in meters
= station approach speed in m/s
va
vmax = maximum line speed in m/s
= braking safety factorworst case service
K
typically 75%
= separation safety factorequivalent to
B
number of braking distances plus a margin,
(surrogate for blocks) that separate trains
= time for overspeed governor to operate
tos
= time lost to braking jerk limitation
tjl
(seconds) typically 0.5 seconds
= operator and brake system reaction time
tbr
= dwell time (seconds)
td
tom = operating margin (seconds)
as
= service deceleration rate in m/s2
ds
Typical values will be used and this equation solved for the
approach speed under two circumstances:
34
B = 1.2 for cab control, 2.4 for 3 aspect signaling
Table 3.4 Optimum approach speeds
Table 3.5 Headways with dwell and operating margins
1. three-aspect signaling system (B = 2.4)

2. multiple command speed cab controls (B = 1.2)
A 45-sec dwell time is usedtypical of the busiest stations on
rail transit lines operating at capacitytogether with an
operating margin time of 20 sec. The brake system reaction time
will use a moderate level of 1.5 secthis should be higher for
old air brake equipment, lower for modern electronic control,
particularly with hydraulically actuated disk brakes. Other
factors remain at the levels used in the line headway analysis.
(See Table 3.3.) The results of solving Equation 3.15 for
35
The longest dwell station is usually at the maximum load point station and
is so assumed through this report. Reference to Chapter Four, Station
Dwells shows that a high-volume mixed-flow station could have a longer
dwell than the higher volume maximum load point station.
30
Figure 3.6 Speed limits on curves and switches

vsl = (87R(e + f))1/2
where
Figure 3.4 Station headway for lines at capacity
Equation 3-16
vsl = speed limit in km/h

R = radius of curvature in meters
e = superelevation ratio (height the outer rail is
raised divided by track gauge) usually not
greater than 0.10
f = comfort factor (ratio of radial force to
gravitational force0.13 is the maximum
used in rail transit with some systems using as
low as 0.05)
In U.S. customary units, mph and feet, the speed limit is

vsl = (15R(e + f))1/2
Equation 3-17
The results of speed limits due to curves are plotted below for
both flat curves and curves superelevated with the maximum
radial force (e = 0.10). Transition spirals are not taken into
account in Figure 3.6. The vertical bars show the AREA36
recommended speed limit range for lateral and equilateral level
turnouts of size #6, #8 and #10. Note that many operators have
their own speed limits for turnouts that may differ from those
shown.
3.7 SENSITIVITY
Figure 3.5 DistanceSpeed chart
Two factors have not been taken into account in the

determination of minimum headways in the preceding section
grades and fluctuations in traction voltage.
3.7.1 GRADES
distanceusing the performance data of Table 3.3. If a more
restrictive speed limit is within the distance for a given approach
speedplus the length of the trainthen that more restrictive
limit should be used in Equation 3-15 to calculate the minimum
headway.
On existing systems speed limits are usually posted on the
wayside and included in the rule book. On new systems where
speed limits are not known they can be approximated from
The principal effect of grades is where downgrades into stations

increase the braking distance37 and the distance associated with
36
37
American Railway Engineering Association.

Certain modern equipment uses accelerometers to adjust propulsion and
braking to constant levelsindependent of train load or grades. In this
case grade need not be taken into accountup to the point that wheel-rail
adhesion becomes inadequatean unlikely event.
31
Figure 3.8 Headway changes with voltage
Figure 3.7 Effect of grade on station headway

(cab signals, dwell = 45, margin = 20 secs)
Table 3.6 Result of 4% station grades on headway

(cab signals, dwell = 45, margin = 20 secs)
the runaway propulsion factor. A simple method to compensate

for grades is to adjust the service braking and acceleration rates
in Equation 3-15 while holding the component of the equation
that relates to the time for a train to exit a platform constant. The
acceleration due to gravity is 9.807 m/s2. Thus each 1% in
downgrade reduces the braking rate by 0.098 m/s2. The results
are shown in Figure 3.7. Note that most rail transit systems have
design standards that limit grades to 3 or 4%, a few extend to
6% and the occasional light rail grade can extend to 10%. The
impact of grades is greater into a station. The greatest impact is
a downgrade into a station which increases the braking and so
the safe separation distance. Block lengths must be longer to
compensate for the longer braking distances. The absolute and
percentage changes are tabulated in Table 3.6 for the typical
heavy rail maximum grade of 4%.
voltage occurs at locations most remote from sub-stations in the

peak hour when the maximum number of trains are in service.
The lower voltage reduces train performanceat a time when
the heavy passenger load is doing likewise. Both acceleration
and balancing speed are reduced; braking is not affected.
The acceleration of a train is approximately proportional to
the power applied to the motors, which in turn is proportional to
the square of the supply voltage. This is particularly true for
older equipment with switched resistor controls39, less so with
modern electronically controlled equipment.40 Consequently, for
older equipment without on-board motor voltage feedback and
control, the common 10% reduction in voltage will reduce
acceleration to 81% of normal, the very rare 30% drop will
reduce acceleration to 49% of normal.
Reduced acceleration affects the platform clear out
component of the headway calculation. The resultant headway
sensitivity to voltage is shown below. At a typical 15% drop in
voltage (85% in Figure 3.8), headway increases by 3.2 seconds,
a 2.7% change. It is not possible to calculate line voltage at any
instance of time without a complete train performance and
traction supply system simulation. This will automatically occur
if a simulation is used to determine the minimum headway.
Otherwise it is uncertain whether a manual adjustment should be
made based on the above chartwith certain designs of modern
rolling stock the effect of voltage drop can be less than shown.
3.7.3 ACCELERATION
Changes in acceleration affect the time required for a train to
clear the platform and make minor adjustments to the runaway
3.7.2 LINE VOLTAGE

39
Rail transit in North America is supplied by direct current power

at a potential of 600 to 750 volts with the occasional 1,500-volt
system. As more power is drawn through the substations,
feeders and third rail or overhead catenary, the voltage drops.
Voltage is higher in the vicinity of substation feeders and drops
off with distance. Voltage is said to be regulated within a system
specification that is typically +20% to -30%. 38 The lowest
38
Certain newer rail systems have purchased vehicles with electronic motor
controls that are intolerant of voltage drops. Consequently the traction
supply voltage has to be regulated to closer tolerances.
Estimated to be used on about three quarters of the rolling stock in North

America, including all NYCT cars except prototypes.
40
Modern electronically controlled equipment may use accelerometers
which will command the vehicles power conversion unit to compensate
for reduced voltage. Similar feedback systems may attempt to regulate
motor voltageven with reduced line voltage. However such corrective
action defeats the self regulating effect of the reduced line voltagea
rationing of power when demand from the trains exceeds the capability of
the power supply and so increases the likelihood that the power supply
system will trip (disconnect) due to overload. On manually driven systems
lower line voltage is immediately apparent to the driver and serves as an
advisory to reduce demand or, when trains are lined up due to a delay, to
start up in sequential order rather than simultaneously. Consequently,
providing full correction for drops in line voltage is unwise.
32
propulsion safety factor. Headways for a cab signal train control
system are shown with acceleration adjusted to 50%, 75% and
125% of the normal value1.3 m/s2(3.0 mphps). (See Figure
3.9).
3.7.4 BRAKING
Changes in braking rate affect both the braking time and the safe
separation time. Headways for a cab signal train control system
are shown with braking adjusted to 50%, 75% and 125% of the
normal value in Figure 3.10. Changes in the braking rate have a
greater effect on headway than those of acceleration. Note that
the optimum approach speed increases with the braking rate.
The normal rate (100%) is 1.3 m/s2 (3.0 mphps).
Figure 3.11 Headway changes with train length
3.7.5 TRAIN LENGTH

All previous work in this section has used a maximum train
length of 200 m (660 ft). Shorter trains will permit closer train
spacing as shown in Figure 3.11.
3.8 MOVING-BLOCK
THROUGHPUT
Figure 3.9 Headway changes with the acceleration rate
Moving-block signaling systems can use a fixed safety

separation distance, plus the calculated braking distance, to
separate trains, or a safety distance that is continually adjusted
with speed and grades. In this section both approaches will be
developed and compared.
3.8.1 FIXED SAFETY DISTANCE

The minimum station headway for the close-in operation is
expressed in Equation 3-15. For a moving-block signaling
system there is no requirement for a train to travel its own length
and vacate the station platform before freeing up a block for the
following train. Rather, the moment a train starts from a
platform the distance so freed is added to that available for the
following train to proceed.
Figure 3.10 Headway changes with the braking rate
The term for the time to clear the platform block can be
removed. The safety separation constant Ba surrogate for the
number of blocks between trains can be set to zero. The fixed
safety distance can be added to the train length to produce a
term that represents the time to travel both the train length plus
the fixed safety distance. The overspeed acceleration time
equivalent and time constant terms can be removedallowance
for runaway propulsion is included in the fixed safety distance.
The overspeed time can similarly be deleted.
33
The other factors in the equation should remain. The braking reaction time can be adjusted for the specific equipment. The station
headway Equation 3-15 is shown below with the main components identified
where
H(s) = station headway in seconds

= length of the longest train in meters
L
= distance from front of stopped train to
D
start of station exit block in meters
= station approach speed in m/s
va
vmax = maximum line speed in m/s
= braking safety factorworst case service
K
typically 75%
= separation safety factorequivalent to
B
number of braking distances (surrogate for
blocks) that separate trains
= time for overspeed governor to operate on
tos
automatic systemsto be replaced with
driver sighting and reaction times on
manual systems (seconds)
= time lost to braking jerk limitation
tjl
typically 0.5 seconds
= brake system reaction timeolder air
tbr
brake equipment only (seconds)
= dwell time (seconds)
td
= operating margin (seconds)
tom
as
ds
The final four time constants can be abbreviated so that

t = tjl + tbr + td + tom
Equation 3-18
The adaptation of Equation 3-15 for a moving-block signaling

system with fixed safety separation becomes
Equation 3-19
where
The resultant throughput is high and becomes limited by station

dwells, junctions and issues of operational allowances.
Safety distances for more conventional equipment are triple or
quadruple, particularly if there are significant grades. In these
circumstances a variable safety distance will increase the
throughput.
This alternate approach develops an approximation for a
safety distance that adjusts with circumstances. In this case the
assumption is made that the safety distance comprises the
braking distance (i.e., B = 1) plus the runaway propulsion
components and a positioning error distanceall adjusted for
any downgrade into the headway critical station.
Discounting grades for the moment the station headway can
be represented by:
Equation 3-20
where
Pe = positioning error
B = 1
Adjusting for a the grade into a headway critical station, the

service acceleration should be increased by one hundredth of the
force of gravity for each percentage of grade, and the service
braking rate reduced similarly. Thus the acceleration rate is
multiplied by (1 - gG/100) where g is the acceleration due to
gravity (9.807 m/s2) and G is the percentage gradenegative
for downgrades. This adjustment approximates to (1 - 0.1G).
The result becomes
Smb = moving-block safety distance
The calculation of the appropriate safety distance is described

by Motz(R47). The process is complicated and requires judgment
calls on how to represent the worst case situation. The final
figure may involve compromises involving decisions of the
appropriate government regulatory body (if any) and/or the rail
transit system executive.
The Vancouver SkyTrain moving-block signaling system uses
a short safety distance of 50 m (165 ft), reflecting the short
trains and high levels of assured braking from magnetic track
brakes and motor brakingboth independent of traction power.
Equation 3-21
The results of this equation are shown in Figure 3.12 using data
from Table 3.3 with B = 1 and a positioning error of 6.25 m (21
ft). The resultant minimum headway of 97 sec occurs at an
approach speed of 56 km/h (mph). The respective curves for a
conventional three-aspect signaling system and a cab control
system are included for comparison. As would be expected, a
34
Figure 3.12 Moving-block headways with 45-sec dwell and

20-sec operating margin compared with conventional fixedblock systems
moving-block system with a speed variable safety distance

shows the lowest overall headway. The difference between the
two methods of determining the safety distance represents an
eight second difference in the minimum headwaypointing out
the importance of selecting the best method when a close
headway is required.
The elasticity of moving-block headways with respect to
voltage fluctuations will be negligible as the time to clear the
plat-form is not a component in calculating the moving-block
signaling system headway. The effect of grades is shown in
Figure 3.13.
Downgrades (negative) into a station significantly reduce the
minimum headway while positive grades have little effect.
Figure 3.13 Effect of grades on a moving-block signaling

system with variable safety distance
Figure 3.14 Terminal station track layout 42
train (lower left) being held at the cross-over approach signal

while a train departs. It must, moving from a stop, traverse the
cross-over and be fully berthed in the station before the next
exiting train (lower right) can leave. The distance involved is
Da = P + T + CS
3.9 TURN-BACK
THROUGHPUT
where
Correctly designed and operated turn-backs should not be a

constraint on capacity. A typical minimal terminal station
arrangement with the preferred41 center (island) platform is
shown in Figure 3.14. The worst case is based on the arriving
Da
P
T
S
C
=
=
=
=
=
Equation 3-22
approach distance
platform length
distance from cross-over to platform
track separation ( platform width +1.6m)
switch angle factor
5.77 for #6 switch
6.41 for #8 switch
9.62 for #10 switch
The time for this maneuver is expressed as

41
While side platforms reduce the track to track centers and so reduce the
maneuver time, they require passengers to be directed to the correct
platform for the next departing train. This is inherently undesirable and
becomes more so when a train cannot depart because of a defect or
incident and passengers must be redirected to the other platform.
42
The diagram shows no run-on space beyond the station platform. Where
there is little or no such space, mechanical or hydraulic bumpers should be
provided.
Equation 3-23
where
ta = approach time
as = initial service acceleration rate in m/s2
ds = service deceleration rate in m/s2
35
The distance to exit the station, a straight run, is shorter but the
initial acceleration rate will start to taper off. Leaving the travel
distance the same to compensate for this, the time for the exiting
train to clear the cross-over can be approximated as:
Equation 3-24
In between these two travel times is the terminal time that
includes the dwell for alighting and boarding passengers, the
time for the train operator to change ends and conduct any
necessary inspections and brake tests, the time for the crossover
switches to move and lock plus any desired schedule recovery
time.
With two terminal tracks, the headway restriction is half the
sum of these time components, expressed as:
Equation 3-25
where
H(t)
ta
te
tt
ts
= terminal headway time

= terminal approach time
= terminal exit time
= terminal layover time
= switch throw and lock time
(all in seconds)
Determining the terminal layover time is difficult. An approach

is to look at the maximum terminal layover time for a given
headway by transposing Equation 3-24.
tt 2(H(t) - ts) - te -ta
Equation 3-26
The maximum terminal layover time can then be calculated.

With the following typical worst case parameters:
where
the headway = 120 sec

train length = 200 m
track separation = 10 m
distance from cross-over to platform = 20 m
initial service acceleration rate = 1.3 m/s2
service deceleration rate = 1.3 m/s2
switch is #10
switch throw and lock time is 6 sec
the terminal time tt 175 sec. This would increase by 9 sec if

the incoming train did not stop before traversing the cross-over.
While this is not a generous amount of time, particularly to
contain a schedule recovery allowance, many systems maintain
such close headways with minimal delays.
This maximum permitted terminal time can be calculated for
the specific system and terminal parameters. Where the time is
insufficient there are numerous corrective possibilities. These
include moving the cross-over as close to the platform as
possible note that structures can restrict the cross-over
location in subways.
The full terminal layover time is available for station dwell. If
passenger movement time is a limiting factor then this can be
reduced with the use of dual-faced platforms. At terminals with
exceptionally heavy passenger loading, multiple track layouts
may be needed. An atypical alternative, used at SEPTAs 69th
Street; PATHs World Trade Center termini; and the Howard,
Desplaines, and 54th St. CTA Stations is the use of loopswith
the exception of several examples in Paris this is rare for rail

transit.
Crew turnaround time can be expedited with set-back
crewing. At a leisurely walking pace of 1 m/s, it would take 200
sec for a driver to walk the length of a 200 m train, more if the
driver were expected to check the interior of each car for left
objects or passengers. Obviously this could not be
accommodated reliably in a 175-sec terminal layover time.
Terminal arrangements should accommodate some common
delays. An example would be the typical problems of a train
held in a terminal for a door-sticking problem; waiting for police
to remove an intoxicated passengereuphemistically termed a
sleeper; or for a cleaning crew. Alternately one track may be
preempted to store a bad order train. On these occasions the
terminal is temporarily restricted to a single track and the
maximum terminal layover time is reduced to 61 sec with the
above parameters (70 sec without an approach stop). This may
be sufficient for the passenger dwell but cannot accommodate
changing ends on a long train and totally eliminates any
schedule recovery allowance.
More expensive ways to improve turn-backs include
extending tracks beyond the station and providing cross-overs at
both ends of the station. This permits a storage track or tracks
for spare and disabled trainsa useful, if not essential, failure
management facility. With cross-overs at both ends of the
station, on-time trains can turn-back beyond the station with late
trains turning in front of the stationproviding a valuable
recovery time of some 90 sec at the price of additional
equipment to serve a given passenger demand.
The above analysis has assumed that any speed restrictions in
the terminal approach and exit are below the speed a train would
reach in the calculated movementsapproximately 21 km/h (13
mph) on a stop-to-stop approach, 29 km/h (18 mph) as the end
of the train leaves the interlocking on exit. For safety reasons,
some operators have imposed very low entry speeds,
occasionally enforced with speed control signaling.
Slow terminal approaches are common on manually driven
rail transit systems in the United States. In some cases this
approach could be a greater restriction than the start from stop at
the approach cross-over represented in Equation 3-24. If an
approach speed restriction exists that is less than (taas/2) (m/sec)
then the above methodology should not be used.
3.10 JUNCTION
THROUGHPUT
Correctly designed junctions should not be a constraint on
capacity. Where a system is expected to operate at close
headways, high use junctions will invariably be grade separated.
At such flying junctions, the merging and diverging movements
can all be made without conflict and the only impact on capacity
is the addition of the switch throw and lock times, typically 3 to
6 sec. Speed limits, imposed in accordance with the radius of
curvature and any superelevation, may reduce the schedule
speed but should not raise the minimum headwayunless there
is a tight curve close to a headway limiting station.
The capacity of a flat junction can be calculated in a similar
manner to the terminal station approach. The junction
36
Figure 3.15 Flat junction track layout
arrangement is shown in Figure 3.15. The worst case is based on

a train (lower left) held at signal A while a train of length T
moves from signal B to clear the interlocking at C. The
minimum operable headway is the line headway of train A
(from Figure 3.3) plus the time for the conflicting train to clear
the interlocking plus the extra time for train A to brake to a
stop and accelerate back to line speed. Ignoring specific block
locations and transition spirals, this can be expressed approximately as:
Equation 3-27
where
H(j)
H(l)
T
S
C
as
ds
vl
ts
tom
= limiting headway at junction (seconds)

= line headway (Figure 3.3) (seconds)
= train length in meters
= track separation in meters
= switch angle factor
5.77 for #6 switch
6.41 for #8 switch
9.62 for #10 switch
= line speed in m/s
= switch throw and lock time (seconds)
= operating margin time (seconds)
The limiting headway at the junction can then be calculated with

the following typical parameters:
where
line headway = 32 sec

line speed = 100 km/h
train length = 200 m
track separation = 10 m
initial service acceleration rate = 1.3 m/s2
service deceleration rate = 1.3 m/s2
switch is #10
switch throw and lock time is 6 sec
The result is a junction limiting headway of 102 sec plus an

operating margin. While in theory this should allow a 120-sec
headway with a flat junction, it does not leave a significant
operating margin and there is a probability of interference
headways. General guidance in rail transit design is that
junctions should be grade separated for headways below 150 to
180 sec.
An exception is with a moving-block signaling system
incorporating an automatic train supervision system with the
capability to look forwardand so adjust train performance and
station dwells to avoid conflicts at the junction, i.e. trains will

not have to stop or slow down at the junctionother than for the
interlockings civil speed limit. In this case, the junction
interference headway drops to 63 sec, allowing 120 sec, or
slightly lower, headways to be sustained on a flat junctiona
potentially significant cost saving associated with a movingblock signaling system.
A real-life example of the restrictions created by junctions is
contained in a NYCTA study.43 This capacity analysis of
NYCTA operations focused on the backbone of services in
Queensthe Queens Boulevard line to 179th Street. The
analysis determined headway constraints due to train
performance, the signaling system, and station dwell times. An
analysis of the partially flat junction at Nostrand Avenue
indicated a throughput that was four trains per hour per single
track lower than the 29 to 31 trains per hour that is typically the
NYCTA maximum.
3.11 SUMMARY
Using as few approximations as possible, the minimum
headway has been calculated for a range of train control systems
with a wide number of variables. Table 3.7 summarizes the
results including the raw minimum headway with the dwell and
operating margin times stripped away.
The spreadsheets contained on the available disk allow the
user to change most variables and obtain the minimum headway
under a wide range of circumstances.
CAUTION This table and the spreadsheet make assumptions
and approximations. The results are believed to be a reliable
guide but are not a substitute for a full and careful simulation of
the train control system in conjunction with a multiple train
performance simulation. To these times approximately 6
seconds should be added for a 4% downgrade into the headway
critical station. Three to four seconds can be added to allow for
voltage drops at peak times on systems at full capacityexcept
for the moving-block signaling system.
The results of this chapter concur with field data and agree or
are close to the calculations of most other headway
determination
43
As reported by panel member Herbert S. Levinson from the study: BOOZ

ALLEN and HAMILTON INC., in association with Abrams-Cherwony;
Ammann & Whitney; George Beetle: Merrill Stuart, Queens Transit
Alternatives Technical Appendix, Part 5a, Operations/Capacity Analysis,
NYCTA, New York, January 1981.
Table 3.7 Headway result summary in seconds with 200 m

(660 ft) (8-10 cars) VSD = variable safety distance
44
Perversely, the operating margin should be increased as the dwell time

increases
37
methods reviewed in Appendix One. Typical cited minimum
headways, without dwell or operating margin times, are in the
range of 50 to 60 sec for conventional train controlcompared
to the 51 to 57 sec in the above summary.
Auer(R09) estimates that a moving-block system should
increase system capacity by 33% based on a 20-sec dwell45 and
10-sec operating margin. With these quantities the headway of
the VSD moving-block signaling systems is 62 secproviding a
capacity increase of 30% over the cab control signaling system
value of 81 sec.
This reflects a slightly conservative approach in calculating
the moving-block signaling system headway with the safety
separation factor B set at a full braking distance. B can be
reduced to less than one. Auers capacity gain is achieved if B
is set to 0.77.
The value of B can be adjusted for the three types of
signaling to calibrate the equations of this chapter with actual
field experience or system simulation.
The components of headway for the above mid range
cabcontrol data are shown in the Figure 3.16 with a station dwell
of 45 sec and operating margin of 25 sec.
The components are shown in the order of Equation 3.15 with
terms running from the bottom upwards. Dwell is the dominant
component and the subject of the next chapter.
Figure 3.16 Headway components for cab control signaling

that comprise the typical North American minimum
headway of 120 sec
45
Note that many of the referenced headway analyses use a fixed dwell of
20 or 30 sec. This is rarely adequate. On heavy rail transit systems with
long trains running at or below headways of 120 sec the dwell at the
headway controlling stations will often reach into the range of 40-50
secand so become the largest headway component.
38
4. Station Dwells
4.1 INTRODUCTION
In Chapter Two, Capacity Basics, station dwells were
introduced as one of three components of headway. Dwells are
the major component of headways at close frequencies as shown
in Figure 4.1based on a heavy rail system at capacity,
operating 180-m-long trains with a three-aspect signaling
system. The best achievable headways under these
circumstances are in the range of 110 to 125 sec.1 In Chapter
Two the concept of controlling dwell was also introduced.
Controlling dwell is the combination of dwell time and a
reasonable operating marginthe dwell time during a normal
peak hour that controls the minimum regular headway.
Controlling dwell takes into account routine perturbations in
operationsbut not major or irregular disruptions. The sum of
controlling dwell and the train control systems minimum train
separation time produces the maximum train throughput without
headway interference.
In this chapter the components of dwell time are examined.
The major componentpassenger flow timeis analyzed. and
methodologies developed for determining passenger flow times
and dwell times.
The second category analyzed dwell time relative to the

number of passengers boarding and alighting. This group
concluded that linear regression provided the most suitable fit
for both rapid transit and light rail with high- and low-level
loading for specific systems.2 Three references improved the
data fit by including
Table 4.1 List of dwell time references
Alle, Improving Rail Transit Line Capacity Using
Computer Graphics
Anderson, Transit Systems Theory
Auer, Rail-Transit People-Mover Headway
Comparison
Barwell, Automation and Control in Transport
Canadian Urban Transit Association, Canadian
Transit Handbook
Celniker, Trolley Priority on Signalized Arterials in San
Diego
Chow, Hoboken Terminal: Pedestrian Planning
Gray, Public Transportation Planning, Operations and
Management
Jacobs Transit Project--Estimate of Transit Supply
Parameters
Janelle, Interactive Hybrid Computer Design of a
Signaling System
Klopotov, Improving the Capacity of Metropolitan
Railways
Koffman, Self-service Fare Collection on the San
Diego Trolley
Kraff, Evaluation of Passenger Service Times
Levinson, Some Reflections on Transit Capacity
Levinson, ITE Transportation Planning Handbook
Chapter 12
Levinson, Capacity Concepts for Street-Running Light
Rail Transit
Lin, Dwell Time Relationships for Light Rail Systems
Miller, Simulation Model of Shared Streetcar Right-ofWay
Motz, Attainable Headways Using SELTRAC
Pushkarev, Urban Rail in America
Schumann, Status of North American LRT Systems
TRB, Collection and Application of Ridership Data on
Rapid Transit
TRB, Highway Capacity Manual, Chapter 12
US DoT Characteristics of Urban Transportation
Systems
Vuchic, Urban Public Transportation Systems and
Technology
Walshaw, LRT On-Street Operations: The Calgary
Experience
4.2 LITERATURE REVIEW

The literature review produced 26 dwell time references listed
in Table 4.1. The full listing is contained in Chapter Twelve,
Bibliography and a summary of each reference is contained in
Appendix One. These references can be divided into three
categories. The largest category discussed dwell as a component
in calculating train throughput.
Figure 4.1 Typical headway components in seconds

1
Some European systems operate three or more aspect signaling systems

with headways down to 90 sec by strict control of dwellson occasion,
closing doors before all passenger movements are complete. This is not an
acceptable practice in North America.
Lin and Wilson (R44) indicate that crowding may cause a non-linear increase
in dwell time during congested periods. Koffman, Rhyner and Trexler, (R33)
after testing a variety of variables, including various powers, exponentials,
logarithms and interaction terms, conclude that a linear model produced
the best results for the specific system studied.
39
the number of passengers on-board a car as a variable. One
paper, by Koffman, Rhyner and Trexler(R33), evaluated a variable
to account for passenger-actuated doors on the San Diego
trolley.
In the third category, a single paper (Alle(R02)) answered two
key questions: How many trains can realistically pass a point in
one hour? and What is the impact of station dwell times on
this throughput?
Using an at-capacity section of the MTA-NYCT E & F lines,
Alle analyzed the actual peak-hour dwells at Queens Plaza
Station in New York by trapping 85% of the area under the
normal distribution curve. The upper control limit becomes the
mean plus one standard deviation with a 95% confidence
interval. The results determined that this specific single track,
with the given set of dwells, can support trains every 130 sec
almost identical to the actual throughput of 29 trains per hour
(124 sec).
Alles methodology is based on measurements of actual
inservice dwell times, and so it is unsuitable for determining
controlling dwells of new systems or new stations added to
existing systems where such information would not be available.
With the above exception, the literature offers only methods
to determine passenger flow times; no material was found that
adjusts these flow times to either the full station dwell time or a
controlling dwell time. Many reports, and even some
simulations, use a manually input average dwell time, a worst
case dwell time, or merely a typical dwell timeoften quoted at
15 to 20 sec per station with 30 sec or more for major stations.
These gross approximations usually produce a throughput of 40
to 50 trains an hour and so require applying one or more factors
to adjust the resultant throughput to the actual North American
maximum of 30 to 32 trains an hour.
This situation required the authors to make a fresh start at
developing a methodology for calculating dwells. Much of the
field data collection involved timing dwells and passenger
flows.
4.3 DWELL CONSTITUENTS

Dwell is made up of the time passenger flow occurs, a further
time before the doors are closed and then a time while waiting to
depart with the doors closed. Figures 4.2 through 4.5 show these
dwell components for the peak period of four selected systems.
Each of the systems has a different operating philosophy. BART
is automatically driven with door closure and departure
performed manually; the latter subject to override by the
automatic train control. NYCT is entirely manual, subject only
to a permissive departure signal. BC Transit is an entirely
automatic system with unattended cars; door closing and
departure times are preprogrammed. Station dwells are
contained in a nonvital table of the train control system and are
adjusted by station, destination, time-of-day and day-of-week.
The Toronto Transit Commission is also entirely manual but,
unlike New York, has recently implemented a safety delay
between door closure and train departure on the Yonge subway.
The data collection did not time any delays between a train
stopping in a station and the doors opening. Although there were
such minor delays, few were long enough to possibly annoy
passengers. Delays do occur with passenger-actuated doors used
on many light rail systems. These are discussed separately in
section 4.4.2 of this chapter.
Figure 4.2 BART Montgomery Station

components p.m. peak February 9, 1995
dwell
time
The preprogrammed nature of the BC Transit observations are

very evident. There are two services in the data set. The short
turn service has shorter dwells until it endsjust over halfway
down the chart. Minor variants in the total dwell time for each
service are due to observation errors. Data were collected at the
heaviest used doorway(s) on the train. While it was not always
possible to guarantee that this was selected, it is still surprising
that the proportion of dwell time productively used for
passenger movements is so small, ranging from 31 to 64% of the
total dwell. Only New York fares well in this regard with a
percentage of productive time double the other examples.
However, there were major variations in the percentage of
productive time between stations on the same system (See Table
4.3).
These four charts are representative of 61 data sets of door
flows collected in early 1995 for those few systems operated at,
or close to, the capacity of their respective train control systems.
40
Figure 4.3 NYCT Grand Central Station dwell time

components a.m. peak February 8, 1995
The data represent the movement of 25,154 passengers over 56

peak periods, two base (inter-peak) and three special event
times, at 27 locations on 10 systems. All data sets are contained
on the computer disk. Table 4.2 summarizes the results. The low
percentage of dwell time used for passenger flow at the heaviest
use door presents a challenge in determining dwell times from
the passenger volumes in section of this chapter.
In Chapter Three, Train Control and Signaling, it was
suggested that automatic drivingwhen compared with manual
drivingshould permit a train to run closer to civil speed limits
and not commence braking until the last moment, thus reducing
train separation by 5 to 15% and increasing capacity by a like
amount and improving regularity.
There was insufficient data to confirm this, although Figure
4.5, shows BC Transits automated operation with a short-turn
service integrated into two other services at a very consistent 90sec separation.
Figure 4.4 Toronto Transit Commission King Station S/B

dwell time components: am peak February 6, 1995
However, the project observers, timing dwells and counting a

total of over 25,000 passengers at various locations on 10
systems noted a wide variation in operating practices that ranged
from efficient to languid, with automatically driven systems
predominantly in the latter group. It would appear that any
operating gains from automatic driving may be more than offset
by time lost in station dwell practices.
Several light rail and heavy rail systems were notably more
expeditious at station dwells than their counterparts,
contributing to a fasterand so more economic and attractive
operation. Most automatically driven systems had longer station
dwells extending beyond the passenger movement time.
This inefficiency is extending to some manually driven systems where safety concerns have resulted in the addition of an
41
Table 4.3 Summary of all door observations through a single
double-stream door during the peak period (1995)
Figure 4.5 BC Transit SkyTrain Burrard Station inbound

dwell time components am peak April 5, 1995
Table 4.2 Summary of door observations through one

double-stream door during the peak periodfour rail
transit systems operating at or close to capacity (1995)
artificial delay between the time the doors have closed and the
train starts to move from the platform.
A companion Transit Cooperative Research Program project
A-3, TCRP Report 4, Aids for Car Side Door Observation, and
its predecessor work, National Cooperative Transit Research &
Development Program Report 13, Conversion to One-Person
Operation of Rapid-Transit Trains, address some of these issues
but do not examine overall door-platform interface safety or the
wide differences in operating efficiency between various light
and heavy rail systems. This issue is discussed further in
Chapter Eleven, Future Research.3
4.4 DOORWAY FLOW TIMES

4.4.1 FLOW TIME HYPOTHESES
Flow time is the time in seconds for a single passenger to cross
the threshold of the rail transit car doorway, entering or exiting,
per single stream of doorway width.
3
Dwells may be intentionally extended to enable cross-platform connections

between local and express trains.
42
In the course of conducting this study, several interesting
conjectures and educated guesses were encountered relating to
flow times and rail transit vehicle loading levels. Certain of
these suggest the attractiveness of air-conditioned cars on hot
days may decrease both doorway flow times and increase the
loading level. Similarly with warm cars in cold weatherwith
loading levels offset by the bulk of winter clothing. While there
is some intuitive support for these hypotheses no data were
obtained to support them.
Other hypotheses related to different flow times between old
and new rail transit systems, for example, that after delays and
under emergency operation passengers will load faster and
accept higher loading levels. Similar circumstances apply when
rail transit is used to and from special eventssuch as sporting
venues.
4.4.2 FLOW TIME RESULTS

Part of the dwell time determination process involves passenger
flow times through a train doorway. Data were collected from a
representative set of high-use systems and categorized by the
type of entrylevel being the most common, then light rail with
door stairwells, with and without fare collection at the entrance.
These data sets were then partitioned into mainly boarding,
mainly alighting and mixed flows. The results are summarized
in Figure 4.6. The most interesting component of these data is
that passengers enter high-floor light rail vehicles faster from
street level than they exit. This remained consistent through
several full peak period observations on different systems.
Hypotheses include brisker movement going home than going to
work, entering a warm, dry car from a cold, wet street and, in
the Portland light rail case, caution alighting onto icy sidewalks.
Balance may also be better when ascending steps than when
descending.
The fastest flow time, 1.11 sec per passenger per single
stream, was observed on PATH boarding empty trains at Journal
Square station in the morning peak. These flow data are
consolidated and summarized by type of flow in Figure 4.7. The
results show that, in these averages, there is little difference
between the high-volume, older East Coast rail rapid transit
systems, and the medium-volume systemsnewer light rail and
rail rapid transit. Doorway steps approximately double times for
all three categories: mixed flow, boarding and alighting. Light
rail boarding up steps, with exact fare collection, adds an
average of almost exactly 1 sec per passenger.4
While most field data collection on doorway flow times is
from the peak periods, the opportunity was taken on BC
Transits rail rapid transit system to compare peak-hour with
off-peak and special event flows, as summarized in Figure 4.8.
Project resources prohibited significant data collection at special
events and outside peak periods. However, four field trips were
made to survey flows and loading levels on BC Transit. One
was before a football game, the second before a rock concert. In
both cases a single station handled 10,000 to 15,000 enthusiasts
in less than an hour. The other data collection trips surveyed a busy
4
No data were collected for light rail fare payment alighting down stepsa
situation unique to Pittsburgh.
Figure 4.6 Selection of rail transit doorway flow times (1995)
Figure 4.7 Summary of rail transit door average flow times
43
Figure 4.9 Mixed flow times versus door width

Figure 4.8 BC Transit doorway flow time comparisons
(1994-5)
suburban station in the early afternoon base (inter-peak) period.

The resultant data are contrary to the supposition that special
event crowds move faster and that off-peak flows are slower
than in the peak hour.
The results showed an increase in alighting flow times before
special events. However, loading densities were 20 to 30%
higher than during a normal peak hour. This higher level of
crowding, together with the fact that many special event
passengers are not regular riders, may account for the slower
alighting time. Separate BC Transit analysis(R27) has measured
car occupancy differences between normal peak-hour operation
and after service delays. Standing density increased from a mean
of 2.8 passengers per m2 to 5 passengers per m2. The equivalent
standing space occupied declined from 0.36 m2 per passenger to
0.2 m2 per passenger (3.9 to 2.2 sq. ft. per passenger).
Off-peak flows are invariably mixed. The BC Transit off-peak
data, an average of 21 trains over a 2-hour period, show faster
movement than comparable peak hour mixed flows. However,
these data are insufficient to draw firm conclusions.
4.4.3 EFFECT OF DOOR WIDTH ON

PASSENGER FLOW TIMES
Figures 4.9, 4.10 and 4.11 plot the relationship between flow
times in seconds per passenger per single stream against door
width. A variety of statistical analyses failed to show any
meaningful relationship between door width and flow time. The
only conclusion can be that, within the range of door widths
observed, all double-stream doors are essentially equal.
Field notes show that double-stream doors frequently revert to
single-stream flows and very occasionally three passengers will
move through the doorway simultaneously when one is in the
middle and two moveessentially sidewayson either side. At
some width below those surveyed a doorway will be effectively
single stream. At a width above those surveyed a doorway will
Figure 4.10 Boarding flow times versus door width
routinely handle triple streams. There are no singleor triplestream doors on any modern North American rail transit vehicle
although they exist on AGT and in other countries. JR East in
Tokyo is experimenting with a quadruple-stream doorway
shown in Figure 4.12. Wide doors have been a characteristic of
the AEG5 C100 AGT used in many airports and on Miamis
Metromover. This four-stream 2.4-m (8-ft) door is shown in
Figure 4.13.
4.5 ANALYZING FLOW TIMES

Procedures must be developed that will translate station passenger volumes and flow times per passenger into total doorway
use times and then into dwell times. Other work has developed
5
Previously Westinghouse Electric Corporation.
44
Figure 4.14 Histogram of flow time
Figure 4.11 Alighting flow times versus door width
Figure 4.12 Quadruple-stream doorway in Tokyo
relatively simply linear regression formulae with slight

improvements in fit using quadratic terms and the number of
passengers remaining on-boarda relatively crude surrogate for
the level of doorway congestion. Most work in this area has
been restricted to limited amounts of data from a single system.
Linear regression would also be possible for the more
extensive data collected during this project. However an
examination of these data indicated that separate regression
equations would be required for each systemand even for
different stations and different modes, alighting, boarding and
mixed, on a single system. This is undesirable and unsuitable for
determining the capacity of new rail transit systems where
regional transportation models provide an estimate of hourly
passenger flow by station, from which dwell times must be
estimated.
tThe projects statistical advisory team pursued the goal of a
single regression formula for all systems with level loading,
accepting the need for variations between mainly alighting,
mainly boarding and mixed passenger flows. The result, in the
following sections of this chapter, involves relatively erudite
statistical analysis. The only satisfactory results required
logarithmic transforms. Readers may elect to skip the remainder
of this chapter. Section 7.5.3 in Chapter Seven offers simpler
methods to estimate station dwell times and presents the results
of the following work in a simplified manner. The computer
spreadsheet allows the calculations to be carried out without any
knowledge of the underlying methodologies.
4.5.1 DATA TRANSFORMATION
Figure 4.13 Quadruple-stream doorway, Miami Metromover
To
assess
the
distribution
of
the
flow
time
(seconds/passenger/single stream), the explicit outliers (5 zero
times and one time of 36.0) were removed. The histogram in
Figure 4.14 shows a clear skewing. In the next step logarithmic
transformations were made of the flow times to obtain a
normally distributed set of data.
This is achieved by a power transformation technique due to
Box and Cox, which raises the flow time to a power determined
by an algorithmic procedure. The procedure chooses the power
to get a best fit (i.e., minimize the residual sum of squares due
45
Table 4.4 Overall data set summary (seconds)
Table 4.5 System comparison summary ln(flow time

(secs))
Figure 4.15 Residual sum of squares
There are highly significant differences between the cities

(p<0.0001) which are enumerated in the following table. An x
indicates a difference significant at the 5 percent level between
the cities.
Table 4.6 Significant differences between systems
Figure 4.16 Expected flow time cumulative probabilities

versus observed cumulative probabilities (abscissa)
to error) in a typical regression. The results of these calculations

are shown in Figure 4.15.
This graph indicates that a power of -0.25 or 0 is appropriate.
For ease of interpretation a power of zero, which corresponds to
a natural logarithm (ln) transform, is preferable. Further
calculation shows that this transformation is statistically
warranted. Confirmation of this decision can be seen by
comparing the normal probability plots obtained from
regressions of flow time and ln(flow time) against time of day,
shown in Figure 4.16.
mean value of ln(flow time) between different levels of a

variable (e.g., by system).
RESULTS
Overall Descriptive statistics for the overall data set are as
follows: where SD or Std Dev = standard deviation, No. =
Number of observations or Cases, ln = natural logarithm.
4.5.2 COMPARISONS
Box plots are the easiest way to visually compare the natural log
transformed flow time data between cities, time of day, loading
levels and event types. These plots enable the researcher to
quickly compare the central values (the mid box horizontal line
is the median) and gauge the spread of the data (the box
represents the interquartile range; i.e., the top is the 75th
percentile and the bottom is the 25th percentile).
Analysis of variance is used to examine differences in the
City/system comparison In this comparison all data are used

and the descriptive statistics for the eight systems are as follows
(Table 4.5): There are highly significant differences between the
cities (p<0.0001), which are enumerated in the Table 4.6. An x
indicates a difference significant at the 5 percent level between
the cities.
Alighting/boarding comparison All trains with greater than or
equal to 70% boarding passengers were declared to be boarding
46
Table 4.7 Alighting/boarding comparison
times, attention was restricted to the Vancouver Sky Train

(Table 4.10). The special event log flow time was significantly
(p<0.0001) lower than that during normal peak times. Figures
4.17 through 4.21 show the comparison box plots with the
following key.
The mean natural log of the flow time was significantly

(p<0.0001) less for alighting.
Table 4.8 Time of day comparison ln(flow time (secs))
The morning mean natural log of the flow time was mildly
significantly (p=0.02) higher than that in the afternoon.
Table 4.9 Loading level comparison ln(flow time (secs))
There were significant differences in the mean natural log of the

flow times between each pair of loading levels (p<0.05).
Table 4.10 Event time comparison ln(flow time (secs))
The special event log flow time was significantly (p<0.0001)

lower than that during normal peak time
and similarly those with greater than or equal to 70% alighting

passengers were declared to be alighting. This reduced the data
set to 1047 cases with descriptive statistics as follows (Table
4.7): The mean natural log of the flow time was significantly
(p<0.0001) less for alighting.
Time of day comparison All data were used in comparing am
and pm natural log flow times. The descriptive statistics are as
follows (Table 4.8): The morning mean natural log of the flow
time was mildly significantly (p=0.02) higher than that in the
afternoon.
Loading level comparison In order to have a homogeneous
dataset for comparing the effect of boarding levels, attention
was restricted to the SF Muni datasets. The following
descriptive statistics were calculated (Table 4.9). There were
significant differences in the mean natural log of the flow times
between each pair of loading levels (p<0.05).
Event Time Comparison In order to have a homogeneous
dataset for the comparison of the normal and special event
4.5.3 PREDICTION OF DOOR MOVEMENT

TIME USING BOARDING AND ALIGHTING
Preliminary regressions indicate that it is preferable to use the
natural logarithm of the door movement (DM) time. This is
illustrated in Figure 4.22, where the normal plot for the
transformed DM time is much closer to the line of identity, that
indicates normality. So, as with the flow time, the natural
logarithm of the door movement time is modeled, and the
resulting prediction is transformed back to the raw scale by
exponentiation. There is evidence (p=0.02) that separate fits are
warranted for mainly boarding (i.e. > 70% boarding), mainly
alighting (i.e. > 70% alighting) and mixed.
A number of parameterizations and combinations of the two
independent variables, number boarding (B) and number
alighting (A) are possible. The coefficients of determination for
the various models are shown in the following table. The
coefficient represents the proportion of variation in the data that
is explained by the model. In addition to these
parameterizations, the natural logarithm of the numbers
boarding and alighting were considered, and dummy variables
were used to model the levels resulting from a discretization of
the variables. However, these latter approaches did not provide
better fits than those above and so were not considered further.
The models were applied to the overall dataset and the three
mutually exclusive subsets of mainly boarding (i.e. > 70%
boarding), mainly alighting (i.e. > 70% alighting) and mixed;
results are shown in Table 4.11. From the table, it can be seen
that there are gains of up to 16% in the proportion of variation
explained by considering separate models for the subsets of
mainly boarding, mainly alighting and mixed. The gains in
considering more complex models than the simple additive
linear model (Model 1) are less clear.
There is little gain from introducing a term for the interaction
between the number boarding and the number alighting as in
model 2. However, there is an approximate gain of 10 percent,
resulting from the introduction of quadratic terms in model 3,
but no further gain from adding an interaction to this as in model
4. Similarly, there is no gain from higher order terms and
interactions, which also tend to make the prediction more
unstable. Hence the quadratic model (Model 3) is chosen as the
best fit, explaining 50% to 80% of the variation in the data.
Residual plots from the regression with this quadratic model
show an inverse fanning indicating that the residuals are
inversely proportional to the logarithms of the flow times. While
this could be transformed toward an identical error structure, in
the interests of parsimony, no reparameterization of the
logarithm of the flow time is attempted. The Durbin-Watson
statistic ranges between 1.3 and 1.6 indicating significant firstorder positive auto correlation among the residuals and so
standard errors for parameters and associated tests must be
viewed with some caution.
47
Figure 4.17 City/company comparison
Figure 4.20 Loading level comparison
Figure 4.18 Alighting/boarding comparison
Figure 4.21 Event time comparison
Figure 4.19 Time of day comparison
KEY
FT = Flow Time = the time in seconds for a single
passenger to move through a single-stream doorway
DM Time = Doorway Movement Time, the time in
seconds a single doorway is used for all continuous
passenger movements during a single dwell
A = number of passengers alighting and;
B = number of passengers boarding through a single
stream level loading rail transit car doorway
SN = number of standing passengers on-board the
surveyed car at the end of the dwell
48
Table 4.13 R2 data for tested models 5-7
Figure 4.22 Expected cumulative probabilities versus

observed cumulative probabilities of door movement time
and ln(door movement time)
Table 4.14 Doorway movement regression results, model 5
Table 4.11 R2 data for tested models 1-4
4.5.4 PREDICTION OF DOORWAY

MOVEMENT TIME USING NUMBER
BOARDING AND ALIGHTING PLUS THE
NUMBER STANDING
Table 4.12 Flow time regression results for model 3
The final regression models are presented in Table 4.12. All

coefficients are highly significant (p<0.001), except for A2 in the
mainly boarding dataset (p=0.2), and B2 (p=0.6) in the mainly
alighting dataset. Expressed as equations these are
ln(flow time overall) = 1.514 + 0.0987B + 0.0776A - 0.00159B2
- 0.000985A2
ln(flow time mainly boarding) = 1.380 + 0.124B + 0.0722A
- 0.00214B2 - 0.000857A2
ln(flow time mainly alighting) = 1.440 + 0.0979B + 0.0922A
- 0.00103B2 - 0.00116A2
2
ln(flow time mixed) = 1.368 + 0.112B + 0.0948A - 0.00225B

- 0.00184A2
The above quadratic model for the logarithm of the DM time

was augmented with the number standing standardized for the
floor area of the car (SN) to give model 5. Models 6 and 7
introduce quadratic terms in SN and its interactions with B & A.
Data from BART, MUNI and PATH were not used, thus
reducing the car numbers to half of those in the previous section.
Table 4.13 presents the coefficients of determination for these
models. In comparing these models to model 3 of the previous
section, there appear to be gains for the mainly boarding and
mixed models. However, there is no point in considering more
complex models than model 5 which is linear in SN. The
residual analyses show similar characteristics to the model
without the standardized number standing, so once again all
standard errors must be viewed with some caution. The final
regression models are presented in Table 4.14. All regression
coefficients are highly significant (p<0.001) except for B
(p=0.006), B2 (p=0.6) and SN (p=0.009) in mainly alighting
dataset. Expressed in equation form the models are
ln(flow time overall) = 1.412 + 0.0845B + 0.0890A - 0.00131B2
+ 0.00149A2 + 0.0460SN
ln(flow time mainly boarding) = 1.0724 + 0.124B + 0.104A
- 0.00194B2 - 0.00153A2 +
0.0782SN
49
ln(flow time mainly alighting) = 1.302 + 0.147B + 0.105A
- 0.00511B2 - 0.00165A2
+ 0.653SN
ln(flow time mixed) = 1.363 + 0.106B + 0.0864A - 0.00235B2
- 0.00159B2 + 0.0563SN
where B and A are the numbers boarding and alighting and SN is
the number standing normalized for floor area.
This model, with examples, is demonstrated in the computer
spreadsheet. The model has limitations and becomes inaccurate
with values of A or B > 25.
4.5.5 PREDICTION OF DWELL TIME FROM

DOORWAY MOVEMENT TIME
As shown in Figure 4.23 it is desirable to transform the dwell
time using natural logarithms, since the normal plot is
considerably straighter, indicating a progression toward
normality. The dwell time is modeled using its natural logarithm
and exponentiated back to the raw scale. Examination of
interaction terms shows no evidence (p=0.5) of a need to
consider separate predictions for the automatic systems (BART
and Vancouvers Sky-Train). The coefficient of determination
has a value of 0.34 with a linear model and there is no gain
evident from considering quadratic terms.
Residual analysis indicates an inverse fanning that will not be
corrected for so as to keep the model simple. However, the
Durbin-Watson statistic is 1.2 indicating strong positive serial
auto correlation, so that all standard errors and associated tests
must be viewed with some caution. The final regression model
for the natural logarithm of the dwell time is shown in Table
4.15. It is noted that this relationship is not as strong as those in
Figure 4.24 Scatterplot of ln(dwell time) versus DM time
Table 4.16 Mean doorway movement and dwell times

(with standard deviations) for all data sets of selected
systems (s)
the previous section. The association is displayed in the scatterplot of Figure 4.24. The mean dwell and DM times, together
with their standard deviations, are displayed in Table 4.16.
4.5.6 ESTIMATING THE CONTROLLING

DWELL
Figure 4.23 Expected cumulative probabilities (ordinates)

versus observed cumulative probabilities (abscissa)
Table 4.15 Modeling dwell time on

doorway movement time
It is usually the longest dwell time that limits the capacity of a

rail transit system. This controlling dwell is determined at the
most heavily used doorway on the peak-15-min train with the
highest loading and is typically at the busiest station on the line
being examined. Occasionally the controlling dwell may be at
other than the busiest station on a line. This can be due to speed
restrictions that increase the other headway components at this
station or to congestion that increases the passenger doorway
movement timefor example platform congestion due to
inadequate platform exits, platform obstructions or, at stations
with multiple routes, due to passengers waiting for other trains.
There are a number of possible methods for estimating the
controlling dwell. In essence, all these methods seek to determine
50
an upper bound for the dwell time below which the bulk of the
population falls.
Examples of these methods, comparison with actual field data
and suggestions of the most appropriate method to use in
different circumstances are discussed in the application chapter:
Chapter Seven, Grade Separated Rail Capacity Determination,
Section 7.5.3 Determining the Dwell Time.
Table 4.17 Controlling dwell data limits (seconds)
ALLES METHOD(R02)
This approach focuses on providing a prediction interval for the
mean. In other words, in the long run all sample means should
fall within these limits 95% of the time. However, it is really a
prediction for a typical dwell time that is desired as this will
provide the reference limit or bound that is required. As such,
Alles formula seems inappropriate. Moreover it is a
nonstandard approach which consists of adding the 95%
confidence widths for the distribution of the sample mean and
the sample standard deviation. The rationale for adding the
confidence width of the sample standard deviation is not clear.
The prediction interval for the sample mean is a random
variable itself, and as such, it is possible to construct a
confidence interval around it, which may have been the intent. If
one were considering the limits for the dwell time of a typical
new train, then the variance of the upper prediction limit is
approximately 3s2/n where s is the sample standard deviation
and n is the sample size. As Alles method considers a limit for
the mean and not a typical unit, it is not considered further.
MEAN PLUS STANDARD DEVIATIONS

This is the traditional approach derived from control theory. It
provides a prediction interval for a new train as opposed to one
for the mean of all trains. Since it is maximum capacity that is
the ultimate objective, only the upper limit is of interest.
A dwell based on the statistical mean plus one standard
deviation ensures that 83% of the observed data would be equal
to or less than this value. A dwell based on the statistical mean
plus two standard deviation ensures that 97.5% of the
observed data would be equal to or less than this value.
Both one and two standard deviations have been used in other
work. In either case it is necessary to ensure that the calculated
controlling dwell contains sufficient operating margin or
allowance to compensate for minor irregularities in operation.
With the addition of one standard deviation some additional
allowance for operational irregularities is necessary. With two
standard deviations the need for any additional allowance is
minor or unnecessary
DWELL TIME PLUS AN OPERATIONAL ALLOWANCE

OR MARGIN
In many situations, particularly new systems, sufficient data is
not available to estimate the dwell standard deviation over a one
hour or even a 15 min peak period. In these cases or as an

alternate approach an operational allowance or margin can be
added to the estimated dwell time due to a specific volume of
passenger movements. The figures for the controlling dwell are
listed in Table 4.17 using both the mean plus one or two
standard deviations and the mean plus operational allowances of
15 and 25 sec.
Chapter Six, Operating Issues, discusses the need for, and
approaches to, estimating a reasonable operating margin.
Application Chapter Seven, Grade Separated Rail Capacity
Determination, Section 7.5.4, discusses how to select an
operating margin in specific cases.
4.6 SUMMARY
The analysis in this chapter has produced methodologies
whereby the passenger doorway flow time can be determined
from four logarithmic modelsoverall, mainly boarding,
mainly alighting and mixed flowusing as input the number of
passenger movements, without reference to a specific mode,
system or city.
A fifth model, also logarithmic, but considerably simpler,
determines dwell time from passenger doorway flow time.
Three alternative methods are then examined to convert the
resultant dwell time to the controlling dwell time. The first two
methods, traditional dwell plus two standard deviations, which
most closely matched the field data, and Alles method both
require information on dwells over the peak hour. This
information is not readily available when trying to estimate the
capacity of new or modified rail transit systems, leaving the
third method, adding an estimated operating margin to the
calculated maximum dwell.
These methodologies are deployed in Chapter Seven, Grade
Separated Rail Capacity Determination and in the spreadsheet
as one of several complete methods to calculate system capacity.
51
5. Passenger Loading Levels

5.1 INTRODUCTION
Establishing the loading level of rail transit is usually the final
step in determining capacityand one of the most variable.
After the maximum train throughput has been calculated from
the inverse of the sum of signaling separation time, dwell time
and operating margin, then capacity is based only on train length
and loading level.
It is important to remember the feedback processes; that train
length significantly changes the signaling separation time and
that loading levels affect dwell times.
The existing loading levels on North American rail transit
vary from the relaxed seating of premium service (club cars)
operated on specific trains of a few commuter rail lines to the
densest loading of an urban subway car in Mexico Citya
range of 1.5 to 0.17 m2 per passenger (16 to 1.8 sq ft).
This wide range is more than eight to one. A more normal
loading level range, discounting Mexico City and commuter rail,
is two or three to one. This range makes the precise
determination of loading level difficult. The main factor is a
policy issue, the question of relative comfortheavily
restrained by economic issues.
Notwithstanding Torontos subway and PATCOs
Lindenwold line, the first new rail transit network in North
America in the last half century was BART. In the early 1960s,
planning for this networkmore a suburban railway than an
inner-city subwaywas based on the provision of a seat for
every passenger. Subsequently economic reality has forced
acceptance of standing passengers, particularly for shorter trips
in San Francisco and through the Transbay tube. Nevertheless,
BART remains an example of a system that was designed to,
and succeeded in, attracting passengers from alternate modes.
More so now, entering the twenty-first century, than 30 years
ago, rail transit is being planned as an alternative to the
automobile. While additions to existing systems can be expected
to follow existing standards, new systems have to determine
their service standards. The principal standards include speed,
frequency of service at peak and off-peak timesoften termed
policy headwaysand loading levels. Schedule speed is fixed
when the alignment, station spacing and equipment
specifications are set; headways are usually closely tied to
demand, although unmanned trains, as used on Vancouvers
SkyTrain and Miamis Metromover, make short, frequent trains
over much of the day more affordable. Loading level is the
remaining variable. Loading levels and headways interact as
more comfortable standards require either longer or more
frequent trains.
Demery(R22) states:
Long before crowding levels......reached New York
levels, prospective passengers would choose to travel
by a different route, by a different mode, at a different
time, or not at all.
and
Outside the largest, most congested urban areas, the
level of crowding that transit passengers appear
willing to tolerate falls well short of theoretical
design or maximum vehicle capacity.
These are important issues to consider in establishing loading
standards.
In the next section, existing loading standards are reviewed.
The remainder of the chapter determines a range of loading
standards that can be applied in specific circumstances for each
mode.
It is possible to determine the interior dimensions of a rail
transit vehicle; subtract the space taken up by cabs, equipment
and, for low-loading light rail, stairwells; then assign the
residual floor space to seated and standing passengers on the
basis of selected densities. This approach is one of several
followed in this chapter. However, the recommended method is
simply to apply a passenger loading per unit of train length.
5.2 STANDARDS
A 1992 New York City Transit policy paper, Rapid Transit
Loading Guidelines,(R48) gives the loading and service standards
that have been applied, with minor modifications, to the New
York subway system since 1987. The guidelines provide for
slightly more space per passenger than those in effect until
1986. Modifications have allowed for a relaxation in the
nonrush hour passenger loading guideline to allow for the
operation of short trains.
The loading guidelines were established from test loadings of
different car types, loading surveys of revenue service at the
peak load point and comparisons with the policies of other rail
transit operators. Additional concerns such as passenger
comfort, dwell time effects, uneven loading within trains, and an
allowance for slack capacity in the event of service irregularities
and fluctuations in passenger demand were also considered. A
rush hour standard of 3 sq ft per standing passenger (3.6
passengers per m2) was generated from this work. The policy
recognizes that this condition is only to be met at the maximum
load point on a route and so is effective for only a short time and
small portion of the overall route. For comparison, the agencys
calculations of the maximum capacity of each car type are based
on 6.6 - 6.8 passengers per m2.
Figure 5.1 compares the loading standards of the older North
American subway systems. NYCT standards for loading in the
nonrush hours are more generous, with a seated load at the
maximum load point being the general standard. If this would
require headways of 4 min or less, or preclude operation of short
trains, a standard of 125% of seated capacity applies. This
52
Table 5.2 Passenger space on selected US systems(R22)
Figure 5.1 Scheduled loading guidelines (passengers/m2)
Table 5.3 Changes in NYCT peak-hour car loading(R22)
Figure 5.2 New York loading guidelines (passengers/m2)
Table 5.1 New York policy service levels
consideration of passenger comfort also extends to rush hour

service on lines where the headway is longer than 4 min. In
these cases a sliding scale is used to ensure lower standing
densities on routes with longer headways, as shown in Figure
5.2. Minimum headways for each day and service period are
shown in Table 5.1. The NYCT standard of 3.6 passengers per
m2 can be compared with the average occupancy into the CBD
over the peak period as shown in Table 5.2. Table 5.3 tabulates
and compares daily and peak-hour ridership and passengers per
vehicle for 19 New York CBD trunks for 1976 and 1991. This
decrease in NYCT car loadings partly reflected the improvement
Average and Median include additional data sets.
in service standards of 1987, among other factors. Several

trunks continue to operate at or near capacity.1
Care should be taken in comparing and applying the service
standards with hourly average loadings. Service standards are
usually based on the peak within the peak15 min or less.
A loading diversity factor equating 15-min and peak-hour
flows was introduced in Chapter One, Rail Transit In North
America. Section 5.6 of this chapter discusses the issues of loading diversity, provides data on existing factors by system and
mode, and recommends factors for use in capacity calculations.
The loading diversity factor for New York trunk routes, shown
1
Similar comparisons can be made for other cities and earlier years using
data from this report and from the TRBs Highway Capacity Manual,
Chapter 12 and appendices. Ridership and loading level information in the
HCM are based on data to 1976 plus some historic data.(R67)
53
in Figure 5.3, ranges from 0.675 to 0.925 with an average of
0.817. This diversity must be taken into account to determine
peak-hour capacity from a given service standard. NYCTs
standard of 3.6 passengers per m2 over the peak-within-the-peak
becomes 3.6 0.82 or 2.95 (3.65 sq ft per passenger) on
average, over the peak hour.
Outside New York the peak-within-the peak tends to be more
pronounced and the peak-hour diversity factor is lower.3 In part
this is due to the long established Manhattan program to stagger
work hours and the natural tendency of passengers to avoid the
most crowded periodparticularly on lines that are close to
capacity.
Space occupancy during the peak period on other North
American rail transit systems varies widely from below 0.3
passengers per m2 (3.2 sq ft) to over 1.0 m2 (11 sq ft) on some
commuter rail lines, as shown in Figure 5.4. Note that the
highest capacity entry (labeled NYCT) represents two tracks
that combine local and express service.
In analyzing this data Pushkarev et al.(R51) suggest a standard
of 0.5 m2 (5.4 sq ft) per passenger. This will be discussed in the
next section. In addition to standards or policies for the
maximum loading on peak-within-the-peak trains and for
minimum headways (policy headways) at off-peak times, some
operators specify a maximum standing time. This is more often
a goal rather than a specific standard20 min is typical.
Commuter Rail Loading levels for commuter rail are unique

and uniform. Although standing passengers may be accepted for
short inner-city stretches or during times of service
irregularities, the policy is to provide a seat for all passengers.
Capacity is usually cited at 90 to 95% of the number of seats on
the train.
5.3 SPACE REQUIREMENTS

The surveyed literature contains many references to passenger
space requirements. The Batelle Institute(R12) recommends
comfort levels for public transport vehicles. The passenger
standing density recommendations are
COMFORTABLE 2-3 passengers per m2
UNCOMFORTABLE 5 passengers per m2
UNACCEPTABLE >8 passengers per m2
In contrast, Pushkarev et al.(R51), suggesting gross vehicle floor
area as a readily available measure of car occupancy,
recommends the following standards:
ADEQUATE 0.5 m2provides comfortable capacity per
passenger space
TOLERABLE WITH DIFFICULTY 0.35 m2lower limit
in North America with some touching
TOTALLY INTOLERABLE 0.2 m2least amount of
space that is occasionally accepted
Batelle(R12) also provides details of the projected body space of
passengers in various situations. The most useful of these for
3
Shown in Chapter One, Figures 1.4 and 1.6.
Figure 5.3 15-min peak-within-the-peak compared to full

peak-hour ridership on New York subway trunks
54
Table 5.6 International transit space use (R30)
Figure 5.4 Peak-hour space occupancyall U.S. systems(R51)

Figure 5.5 Passengers per length of car versus % seated
Table 5.4 Passenger space requirements (R12)
Table 5.5 Passenger space requirements (R30)
rail transit capacity are shown in Table 5.4. The tight double
seat corresponds closely to the North America transit seating
minimum of 34- to 35-in.-wide double seats on a 27- to 33-in.
pitch (0.88 m by 0.76 m) 3.6 sq ft or 0.33 m2 per seat.
Jacobs et al. (R30) contains a comprehensive section on vehicle
space per passenger, stating that while 53% of U.S. rapid transit
lines enjoyed rush hour loadings of 0.5 m2 per passenger or
better, the space requirements shown in Table 5.5 are recom-
mended and actual values for the stated conditions. The report is
one of the few to discuss the diversity of standing densities
within a carhigher in doorways/ vestibules, lower in aisles and
at car ends (unless the car has end doors). Table 5.5 is
particularly interesting in that the design space allocation for
light rail is slightly lower than for heavy rail.
Klopotov(R32) cites typical average peak-hour space
requirements from an international survey (Table 5). Lang and
Soberman(R39) discuss seating provisions relative to
compromises between capacity and comfort. They suggest that
all rapid transit cars are substantially similar in width. The
report compares passengers per square foot with the percentage
seated. This ranges from 0.3 passengers per square foot with
50% seated to 0.6 passengers per square foot with 15% seated.
This is then translated into passengers per linear foot of train, as
shown below in Figure 5.5. The maximum vehicle capacity is 4
passengers per linear footapproximately 2.5 square feet per
passenger. Lang and Soberman also discuss the importance of
ease of ingress and egress, recommending minimum distances
between seats and doorways and discouraging three abreast
seating. Comfort levels are discussed relative to smoothness of
operation and the issue of supply and demand. Where systems
are oversubscribed and few attractive alternate forms of
transportation are available, high levels of crowding will be
tolerated. Where systems wish to attract passengers, higher
comfort levels, i.e., less crowding, are desirable.
Levinson et al.(R43) and also the Transportation Research
Boards Highway Capacity Manual(R67) introduce the concept of
loading standards A through F (crush) similar to the alphabetized
level of service for road traffic. The suggested schedule design
capacity is 2.8 to 3.3 passengers per m2, 25% below the crush
capacity. The peak-hour factor is discussed for 15-min peak-
55
within-the-peak. A range of 0.70 to 0.95 is suggested,
approaching 1.0 in large metropolitan areas.
Vuchic(R71) suggests passenger space requirements of 0.30 to
0.55 m2 per seat and 0.15 to 0.25 m2 per standee. Vehicle
capacity in passenger spaces per vehicle is shown as:
Equation 5-1
Where:
= vehicle floor area loss factor for walls

Ag = gross vehicle floor area
Al = vehicle floor area used for cabs, stairwells
and equipment
m = number of seats
= floor area per seat
= floor area per standing passenger
Young(R76) discusses a wide range of topics dealing with

passenger comfort. He cites the typical transit vehicle as
allowing 0.40 m2 (4.3 sq ft) per seated passenger and 0.22 m2
(2.4 sq ft) per standing passenger. The seating ratio is tabulated
for a range of North American and European heavy rail and
light rail systems. Heavy rail ranges from 25% to 100% seated
and light rail from 40 to 50% in North America to 20 to 44% in
Europe. Minimum seating pitch is recommended as 0.69 m (27
in.), 0.81 m (32 in.) to a bulkhead.
Several reports suggest vehicle passenger capacity can be
stated as a multiple or percentage of the number of seats.
Chapter 12 of the Highway Capacity Manual (R67) develops a
measure of seated and total passengers per linear foot of car
length, introduced in section 5.5 of this chapter.
Recommendations for a range of loading standards are
developed in later sections of this chapter and applied in
Chapter
Seven,
Grade
Separated
Rail
Capacity
Determination, and the reports spreadsheet.
Wheelchairs There was no reference to wheelchair space

requirements in the literaturemuch of which predates the 1991
Americans with Disabilities Act. Although wheelchairs come in
several sizes, a common space allowance is 0.55 m2 (6 sq ft),
more for electric chairs and those whose occupants have a
greater leg inclination, less for compact and sports chairs.
However, it is not the size of the chair that is a concern as
much as the maneuvering and stowage space. Typically a chair
occupies the space of a double seat whose seat squab folds up.
Restraints and seat belts may be provided but the smoothness of
the ride allows most rail transit systems to omit these. In certain
vehicle layouts additional seats have to be removed to allow
access to the designated wheelchair location.
In optimum designs wheelchair space occupancy should be
assigned as the space of a double seat0.8 m2 (8.6 sq ft) with a
50% increase considered as an upper limit1.2 m2 (13 sq ft) No
further allowance is necessary for maneuvering space as this
will be occupied by standing passengers when circumstances
dictate.
In several rail transit vehicle designs, capacity has actually
increased with the removal of seats to provide a designated
space for wheelchairs, or, selectively, bicycles. Where the
designated space does not involve a fold-up seat the empty
space is frequently used by standing passengers or to store

baggage, baby strollers etc. Providing locations to store such
potential obstacles away from doorways and circulation areas
can assist in reducing dwell times.
Wheelchair effects on dwell times are discussed in Chapter
Four, Station Dwells, and Chapter Eight, Light Rail Capacity
Determination.
5.4 VEHICLE CAPACITY

In estimating the capacity of a rail transit vehicle one of the
following approaches should be selected.
5.4.1 COMMUTER RAIL

Commuter rail capacity is based on the number of seats. Table A
3.5 in Appendix Three lists the dimensions and seating of all rail
transit vehicles in North America. A summary extracted from
this table is shown in Table 5.7. Commuter rail seating per car
ranges from a maximum of 185 to below 60 on certain club cars
and combination cars.5 Seats will be reduced where staff, toilet,
wheelchair, baggage or bicycle space is provided. The highest
seating densities use 3+2 seating. Although suitable for shorter
runs, 3+2 seating is not popular with passengers. The middle of
the three-seats is often under utilized and capacity should be
factored down accordingly by a suggested further 5%.
Table 5.7 Commuter rail vehicle summary data
Bi-level cars are sometimes designated as tri-levelas there is an

intermediate level at each end over the trucks.
Not tabulated. Cars with baggage space, crew space or head-end (hotel)
power.
56
Table 5.8 Light Rail Equipment Summary
Table 5.9 Heavy rail equipment summary
Total passengers based on the agencys or manufacturers nominal crush load.
Commuter capacity should be calculated as 90 to 95% of the

total seats on a train, after allowing for cars with fewer seats due
to other facilities. Where there are high incremental passenger
loads for relatively short distancesfor example the last few
kilometers into the CBDa standing allowance of 20% of the
seats may be considered. However, this is unusual and standing
passengers should not normally be taken into account on

commuter rail.
5.4.2 EXISTING SYSTEMS

The vehicle capacity on existing systems should be based on
actual loading levels of a comparable service. Actual levels on
57
a specific system or line should be adjusted for any difference in
car size and interior layoutparticularly the number of seats
as outlined in section. If the average occupancy over the peak
hour is used then the loading diversity factor should be omitted.
If the higher peak-within-the-peak loading is used, then the
loading diversity factor should be applied to reach an hourly
achievable capacity.
Particular care should be taken in applying any passenger
loading level based on car specifications. The often cited total,
maximum, full or crush load does not necessarily represent a
realistic average peak hour or peak-within-the-peak occupancy
level. Rather it reflects the specifier or manufacturer applying a
set criteriasuch as 5 or occasionally 6 passengers per square
meterto the floor space remaining after seating space is
deducted. Alternately it can represent the theoretical, and often
unattainable, loading used to calculate vehicle structural strength
or the minimum traction equipment performance.
Tables 5.8 and 5.9 provide dimensions and capacity
information of selected, newer, heavy rail and light rail
equipment in North America.
Table A 3.5 in Appendix Three lists the dimensions and
seating of all rail transit vehicles in North America.
5.4.3 VEHICLE SPECIFIC CALCULATIONS

Detailed calculations of vehicle passenger capacity are possible,
however, given the wide range of peak hour occupancy that is
dependent on policy decisions, elaborate determination of
interior space usage is generally overkill. Reasonably accurate
estimation of vehicle capacity is all that is needed. The
following procedures offer a straight forward method.
Converting Exterior to Interior Dimensions
Rail transit vehicle exterior dimensions are the most
commonly cited. Where interior dimensions are not available,
or cannot be scaled from a floor plan, approximate interior
dimensions can be estimated.
Typically the interior width is the exterior width less the
thickness of two walls0.2 m (8 in.). Heavy rail
configurations are most commonly married pairs with one
driving cab per car. The typical exterior length is quoted over
the car anticlimbers. Although cab sizes vary considerably,
the interior length can be taken to be 2.0 m (6.7 ft) less than
the exterior length. This reduction should be adjusted up to
2.5 m if the exterior dimension are over the couplers and
down to 1.5 m if only half width cabs are used, or 0.5 m if
there is no cab.
Beware of rare pointed or sloping car ends which require
this deduction to be increased. Curved side cars are measured
from the widest pointwaist levelallowing seats to fit into
the curve and so increasing the aisle width. This maximum
waist width should be used, not the width at floor level.
The first step after obtaining the interior car dimensions is to
determine the length of the car side that is free from doorways.
Deducting the sum of the door widths, plus a set-back allowance
of 0.4 m (16 in.)7 per double door, from the interior length gives
the interior free wall length.
7
A lower set-back dimension of 0.3 m (12 in.) may be used if this permits
an additional seat/row of seats between doorways.
Seating can then be allocated to this length by dividing by the

seat pitch:
0.69 m (27 in.)8 for transverse seating
10.43 m (17 in.) for longitudinal seating
The result, in lowest whole numbers9, should then be multiplied
by two for longitudinal seating or by 3, 4, or 5, respectively, for
2+1, 2+2 or 2+3 transverse seating. The result is the total
number of seats. A more exact method would be to use the
specific length between door set-backs. Articulated light rail
vehicles should have the articulation width deducted. Four seats
can be assigned to the articulation, if desired.
The floor space occupied by seats can then be calculated by
multiplying transverse seats by 0.5 m2 (5.4 sq ft) and
longitudinal seats by 0.4 m2 (4.3 sq ft). These areas make a
small allowance for a proportion of bulkhead seats but otherwise
represent relatively tight and narrow urban transit seating. Add
10 to 20% for a higher quality, larger seat such as used on
BART.
The residual floor area can now be assigned to standing
passengers. Light rail vehicles with step wells should have half
the step well area deducted. Although prohibited in many
systems, passengers will routinely stand on the middle step,
squeezing into the car at stops if the doors are treadle operated.
Articulated light rail vehicles should have half the space
within the articulation deducted as unavailable for standing
passengers, even if the articulation is wider. Many passengers
choose not to stand in this space.
Standing passengers can be assigned as follows:
5 per square meter (0.2 m2, 2.15 sq ft per passenger)an
uncomfortable near crush load for North Americans10 with
frequent body contact and inconvenience with packages
and brief cases; moving to and from doorways extremely
difficult.
3.3 per square meter (0.3 m2, 3.2 sq ft per passenger)a
reasonable service load with occasional body contact;
moving to and from doorways requires some effort
Increase to 0.8 m (32 in.) for seats behind a bulkhead

For more accurate results the sidewall should be divided into the lengths
between each set of doors (and, when appropriate, between the door and
any articulation) and checked, or adjusted, to ensure that an integer of the
seat pitch is used. The computer spreadsheet carried this out by dividing
the interior free wall length by the number of doorways plus one. The
number of integer seat pitches in each space is then determined and used
to calculate the total vehicle seating. The appropriate seat pitch is used
automatically, 0.43 m for N=2, 0.69 m for N>2.
However, this approach can result in the seating changing radically with a
small change in vehicle length, articulation length or door width, any of
which are sufficient to add or remove a row of seats between each set of
doors. On a four door car with 2+2 seating this results in the seating
adjusting up or down by 20 seats at a timefive rows of four seats.
Neither Equation 1.3 nor the computer spreadsheet can substitute for a
professional interior design, which can optimize seating with a
combination of transverse and longitudinal seats. Other design criteria can
also be accommodated, including the provision of wheelchair spaces and
maximizing circulation space around doorways.
10
Loading levels of over 6 passengers per square meter are reported on
Mexico Citys metro, lines 1 and 3. These are a unique exception in North
America.
9
58
Figure 5.6 Schematic of rail car showing the dimensions of Equation 5.2
2.5 per square meter (0.4 m2, 4.3 sq. ft. per passenger)11
a comfortable level without body contact; reasonably easy
circulation, similar space allocation as seated passengers.
The middle level above is slightly relaxed from the often stated
standard of four standing passengers per square meter. So-called
crush loads are frequently based on 5 or 6 passengers per square
meter, the latter being more common in Europe. Asian standards
for both maximum and crush loads reach 7 or 8 standing
passengers per square meter.
The resultant sum of seated and standing passengers provides
a guide for the average peak-within-the-peak service loading
level for the specific vehicle. Peak-hour loading should be
adjusted by the vehicle loading diversity factor. No specific
allowance has been made for wheelchair accommodation or for
reduced standing densities away from doorways. The above
range of standing densities makes such small adjustments
unnecessary. Cars intended for higher density loading should
have a greater number of doors. Space inefficiencies at the
extremities of a car are unavoidable unless the London
Underground arrangement of doors at the very end of each car is
adopted.
The above process can be expressed mathematically as
N = seating arrangement
2 for longitudinal seating
3 for 2+1 transverse seating
4 for 2+2 transverse seating
5 for 2+3 transverse seating 13
Sa = area of single seat
0.5 m2 (5.4 sq ft) for transverse
0.4 m2 (4.3 sq ft) for longitudinal
Dn = number of doorways
Dw = doorway width
Sb = single set-back allowance
0.2 m (0.67 ft)or less
Sw = seat pitch
0.69 m (2.25 ft) for transverse
0.43 m (1.42 ft) for longitudinal
Figure 5.6 shows these car dimensions.
The equation can be worked in either meters or feet. An
expanded version of this equation is included on the computer
spreadsheet. The spreadsheet calculation automatically applies
the Sw seat pitch dimension through an IF statement acting on N,
the seating arrangement factor, using the longitudinal dimension
if N=2.
Offset Doors A small number of rail vehicle designs utilize
offset doors. These do not merit the complexity of a separate
equation. Provided that each side of the car has the same number
of doors Equation 5.2 will provide an approximate guide to
vehicle capacity with a variety of seating arrangements and
standing densities.
Equation 5-1212
where
11
Vc
Lc
La
Ws
Wc
Ssp
= vehicle capacitypeak-within-the-peak
= vehicle interior length
= articulation length for light rail
= stepwell width (certain light rail only)
= vehicle interior width
= space per standing passenger
0.2 m2 (2.15 sq ft) maximum
0.3 m2 (3.2 sq ft) reasonable
0.4 m2 (4.3 sq ft) comfortable
This upper level is a peak-within-the-peak occupancy level for standing

passengers. Over the peak hour, it corresponds closely to Pushkarev(R50)
and Jacobs(R30) estimates of a United States rush-hour loading average of
0.5 m2 per passengerboth seated and standing. It also corresponds to
Pushkarev and Batelles(R12) recommendation for an adequate or
comfortable loading level.
12
= expression rounded down to nearest integer (whole number).
Fast Alternative A fast alternative method is to divide the gross

floor area of a vehicle (exterior length x exterior width) by 0.5
m2 (5.4 sq ft) and use the resultant number of passengers as the
average over the peak hourwithout applying a vehicle loading
diversity factor. An average space over the peak hour of 0.5 m2
(5.4 sq ft) per passenger is the U.S. comfortable loading level
recommended in several reports and is close to the average
loading on all trunk rail transit lines entering the CBD of U.S.
cities.
5.4.4 RESULTS OF THE CALCULATION

Light Rail Applying the calculations of section produces passenger loading levels for typical light rail vehicles as shown in
13
2+3 seating is only possible on cars with width greater than 3 meters, not
applicable to light rail or automated guideway transit.
59
Table 5.10 Calculated light rail vehicle capacity
Table 5.11 Calculated heavy rail vehicle capacity
Table 5.10. Two articulated light rail vehicles are shown, the
common Siemens-Dwag car used in nine systems (with some
dimensional changes) and the largest North American light rail
vehicle used by the MTA in Baltimore. The resulting capacities
are for a generic version of these cars. Reference to Table 5.9,
Light Rail Equipment Summary, shows that the actual number of
seats in the Siemens-Dwag car varies from 52 to 72 while rated
total capacity varies from 96 to 201. This stresses the wide,
policy related, car capacity issue.
The calculation cannot encompass all options. However, the
calculation provides a policy surrogate in the form of the
allocated standing space,0.2, 0.3 or 0.4 m2 per passenger.
Seating should be adjusted accordingly. A need for high
standing levels would suggest longitudinal seats, low standing
levels, the 2+2 transverse seats.
Heavy Rail Applying the calculations of section produces

passenger loading levels for typical heavy rail vehicles as shown
in Table 5.11. Data is shown for a generic 23 meter heavy rail
car with variations of seating arrangements and standing space
allocations. Two data sets follow for the smaller cars used in
Vancouver and Chicago.
5.5 LENGTH
In this section the above calculations are converted to the
passengers per unit length method suggested by Lang and
Soberman(R39) and others, stratified into classes, then compared
with actual peak-within-the-peak loading levels of North
American rail transit. Given the variation in loading levels that
depend on policythe standing density used and seat spacing
this simplified method is appropriate in most circumstances. It is
the recommended method of estimating peak-within-the-peak
car capacity except for circumstances and rolling stock that are
out of the ordinary.
Light Rail Applying the calculations of section produces passenger loading levels for typical light rail vehicles as shown in
60
Figure 5.7 Linear passenger loading of articulated LRVs
three variations, 4 and 3 door versions, and transverse or

longitudinal seatingwith a range of 7.0 to 11.5 passengers per
meter of car length. The lower end of the range of seven to eight
passengers per meter lengthwith a standing space per
passenger of 0.4 to 0.3 m2is an appropriate range for higher
use systems. A lower figure of six corresponds closely with the
recommended quality loading of an average of 0.5 square meters
per passenger and is appropriate for a higher level of service on
new systems. In both cases a reduction by one should be used
for smaller, narrower cars.
These calculated linear loading levels can be compared with
actual levels on major North American rail transit lines shown in
Table 5.12 and summarized in Table 5.13.
Heavy Rail outside New York shows a level comparable with
the recommended comfortable level of 6 passengers per meter
of train length. New York is higher by some 25%, averaged over
11 trunk routes. Commuter rail, with most passengers seated,
has an average only 13% lower than the average of heavy rail
outside New York. Only two light rail lines are running close to
capacity and peak-within-the-peak ridership is not available for
these.
5.6 LOADING DIVERSITY
Figure 5.8 Linear passenger loading of heavy rail cars
Table 5.10 and as passengers per unit length in Figure 5.7. As

would be expected, the wider and longer Baltimore car has
proportionately higher loadings per meter of length. The typical
Siemens-Dwag car used on nine systems (with some
dimensional changes) has a range of 5.0 to 8.0 passengers per
meter of car length. The lower level of five passengers per meter
lengthwith a standing space per passenger of 0.4 m2
corresponds closely with the recommended quality loading of a
an average of 0.5 m2 per passenger.
Heavy Rail Applying the calculations of section 5.4.3 produces

passenger loading levels for typical heavy rail vehicles as shown
in Table 5.11 and, as passengers per unit length, in Figure 5.8.
As would be expected, the smaller and narrower cars in
Vancouver and Chicago have lower loadings per meter length.
The more generic 23-m-long cars used in over 12 North
American cities have a remarkably close data set for each of the
Passengers do not load evenly into cars and trains over the peak
hour. This unevenness is the diversity of passenger loading.
There are three different types of loading diversity: unevenness
of passenger loading within a car; unevenness of passenger
loading within cars of a train; unevenness of passenger loading
within peak-hour trains. The loading diversity factor developed
in this section essentially encompasses all three.
In individual cars, the highest standing densities occur around
doorways, the lowest at the ends of the cars. Several European
urban rail systems add doors, sometimes only single stream, at
the car ends to reduce this unevenness. London Transports
underground system is the most notable with this feature on
most rolling stock,14 except at car ends with a driving cab. The
end door on the low-profile cars are 0.75 m (2.5 ft) wide
compared to the main doors of 1.56 m (5.1 ft). These
exceptionally wide doors, with their 0.17 m (6.8 in) set-backs
often accommodate three streams of passengers.
No data exist to determine such loading diversity within a car
and the variations are accommodated in the average loadings of
the previous sections. It is important in cars designed for high
occupancies to minimize this effect by using wide aisles,
uncluttered vestibules and suitable hand holds that encourage
passengers to move into the extremities of a car. Very little
information was found on car interior design efficiency in the
literature search with the exception of Young(R76) Passenger
Comfort in Urban Transit Vehicles.
A second level of diversity occurs in uneven loading among
cars of a train. This second level is also included in the average
loading data of the previous sections and in the application
chapters. Cars that are closer to station exits and entrances will
be more heavily loaded than more remote cars. This inefficiency
can be minimized by staggering platform entrances and exits
14
Londons Docklands Light Railway does not have end doors.
61
Table 5.12 Passengers per unit train length, major North American trunks
Table 5.13 Summary of linear passenger loading (per meter)

Additional passenger loading per unit length data are compiled
in Tables 7.4, 7.5 and 7.6 of Chapter Seven.
between ends, centers and third points of the platforms. This is

not always possible or practiced. The busiest, most densely
occupied rail lines in North America, lines 1, 2 and 3 of Mexico
Citys metro all have stations with center entrances/exits. Even
so, relatively even loading occurs both here, and on rail transit
lines at or near capacity elsewhere, due to the duress factor that
encourages passengers to spread themselves along the platform
during heavily traveled timesor risk being unable to get on the
next arriving train.
Few systems count passengers by individual cars when these are
crush loaded. This is difficult to do with any accuracy and the
results differ little from assigning a set full load to each car
Figure 5.9 Vancouver, Broadway Station inbound peak-hour

passenger distribution between cars of train. October 27
1994, 50 trains, 12,173 passengers
of a fully loaded train. Data are available from two Canadian

properties.
BC Transit operates four car trains on headways down to 90
sec. Pass-ups are routine at the busiest suburban station, Broadway with an end and two third-point entrances/exits. The relative
loading of the four cars is shown in Figure 5.9. The main
entrance/exit is provided with escalators and lies between the
62
Figure 5.10 TTC Yonge Subway, Wellesley Station

southbound,
a.m.
peak-period
average
passenger
distribution between cars of train. Jan 11, 1995, 99 trains with
66,263 passengers
second and third cars of the train. While the second car is the
most heavily loaded, the third is the lightest loaded indicating
the influence of entrance/exit locations at other major stations.
There is no significant variation in the average loading
diversity between the peak hour and the peak-period both of
which remain within the range of +5% to -6%. The unbalance
for cars on individual trains ranges from +61% to -33%. The
uniformity of loading can be attributed to four factorsthe short
trains, wide platforms, close headways and dispersed
entrance/exit locations between the stations of this automated,
driverless system. The Toronto Transit Commissions Yonge
Street subway shows a more uneven loading between cars in
Figure 5.10. In the morning peak period the rear of the train is
consistently more heavily loaded reflecting the dominance of the
major transfer station at Bloor with the interchange at the
northern end of the Yonge platform. As would be expected,
there is little variation in the average car loading diversity
between the peak hour and the peak period due to the pressures
on passengers to spread along the platforms at busy times. The
average diversity of individual car loading over the peak period
has a range of +26%to -39%. The unbalance for cars on
individual trains ranges from +156% to -89%. 15 In the afternoon
peak period shown in Figure 5.11, the reverse occurs with the
front of the train most heavily loadeddespite the principal
entrances at the two major downtown station being toward the
rear of the train. There is less variation in the average car
loading diversity between the peak hour and the peak period
than in the morning. The average diversity of individual car
loading over the peak period has a range of +13% to -28%. The
unbalance for cars on individual trains ranges from +113% to 72%. These ranges are lower than in the morning reflecting the
less intense peak-within-the-peak in the pm rush hour.
It is this peak-within-the-peak that provides the third and most
important diversity factor, termed the peak-hour loading
diversity factor and defined by:
Figure 5.11 TTC Yonge Subway, Wellesley Station

northbound, p.m. peak-period
average
passenger
distribution between cars of train. Jan 11, 1995, total 69,696
passengers on 108 trains
where
= Diversity factorpeak hour

Dph
Rhour = Ridership in peak hour
R15min = Ridership in peak 15 min
Passengers do not arrive evenly and uniformly on any rail transit

system as shown dramatically over the extended peak period in
Table 5.14 Diversity of peak hour and peak 15 min16
Equation 5-1
16
15
One car of one train was completely empty (-100%), possibly due to an
incident or defective doors. This outlier was excluded from the data set.
17
This peak-hour diversity factor is the same as the peak-hour factor (phf) in
the Highway Capacity Manual(R47).
63
Figure 5.12 Individual train loads, TTC Yonge Subway, Wellesley Station southbound Jan. 11, 1995 (5-min tick marks)
Figure 5.13 Individual train loads TTC Yonge Subway, Wellesley Station northbound Jan. 11, 1995 (5-min tick marks) Note
cluster of low occupancy trains at 14:24 to 14:44h following a crush load train after a 29-min gap in service.
Figure 5.14 Individual train loads Vancouver, Broadway Station inbound October 27, 1994 a.m. peak (1-min tick marks)18
18
The courtesy of the Toronto Transit Commission and British Columbia Rapid Transit Company in providing car by car and train by train checker data is
acknowledged. The willingness of the Toronto Transit Commission to allow use of data with unusual erratic headway operation is particularly appreciated.
64
Figures 5.12 and 5.13 for the Toronto Transit Commissions
Yonge subway.
These figures do not show the smooth peaks-within-the-peak
often displayed in texts but rather the realities of day-to-day rail
transit operation. The morning peak-within-the-peak has a
pronounced abnormality at 8:35h following a short gap in
service.
The afternoon peak actually occurs at 14:24h following a 26min delay due to a suicide. Next are two abnormally low troughs
as the delayed trains move throughand the commissions
control center strives to normalize service prior to the start of
the real peak hour.
In both charts the different loading, train by train, is striking
and it is difficult to visually pick out the peak hour or the 15 min
peak-within-the peak. This entire data set of car by car loadings
and headways, representing 1,242 individual car counts of
135,000 passengers, is contained on the computer disk.
Figure 5.14 shows an a.m. peak-period for BC Transit that,
although without major delays, shows the irregular loading from
train to train due to the interlacing of short-turn trains with
regular service from 07:30h onwards. The loading diversity
factor was obtained for most systems. The principal data

deficiency was for light rail where few systems count
passengers by train.
The diversity of train loading over the peak hour is shown in
Table 5.14. Note that the values can be strongly affected by the
level of service provided. This is particularly true of infrequent
commuter rail lines. (Infrequent service on two of GO Transits
lines contributes to GOs relatively low average.) Rail rapid
transit (RT) is generally the most frequent mode and so has
relatively low values for the diversity factor. Values for light rail
transit are intermediate.
Diversity of loading within a car and among cars of a train are
included in the recommended peak-within-the-peak loading
levels. The peak-within-the-peak loading diversity factor is not
so included and must be used to adjust passenger volumes from
the estimated design capacity to a more practical achievable
capacity. This important loading diversity factor is discussed
further in Chapter Seven, Grade Separated Rail Capacity
Determination, and subsequent mode specific chapters. Here
suitable values are recommended for use in calculating the
maximum achievable capacity.
65
6. Operating Issues
6.1 INTRODUCTION
The previous three chapters have introduced the three major
components that control rail transit capacity. Chapter Three,
Train Control and Signaling, describes the capabilitiesand
determination of separationfor a range of train control
systems. The minimum separation of the train control system
can be calculated with some precision once the weak link has
been determinedusually the maximum load point station.
Whether a train will achieve this minimum separation is an
operating issue. Is the equipment performing to specification?
On manual systems, is the train driven at or close to the optimal
envelope? The answer to both questions is not always yes. To
operate a rail transit at its maximum achievable capacity without
interference between trains, an allowance has to be made for
these operating variables.
Chapter Four, Station Dwells, analyzed and developed
alternate methodologies to estimate dwells. Dwells cannot be
estimated with precision. They are affected by many day-to-day
circumstances. While some variables are accommodated in the
methodology it is not possible to make allowances for all. An
additional allowance is required to handle some of the day-today
irregularities. This is an operating issue. Dwells can also be
optimized by the design of stations, vehicle interiors and
schedulinganother operating issue.
Chapter Five, Passenger Loading Levels, offered two routes
to estimate the number of passengers. One is how many
passengers will physically crowd onto a trainproviding the
maximum achievable capacity. The other requires a policy
decision to establish a more comfortable peak-within-the-peak
loading level, appropriate to todays modern rail transit and
attractive to passengers. Either level is capable of handling an
overload of passengers when situations dictate. This again is an
operating issue.
Each of these operating issues will be discussed in this
chapter, concluding with recommendations on the range of
operating margins that should be included in the minimum
headway that, in turn, produces the maximum achievable
capacity that is the goal of this project.
6.2 TRAIN PERFORMANCE

Much has been made of the uniformity of performance of the
electrical multiple-unit trains that handle over 90% of all North
American rail transit. There is indeed a remarkable uniformity in
the rates of braking and acceleration due to the dictates of
passenger comfort. Variations in the reduction of acceleration
with speed increase and different maximum or balancing speeds
have been accommodated in the calculations of minimum train
control system separation in Chapter Three, Train Control and
Signaling. These calculations also accommodate fluctuations in
line voltage.
Although the wide spread introduction of electronic controls

has improved the uniformity of actual to specified performance,
there still can be differences between individual cars and trains
due to manufacturing tolerances, aging of components and
variance in set-up parameters.
The result can be up to a 10% difference in performance
between otherwise identical cars. Any impact is diluted when
the under-performing car is coupled in a train. One such car in a
ten-car consist will make a negligible difference. In a two-car
train the results are noticeable. In many systems, underperforming cars or trains are colloquially called dogs. Often
such trains cannot keep schedule and become progressively late.
As discussed later in this chapter, this situation can reduce
system capacity. This is a sufficiently common situation that an
allowance should be made in determining achievable capacity
and under-performing trains are one component, albeit minor, in
determining an appropriate operating margin.
There is a trend to design rail equipment not only to fail safe
but also to fail soft. Certain electronic-monitored rail transit cars
are designed to drop to lower performance rates if motor or
control equipment exceeds a set temperature, or if the line
voltage drops below a certain level. This performance drop may
be sudden or can be progressive but has to be significant,
typically 25% to 50%, to achieve the desired effect. Once a
single car on a train has reduced performance, the remaining
cars become overloaded and it is easy for an avalanche effect to
disable the entire train. This level of performance reduction
cannot reasonably be compensated for in the operating margin.
Automatic warning of the reduction is usually provided and
rapid removal of the equipment by train or control room
operators is needed to avoid service disruptions.
Lower braking performance will also affect capacity.
However the minimum train separation calculations, for safety
reasons, have already compensated for this by assuming a
braking performance set at a proportion of the normal
specification of 1.3 m/s2. The equations of Chapter Three allow
a user-specified value to be inserted for this percentage. The
recommended value is 75%.
Brake system failures are not regarded as a capacity issue.
Trains with one or more sets of cut-out brakes are invariably
immediately removed from service.
Performance differences are minor compared to the effect of
component failures. Failure management procedures have been
a feature of the industry from the earliest daysusually
allowing a defective motor to be cut-out so that the affected car
or train can continue in-service, or if significantly crippled, limp
home. This practice can also extend to motor control equipment
and other subsystems. Air and low voltage power are invariably
train-linedthat is, shared between coupled carsso that the
failure of a compressor, battery, motor-generator set or inverter
should have no effect on performance.
Redundant components are also becoming more common for
motor and train control equipment. These features, combined
66
with automated, and sometimes remote, diagnostics, and
effective preventive maintenance programs have resulted in
increases in the mean distance traveled between disruptive inservice failures. It is not uncommon for many classes of modern
rail equipment to achieve 100,000 km (60,000 mi) between inservice failures and a few car series on a handful of systems
have reached double this level.
The typical rail transit car travels 80,000 km (50,000 mi) each
yearsomewhat less for light rail vehicles. Some 20% of this
travel occurs during the peak hours. Each car therefore has a
potentially disruptive peak-hour failure approximately once
every 5 years. With multiple-unit trains the chance of a failure is
proportionate to the number of cars. Counteracting this is the
fact that a failure that could be chronic for a single car is rarely
so on longer trains. It is not uncommon for an eight- or ten-car
train to include one car with a totally inactive propulsion
system.
Consequently, it is neither appropriate nor practical to
compensate for major equipment failures in determining the
achievable capacity of a rail transit line. Operations planning
should ensure that such failures can be managed with the least
disruption. Unfortunately, operations planning is often given
scant attention in the initial design of a rail transit system. Thus
senior operating staff arrive to find many operating failure
management options have not been provided. These include
periodic pocket or spur tracks to accommodate bad-order
equipment, or spare equipment to plug gaps in service; frequent
cross-overs and bidirectional signaling to permit operating
around failed or derailed trains, failed switches, line-side fires
and suicides; and terminal station layout allowing forward and
rear train reversals and storage of spare or bad-order
equipment.
Poor or nonexistent operations planning may result in a
system that is unable to reach its achievable capacity or to
sustain this capacity reliably. This is an important issue as this
project has striven to determine a rail transit capacity that is
both achievable and sustainable. Attempting to quantify the
impacts of the more significant equipment failures on capacity
is beyond the scope of the study. Eleven references in
Appendix One, Literature Summaries, discuss operations
simulation and modeling that allow some failure scenarios to
be considered and the temporary reduction in capacity
determined.
Abramovici(R01), in Optimization of Emergency Crossovers
and Signals for Emergency Operations in Rail Rapid Transit
Systems, calculates the impact of forced single track working
on capacity for a typical rail rapid transit system with crossovers approximately 3 km (2 mi) apart. Achievable capacity
is reduced to 33% of normal with uni-directional signaling
and 60% of normal with bi-directional signaling. However,
with optimized cross-overs and bi-directional signaling,
emergency operation at 80-90% of normal capacity can be
obtained.
Retaining so high a proportion of capacity during a serious
failure carries a pricebut a price that is reducing as the
industry moves to train control systems with inherent bidirectional capability. New systems that are being designed for
high capacity or have links that preclude rerouting passengers
on other routes, should examine the cost effectiveness of
retaining an emergency situation capacity that is a high
proportion of normal achievable capacity.
6.3 OPERATING VARIATIONS

Differences among train operators can have an effect on
capacity because of operating below the maximum equipment
performance envelope and civil speed restrictions; an
understandable situation, particularly with inexperienced
operators who want to avoid triggering the automatic overspeed
emergency brake.
The result is twofold. The signaling system minimum train
separation will be increased and the train will fall behind
schedule. As discussed in Chapter Three, other workers have
suggested that automatically driven trains can achieve a
throughputand so achievable capacity that is 5 to 15%
higher than manually driven trains. The project has been unable
to obtain any data to support this, and the station dwell field
survey suggests that any such gain is more than lost in the
relatively slow station-door opening and departure procedures
that were noted, predominantly on automatically driven systems.
A train that is late due to operator performance is no different
from one that is late due to equipment under-performance, as
discussed in the previous section. At close headways, passengers
tend to arrive uniformly on station platforms with surges at
interchange stations due to the arrival of connecting buses or
trains. The result is that a late train will have additional
passenger movement, will have a longer station dwell and will
become progressively later until it interferes with the schedule
of the following train.
The same situation occurs if the train ahead runs fasttermed
running sharp on many systems. More passengers accumulate
on the platforms and the following train has longer dwells.
To accommodate these routine irregularities, two allowances
are made in operations planning and scheduling. An operating
margin is added to the minimum train separation time and
maximum load point station dwell to create a minimum headway.
This operating margin is, in effect, the amount of time a train can
run behind schedule without interfering with the following trains.
The operating margin is an important component in determining
the maximum achievable capacity and an analysis of existing
margins and recommendations for estimating margins are the
subjects of the next section in this chapter.
The second allowance is schedule recovery, an amount of
time added to the terminal turn-around time and dwell that
allows for recovery from the accumulated delays on the
preceding oneway trip. Schedule recovery time has some effect
on achievable capacity and also has economic implications as it
can increase the number of trains and staff required to carry a
given volume of passengers. The methodology for calculating
turn-around times was presented in Chapter Three. The amount
of schedule recovery time needed to avoid constraining capacity
cannot be calculated. The best guidelines are that it should be at
least half a headway at headways below every 5 min moving
toward a full headway as frequency drops toward the minimum
train separation. Chapter Three discussed ways to provide
schedule recovery at terminal station by turning on-time trains
behind the station. Late trains can then be turned in front of the
station gaining 90 to 120 sec but an at economic cost.
Experience on some rail rapid transit systems, operating at
their closest design headway, has shown that removing one train
from service, that is, running 29 trains an hour instead of the
rated capability of 30 trains an hour, can sufficiently reduce
67
accumulated delays such that the 29 trains run closer to schedule
and actually carry more passengersand at a lower cost.
Due to equipment unavailability or failure early in the peak
period, or to staff absenteeism that cannot be made up from the
spare board, runs are periodically missed on rail transit
systemsparticularly the larger ones. This situation creates a
gap in service. Dispatchers or supervisorsand certain
automatic train supervision systemswill strive to close the gap
or at least arrange for it to fall outside the peak-within-the-peak
at the maximum load point station. Nevertheless the remaining
trains must handle the passengers from the missing train(s).
Their dwells will increase and the achievable capacity will be
reduced.
There is no way to determine the probability or quantity of
missed runsor their effect on achievable capacity. Such
irregularities can only be accommodated in the conservative
assignment of loading levels and operating margins. Where
achievable capacity has been based on the bare minimum of
these discretionary components then missed runs will create
significant peak-period perturbations.
6.4 OPERATING MARGINS

As a starting point for recommending suitable operating margins
to incorporate into the determination of the maximum
achievable capacity, an attempt was made to survey existing
operating margins.
In general operating agencies were unable to quote specific
data. Rail transit planners and schedulers discuss the desirability
of both operating margins and schedule recovery but generally
operating margin is as much accidental as planned. It is the
amount of time between the closest headway and the sum of the
minimum train separation and the maximum load point station
dwell. As headways widen, operating margin increases. When
headways are pushed to their limit it diminishes, sometimes
almost to zero. As a result service irregularities increase. Some
operators accept this as the price of obtaining maximum
capacity and will even push a train into service on a line that is
theoretically at capacityand then usually remove it
immediately after a single one-way peak-direction trip. More
passengers have indeed been carried and line staff are left to sort
Figure 6.1 CTS 3rd St. SW E/B
out any erratic performance at the end of the peak period when a
few gaps or bunching in service are less critical.
This approach is counter to the suggestion of the previous
section that capacity could be increased by removing a peakhour train. This is very much a system-specific operating issue.
It involves minutiae that cannot easily be simulated and is
beyond the scope of this study. On a system that is at or close to
capacity, the only realistic way to find out if adding or
subtracting a train will increase capacity, and/or improve
headway regularity, is to try it for a period of time.
To determine operating margins on existing systems,
maximum-load-point station dwell and headways were recorded
during both morning and afternoon peak periods on 10 North
American systems. The results are shown graphically on the
following page. This is truly a case where a pictureor chart
tells a thousand words. There are many possible reasons for
irregular headways (shown as spikes), where known, for
example a passenger holding a door, these are tabulated in the
main data spreadsheet, provided on disk with this report.
Unknown reasons can include technical failures, trains holding
for a meet or trains coming into or going out-of-service.
Light rail headways on observed systems were generally
sufficiently long that any irregularities reflected problems other
than schedule interference between trains. The closest observed
on-street headway was in Calgary, shown in Figures 6.1 through
Figure 6.3 Note that the headways are all multiples of the 80-sec
traffic light cycle. This multiple of light cycles is pursued in
Chapter Eight, Light Rail Capacity Determination. Although
one train per cycle is often possible, the recommendation is that
achievable capacity should be based on one train every other
cycle. The seemingly erratic headways in Calgary are
misleading as three routes, forming two interlaced services share
this downtown bus and light rail mall.
The other light rail representative in the headway regularity
charts on the following pageis San Francisco Muni operating in
the Market Street subwayFigure 6.8. This operation is
effectively high-level rail rapid transit with the complication that
individual cars on trains from five surface routes are coupled
into longer trains for operation in the subway after lengthy
sections of on-street operation. Regularity of arrival at the
coupling points is difficult to achieve and, with different cars of
the same train
Figure 6.2 CTS 1st St. SW W/B
Figure 6.3 CTS City Hall E/B
68
Figure 6.4 BART Embarcadero W/B
Figure 6.8 Muni Montgomery W/B
Figure 6.12 PATH Journal Square W/B
Figure 6.5 BC Transit Burrard W/B
Figure 6.9 NYCT Grand Cen. S/B Exp.
Figure 6.13 PATH Exchange Place E/B
Figure 6.6 BC Transit Broadway E/B
Figure 6.10 NYCT Grand Cen. N/B Exp.
Figure 6.14 TTC King S/B
Figure 6.7 BC Transit Metrotown E/B
Figure 6.11 NYCT Queens Plaza W/B
Figure 6.15 TTC Bloor N/B
69
Table 6.1 Data summary of surveyed North American rail rapid transit lines at or close to capacity (seconds)
1
2
Adjusted to remove long delay at beginning of peak-period.

Only off-peak data. Included for comparison. Excluded from averages.
going to different destinations, dwells can be extended when

passengers must move around a crowded platform to locate their
specific cara relatively rare occurrence as the trains are
usually made up in the same order. Destination signs at each
platform berth, and on the side of each car, assist passengers in
finding their specific car or train.
Figures 6.1 to 6.15 are shown in small scale allowing them to
fit on a single page for easy visual comparison. The overall
impression is of many irregularities in operation. The data is from
a random sample of normal days, or a consolidation of 2 adjacent
days. Only when there were major service disruptions was the
data survey abandoned and rescheduled for another peak period.
Although much has been made of the uniformity of rail rapid
transit operation that allows generic calculations of minimum
train separation and dwell times, headway irregularities are a
factor of life and must be accommodated in estimating the
achievable capacity of a line through use of conservative
loading levels, realistic dwells and the addition of an operating
margin.
Data are summarized in Table 6.1 with calculations of dwell
and headway means and standard deviation.
The operating residual is the result of removing the minimum
train separation and the mean dwell plus two standard deviations
(see section 4.5.7) from each mean headway. Minimum train
separation is estimated at 50 to 55 sec for three aspect signaling
system, 40 sec for BC Transits moving-block signaling system
and 80 sec for Calgary based on the traffic light cycle times
along the downtown mall. BART has regulatory and
powersupply constraints that limit the number of trains
simultaneously in the Trans-bay tunnel. A nominal minimum
headway of 90 sec is used. This should be possible with the
planned future train control improvements.
The results are shown in the last column and in Figure 6.16
with the operating residual as the top component of each bar.
The bars are arranged in order of increasing headway. Note that
the bar furthest to the right is the only off-peak data set. It is
included only for comparison and shows the large operating
residual available when a system is not at capacity.
The operating residuals range widely and bear little
relationship to system, technology or loading levels. They
indicate whether adequate operating margin can be
accommodated. The most generous ones are on BC Transits
SkyTrain due to the closer minimum train separation of the
moving-block signaling system. Torontos King station has a
higher operating margin than expected due, in great part, to the
very short dwell with all alighting passengers. At Bloor station
on the same line, larger volumes of mixed-flow passengers
almost double the dwell time reducing the operating residual to
17 sec. Bloor station is the constraint on the line. At one time,
the Toronto Transit Commission had planed to rebuild Bloor
Station with dual platforms.
A proxy for service reliability is the headway coefficient of
variationthe standard deviation divided by the mean.
Discounting the high values for Calgarys light rail caused by
traffic light cycles, this ranges from a high of some 0.5 on the
TTC and BART to approximately half this on and NYCT and
PATH. BC Transits sophisticated automatic train supervision
and driverless trains show their capability and produce the
lowest and best figure. These results are somewhat incongruous
as there are automated and traditional, manual operations at both
the top and bottom of the listing. Ideally there should be a
relationship between the operating residual and the headway
coefficient of variation. However, as shown in Figure 6.17, there
is no reasonable relationship.
70
Figure 6.16 Headway components of surveyed North American rail rapid transit lines at or close to capacity (seconds)
necessary to provide higher capacity, a handful of rail transit

lines in New York and Mexico City all but eliminate the
operating margin with times below 10 sec.
It is recommended that a range be considered for an operating
margin. A reasonable level for a system with more relaxed
loading levels, where the last ounce of capacity is not needed,
should be 35 sec. Where that last margin is needed then a
minimum level of 10 sec can be used in the clear understanding
that headway interference is likely.
In between these extremes is a tighter range of 15-20-25 sec
that is recommended. This range is used in estimating
achievable capacity with the simple procedures and
recommended as a default value in the computer spreadsheet.
Figure 6.17 Relationship between operating residual and the

headway coefficient of variation
6.5 ESTIMATING MARGINS

Although there is no clear relationship between existing
operating margins and other operating criteria, this does not
allow this important factor, and the related terminal recovery or
layover time, to be discounted. The inevitable headway
irregularities and the need for reasonable operating flexibility
require the greatest possible operating margin and recovery time
to ensure reasonably even service and to achieve maximum
capacity.
Taking the operating residual as a surrogate for operating
margin, the average of the near capacity systems, discounting
Calgary and off-peak data, is 39 sec. The lower quartile is 25 sec
and the lower half is 32 sec.
Selecting a recommended operating margin is a dilemma; too
much reduces achievable capacity, too little will incur sufficient
irregularity that it may also serve to reduce capacity. Yet, when
6.6 OPERATING WITHOUT

MARGINS
It is reasonable to ask how several rail transit lines in other
countries operate at much closer headways than in North
America and yet achieve substantially higher capacities with
excellent on-time performance and reliability.
The four highest capacity double-track rail transit lines in the
world are believed to be Tokyos Yamanote line; sections of the
Moscow and St. Petersburg metros that operate at 90-sec
headways; and Hong Kongs Mass Transit Railway Corporation
which carries 75,000 passengers per peak-hour direction in 32
trains on the lower Kowloon section of the Tsuen Wan line.3
All systems have been visited by the Principal Investigator.
The Russian 4 systems appear to have a high level of staff
3
The MTRC has a capacity constraint where the Kwun Tong subway
terminates so as to deposit entire train loads at the peak point of another
line. MTRC is presently installing the SACEM quasi moving-block
signaling system to increase the system capability from 32 to 34 trains an
hour. Only so small an increment is needed as the capacity constraint will
be relieved by the new airport subway line presently under construction.
Similar operating arrangements occur on the Russian-designed metros in
Warsaw and Prague.
71
Figure 6.18 JR East high capacity car with six double doors
and longitudinal seats that are locked up against the wall in the
morning peak. The small number of seats are automatically
unlocked at about 10.00h.
discipline and surprising equipment reliability. The close

headways are maintained by strict control of dwell times. Each
station headwall has a clock showing the time from the
departure of the previous train. As the 90-sec headway time
approaches the doors are closedoften irrespective of whether
passenger movement had finished.and the train departs
precisely 90 sec behind the previous train. Any delay to a train
consequently rebounds down the linebut trains behind the
delay remain perfectly spaced. This approach is also partially
responsible for the high capacity of many double-track lines in
Japan but here other factors play a role.
The Japanese systems maintain the worlds highest passenger
throughput despite an intricate combination of through worked
services combining trains from different companiesboth
public and privatein multiple operating patterns: non-stop,
express, limited express, skip-stop and local.
Six factors5 combined to maintain these high capacities. First
is the very high loading levels that would not be acceptable in
the west (these levels are increasingly a concern in Japan as an
affluent population demands better commuting quality). Despite
this concern, the JR East has just introduced a high capacity car
with almost no seats, illustrated in Figure 6.18.6 Second is an
aggressive management of station dwells using more or wider
doors, large interior off-sets, and clearly marked door positions
and queuing areas on each platform. A trial car with wide doors
and platform markings is shown in Figure 4.12.
This dwell management is completed by familiar platform
managers and their white-gloved assistants. Contrary to popular
belief, the manager will rarely handle a passenger; the assistants
5
Based on discussions held by the Principal Investigator with executives

from several Japanese subway and suburban railway companies on an
October 1994 transit study tour.
6
The significant use of urban rail transit in Japan can be put in context with
the 1993 daily rail ridership in the greater Tokyo region of 35.96 million
passengers, about double the total daily ridership in all three North
American countries. Tokyo is served by the partly privatized JR East
railway; two subway companies, one public and one private; and seven
private suburban railwaysthe largest two of which, Odaky. and Tbu
together carry 50% more passengers a day than the NYCT.
are not trying to push more passengers onto the train but to close
the doors and avoid delays.7
The third factor is the precision of driving. Most drivers are
recruited to this prestigious job from railway high schools where
they have already been indoctrinated. Driver training can take
six months at special schools before the recruit gets extensive
line experience under the supervision of a senior operator. Some
schools have simulators with every meter of each line
videotapedparticularly important as even some of the high
capacity lines have grade crossings. Many grade crossings are
protected by a criss-cross array of infra-red presence detectors
that control an approach signal. The nerve and precision to drive
at these, still red, signals at maximum line speed is remarkable.
Equivalent discipline applies to vehicle and system
maintenance. Federally enforced levels of inspection and
preventive maintenance ensure exceptionally high equipment
availability. These levels would be uneconomic in North
America and the cost is being questioned by some Japanese rail
transit operators.
The fifth factor is the extensive use of off-line stations,
intermediate stations with four tracks, and terminal stations with
multiple tracks.
The final factor is the reliability built into the equipment
through redundancy and use of over-designed components.
Japanese urban rail rolling stock is heavy, in part due to these
design practices and in part due to government buffing strength
regulations. This also carries a high price and one Japanese
railway has recently specified a series of throw-away cars.
Vehicles are designed and built to have half the life of
conventional stock, thus avoiding the cost of the exceptionally
thorough and expensive rebuilds periodically required on
conventional equipment by central government regulations.
Hong Kongs high capacity MTR shares only a few of the
Japanese featuresmainly very high levels of crowding.
Coincidentally, Hong Kong handles the same number of peakhour passengers on two tracks as NYCT does on its busiest fourtrack Manhattan trunk.
Dwell control is a feature of other systems, but its methods
would not be acceptable in North America and are steadily
falling out of use elsewhere. The omission of doortraction/brake interlocks allows train doors to open before a
train has stopped and to close as the train is moving away from
the platform. If this feature is cautiously employedas once
common in Paris and Berlindwells can be reduced. On the
Buenos Aires metro the practice extended to doors that might
not close at all between stations.
6.7 SKIP-STOP OPERATION

Certain high-capacity operations in Japan use skip-stop service,
as employed in Philadelphia and New York, and until recently,
in Chicago. Skip stops, in themselves, provide faster travel times
for the majority of passengers with less equipment and staff. In
themselves skip stops rarely increase capacity as the constraint
remains the dwell at the maximum load point station at which,
by definition, all trains must stop. In fact capacity can be slightly
reduced as the extra passengers transferring between A and B
7
Platform attendants/managers also exist on North American systems.
72
trains at common stations, can increase dwells. Conversely a
balanced skip-stop operation can equalize train loadings and
reduce extreme dwells.
The common stations on the Japanese skip-stop operations
have multiple platforms, typically two-island platforms allowing
passengers to transfer across the platform between A and B or
between local and express trains.
Skip-stop operation is only applicable if the headways are
sufficiently short that the up to two-headway wait at minor
stations is acceptable to passengers.
Light rail operations may also skip stations when an ondemand operating policy is adopted. This requires on-board
passenger stop signals that can range from the traditional pullcords to use of the passenger-actuated door controls on
stanchions at each doorway. Drivers must observe whether there
are any intending passengers as they approach each station. This
is a particularly efficient way to increase line schedule speed
and reduce operating costs. However, at higher capacity levels,
all trains will stop at all stations and the practice has no effect on
achievable capacity.
Demand stops are common on the eastern light rail operations
that have evolved from traditional streetcar services but are
surprisingly rare elsewhere, even where there are clearly lowvolume stations and quiet times which could contribute to lower
energy, lower maintenance costs and a faster, more attractive
service.
Off-line stations can greatly increase capacity. They are used
in other countries but are unknown in North America except on
AGT systems. AGT off-line capacity is discussed in Chapter
Ten.
6.8 PASSENGER-ACTUATED
DOORS
The majority of new North American light rail systems have
elected to use passenger-actuated doors. The rationale is
increased comfort as interior heat or air conditioning is retained,
and wear and tear on door mechanisms is reduced. The practice
can extend dwells but is of little value at higher capacities or
busy stations where all doors are generally required.
Consequently some systems use the feature selectively and
allow the train operator to override and control all doors as
appropriate.
A typical rail rapid transit car door will cycle in 5 sec. Certain
doors on light rail systems, associated with folding or sliding
steps, can take double this time. Obviously a cycle initiated at
the end of the dwell will extend the dwell by this cycle time plus
the passenger movement time.
The problem is a contrariety as a system approaching
achievable capacity could not tolerate such dwell extensions but
would, in any event, be using all doors which might just as well
be under driver controlavoiding any last minute door cycling.
6.9 OTHER STATION

CONSTRAINTS
Many station-related factors can influence demand. Poor
location, inconvenient transfers to connecting modes, inadequate
or poorly located kiss-and-ride or park-and-ride facilities

may deter usage. Inadequacies in passenger access to a
station may reduce demand but not capacity. The only factor
that has a potential effect on the achievable capacity of a
system is the ease of exiting from a platform. Adequate
passageways, stairways and escalators must be provided to
ensure that a platform can clear before the arrival of the next
train.
Station exiting requirements are specified by the National Fire
Prevention Association 130 rapid transit standards. Exits,
emergency exits and places of refuge must be adequate to allow
a platform with one headways worth of passengers plus the
entire complement of a full-length fully loaded train to be able
to be evacuated to a safe location within four minuteswithout
using elevators and treating escalators as a single-width
stairway.
These regulations ensure that, in all but the most unusual
circumstances, where there is a disproportionate reliance on
emergency exits, full capacity loads can leave the platform
before the next train arrives.
On older systems NFPA 130 requirements may not be met.
Additional exits must be provided to ensure that achievable
capacity is not constrained by platform back-ups. Rates of flow
are established for passageways, up and down stairs and
escalators according to width.
In emergencies, exit-fare payment devices can be placed in a
free passage mode. This is not the case in normal operation and
adequate exit-fare control must be provided. The nominal rate
for a single-coin or magnetic-ticket-actuated fare gate or
turnstile is 60 passengers per minute. This is an optimistic rate.
Actual usage will range between 30 and 40 passengers per
minute, possibly longer at stations with a large proportion of
tourists or other non-regular transit users. The exit-fare gate rate
is also reduced by failure rates and, on systems with distancerelated fares, by tickets with inadequate stored value. Typically
10% of fare gates should be assumed to be out-of-service at any
time. About one in 4000 transactions will fail with magnetic
tickets. Proximity cards are reported to have failure rates two to
three times better but there is insufficient use to confirm this.
Add-fare requirements can be as low as one in a 100 depending
on operator policyseveral systems allow a passenger to
underpay, on the final ride on higher value stored value tickets,
as a form of random discount.
Whether due to a failure to read a ticket or the need to add
fare to a card, the existing fare gate can be obstructed for a
considerable period, particularly if the passenger repeats the
ticket insertion. It is essential that adequate exiting fare
equipment be provided at high capacity stations to ensure that
passengers do not back-up onto a platform.
Stations with high mixed flows must also have platforms of
adequate width to accommodate the flows. Width is also a factor
in making it easy for passengers to distribute themselves along
the length of a train and so improve the loading diversity
factor.
Fare payment is a particular factor on the few light rail systems
that still use on-board payment and checks. The flow rate analysis
showed that flat fare payments added almost exactly 1 sec per
boarding passenger, about 25% to an upstairs board, 50% to a
level board. This can significantly impact running time over
many stations. These factors however cannot be applied to the
dwell time calculations of Chapter Four, Station Dwells, as the
73
Figure 6.20 Tri-Mets Siemens-Dwag partial low-floor car
Figure 6.19 Wheelchair loading platform and ramp

far more drastic impact is the restriction of boarding to the
manned door, rather than spread along all doors of the train.
The Toronto Transit Commission has recently followed the
practice of most new light rail systems and introduced a proof of
payment fare collection system on its Queen St. streetcars. San
Francisco and Philadelphia have station collection in the subway
portion of their lines. MUNI has long term plans to move its
entire light rail fare collection to the faster and less expensive
proof of payment systemtwo surface stations have already
been converted.
If on-board manual fare collection is used, dwells must be
increased by the above percentages to arrive at achievable
capacity. The computer spreadsheet does not compensate for
this.
6.10 IMPACT OF
AMERICANS WITH
DISABILITIES ACT (ADA)
With dwell times being one of the most important components
of headway, the time impact of persons using wheelchairs was
examined. In addition to the modest number of field
observations that could be timed, data were obtained from those
systems that have actual rather than anecdotal movement and
delay times. The facts to date, while sparse, do tell a coherent
story. Actual measured lift times are shorter than anecdotal
claims, running 2-3 min with some as low as 60 sec. Level
wheelchair movements are generally faster than walking
passengers except where the car or platform is crowded. One
movement at a new San Francisco loading platform on the K
line was measured at 13 sec from doors fully opened to train
moving.8 An example of this mini-high or high-block loading
arrangement is shown in Figure 6.19.
8
However, this is one of the arrangements where the car/train must stop
twice, once for physically challenged passengers, then again for regular
passengers.
Figure 6.21 Profiled light rail platform showing slide out or

fold down step that avoids any internal steps
Figure 6.22 Profiled light rail platform Provides two steps

into all doors, except the front door which is wheelchair
accessible. All slopes are a maximum of 8.5 to meet ADA
requirements. Most of the platform is only slightly higher than a
sidewalk. Additional details on light rail wheelchair facilities
with city specific information are contained in Chapter Eight,
Light Rail Capacity Determination.
San Francisco has one of the best of the high-block loading

arrangements although requiring a second stop. The loading
takes place at the parallel second, rather than tapered first door.
An elastic filler covers most of the gap between the platform
and door threshold. No bridge is required, the driver does not
have to leave the cab, relying on wayside markings to position
the train with the second door at the wheelchair loading
platform.
Most rail transit wheelchair users are very agile. These are the
people who want the mainstream option and use it. They seem
to be particularly sensitive to not causing delays.
As well as being the preferred arrangement for meeting ADA
regulations, high-platform loading also provides the maximum
capacity. Dwells are reduced and no interior car capacity is lost
to the stepwells or to interior stepsa feature of high-floor cars
with low-level boarding and some low-floor cars. Low-floor
cars will offer much of the speed and easy access of highplatform loading. The first low-floor car to be introduced in the
United States (Figure 6.20) will be running in 1997 in Portland.
Level high-floor loading may be problematic in many
systems. The options range from the interior folding steps used
in San Francisco to the outboard folding steps used in San Diego
or the Manchester style profiled platform, shown in Figures 6.21
and 6.22. Such a platform has an intermediate height and is
74
profiled up to a short stretch that is level with one doorway for
wheelchair use. Where the street arrangement permits, the
profiled platform can be raised so that its mid-sectiontaking
up most of the lengthis raised one step providing a single-step
entry to most doors.
Another option to meet the ADA requirements is the separate
wheelchair ramps that are used in Baltimore, Sacramento and
San Francisco, among others. In this arrangement, shown in
Figure 6.19, a car-floor-level platform, sized for one wheelchair,
is accessed by a ramp at one end, preferably the front end of
each light rail stop. This arrangement is often termed high-block
or mini-high loading. These are less popular with the physically
challenged community and present a greater physical and visual
intrusion into the street scene. However there are numerous
examples, particularly in Sacramento, of carefully integrated
and relatively unobtrusive arrangements. These high-block
platforms have advantages over car- or platform-mounted lifts in
reducing delays. The platforms also save the need for
maintenance and repair of mechanical lift equipment.
One of the most salient issues is the number of persons using

wheelchairs that will elect to use mainstream rail transit when
all ADA measures have been implemented. In the project
survey over 25,000 passengers were counted at one doorway
out of the eight to 40 doorways on each monitored train. Out
of an estimated 100,000 peak-period passenger movements
observed on those systems that are fully wheelchair accessible,
five wheelchairs were seen and timed. This represents one
wheelchair per 20,000 passengers. Other systems have
estimated ratios that range from one in 5,000 to one in 10,000.
However the usage of lifts is some three to five times higher
than this due to use by passengers other than those in
wheelchairs.
During the survey, doorway delays were observed quite
frequently due to passengers, not in wheelchairs, who were
otherwise physically or mentally challenged; elderly; with
children; carrying packages; or accompanied with push-chairs,
shopping trolleys, crutches and walking frames. Most of the
latter, on light rail with steps declined to use the lift and created
the longest doorway times for a single passenger. ADA
requirements will reduce such delays as systems move away
from mechanical lifts at single doors to multiple door level
loadingwhether high or low floor.
Many delays were also due to passengers hesitating at a
doorway, possibly uncertain that this was the correct train to
boardor the right station to exit. The ADA requirement to
clearly delineate the platform edge, and to visually and aurally
indicate the train arriving at a platform and, once on-board, the
next station should reduce delays due to such confusion.
Others have raised the potential problem of a wheelchair user
attempting to board a heavily loaded train or light rail car. In
theory operating staff should ask standing passengers to vacate
the car to accommodate the wheelchair. This obviously has the
potential for lengthy dwell extensions.
However, very few such situations occur. The average rail
transit car loading in North America through the peak hour is
0.5 m2 per passenger (5.4 sq ft) At this loading a wheelchair
could be accommodated in any vestibule, on any train, without
impeding other passengers or delaying the train. Passengers not
Figure 6.23 Wheelchair user in designated space BC

Transit
only move aside to accommodate a boarding wheelchair but

often will assist the wheelchair user reaching a designated space.
Once on-board there is the issue of any capacity reduction due
to the space taken by the wheelchairequivalent to three to six
standing passengers, depending on the loading density. Given
the average peak-period space occupancy cited in the last
paragraph, there is clearly no impact on most systems, although
NYCT and the San Francisco Muni, for example, might be
affected. It is possible that the location of designated spaces
relative to doorways and the positioning of wheelchairs could
disrupt interior passenger circulation on narrow rail transit cars.
However, Figure 6.23 shows a wheelchair user on a BC
Transit car, one of the narrowest rail rapid transit car designs on
the continent. The wheelchair users legs extend slightly into the
aisle but are less of an obstruction than the other passengers
sitting on the longitudinal seats in the foreground of the photograph. On these cars the wheelchair-designated space is immediately adjacent to and parallel to the door. There are no restraints.
Special handholds are provided and an interior wallon the far
side of the wheelchairprevents wheelchair movement in the
event of emergency braking. A seat folds down when the space
75
is not occupied. The only non-standard feature of the location
are a lower height passenger intercom and the omission of the
dual stanchion in the center of the vestibule that would interfere
with wheelchair maneuverability.
There was insufficient information obtained from operating
agencies or the survey to quantify any impact of ADA on the
achievable capacity of rail transit systems. There were sufficient
numbers and varieties of boardings and alightings observed for
the study team to conclude that, with full implementation of
ADA, and the elimination of lifts on close headway rail systems,
wheelchairs generally will have no or little impact on

capacityeven allowing for substantial increase in use and for
rare incidents, such as one observation, where the front wheels
were briefly stuck in the platform-door gap.
In the interim, wheelchair-lift use may cause delays but these
are generally on systems with long headways (6 min and above)
and have minimal impact at these levels. In the longer term
other requirements of ADA may sufficiently improve boarding
and alighting movements to off-set any negative impact of
wheelchair useif indeed there is such an impact.
76
7. Grade Separated Rail

Capacity Determination
7.1 INTRODUCTION
The preceding four chapters developed the methodologies for
each of the components in calculating capacity. This chapter
brings these methodologies together for the principal category of
grade separated rail, which includes over 90% of rail transit in
North America:
grade separated rail transit is operated by electrically
propelled multiple-unit trains on fully segregated,
signaled, double-track right-of-way.
This category encompasses all rail rapid transit, all automated
guideway transit (AGT), some of the heaviest volume commuter
rail lines and sections of most light rail systems.
AGT systems use proprietary technology and often have train
control separation times and vehicle loading levels that are
atypical of conventional rail transit. These atypical situations
and the capacity of AGT with off-line stations are dealt with in
Chapter Ten, AGT Capacity Determination.
Light rail operates in a variety of rights-of-way, each of which
has specific achievable capacities. Chapter Eight, Light Rail
Capacity Determination, contains the procedures to determine
capacity for light rail operating on other than double-track grade
separated sections. Single-track sections, if present, are usually
the capacity limitation. However these are rare and in all of the
light rail systems examined, the achievable capacity was
controlled by the signaling throughput of grade separated
sectionsdetermined by the procedures of this chapter.
This is due to two reasons. Several light rail systems converge
surface routes into a signaled grade separated section operating
at, or close to, capacity. Other, less busy systems, have the
signaled grade separated sections designed economicallynot
for minimum headways down to 2 min. Typically this signaling
is designed for 3- to 4-min headwaysmore restrictive than the
headway limitations of on-street operation, with or without
varying forms of pre-emption. However signaled grade
separated sections may not always be the prime headway
limitation. Chapter Eight explains how to calculate and
determine the weak link in the capacity chain for light rail.
Determining the weak link in the capacity chain is also the
starting point in this chapter with respect to this main
categorygrade separated rail transit.
7.2 THE WEAKEST LINK

Chapter Three, Train Control and Signaling, developed the
methodology for the train control system maximum throughput
in three situations:
1. The close-in time at the busiest station,

2. Junctions, and
3. Turn-backs.
In new systems it is poor design that capacity should be limited

by junctions or turn-backs. Both can be designed to avoid
constraints. Chapter Three, section 3.10, shows that a flat
junction can handle 200-m (660-ft) trains with standard rail
transit performance, under fixed-block train control, on noninterference headways down to 102 sec plus an operating
margin. The equivalent time for the same length trains with a
moving-block signaling system is 63 sec plus an operating
margin. Chapter Three recommends that junctions controlled by
a three aspect signaling system should be grade separated where
trains combine to a joint headway below 3 min. Only where
there are flat junctions with headways for their respective train
control systems below these levels, plus a 20-sec operating
margin, is it necessary to utilize Equation 3-26 to determine the
junction throughput limitation.
Section 3.9 of Chapter Three similarly shows that a two-track
terminal station can turnback 200-m trains every 120 sec with a
terminal time of 175 secthat is the time for passenger flows
and for the driver to change ends. Section 3.9 and Chapter Six,
Operating Issues, suggest that where passenger flows are heavy,
dual-faced platforms be provided; where changing ends is a
limitation that crew set-backs be used; that greater operational
flexibility and improved failure management is obtainable by
providing turn-back capability both ahead of and behind the
station with a storage track for spare or bad-order rolling stock;
and, finally, that a three-track terminal station can handle
exceptional passenger flows from trains on headways below 90
sec.
On new systems, turn-backs can be disregarded as a capacity
constraint unless economic circumstances or labor practices
prevent an optimal terminal design. Only in such exceptional
circumstances is it necessaryafter determining the minimum
headway from this chapterto apply Equations 3-21 and 3-25
to ensure that adequate terminal time is provided to allow for the
anticipated passenger flows and changing ends.
On older systems, terminal station design may be sub-optimal
and Equation 3-25 should be checked with the actual station
cross-over geometrics to ensure there is adequate terminal time.
This calculation should then be cross-checked with actual field
experience.
In either case a turn-back constraint is only likely if all trains
use the terminal station. If peak-period short turns are operated
such that only a proportion of trains use the terminal station then
a systems capacity limitation can be assumed to be the closein
movement at the busiest station.
77
7.3 GROWTH AND

ACHIEVABLE CAPACITY
The achievable capacity as defined in this report is not the
capacity at which a rail transit will openor reach after a
decade. It is the maximum achievable capacity when the system
is saturated and provided with a full complement of rolling
stock. It can be looked at as the long-range design capacity after
decades of growth.
A difficult question is what ultimate capacity a system should
be designed for. With good data, a constancy of historical trends
some transportation models can be calibrated to predict
passenger demand with reasonable accuracy. However
predictions beyond 10 to 15 years are of decreasing accuracy
particularly in areas without an existing rail transit system or
good transit usage which makes the modal split component of
the model difficult to calibrate.
When modeling does not provide a reasonable or believable
answer it is possible to fall back on an old rail transit rule of
thumb, namely, to design for three times the initial mature
capacity. Mature capacity occurs 5 to 10 years after a system
opens, when extensions and branches are complete, modal interchangesbus feeders and park and ridehave matured, and
some of the rail transit initiated land-use changes, including
development and densification around stations, have occurred.
The achievable capacity determined from this report can be
used to establish the train and station platform lengths and the
type of train control that will allow this long-term demand to be
metwhether obtained from a long-range model or by rule of
thumb. This long-term demand may be 30 to 50 years ahead. If
this suggests that 180-m-( 600-ft-) long trains and platforms will
be required then it does not mean they have to be built initially.
Stations can be designed to have platforms expanded in the
future. However, underground stations should have the full
length cavity excavatedotherwise it can be difficult and
expensive to extend platforms while the rail line is operating.
7.4 SIMPLE PROCEDURE

Taking advantage of the relative performance uniformity of
electric multiple-unit trains in urban rail transit service allows
the use of this simple procedure to estimate a range of achiev
able peak hour passenger capacities for grade separated lines at
their maximum capacity.
The necessary choices are only two, the type of train control
system and the train length. The range is provided by assigning
1) a range centered around a typical dwell time plus operating
margin, and 2) a small loading range centered around the
recommended peak-hour average space per passenger of 0.5 m2
(5.4 sq ft). As this is a peak-hour average, no loading diversity
factor is required.
This simple procedure assumes system and vehicle characteristics that are close to the industry norms listed in Table 7.1. It
also assumes that there are no speed restrictive curves or grades
over 2% on the maximum load point station approach and that
the power supply voltage is regulated within 15% of specifica-
Table 7.1 Simple method performance assumptions
tions. The procedure, as does the study as a whole, assumes an

adequate supply of rolling stock. If any of these assumptions are
not met then the simple procedure may be used only as a
guideline and the complete procedure of section should be used.
This procedure does not apply to locomotive-hauled commuter
rail or to automated guideway transit using a proprietary system
with small, narrow vehicles.
This simple procedure is contained on the computer disk but a
computer is not required. The result can be calculated in the
time it takes to load the spreadsheet program or, if the
recommended medium-comfort loading levels are accepted,
directly and simply from Figure 7.5 (cab control signaling) or
Figure 7.6 (moving-block signaling) at the end of this section.
The range of trains per hour are shown in Figure 7.1 for the
above assumptions for cab control systems and in Figure 7.2. for
moving-block signaling systems. New systems that are designed
for maximum capacity would not use the more limited and more
expensive three-aspect signaling system. Such a system may be
used for systems designed for less than maximum throughput
in which case this procedure is not applicable. Consequently the
choice of train control system is limited to cab control and
moving-block.
This is a method to determine the maximum capacity of a rail
transit system. Consequently, train lengths are shown for typical
maximum lengths of 200 and 150 m (trains of 8 and 6 heavy rail
cars) and 120, 90 and 60 m (trains of 4, 3 and 2 articulated light
rail vehicles respectively). The maximum number of trains per
hour can be selected from Figures 7.1 and 7.2, rounded down
and multiplied by the selected train loading level obtained from
Chapter Five, Passenger Loading Levels, section 5.5. Figure 5.8,
reproduced again as Figure 7.3, shows a range of linear loading
for heavy rail cars from 71 to 11 passengers per meter of length.
Figure 5.7, reproduced again as Figure 7.4, shows a range of
linear loading levels for light rail cars from 5 to 9 passengers per
meter of length. These linear loading levels represent the
peak-within-the-peak and a loading diversity factor should be
1
The lower ranges for the short cars in Vancouver and Chicago should not
be used in the simple procedure method. This is based on 6 to 8 car trains
of 23-m-long cars.
78
Figure 7.1 Cab control throughput in trains per hour with a

range of dwell times plus an operating margin from 45 sec
(lower bound) to 70 sec (upper bound)
Figure 7.3 Linear passenger loading of heavy rail cars
Figure 7.2 Moving-block throughput in trains per hour with a

range of dwell times plus an operating margin of 45 sec (lower
bound) to 70 sec (upper bound)
Figure 7.4 Linear passenger loading of articulated light rail

cars
7.5 COMPLETE PROCEDURE

applied if loading levels in the upper ranges of these charts are
selected. When calculating diversity on the capacity of a line in
a city with existing rail transitof the same modethe existing
loading diversity factor or near equivalents should be obtained
from Chapter Five, Passenger Loading Levels, section 5.6. For
new systems, a loading diversity factor of 0.8 should be used for
heavy rail and 0.7 for light rail. For example the typical median
light rail level of 6 passengers per meter of car length would
reduce to 4.2 applying the suggested loading diversity factor of
0.7.
Applying these loading levels to the throughput ranges above
provides a direct range of passengers per peak hour direction per
track versus train length, shown in Figures 7.5 and 7.6.
The complete procedure to estimate the peak-hour capacity of

grade separated rail transit requires sequential steps.
The first step is to determine the capacity-limiting constraint,
either the station close-in and dwell time, or junction or
turnback throughput. The approach in section, The weakest link,
should be followed. If necessary, the junction or turn-back
throughput can be calculated from the methodologies and
equations of Chapter Three. Should a junction or turn-back
appear to be the limitation on train throughput then the first
recourse is to consider design or operating practice changes that
will remove or mitigate such limitations.
In all but the most exceptional situation, the limitation will be
the close-in, dwell and operating margin time at the maximum
79
Figure 7.6. load point station. The complete procedure requires
that the following values be calculated:
1. the close-in time at the maximum load point station
2. the dwell time at this station
3. a suitable operating margin
4. the peak-within-the peak train passenger load
5. the loading diversity factor to translate from peak-withinthe peak to peak hour.
Figure 7.5 Achievable capacity with multiple command

cabcontrol signaling system and peak-hour average loading
of two passengers per square meter for one track of a grade
separated rail transit line
These procedures can be calculated manually, or by experienced

users developing their own computer spreadsheet. The spreadsheet on the computer disk allows the many variables to be
inserted to produce passengers per peak hour direction per track.
However this spreadsheet cannot and does not assist in
determining the weakest link or the maximum load point station.
Nor does it solve the issue of how much operating margin
should be provided or the appropriate loading level.
When there is uncertainty about these factorsfully
described in Chapter Four, Station Dwells, Chapter Five,
Passenger Loading Levels and Chapter Six, Operating Issues
or where several of the performance variables are unknown, for
example the technology or specific vehicle has not been
selected, then following the complete procedure is not
recommended. The simple procedure above provides a generic
achievable capacity range with less effortand potentially as
much accuracy as the complete method where one or more input
factors will have to be guessed at.
7.5.1 DETERMINING THE MAXIMUM

LOAD POINT STATION
Traditionally the maximum load point station is the principal
downtown station or the downtown station where two or more
rail transit lines meet. This is not always the case. With
increasingly dispersed urban travel patterns some rail transit
lines do not serve the downtown. Los Angeles Green Line and
Vancouvers proposed Broadway-Lougheed line are examples.
The regional transportation model will usually produce
ridership data by station, both ons and offs and direction of
travel. Such data are usually for a 2-hour peak period or peak
hour and rarely for the preferable 15 min peak-within-the-peak.
Depending on the number of zones and nodes in the model, data
accuracy at station level can be poorparticularly if there is
more than one station in a zone. Nevertheless this is often the
sole source of individual station volumes and without it
selection of the maximum load point station requires an
educated guess for new systems.
Figure 7.6 Achievable capacity with moving-block signaling

system and peak-hour average loading of two passengers per
square meter for one track of a grade separated rail transit
line Note: The number of trains per hour vary with train length,
refer to Figures 7.1 and 7.2. With the exception of San
Franciscos MUNI metro, signaled grade separated light rail
lines are rarely provided with the minimum headway
capabilities represented by the capacity ranges in Figure 7.5 and
7.5.2 DETERMINING THE CONTROL

SYSTEMS MINIMUM TRAIN SEPARATION
Chapter Three, Train Control and Signaling, developed the
methodology for minimum train separation with three types of
train control systems, each with progressively increased
throughput:
80
Table 7.2 Minimum train separation parameters
Figure 7.7 Minimum train separation versus length
1. three-aspect signaling system

2. multiple command cab control
3. moving-block signaling system.
Although the equations appear long, the arithmetic is simple and
can be implemented in a spreadsheet with basic functions if the
reports computer disk is not available. Before going to this
effort, check the availability of the required input parameters in
Table 7.2. Parameters can be adjusted for system specific values
or left at their default value. Train length is the most important
variable. However if most parameters are left at their default
values then it would be simpler to refer to Figure 7.7 which
shows the minimum train control separation against length for
the three types of train control system. The equation for threeaspect and cab-control signaling systems, derived from Equation
3-15 of Chapter Three with dwell and operating margin
components removed and grade and voltage factors added, is
Equation 7-1
The equation for moving-block signaling systems with a fixed
safety-separation distance, derived from Equation 3-18 of
Chapter Three with dwell and operating margin components
removed is
Equation 7-2
Note that this equation is not affected by either line voltage or

station grade. Lower voltages increase the time for a train to
clear a station platform. In moving-block systems this time does
not affect throughput. When a train starts to leave a station the
target point of the following train is immediately advanced
accordingly. The worst case approach grade is included in the
determination of the safety distance. This can result in
suboptimal minimum train separation.
Higher throughput is usually obtained with a moving-block
signaling system with a variable safety distance comprised of
the braking distance at the particular speed plus a runaway
propulsion allowance. The equation for such a system, derived
from Equation 3-20 of Chapter Three with dwell and operating
margin components removed and a line voltage factor added, is
Equation 7-3
The appropriate one of these equations must be solved for the
minimum value of T(s). The approach speed va that produces
this minimum value must then be checked against any speed
restrictions approaching the station from Figure 7.8. The dotted
line example in Figure 7.8 shows that at 120 m2 from a station,
the approaching train will have a speed of 64 km/h. If there is a
speed limit at this point that is lower than 64 km/h then the
minimum train separation T(s) must be calculated with the
approach speed va set to that limit.
Finally, whether using the spreadsheet or individual calculations, check the results with Figure 7.7. The minimum train
2
Distance from the front of the approaching train to the stopping point.
81
Three methods of estimating dwell or controlling dwell are
provided in this section. The first method is the one used in the
simple procedure of this chapter and by most of the literature
referencessimply assigning a reasonable figure to the
headway critical station. The second method uses field data
from this study allowing the selection of a controlling dwell
(mean dwell plus 2 standard deviations) from the headway
critical station of systems with similarities to the one being
analyzed.
The fourth and final method uses the statistical approach of
Chapter Four of determining dwells based on peak-hour
passenger flows. This method is complex and still requires an
estimate of the ratio of the busiest door to average door flow.
None of these methods are entirely satisfactory. It is
regrettable that the study failed to find a better method of
estimating dwell or controlling dwell times and explains why
other practitioners over a period of three decades have resorted
to simply assigning a reasonable value to dwell.
METHOD ONE Assigning a Value

Figure 7.8 DistanceSpeed braking into a station
separation should be close to or moderately greater than the

values charted. If lower, there is probably an error as the charted
values are the minimums using typical maximum rail transit
performance criteria and without applying any corrections for
grades or speed restrictions into or out of the station.
Existing rail transit systems operating at or close to capacity

have median dwells over the peak hour that range from 30 to 50
sec with occasional exceptional situationssuch as the heavy
peak-hour mixed flow at NYCTs Grand Central Station of over
60 sec. A tighter range of dwell values35 to 45 secis used
in the simple procedure and can be used here together with the
more accurate calculation of the minimum train separation.
METHOD TWO Using Existing Dwell Data
7.5.3 DETERMINING THE DWELL TIME

This section deals with dwell to which both an operating margin
and the minimum train signal system separation must be added
to produce the headway.
The train close-in time at the headway critical station, being
dependent on the physical performance and length of a train and
other fixed system characteristics, can be calculated with some
precision. Station dwell time cannot be determined with the
same exactitude. All but one of literature references to dwell
assigned a set time to dwell. Many simulations do likewise
using typical figures of 15-20 sec for lesser stations and 30-45
sec for major stations. The one methodology to determine
controlling dwelldwell plus operating marginrequires
knowledge of dwell times over the peak hourinformation only
available for existing systems or new lines in areas where a
station with similar passenger volumes can be analyzed.3
Chapter Four, Station Dwells, describes the main constituents
of dwell:
Dwell data from the projects field survey are summarized in

Table 7.3. Data were usually collected at the highest use station
of lines that were at or close to capacity. As none of the newer
light rail systems are approaching capacity4 the busiest systems
Table 7.3 Peak-period dwells for heavily used systems
Passenger flow time at the busiest door

Remaining (unused) door open time
Waiting to depart time (with doors closed)
3
ALLE, P., Improving Rail Transit Line Capacity Using Computer

Graphics. The methodology for calculating controlling dwell is contained
in full in Appendix One and can be used in the rare case that the dwell
determination can be based on existing dwell time data. No operating
margin should be added when controlling dwell is calculated.
Maximum design capacitythat is without limitations of single-track

sections or line sections signaled for lower throughput than the maximum
capabilities of the signaling system.
82
were surveyed. Selection of a dwell from this table is less
arbitrary than method one and allows some selectivity of mode
and the opportunity to pick systems and stations with similar
characteristics to those of the one under examination. The
selected median dwells range from 27.5 sec to 61.5 sec. The
highest data, with the exception of the TTCs King Station, are
mainly alighting and mixed flow records from manually
operated systems with two-person crews. Most dwells in Table
7.3 fit into the 35 to 45 sec range suggested in the previous
method.
Where comparable field data also allows the calculation of
standard deviation the controlling dwell can be selected as the
mean dwell plus two standard deviations. Refer to Table 4.17
for examples. When the controlling dwell is so estimated any
additional operating margin (section 7.5.4) can be reduced or
eliminated. Alternately the greatest of the mean dwell plus two
standard deviations or the mean dwell plus the operating
margin (from section 7.5.4) can be used.
METHOD THREE Calculating Dwells from Station Hourly

Passenger Flows
This method involves complex mathematics. It is applicable to
new systems5 where Method two is not appropriate and where
data on hourly, directional flow at each station is available from
a regional transportation model. Use of the Excel version of the
spreadsheet is recommended for this method and a simplified
guide is contained in the spreadsheet. Other readers may wish
to skip this section and jump to 7.5.4.
Chapter Four developed regression equations to relate passenger
flow times to the number of boarding, alighting or mixed flow
passengers, and, in turn, to convert this flow time to dwell time.
These regression equations can be used to estimate the dwell
time from hourly passenger flows into the maximum load point
station. However the best regression fit involves logarithmic
functions and the estimation of a constant for the ratio between
the highest doorway and the average doorway passenger flow
rate.
The mathematics are complex and it is uncertain if the results
provide any additional accuracy that merits this complexity
particularly if the hourly station passenger volumes by direction
are themselves somewhat uncertain. This method is best suited
to new lines in locations without rail transit and with a
sufficiently refined and calibrated regional transportation model
that can assign hourly passenger flow, by direction, to individual
stations.
The first step in the process is to obtain the hourly passenger
flow from the regional transportation model. Many models
produce 2-hour am peak flows. In this case, use either the
models peak-hour conversion factor or a typical value of 60%
to arrive at an approximate peak-hour passenger figure.
Then, from the model select the station with the highest
passenger volume, either into or out of the station, and classify
the flow as, mainly boarding, mainly alighting or mixed. Most
models deal with the morning peak period. If the maximum load
5
This method can also be used on existing systems to estimate the change
(increase) in the controlling dwell at stations where new development, or
interchange with a new rail transit line, significantly increases the stations
passenger volume.
point station is downtown it is likely that the flow will be

primarily alighting. If the station is also an interchange with
another rail transit line then flows could also be mixed.
Unless station flows are also available for the afternoon rush
this process assumes that the morning peak defines limiting
head-wayand so maximum capacity. This is usually the case.
Morning peaks tend to be sharper, afternoon peaks more
dispersed as a proportion of passengers pursue diversions
shopping, banking, visiting a bar, restaurantor theaterbetween
work and the trip home. This more spread peak should override
the fact that boarding is slightly slower than alighting.
As the controlling dwell time will occur during the peakwithin-the-peak, the next step is to adjust the flow to the peakwithin-the-peak 15 min rate using a loading diversity factor.
Equation 7-4
where
= diversity factorpeak hour

Dph
Rhour = ridership in peak hour
R15min = ridership in peak 15 min
The factor should be selected based on the rail mode and the
type of system. Section 7.5.6, later in this chapter, describes
how to select an appropriate diversity factor.
The peak 15-min movement of passengers on a single-station
platform, P15min, can be expressed as
Equation 7-5
where
Phour = peak-hour movement of passengers on a

single station platform (obtained from
regional transportation model)
The number of double-stream train doors available in that 15min period, D15, is
Equation 7-6
where
T(s)
td
tom
Dn
Nc
=
=
=
=
=
train control separation in seconds

dwell time in seconds
operating margin in seconds
number of double stream doors per car
number of cars per train
The passenger flow at the busiest, i.e., controlling, door of the

train in the peak-within-the-peak, Fmax is
Equation 7-7
where
R = Ratio of busiest door usage to average door

usage
This ratio is close to unity for heavily loaded rail transit lines
operating at capacity as passengers are forced to spread
themselves relatively evenly along the platform. Under lighter
conditions the ratio will increase. As capacity is being calculated
at the maximum load point station during the peak-within-thepeak, a ratio of 1.2 is recommended for heavy rail and 1.5 for
light rail.
The regression equations of Chapter Four, Station Dwells,
83
section 4.6.4, can be simplified by omitting the reverse flow
terms and are expressed for all alighting, all boarding or mixed
flow as:6
Table 7.4 Values used to compute Figure 7.9
Equation 7-8
Equation 7-9
Equation 7-10
where
FTmax = Flow time for the respective type of flow,

alighting, boarding or mixed, at the
maximum use door (seconds)
Section 4.6.6 determined dwell time relative to the respective

maximum doorway flow time as:
Equation 7-11
Substituting in Equation 1-11 for FTmax and Equation 1-8 for
Fmax produces:
ln(td) = 3.168
+ 0.0254e
Equation 7-12
Equation 7-12 is solely for the expected dominant am peak and
mainly alighting case. Similar expressions can be derived for
mainly boarding flows and for mixed flows.
These equations have to be solved for the value of the dwell
time td, contained as both a natural logarithm and as an
exponential. The equations are not solvable in closed form and
the preferred solution is the simplest, using recursive numeric
assumptions.
The recursive numeric assumption approach is carried out in
the spreadsheet on the computer disk. The dwell is shown to the
nearest integer. This seeming accuracy should not be allowed to
conceal the uncertainties of some of the equation components.
At best the ensuing accuracy should be in the range of 3-4
seconds, not necessarily better than the alternative, simpler
methods of estimating or assigning a dwell timebut the only
method that relates dwell time to the hourly, directional station
passenger volumes. The results for all alighting passengers
based on the values of Table 7.4 are shown in Figure 7.9.
The Excel version of the spreadsheet contains a simplified stepby-step guide to utilize this method of estimating dwell times.
The results show the expected trend. Dwell time increases with
the hourly passenger movement. The resultant achievable capac6
Chapter 4, section 4.6.4, also developed regression equations with slightly

improved explanation of variance by including the number of passengers
standinga surrogate for impedance to passengers when boarding or
alighting from the car. As the number of standing passengers cannot be
reasonably known when estimating achievable capacity, these slightly
improved equations are not used.
Figure 7.9 Dwell time and achievable capacity at a

maximum load point station versus hourly alighting
passengers at a single platformunit values from Table 7.4.
ity decreases at a lesser rate. Capacity is reduced by a

comparable amount if either the number of doors per car is
reduced from four to three or an uneven spread of passengers
along the platform results in the ratio of the maximum to
average door flow increasing from 1.2 to 1.5.
Although the regression analysis is based on data from heavy
volume stations, the results become increasingly inaccurate at
extremes. Neither equation 7-12, nor its implementation on the
computer spreadsheet, should be used with the maximum doorway flow greater than 25-30 passengersequivalent to approximately 20,000 passengers per peak-hour direction per platform
with the default values of Table 7.4. It is unlikely that a single
station would handle half the total capacity of a line. Where this
84
does happensuch as a single downtown terminal station
multiple platforms or dual-faced platforms will be required.
Although the analysis can be adjusted for the number of
provided platform faces at through stations, the estimation of
dwell times based on hourly passenger flow is not applicable to
terminal stations where other factors dictate the layover time.
This method is particularly valuable to estimate the changes
in headwayand capacityfrom increased passenger volumes
at an existing station. If land use changes or area growth
increase the estimated hourly usage of a station significantly, for
example, an additional 5,000 passengers per peak hour
directionthen the value of R (the ratio of busiest door usage to
average door usage) can be calculated rather than estimated
from the current dwell time. The difference between the
calculated dwell before and after the passenger growth can be
added to the existing peak dwell with potential accuracy within
2 seconds.
7.5.4 SELECTING AN OPERATING MARGIN

Chapter Six, Operating Issues, introduced the need to add an
operating margin to the minimum train separation and dwell
time to create the closest sustainable headway without
interference.
Ironically, the closer the trains operate, and the busier they
are, the more chance there is of minor incidentsdelaying service
due to an extended dwell, stuck door or late train ahead. It is
never possible to ensure that delays do not create interference
between trains nor is there any stated test of reasonableness for a
specific operating margin.7 A very small number of rail transit
lines in North America are operating at capacity and can
accommodate little or no operating margin. On such lines
operations planners face a dilemma of scheduling too few trains
to meet the demand, resulting in extended dwells and erratic
service, or adding trains to the point that they interfere with one
another. Striking a balance is difficult and the tendency in
practice is to strive to meet demandequipment availability and
operating budget permitting. While the absolutely highest
capacity is so obtained, it is poor planning to omit such an
allowance for new systems.
The more operating margin that can be incorporated in the
headway the better; systems running at maximum capacity have
little leeway and the range of operating margins used in the
simple procedure20 to 25 secremains the best guide. The
recommended procedure is to aim for 25 sec and back down to
20 or even to 15 sec if necessary to provide sufficient service to
meet the estimated demand. Where demand is unknown or
uncertain in the long term futurewhen a system in planning
reaches maximum capacitythen 25 sec, or more, should be
used.
When the controlling dwell has been estimated as the mean
dwell plus two standard deviations the operating margin can be
reduced to 10 seconds or less, or eliminated. Alternately the
greatest of mean dwell plus two standard deviations or mean
dwell plus operating margin can be used.
7
The principal investigator has discussed the concept of a goal with rail
transit planners based on an average of one disturbed peak period per ten
weekdays (two weeks) but has never seen such goals documented.
7.5.5 SELECTING A PASSENGER LOADING

LEVEL
Chapter Five, Passenger Loading Levels, discusses the wide
range of loading levels used in North America. Selecting a
loading level is a policy issue and the process for this complete
procedure is the same as that of the simple procedure. Use of the
passenger occupancy per linear meter of train is recommended.
In selecting a loading level take into account that this is for the
15-min peak-within-the-peak and that the average over the peak
hour and peak-period will be more relaxed.
If the line for which capacity is being determined is an
addition to an existing system then existing occupancy levels or,
where available, existing loading policies can be used. Some
cities have a wide variation of peak-within-the-peak loading
levels from line to line. Mexico City is probably the most
extreme example in North America. Where this variety exists
then the loading level should be selected based on the closest
matching linefor example, a heavy trunk serving downtown
or a cross-town feeder line.
Figure 7.3 and Figure 7.4 provide a range of loading levels
from 5 to 9 passengers per meter of car length for light rail and
from 7 to 11 for heavy rail. For new systems where attempts
are being made to offer a higher quality of service, the
recommended approach is to base the loading level on the
commonly suggested medium comfort level for new rail
transit systems of 0.5 m2 per passenger, averaged over the
peak hourthat is, no loading diversity factor is required.
This provides a recommended linear loading level of 6
passengers per meter of train length for heavy rail and 5 for
light rail.
An alternative approach is to base the loading levels on either
the nominal capacity of a vehicle or the actual peak-hour use.
The nominal capacity of a range of vehicles is shown in Table
7.5. Note that as previously discussed in this report the nominal
rated capacity can be an artificial and impractical crush level.
Table 7.5 is sorted in descending order of loading level. The
upper range should be discounted. A tone is applied over those
data that may be applicable for use in the complete method of
determining capacity. Note that the upper ranges of these levels
are still relatively high and the comfort accordingly low.
Table 7.5 also demonstrates the difficulty in determining
capacity when five essentially identical Siemens-Dwag light
rail vehicles from four different operators are examined. The
nominal capacities of these cars, highlighted with boxes, range
from 6.9 to 3.9 passenger per meter. This is a ratio of 1.8:1
despite the cars having almost the same dimensions and the
same number of seats.
Table 7.6 shows the actual peak-within-the-peak linear
loading levels for major North American trunks, again in
ascending order. Discounting the uniquely high values in New
York the remaining data offer realistic existing levels to apply in
selecting a loading level for a comparable systemor a new line
in the same system with similar characteristics.
It is interesting to note the difference between the actual levels
in Table 7.6 and the nominal (published car capacity) levels for
those systems represented in both tables. These are shown in
Table 7.7. The similarities (CTS-Calgary) and the variances (all
other systems) are a cautionary exercise in the acceptance and
use of published data. However in fairness to certain systems it
85
Table 7.5 Nominal agency or manufacturers car capacity
for heavy and light rail vehicles
Table 7.6 Passengers per unit train length, major North

American trunks, 15-min peak-within-the-peak
Service through NYCTs 53rd St. Tunnel is provided by line E, operating

18.35-m cars, and line F, operating 22.77-m cars. Seats and car loadings
are presented as E/F. The number of passengers per meter given is for
the combined lines; individually this value is 10.7 for the E and 10.0 for
the F.
Table 7.7 Passengers per unit train length, 15 min peakwithin-the-peak, nominal versus actual values (only the
busiest NYCT lines using each car type included)
8
Stated maximum or crush load passenger capacity per vehicle from the
operator or manufacturer. Schedules maximum loads for NYCT. Some
stated values for total passengers are well below realistic crush loading
reflecting an agencys desire to maintain comfortable loading levels.
should be pointed out that the official (nominal) car capacity

could be based on previous decades when higher loading levels
were expected and achieved on heavy rail systems.
7.5.6 DETERMINING AN APPROPRIATE

LOADING DIVERSITY FACTOR
10
The next step is to adjust the hourly capacity from the peakwithin-the-peak 15-min rate to a peak-hour rate using a loading
diversity factor from Chapter Five, Passenger Loading Levels.
The diversity factor is calculated according to Equation 7-4. The
diversity factor was used in Method 4 for calculating the dwell
time. If this method was used then obviously the same diversity
factor must be used. Otherwise the factor should be selected
based on the rail mode and the type of system. Table 7.8
provides existing examples. Unless there is sufficient

similarity with an existing operation to use that specific
figure, the recommended loading diversity factors are 0.80
for heavy rail, 0.75 for light rail and 0.60 for commuter rail
operated by electric multiple-unit trains.
This is the weighted average for scheduled loadings of both car types used
on this trunk. See also note 9.
86
Table 7.8 Diversity of peak-hour and peak 15-min
loading
11
12
Cmax = maximum single track capacity in

passengers per peak hour direction
= train length in meters
L
Pm = loading level in passengers per meter of
train length
Dph = loading diversity factor
The spreadsheet contains this calculation. Given the range of
values that can be calculated, estimated or assigned for certain
of the components in Equation 7-15, it is appropriate that the
results be expressed as a range.
The results should be checked for reasonableness against
typical capacities in Figure 7.10, which is based on the simple
procedure loading levels of 5 passengers per meter for light rail
and 6 passengers per meter for heavy railapproximately 0.5
m2 per passenger. Higher levels are possible only if less
comfortable loading levels have been used. Lower levels imply
either errors or that all seated passengers have been assumed or
an excessive operating margin has been included.
This chart is not an appropriate check for electric multipleunit
(emu) commuter rail whose signaling systems are usually
designed for lower throughput with loading levels based on all
seated passengers. Commuter rail capacity based on train length
is also affected by the common use of bi-level cars, although
few such trains currently fit into the applicable category of
electric multiple-unit operation. Figure 7.10 and an approach to
Grade Separated Rail Capacity Determination are contained in
the Excel version of the spreadsheet. The simplified step-by-step
approach, without charts and equations is also in the generic
version of the spreadsheet. Refer to the spreadsheet user guide at
the front of this report.
Mainly diesel haulednot EMU.

These data are suspicious.
7.5.7 PUTTING IT ALL TOGETHER

The final step in the complete method of determining a grade
separated rail transit lines maximum capacity is to determine
the closest (minimum) headway as the sum of the calculated
value of the minimum signaling system train separation, plus the
calculated or estimated value of dwell time plus the assigned
operating margin.
Hmin = T(s) + td + tom
Equation 7-13
The maximum number of trains per hour Tmax then is

Equation 7-14
The maximum capacity Cmax is the number of trains multiplied
by their length and number of passengers per meter of length,
adjusted from peak-within-the-peak to peak hour.
Equation 7-15
where
Hmin
T(s)
td
tom
Tmax
=
=
=
=
=
minimum headway in seconds

minimum train separation in seconds
dwell time in seconds
operating margin in seconds
train throughput per hour
Figure 7.10 Typical maximum passenger capacities of grade

separated rail transitexcluding all-seated commuter rail.
87
8. Light Rail Capacity

Determination
8.1 INTRODUCTION
This chapter covers methods for determining the capacity of
light rail transit lines. While the approach used in Chapter
Seven, Grade Separated Rail Capacity Determination, will
work in most situations, light rail transit lines often have
characteristics such as street running, grade crossings and single
track, which are not covered in that chapter but which are of
importance in capacity determination. The key to determining
the capacity of a light rail transit line is to find the weakest
linkthe location or factor that limits the capacity of the entire
line.
8.1.1 SELECTING THE WEAKEST LINK

Determining the capacity of light rail transit lines is complicated
by the variety of rights-of-way that can be employed. In the
simplest case, a grade separated right-of-way is used and the
capacity calculation techniques given in Chapter 7 can be
applied. However, most light rail transit lines use a combination
of right-of-way types, which can also include on-street operation
(often in reserved lanes) and private right-of-way with grade
crossings. Other limitations can be imposed by single-track
sections and the street block lengths. The line capacity is
determined by the weakest link; this could be a traffic signal
with a long phase length, but is more commonly the minimum
headway possible on a block signaled section. The first portion
of this chapter discusses the capacity limitations imposed by
right-of-way characteristics.
The capacity constraints are grouped in sections to in order of
decreasing relative importance for most systems. (See Table
8.1). This order is not definitive for all systems, but it is
appropriate for many. System-specific differences, such as short
block lengths on signaled sections, will change the relative
importance of each item.
8.1.2 OTHER CAPACITY ISSUES

Car loading levels for light rail transit for use in the equations in
this chapter should be determined with reference to the passenger
Table 8.1 Light rail capacity constraints
loading standards for light rail transit in Chapter Five,

Passenger Loading Levels. Light rail loading levels are
generally lighter than those for rail rapid transit but not as
generous as the one seat per passenger policy common on
commuter rail.
Light rail train lengths are more restricted than for rail rapid
transit or commuter rail because of lower car and coupler
strengths, and street block and station platform lengths. These
issues are considered in section of this chapter.
One additional issue which is of particular importance to light
rail operations and capacity is the method of access for mobility
impaired passengers. While the speed of each access method
varies, all can have an effect where close headways and tight
scheduling occur. The overall discussion of the impact of the
Americans with Disabilities Act (ADA) is contained in Chapter
Six, Operating Issues. More specific light rail accessibility
issues are dealt with in section of this chapter.
8.2 SINGLE TRACK

Single track is the greatest capacity constraint on light rail lines
where it is used extensively. Single-track sections are used
primarily to reduce construction costs. Some lines have been
built with single track as a cost-saving measure where the rightof-way would permit double track. In other areas single track
has been built because widening the right-of-way and structures
is impossible. Single-track sections can be very short in order to
by-pass a particular obstacle; for example, the San Diego
Trolley had a short single-track segment1 on the East Line in
order to save the cost of building a second overpass over an
Interstate highway. This segment has since been replaced with
double track as part of the double-tracking of the majority of the
San Diego Trolley system. When this program is complete,
single track will be used only on the East Line extension to
Santee.
The Sacramento light rail line, like San Diegos, featured
substantial single-track construction as a way to keep initial
costs low. However, the extensive use of single track has limited
operational flexibility and mandated a minimum headway of 15
min. This long headway has necessitated the use of 4-car trains
to meet the peak-period ridership demand. The length of these
trains is such that they block intersections while stopping at the
downtown stations. As in San Diego, much of the Sacramento
line is in the process of being double-tracked to remove these
constraints.
Tri-Met of Portland is also removing its single-track constraint
at the eastern end of its light rail line in Gresham. A second
1
Actually a gauntlet track with the four rails interlaced, but with the same
operational implications as single track.
88
track is being added on the existing right-of-way in order to
increase operational flexibility and reduce the anxiety train
operators have about arriving late at the single-track meet point.
The latter problem is caused by delays elsewhere on the line,
particularly wheelchair boardings and alightings.
Baltimores light rail transit line includes substantial singletrack construction but ridership demand has not yet been strong
enough to require double-tracking in the existing right-of-way.
While most of these newer light rail lines are moving away
from single-track operation, SEPTA depends on large sections
of single track on its much older Media and Sharon Hill lines.
Careful scheduling is used to allow an approximately 10-min
peak headway of mixed local and express services to operate on
each line. The common eastern portion of these lines is doubletracked.
While determining the extent of single track possible on a
system is possible, the exact layout is highly system specific.
Estimates can be made of the number of track kilometers
required for a certain number of route kilometers once the
intended headway is known.2 While this does not tell the user
where the single-track sections can be used, it can provide
assistance in determining the possible extent of single track for
use in cost estimates.
The distance covered in this time is

Equation 8-2
where
sbs = braking distance to stop
The distance and time covered to reach the maximum singletrack section speed involves specific vehicle characteristics as
the nominal acceleration rateusually identical to the braking
ratedecreases with speed. A reasonable approximation is to
assume that the average acceleration rate to the maximum
section speed is half the braking rate. The total time and distance
from start to stop then become
Equation 8-3
where
tss = time from start to stop

Equation 8-4
where
sss = distance covered start to stop
The time to cover a single-track section becomes
8.2.1 CALCULATING SINGLE-TRACK

HEADWAY RESTRICTIONS
Equation 8-5
where
Single-track sections greater than 400-500 m are potentially the

most restrictive capacity constraint for light rail. The headway
limitation is very simply TWICE the time taken to traverse the
single-track section, plus an allowance for switch throw and
lockunnecessary for spring switches or gauntlet track3plus
an operating margin to minimize the potential wait of a train in
the opposite direction.
This is a very site-specific time; however, a reasonable
approximation can be calculated from the length and maximum
speed on the section, based on the similar performance of
modern light rail vehicles.
The time to brake from the maximum line speed to a stop can
be expressed as
tbs
vmax
ds
tjl
tbr
=
=
=
=
=
=
=
=
=
time to cover single track section (s)

length of single track section (m
number of stations on single track section
average station dwell time on section (s)
Substituting for Sss from Equation 8-4, adding a speed margin to

compensate for the difference between actual and theoretical
performance on a manually driven system, adding the train
length to the section length and adding an operating margin
produces
Equation 8-6
where
Equation 8-1
where
Tst
Lst
Ns
td
time to brake to stop (s)

maximum speed reached (m/s)
deceleration & acceleration rate (m/s2)
jerk limiting time (s)
operator and braking system reaction
time
Tst
Lst
L
Ns
=
=
=
=
td
vmax
ds
tjl
tbr
=
=
=
=
=
SM
tom
=
=
time to cover single track section (s)

length of single track section (m)
train length (m)
number of stations on single track
section
station dwell time (s)
maximum speed reached (m/s)
deceleration4 rate (m/s2)
jerk limiting time (s)
operator and braking system reaction
time
speed margin5 (constant)
operating margin (s)
See Allen, Duncan W., Practical Limits of Single-Track Light Rail Transit
Operation in Appendix One.
3
Gauntlet track interlaces the four rails without needing switches, saving
capital and maintenance costs and potential operating problems due to
frozen or clogged switch points. The disadvantage is that the single-track
section cannot be used as an emergency turn-back (reversing) location.
4
5
Also used as a surrogate for twice the average acceleration from 0 to vmax.
An allowance to adjust for out of specification equipment and train
operators that do not push to the edge of the operating envelope, i.e.,
maximum permitted speed. Typically 1.08 to 1.2, 1.1 is used in the
results.
89
This equation can be readily solved using typical values from
Table 8.2
The value of the maximum single-track section speed should
be the appropriate speed limit for that section. 55 km/h (35 mph)
is a suitable value for most protected, grade separated lines. If
the single-track section is on-street then a speed below the
traffic speed limit should be used. If there are signaled
intersections an allowance of half the signal cycle should be
added to the travel time for each such intersection, adjusted for
any improvements possible from pre-emption.
This equation is included on the computer spreadsheet. A
selection of results is shown in Figure 8.1.
Trains should be scheduled from their termini so that meets are
not close to the single-track sections. Where there is more than
one single-track section this is difficult but not impossible.
Lengthy single-track sections can severely limit headways
and capacity and may require one or more double-track passing
Table 8.2 Data values for single-track travel time
sections in the single-track section. These should, wherever

possible, be of sufficient length to allow opposing trains to pass
on-the-fly and to allow some margin for off-schedule trains.
Obviously trains should be scheduled to pass at this location.
8.3 SIGNALED SECTIONS

Restrictions due to signaled sections are largely covered in
Chapter Seven, Grade Separated Rail Capacity Determination.
However, it should be realized that many light rail lines are not
signaled with the minimum possible headway in mind but more
economically for the minimum planned headway. This can
easily make signaled sections the dominant capacity constraint.
For example the Edmonton light rail line has a peak headway
of 5 min with this also being the minimum headway possible
based on the signaling. At the other extreme is New Jersey
Transits Newark city subway with a peak headway of 2 min
and a minimum headway of 15 sec being permitted by the
signaling. This is made possible with very short advisory signal
blocks, single car trains (PCCs) and multiple-berth platforms at
the terminals. Now only a single route, the city subway no
longer requires the capacity provided by such close blocks
except in unusual circumstances, however, similar arrangements
in Philadelphia are used much closer to capacity.
SEPTA currently schedules trains in the Market Street light
rail subway as close as 60 sec. The closely spaced two-aspect
color-light signaling is for spacing purposes only, that is it is
advisory. A driver can see several signals ahead. A range of
green allows full speed operation with the driver using judgment
to slow down as a red signal approaches. There are no train
stops and the car may pass a red signal and approach the car
ahead on line-of-sight to permit multiple berthing in a single
station.
Equally high capacity is provided at the City Hall terminal,
which is a large loop containing the multiple-berth Juniper
Street station. In past decades as many as 120 streetcars per hour
passed through the tunnel.
These arrangements are not fail-safe and collisions have
occasionally occurred. Multiple-berth stations can be confusing
to passengers but will improve with the ADA-required
information signage. However, with reasonable driver
discipline, these arrangements provide the highest light rail
capacitypotentially over 20,000 passengers per peak-hour
direction per trackand have provided it safely, economically
and at relatively high speeds for over half a century.
8.4 ON-STREET OPERATION

Figure 8.1 Light Rail travel time over single-track section
with speed limit of 55 km/h and various numbers of stations
train length 56 m, dwell time 20 sec, operating margin 20
sec, other data as per Table 8.2. The closest headway with a
single-track section is TWICE the above traverse plus operating
margin time.
Historically, streetcar operation has achieved throughput in

excess of 125 cars per hour on a single track in many North
American locations. Even now the Toronto Transit Commission
schedules single and articulated streetcars at a peak-within-the
peak rate of over 60 cars an hour on Queen Street East where
several car lines share a four block stretch.
Despite this record on-street operation is often raised as a
major capacity constraint for modern light rail systems yet this
90
is rarely the case on contemporary lines. This is particularly true
on most newer lines where light rail trains have exclusive use of
road lanes or a center reservation where they are not delayed by
other traffic making turns, queuing at signals or otherwise
blocking the path of the trains. Exclusive lanes for light rail are
also being instituted on some of the older streetcar systems
where congestion is severe; Torontos King Street is an
example.
Even with these improvements in segregating transit from
other traffic, light rail trains must still contend with traffic lights,
pedestrian movements and other factors beyond the control of
the transit operator. The transit capacity in these situations can
be calculated using the equations presented below.
8.4.1 EMPIRICAL APPROACH

Capacity is the product of train frequency and train capacity.
This can be given as
Equation 8-7
The maximum number of trains per hour can be determined
from Equation 8.2. Note that this should be applied for the
intersection with the longest traffic signal cycle or train dwell
time.
Equation 8-86
where
Cp = trains per hour per track

tc = clearance time between trains is defined as
the sum of the minimum clear spacing
between trains plus the time for a train to
clear a station, with typical values of 25 to
35 sec. (Some transit agencies use the
signal cycle time as the minimum
clearance time.)
D = dwell time at stop under consideration,
typically ranging from 30 to 40 sec,
sometimes to 60 sec.
R = reductive factor to compensate for dwell
time variations and/or uncontrolled
variable associated with transit operations.
R values are tabulated from 1.0 in perfect
conditions with level of service E, to
0.634 with level of service A, assuming
a 25% coefficient of variation in dwell
times. Maximum capacity under actual
operating conditions would be about 89%
of that under ideal conditionsresulting
in about 3,200 effective sec of green per
hour.
g = effective green time in sec, reflecting the
reductive effects of on-street parking and
pedestrian movements as well as any
impacts of pre-emption.
LEVINSON, HERBERT S., Capacity Concepts for Street-Running Light

Rail Transit, Australian Road Capacity Conference, 1994.
C = cycle length in sec. (Cycle lengths should

be divisible into 3600 to allow consistent
train scheduling with headways a multiple
of the cycle length, preferably no less than
two cycles, see tabular example below.)
Cycle Length
60
72
75
80 90 100 120
Cycles per hour
60
50
48
45 40
36
30
Minimum headway 120 144 150 160180 200 240
sec
sec
This empirical approach is often not appropriate for light rail

systems but may have value for traditional streetcar operation.
Note that on-street parking and pedestrian movements can
impact capacity. More details and examples can be found in the
Highway Capacity Manual (R67) and the ITE Transportation
Planning Handbook(R42)(R43).
8.4.2 PRACTICAL ISSUES

It is hard to encompass all the variables which affect on-street
light rail transit operation in a single formula. Note, for
example, the vagueness of the definitions of the R and g
variables in Equation 8.8 as a way to accommodate the less
concrete aspects of on-street operation. Even with these
vagaries, the capacity of on-street light rail is often greater than
on grade-separated, signaled rights of way where higher speeds
and block signals force the separation between trains to be
increased.
Variability due to traffic congestion has been reduced as a
factor as almost all recently built on-street light rail lines operate
on reserved lanes. A number of older systems still have
extensive operation in mixed traffic and so are subjected to the
variability in train throughput this causes by reducing g, the
effective green time for trains. Traffic queuing, left turns and
parallel parking can all serve to reduce light rail transit capacity.
Traffic signals can be a major impediment to light rail transit
operation where they are not designed with the needs of light
rail trains in mind. Poor traffic signaling can make train
operation slow, unreliable and unattractive to potential
passengers. These problems can be addressed through the use of
signal pre-emption and progression.
Signal pre-emption allows the light rail train to extend an
existing green phase or speed the arrival of the next one.
Depending on the frequency of intersections and traffic
congestion, this can have a substantial impact on the flow of
general traffic in the area. As a result, pre-emption in congested
areas is often limited in its scope so as not to have too negative
an effect on other traffic. The degree to which local politicians
and traffic engineers will tolerate the effects of pre-emption
plays a large role in determining the effectiveness of signal preemption schemes.
There is often a misconception of the impact of pre-emption.
At the modest headways typical of new light rail systems, where
trains operate only every few traffic light cycles, the green time
advanced or held for a light rail trains can be restored in the
following cycles with no net loss of cross-street capacity.
Edmonton demonstrated that by tying area traffic signals and the
light rail signaling system into a computer the introduction of
light rail actually increased capacity on both cross-streets and
parallel streets.
91
Signal pre-emption, linked to a central traffic control
computer, is being implemented extensively on the Toronto
streetcar system. Close stop spacing on the streetcar lines gives
pre-emption an edge over progression because of the limited
number of traffic signals between streetcar stops.
The San Diego Trolley originally used signal pre-emption on
its C Street downtown mall but has since switched to signal
progression. Increased light rail service on the mall had exposed
the inadequacy of the pre-emption controllers to deal with high
volumes of bi-directional traffic and resulted in failures. Table
8.3 contains some representative phase lengths for light rail
transit signal pre-emption and progression.
Signal progression has supplanted pre-emption in many
cases where light rail trains operate in congested downtown
areas. This technique gives trains leaving stations a green
window during which they can depart and travel to the next
station on successive green lights. The benefits of progression
increase with greater station spacing as less accumulated time
is spent waiting for the progression to start at each station. The
progression is frequently made part of the normal traffic light
phasing and so is fully integrated with signaling for
automobiles on cross-streets. This reduces delays for transit
and car drivers alike. Station stops are accommodated by the
train missing one light cycle and proceeding on the next.
Ideally the cycle length will be slightly longer than a long
average dwell in order to allow the majority of trains to leave
shortly after passenger activity has ended. Note that the
Calgary timings for progression in Table 8.3 were measured
on the 7th Avenue Mall which is shared with buses; the phases
must therefore be longer to accommodate both transit modes in
the same phase.
It is useful if the train operator can determine when the green
window at the first signal after a station will start as this allows
him to serve more passengers by maximizing the dwell time at
the station. In this way the train operator only closes the doors
when he knows that the train will soon be able to proceed. In
some cases this can be done by observing the operation of the
other traffic signal phases. However, this may not be possible at
some locations. In these cases a special signal display can be
added that counts down the time to the start of the light rail
phase, as at a number of locations on the downtown portion of
the San Diego Trolley.
Operating
heritage
streetcarsvintage
trolleysin
conjunction with light rail service can constrain capacity unless
operated over sections of the light rail (such as downtown San
Jose) where light rail speeds are already low. Figure 8.2 shows a
heritage streetcar on the downtown tracks of Portlands LRT.
Table 8.3 Average phase lengths at light rail transit

crossings (number of crossings observed in parentheses)7
Each crossing was usually monitored for four or more train movements or
until a consistent phase time had been established. Cycle times vary.
Figure 8.2 Heritage streetcar service in Portland. These cars

are accurate reconstructions of the historic Council Crest series
cars. They are built on relatively modern (PCC) trucks and
provide the acceleration, braking and safety featuresbut not
top speedsof modern light rail cars. Equipped with radio and
inductive communications, they operate the pre-emption in the
same manner as the light rail service cars. Even so, operation is
limited to outside weekday peak periods. If necessary to take a
heritage service into account in determining capacity, cars with
modern performance can be treated as the equivalent of a light
rail vehicle. Dwells, particularly with tourists and wheelchairs
can be extended and off-line stations may be necessary, as
provided by Tri-Met at Lloyd Center. The vintage cars may
require specific arrangements that are beyond the scope of this
project.
8.4.3 DETERMINING ON-STREET

CAPACITY
Capacity can be estimated by using Equation 8-8 where blocks
are long and trains are shortfor example a classic streetcar
operation. Where, as is often the case, light rail train lengths
approach the downtown block lengths then the throughput is
simply one train per traffic light cycle, provided the track area is
restricted from other traffic. When other traffic, for example,
left-turn lanes, may prevent a train from occupying a full block
throughput drops as not every train can proceed on receiving a
green light. A common rule of thumb is that the minimum
sustainable headway is double the longest traffic signal cycle on
the at-grade portions of the line.
8.5 PRIVATE RIGHT-OF-WAY

WITH GRADE CROSSINGS
Private right-of-way with grade crossings is the predominant
type of right-of-way for many light rail transit systems. This can
take the form of a route which does not follow existing streets or
one which runs in the median of a road physically separated
from other traffic except at crossings.
92
Capacity on lines with full pre-emption can be determined
using the methods for grade-separated rail transit given in
Chapter 7. However, allowances for any speed restrictions due
to grade crossings must be made. Where full pre-emption is not
available, Equation 8.8 for street running should be used to
determine line capacity since it incorporates the cycle length of
traffic signals, pre-empted or not.
8.5.1 PRE-EMPTION
Light rail transit lines operating on private right-of-way are
generally given full priority at grade crossings by railroad-type
crossbucks, bells and gates, or by traffic signal pre-emption.
Gated, railroad-style crossings are used where train and/or
traffic speeds are high. As shown in Table 8.3, railway-type
gated crossings consistently have the longest phase lengths of
the three main crossing devices. Crossbucks and bells alone, or
pre-empted traffic signals, are used where speeds are lower.
Delays to other traffic are reduced when gates are not used since
the time taken for gates to be lowered and raised is removed as a
factor.
Portlands Eastside MAX line offers an excellent example of
pre-emption. This line features a long section of median running
on a minor arterial street (Burnside Street). Train speed is
limited to the speed limit of the street and signal pre-emption is
used to allow trains to maintain this speed on the line segment.
Traffic signal phase time lost to the cross streets when lights are
pre-empted is returned in subsequent phases. Towards the
eastern end of this line segment the light rail tracks make a very
long, low-angle crossing, of a major arterial with the only
protection being the pre-empted traffic lights. (Figure 8.3) All
pre-empted crossings on the Tri-Met light rail line have signals
in advance to notify the train operator that the train has been
detected and that the signal will become permissive. As can be
seen in Table 8.3, the pre-emption system employed in Portland
is very effective in minimizing the delay to cross traffic while
giving light rail trains almost complete priority.
The SCCTA light rail line in San Jose also uses median
running an arterial street but local traffic engineers have only
given the light rail minimal priority over other traffic,
particularly during rush hours. Where the line runs through the
city of Santa Clara the light rail line has no priority over other
traffic and suffers substantial delays. Similar delays due to a
lack of priority face the Los Angeles Blue Line over the route
section between the end of the downtown subway and the start
of the old interurban right-of-way at the Washington Boulevard
station.
time at the crossing can be predicted consistently and premature

activation of the crossing is not a factor. The train is also either
coasting or braking through the crossing from cruising speed
and so will occupy it for less time.
Stations can be designed to place both platforms on one side
of the crossing or to locate one platform on each side of the
crossing such that trains use the crossing before stopping at the
station. Both arrangements are shown in Figure 8.4. Using
farside platforms is advantageous for the operational reasons
given above, reduced right-of-way requirements, and, for
median operation, allowing left turn bays to be readily
incorporated into the street.
Delays caused by premature activation of crossing gates and
signals at near side stations can be reduced using wayside communication equipment. This can be done with the operator being
equipped with a control to start the crossing cycle before leaving
the station or by an automatic method. The San Diego Trolley
shares some of its trackage with freight trains and uses a communication device that identifies light rail trains to crossing circuits
on the far-side of stations. If the crossing controller identifies a
train as a light rail train, a delay to allow for station dwell is
added before the crossing is activated. This ensures that the
Figure 8.3 Tri-Met light rail train approaching an angled

gated crossing (Note the gate is across the highway). The
potential delay to cross traffic at these crossings is almost three
times longer than with the 100% pre-empted signalized
intersections closer to downtown. At higher train frequencies
these occupancy times will become unacceptable and signalized
intersections would be requiredpotentially reducing light rail
speedsbut not the light rail capacity as the crossing occupancy
time is well within a normal green phase.
8.5.2 GRADE CROSSINGS AND STATION

DWELL TIMES
Grade crossing activation and occupancy times can be affected
by the presence of a station adjacent to the crossing. If the train
must use the crossing after stopping at a station, the activation of
the crossing signals is often premature and the crossing is
unavailable to other traffic for more than the optimum time. In
this case the train is also starting from a stop and so must
accelerate through the crossing, adding to the total delay. Where
the station platform is on the far-side of the crossing, the arrival
Figure 8.4 Light rail platform options at a crossing
93
crossing remains open for cross traffic for most of the time that
the light rail train is stopped in the station. If the controller
cannot identify the train as a light rail train, it assumes the train
is a freight and activates the crossing gates without delay.
Other systems use an inductive link between the light rail
train and wayside to activate pre-emption, switches and, in the
future, ADA-mandated information requirements. The lowest
cost detection approach is the classic overhead contactor.
Trolleybus technology using radio signals from the power
collection pick-up to coils suspended on the overhead wires is
also applicable to light rail but is not used in North America.
This arrangement can permit one light rail train per traffic
signal cycle. However, the possibility of interference with buses
held at a red light suggests the previously referenced maximum
throughput of one train per two signal cycles.
8.6 TRAIN LENGTH AND

STATION LIMITATIONS
8.6.1 STREET BLOCK LENGTH
The length of street blocks can be a major limitation for at-grade
systems which operate on-street. Most jurisdictions are
unwilling to allow stopped trains to block intersections and so
require that trains not be longer than the shortest street block
where a stop is likely. This issue is especially noteworthy in
Portland where unusually short street blocks downtown limit
trains to two cars. The San Diego Trolley also faced this issue
when they operated four-car trains on the East Line for a time.
Since three cars is the maximum that can be accommodated by
the downtown blocks, trains were split in two sections before
entering downtown.
Sacramento is an exception to the street block length rule and
is able to operate 4-car trains in the peak hours. These long
trains block one intersection when stopped. This situation is
almost a necessity as the extensive single-track nature of the
Sacramento line imposes a minimum headway of 15 min on the
service. The capacity limitation of this headway restriction is
therefore partially made up for by the operation of relatively
long trains.
Street block length is also an issue if another vehicle occupies
the same lane used by light rail trains in a block. If this would
cause the rear of the train to protrude into an intersection then
the train must wait for the block to clear before advancing. This
fact provides a strong argument for the provision of an exclusive
light rail transit lane where street running with long trains
occurs. Indeed, operation with mixed traffic is very rare on new
light rail transit systems, likely as a result of this concern.
Where buses and light rail transit trains operate alongside each
other on transit malls in Baltimore and Calgary, the rail stations,
bus stops and lanes are laid out to cause a minimum of
interference between the modes.
as stations have been built with current ridership and service

levels in mind. The relative importance of this constraint is
much greater for commuter rail where platform length is often
constrained for historical reasons.
A more important restriction can be in the design of terminal
stations. Torontos streetcars face terminal design problems
where two or more routes share a common terminal and singletrack turning loop. This is the case at the Broadview and Dundas
West subway stations where there is heavy transferring activity
between the subway and streetcars. The high volumes of transit
vehicles and passengers can cause delays to following streetcars
while passengers board and alight from the preceding car. Any
scheduled recovery time for the streetcar operator is hard to
accommodate in these conditions since the volume of following
cars will practically force cars ahead out of the loop.
The Baltimore light rail line also uses single-track termini but
the level of service (15-min headway) is not high enough for
these to be a capacity limitation. However, the terminals are
designed to allow an arriving train to unload passengers before
the departing train ahead leaves through the use of an extra
platform as shown in Figure 8.5. This arrangement allows the
location of a station in a relatively narrow right-of-way since the
platforms are not adjacent to each other and a wider center
platform is not required. Note that single-track termini, while
inexpensive, have limited flexibility and should generally be
avoided.
8.7 WHEELCHAIR
ACCESSIBILITY EFFECTS
8.7.1 INTRODUCTION
The accessibility of light rail transit to wheelchairs and other
mobility devices (considered together with wheelchairs in this
section) is a major issue for light rail transit systems. The
relative rarity of level loading with high-level platforms on light
rail has resulted in a variety of methods having been devised to
allow wheelchair access to light rail vehicles. Each of the
methods is outlined in the sections which follow. Chapter Six,
Operating Issues, has discussed general capacity issues related
to the ADA, including typical light rail provisions. This section
expands the discussion and adds specific arrangements of
individual operators. The illustrations of wheelchair loading
options, Figures 6.19 through 6.23, are not repeated.
Boarding and alighting times with non-level loading of wheelchairs tend to be highly variable depending on the skill of the
passenger. Experienced users can be remarkably quick. Passenger movement times are often lower than for lift-equipped buses
8.6.2 STATION LIMITATIONS

An obvious limitation to train length is the length of station
platforms. For most light rail transit routes this is not a problem
Figure 8.5 Single-track terminus with separate unloading

platform (Baltimore)
94
as there is more room to maneuver wheelchairs, walkers and
scooters in light rail vehicles. Off-vehicle fare collection also
helps to speed loading for mobility impaired and able-bodied
passengers alike. Some agencies require the passenger and
wheelchair to be strapped in, a time consuming process which is
becoming less common. Some systems have experienced
passenger conflicts over mobility device seating priority when
other passengers occupy the folding seats provided to create
space for wheelchairs and other mobility devices.
It should be noted that both mobility impaired passengers and
transit agencies prefer access methods that do not single out the
mobility impaired passengers for special treatment. Lifts and
special ramps cause delays which reduce the reliability of the
service while isolating those users from other passengers.
Mechanical devices such as lifts can also fail and put a train out
of service. For these reasons, the popularity of lifts and other
special devices for mobility impaired passengers will likely
decrease in favor of more reliable and less exclusionary methods
such as low-floor cars.
Reducing the delays associated with wheelchair boardings
and alightings is an important issue where capacity is
constrained. This is of particular concern on lines with single
track.
8.7.2 HIGH PLATFORMS

High-level platforms allow level movement between the
platform and the car floor. This allows universal access to all
cars of a train and removes the reliability and exclusionary
effects associated with lifts, ramps and special platforms.
Passenger flow is speeded for all passengers since there are no
steps to negotiate on the car. Unfortunately this is not an ideal
access method for light rail as high-platform stations are bulky
and costly to construct on in-street sectionsdefeating two of
the major benefits of light rail, low costs and community
friendly design. Nevertheless high platforms are used
exclusively on a number of systems including Los Angeles, St.
Louis and Calgary.
High-level platforms at stations are also used in Buffalo,
Pittsburgh and San Francisco; in combination with low-level
loading at other stops. Buffalo is unusual in that a subway, with
high-level platforms, serves the outer portion of the line while
the downtown segment is on a transit mall with low-level
loading using fold-out steps and mini high platforms (discussed
below) for wheelchair access. Pittsburgh has separate doors for
each platform level while the San Francisco Muni uses cars
fitted with steps which can be raised to floor height where high
platforms exist.
The profiled platform shown in Figures 6.21 and 6.22 has not
been used in North America but has proved effective in
Manchester offering low cost, low intrusion, fast passenger
movements and mainstream wheelchair loading.
8.7.3 LOW-PLATFORM METHODS
Lifts are mounted in the cars so that the first door on the right
side of every train is lift-equipped. When not in use, the lift is
stored in a vertical position which completely blocks the
doorway to use by other passengers. While the lift initially was
prone to failure, the current installation is quite reliable with a
failure rate of about one-quarter of a percent.8
Boarding and alighting times with the car-mounted lifts are
around 1 min for each passenger movement. However, the need
for the train operator to leave the cab to operate the lift adds to
the time required and can mean the total dwell time extends to
1 or 2 min when the lift is used.
Platform-Mounted Lifts
Platform-mounted lifts are used by the Portland and San Jose
light rail systems. They offer advantages over car-mounted lifts
in that all car doors are left available for other passengers when
the lift is not required, the lift is not subject to car vibration, and
the failure of a lift need not remove a car from service.
Disadvantages include the precise stopping requirements,
increased susceptibility to vandalism and an increase in the
distance that the train operator must walk to operate the lift.
For the SCCTA in San Jose, wheelchair handling is slow
because of their wayside lift arrangement. The lift is stored
vertically in an enclosed housing at the front of each platform.
To operate the lift, the train operator must raise sliding steel
doors on each side of the lift housing, lower the car side of the
lift to floor level, lower the platform side to ground level, have
the passenger board the lift, raise the lift and board the
passenger, store the lift and secure the housing. This procedure
takes 2 to 3 min giving a total train delay (including loading and
unloading) of 4 to 6 min per passenger requiring the lift. These
delays can easily consume the trains scheduled terminal
recovery time. An average of 25 wheelchairs and scooters are
carried each weekday on the SCCTA light rail line but this has
increased to as many as 50 a day for special events.
Tri-Met in Portland uses a different type of wayside lift.
Under normal circumstances the lift is at ground level ready to
receive intending passengers. The presence of the passenger on
the lift signals the passengers intention to board to the train
operator. The train operator then aligns the first door of the train
with the lift and boards the passenger. The cars steps are
bridged by a folding plate on the lift. This configuration speeds
the use of the lift somewhat but does not prevent it from having
an effect on punctuality. The average time required for each
mobility device movement was given as 1 min 50 sec by TriMet staff but this could increase to 4 or 5 min in a worst case
scenario with an inexperienced user. The determination of the
train operator in minimizing dwell in the use of the lift also
varies.
Tri-Met expects to be able to remove the wayside wheelchair
lifts by September 1997 when all trains will include an
accessible low-floor car. Section 6.10 of Chapter Six, Operating
Issues, suggests that other operators will follow Portlands lead,
greatly reducing the potential for wheelchair-related delays in
the future.
Car-Mounted Lifts
8
Car-mounted lifts are used only on the San Diego Trolley, one
of the first light rail transit systems to be wheelchair accessible.
Based on San Diego Trolley data for May 1994. Out of 1,069 lift
passengers carried (2,138 lift cycles) only six failures were recorded
giving a failure rate of 0.28%.
95
Mini-High Platforms
The current trend for wheelchair access to low-loading, highfloor light rail cars is the use of mini-high or high-range
platforms that provide level loading to the wheelchair accessible
door of the train. This method is mechanically simple and
generally uses a folding bridgeplate, manually lowered by the
train operator, to provide a path over the stepwell between the
platform edge and vehicle floor. The mini-high platform is
reached by a ramp or, where space limitations require, by a
small lift. In Sacramento, one of the pioneers of mini-high
platforms, these lifts are passenger operated and the intending
passenger must be on the mini high platform for the train
operator to board them. The Sacramento system handles about
1,200 persons in wheelchairs and five times as many strollers a
month on the mini-high platforms. Mini-high platforms have
been adopted for the new non-level loading light rail lines in
Baltimore and Denver.
The San Francisco Municipal Railway has also installed minihigh platforms at key locations on its surface lines (the
downtown subway is high platform). The cars must make a
special stop to board and alight passengers using the mini-high
platforms as the moveable steps on the car must be raised and
the center door aligned with the platform in order for level
loading to take place. The steps are usually raised before the car
has come to a stop. An elastic gap filler is used between the
platform edge and car doorway. No bridge plate is needed and
the train operator does not have to leave the cab. This
arrangement, aside from the need for a second stop, is very
efficient with the time required for a passenger movement being
under 10 sec. Two of the major surface stops on the Muni
system have been converted entirely to high platforms with
proof-of-payment fare collection to speed general passenger
flows with the additional benefit of making wheelchair loading
and unloading easier.
8.7.4 LOW-FLOOR CARS

Low-floor cars9 offer a straightforward solution to the need for
universal access to light rail vehicles. By bringing the floor
height down to just above the railhead, boarding is simplified
for all passengers as steps are no longer required. Small,
extendible ramps and slight increases in platform edge height
allow passengers in wheelchairs and other mobility devices to
board without the aid of lifts or special platforms10. Low-floor
cars provide much of the benefit of level loading without the
need for high platforms. Typical floor height is 350 mm11 (14
in.), about double the height of a normal curb. Medium or
intermediate height platforms are therefore still required for no
step boarding. Bridging plates with staff attendance are still
required on most designs although it appears that passengers
with pushchairs and many wheelchair users elect to navigate the
gap without this assistance.
9
Note the difference between the terms low-floor car and low-level loading.
The former states that the majority of the floor of the car is slightly above
curb height; the latter describes cars (low-floor cars included) where
passengers can enter from street level, without the need for platforms.
10
Some low-floor car/station platform arrangements require a manually
positioned bridging plate that can extend dwell times.
11
Certain low-floor designs ramp down the doorways to achieve a 280-300mm floor height.
While low-floor cars have operated in Europe for over a

decade, the first North American operation will begin in
Portland in 1997. The use of at least one low-floor car in every
train will allow Tri-Mets existing platform mounted wheelchair
lifts to be removed. Boston has also ordered low-floor cars to
make its Green Line subway-surface routes accessible. As in
Portland, the cars will be compatible with the agencys existing
fleet to allow mixed-train operation. Toronto is also expected to
acquire low-floor cars for use on the Spadina LRT line under
construction but purchase has been postponed because of a
surplus of existing streetcars.
Low-floor cars have some drawbacks which have yet to be
fully resolved. Cost is a problem with any new technology, lowfloor cars included. Cars with a 100% low floor can cost up to
double those with a 70% low-floor design, such as in Portland
which in turn carry a 25-30% cost premium over conventional
high-floor light rail vehicles. With a partial low-floor, the ends
of the car and the driving (end) trucks, and sometimes the
articulation, can be of conventional construction and can retain
component and maintenance commonality with existing light
rail equipment.
Steps inside the car provide access to the high-floor sections.
100% low-floor designs require the use of stub axles, hub
motors and other space-saving components. These items add to
costs and have not yet been satisfactorily proven for high-speed
use or on the tracks typical of North America. As a result, the
cars on order for Portland and Boston will be of the partial lowfloor type. Despite high costs and technical challenges, the
substantial benefits of low-floor cars have made them a popular
choice in Europe and broader North American use will likely
follow for those systems with on-street low level loading.
A published Transit Cooperative Research Program report,
Applicability of Low-Floor Light Rail Vehicles in North
America, deals extensively with this issue.
8.8 CAPACITY
DETERMINATION
SUMMARY
Calculating the capacity of light rail transit lines is a complex
process because of the varieties of rights-of-way that can be
employed for the mode. The basic approach is to find the
limiting factor or weakest link on the line and base the capacity
on this point. The limiting factor for each line could be streetrunning with long traffic signal phases, a section of single track,
or the length of signal blocks where block signaling is used.
The key factors to be considered are as follows:
1. Single track.
2. Signaled sections. Of particular importance where, for cost
reasons, the signaling is not designed to allow minimum
possible headway operation.
3. On-street operation. Capacity effects are strongly related to
the degree of priority given to light rail vehicles relative to
other traffic.
4. Private right-of-way with grade crossings.
96
The first step in the process is to check the headway
capabilities of any single-track section over 500 m (1600 ft) in
length from the procedure in section 1.2 of this chapter. Then
compare this with the design headway of the signaling system
and with twice the longest traffic signal phase of any on-street
section. Select the most restrictive headway in seconds and
convert into trains per hour by dividing into 3600. The simple
procedure provides a reasonable estimate of capacity by using
the range of loading levels shown in Figure 8.6, derived from
Figure 5.7 of Chapter Five, Passenger Loading Levels, with the
incorporation of a loading diversity factor range from 0.70 to
0.90. An example for a typical medium capacity light rail
system has a 400-m single-track section without a station.
Figure 8.1 shows this limits headway to 2 80 sec including an
operating margina total of 160 sec. The system operates fourcar trains on-street. As these are the length of the shortest city
block headway is limited to twice the traffic signal cycle of 80
sec, or 160 sec. Sections of right-of-way are signaled for 3-min
headway180 sec.
Typical of such systems, the right-of-way signaling becomes
the limitation allowing a maximum of 20 trains per hour. Four
car trains of 25-m articulated light rail vehicles at the midpoint
loading of 5 passengers per meter produces an hourly capacity,
inclusive of a loading diversity factor, of 4 25 5 20
10,000 passengers per peak-hour direction. Note that at this
frequency the ability to schedule trains to avoid delays on the
single-track section is unlikely. This will not reduce capacity but
add delays that require more vehicles and crew to carry that
capacity.
Where there are no single-track or on-street constraints and
the signaling system is designed for maximum throughput, the
Figure 8.6 Recommended loading level range for light rail

vehicles in simple capacity calculation, loading diversity
factor 0.70 to 0.90
Figure 8.7 Light rail capacity on segregated right-of-way

with maximum cab-control signaling system throughput
based on range of dwell time plus operating margin of 45 to
70 sec. (headway varies with train length, refer to Fig. 7.1)
maximum capacity can be determined through the procedures of

Chapter Seven, Grade Separated Rail Capacity Determination,
summarized for shorter light rail trains in Figure 8.7. At the
upper end of these levels the system has become a segregated
rail rapid transit system using light rail technology.
No allowance is contained in Figure 8.7 for extended dwells
due to low-level (step) loading, wheelchairs or on-board fare
collection. At minimum headways with cab-control better than
120 sec it is reasonable to expect level loadingwhether high or
lowand off-vehicle fare collection.
Nor is any allowance made for headway constraints due to
junctions or speed restrictions in the maximum load point
station approach. Where any of these situations may apply, the
complete procedures of Chapter Seven, Grade Separated Rail
Capacity Determination, should be followed.
Predominantly segregated and signaled light rail can reach the
achievable capacity of some rail rapid transit systems. At this
upper end of the light rail spectrum achievable capacity
calculations should follow those of rail rapid transit.
Note that no light rail lines in North America exceed a
capacity of 10,000 passengers per peak-hour direction per track.
The exception is Mexico Citys Line Areally a steel-wheeled
metro line with six-car trains on entirely segregated right-ofway. MBTAs Green line trunk is the closest system to 10,000
passengers per peak-hour direction. Achievable capacities to and
above 20,000 passengers per peak-hour direction are reported in
Europe, however, at these levels, the lines, often called premetro or U-bahn, have many or all of the characteristics of rail
rapid transit operated by light rail equipment.
97
9. Commuter Rail Capacity

Determination
9.1 INTRODUCTION
Commuter rail in North America is dominated by the systems in
the New York area where the busiest routes use electric
multiple-unit trains on dedicated tracks with little or no freight
service. Annual ridership is shown in Figure 9.1. The capacity
of such systems can best be determined from the procedures of
Chapter Seven, Grade Separated Rail Capacity Determination.
Care must be taken to take into account the sometimes lower
vehicle performance and lower throughput of signaling systems
where these are based on railroad rather than rapid transit
practices. Elsewhere, with the exception of SEPTAs
Philadelphia lines, Chicagos Metra Electric and South Shore
lines, and the Mont-Royal tunnel line in Montral, commuter
rail uses diesel locomotive-hauled coaches and follows railroad
practices. Electric locomotive-hauled coaches are also being
used by SEPTA and New Jersey Transit (NJT) on routes which
also see electric multiple-unit cars. Dual powered (electric and
diesel) locomotives are used by the Long Island Rail Road
(LIRR) and Metro-North Railroad in the New York area. All
new starts are likely to use diesel locomotive hauled coaches.
For most commuter rail lines the determination of capacity is
at once both simple and inexact. Unlike the grade separated rail
capacity determination, there are no reasonable methodologies
that allow the calculation of the train control throughput and
controlling dwell times to produce the achievable passenger
capacity of a line.
The number of trains that can be operated in the peak hour is
dependent on negotiations with the owning railroad. Many
factors are involved, single or double (or more) track, the signal-
Figure 9.1 Commuter rail ridership (millions per year)
ing or train control system, grade crossings, speed limits, freight

service, switching servicesand the priorities to be accorded to
these. Although railroads are becoming more conducive to
accommodating commuter rail servicesand the revenue and
capital upgrading they producethey have the upper hand and
obtaining slots (alternately called paths or windows) for
commuter trains at a reasonable cost is often a difficult and
protracted business.
There are an increasing number of exceptions where the
operating agency has purchased trackage and operating rights
and so has more say in the operation and the priority of
passengers over freight. The two New York carriers own the
track they operate on while NJT, SEPTA, the MBTA, Metra and
Los Angeles Metrolink, among others, own substantial portions
of the trackage they use. Some agencies, such as SEPTA, have
leverage with the freight railroads as they own track used by the
freight carriers as well as the reverse. However, there may still
be strict limits on the number of trains that can be operated
because of interlockings and grade crossings with other
railroads.
Unlike the capacity determination chapters for other modes,
commuter rail is not provided with both simple and complete
methods for determining achievable capacity. Once the number
of trains that can be operated in an hour has been determined,
the capacity is not dependent on loading standards but on only
the number of seats provided on a train.
9.2 TRAIN THROUGHPUT

Determining train throughput requires consulting the railroad
agreement or the railroad or agency signaling engineers to
determine the maximum permitted number of commuter trains
per hour. Generally these numbers will be based on a train of
maximum length, so the length-headway variations of Chapter
Three, Train Control and Signaling, will not enter into the
picture.
A definitive answer may not always be obtained, particularly
with single-track sections that are shared with freight. Freight
traffic can vary and available commuter rail paths can vary.
Usually the agreement will ensure a minimum number of
commuter rail slots per hour. These may be uni-directional
that is all trains must platoon in one direction in each peak
period. This is generally not a capacity problem but rather an
efficiency issue with respect to equipment and staff utilization.
Uni-directional operation is an issue on lines where reverse
commuting to suburban work sites is important. Indeed,
Chicagos Metra is planning new services aimed specifically at
the reverse commuter.
The number of slots available per hour may range from one
upwards into the double digits. Ten or more trains per hour is at
the upper range of traditional railroad signaling and will exceed it
98
if long, slow freights must be accommodated. At the upper end
of this range, commuter rail is effectively in sole occupancy of
the line for the peak period and can approach 20 trains per track
per houra 3 min headway.1 When electric multiple-unit
commuter trains have similar performance to rail rapid transit,
the capacity calculations of Chapter Seven, Grade Separated
Rail Capacity Determination, can be used as a rough
approximation of railroad signaling throughput by using the
longer train length and adjusting the separation safety factor B
from the suggested value of 2.4 for a rapid transit three-aspect
signaling system to 3 or 4.
However caution should be exercised as some multiple-unit
trains may not have all axles or cars powered; that is, the consist
may be made up of motored and trailer cars. Locomotive-hauled
commuter trains vary in power, length and gearing ratios
making it difficult to cite typical acceleration rates and
impractical to adapt the general calculations used in Chapter 7.
This equation and the associated equation for junction
throughput do not apply in locations and times where freight and
commuter rail trains share trackage or where the signaling
system is designed solely for freight with long blocks.
Additional complications are raised by the variety of services
operated and the number of tracks available. The busier
commuter rail lines tend to offer a substantial number of
stopping patterns to minimize journey times and maximize
equipment utilization. A common practice is to divide the line
into zones with trains serving the stations in a zone then running
express to the station(s) in the central business district. Through
local trains provide connections between the zones. A number of
lines in the Chicago and New York areas are operated this
wayMetras Burlington Northern line to Aurora operates with
five zones in the morning peak; Metro-Norths New Haven line
(including the New Canaan Branch) operates with seven zones.
Such operating practices are made possible with three or more
tracks over much of the route and the generous provision of
interlockings to allow switching between tracks. Grade
separated junctions are also common where busy lines cross or
converge. Commuter rail throughput at complex interlockings
associated with some stations and junctions, for example Harold
Junction on the LIRR, requires specialized analysis that is
beyond the scope of this report.
9.2.1 STATION CONSTRAINTS

Another principal difference between commuter rail and the
other rail transit modes is that commuter rail trains are often
stored at the downtown terminals during the day. This reduces
the need for track capacity in the off-peak direction and allows a
higher level of peak direction service to be operated. MetroNorth, with 462 platform tracks at Grand Central Terminal, is
thus able to use three of its four Park Avenue tunnel tracks in
the peak direction. Even when one of the tunnel tracks was
1
Other typical commuter rail headways can be found in the ITE

Transportation Planning Handbook (R42 and R43).
There is some variation between sources regarding the size of Grand
Central Terminal, Metro-North reports 46 platform tracks. A number of
other sources give the station a total of 67 tracks, including storage and
maintenance tracks.
closed for reconstruction, 23 trains per hour were handled on the

remaining two peak-direction tracks.
The situation at New Yorks Penn Station is less relaxed
where the LIRR has exclusive use of five tracks and shares four
more with Amtrak and NJT. Currently the LIRR operates the
East River tunnels with two tracks inbound and two tracks
outbound with a peak headway of 3 min per track. With limited
station capacity, two-thirds of LIRR trains continue beyond
Penn Station to the West Side Yard. However, not all tracks
used by the LIRR at Penn Station continue to the yard and some
trains must be turned in the station. This can be done in as little
as 3 min in a rush but 5 min is the minimum scheduled.
Capacity into the station could be increased by improving track
connections to the West Side Yard and so further reducing the
number of trains which must be turned in Penn Station; this
change would permit the East River tunnels to be operated with
three tracks in the peak direction and allow the operation of
additional trains.
9.2.2 STATION DWELLS

Station dwell times on commuter rail lines are generally not as
critical as they are on rapid transit and light rail lines as
frequencies are lower and major stations have multiple
platforms. In most cases the longest dwells are at the downtown
terminals where the train is not blocking others while passenger
activity takes place. Passenger flows are generally unidirectional and so are not slowed by passengers attempting to
board while others alight and vice-versa. Exceptions are
locations where major transferring activity takes place between
trains but these are limited. Jamaica station on the LIRR is an
example.
SEPTAs four track regional rail tunnel through Center City
Philadelphia is one of the few locations where commuter trains
run through from one line to another without terminating
downtown. SEPTA schedules provide a very generous time of
10 min for trains to make two station stops over this 2.3 km-line
segment.3
Commuter rail station dwell times are dependent on the
platform level and car door layout. The busiest lines are
equipped with high platforms and remotely controlled sliding
doors, as on rapid transit cars. Single-level cars often use
conventional traps for high- and low-platform stations but these
are time consuming to operate and require a large operating
crew. Cars used on lines with both high and low platforms can
be fitted with conventional trap doors at the car ends and sliding
doors for high-platform use at the center of the car, as on NJT,
the South Shore in Chicago and the Mont-Royal line in
Montral. Most bi-level and gallery cars are designed for low
platforms and have the lowest step close to the platform for easy
and rapid boarding and alighting. Bi-level cars of the type
popularized by GO Transit feature two automatic sliding
double-stream doors per side allowing cars to be emptied in 1 to
2 min. Gallery cars usually feature one exceptionally wide door
(2-m wide) at the center of each side to allow rapid boarding and
alighting with multiple passenger streams.
3
While there are three stations on this segment, the timetables only provide
departure times and so do not include the dwell time at the first Center City
station. Go Transit is the other agency that through routes commuter trains.
99
The estimation process for dwell times in Chapter Four,
Station Dwells, should not be used for other than multiple-unit
equipment with power operated sliding doors. Generally
locomotive-hauled commuter rail equipment (and in some cases
EMUs) have fewer doors, not all of which may be in use. Dwell
times can be extended when passengers have longer to move
within a car or train to an open door.
9.3 TRAIN CAPACITY

Except for a few situations where standing passengers are
accepted for short distances into the city center, commuter rail
train capacity is based solely on the number of seats provided on
each train. A loading diversity allowance of 0.9 or 0.95 is used.
Where the equipment is known, the best procedure is to add
the number of seats in a train. Unless there is an agency policy
of peak-hour occupancy at 95% of total seats, the 0.90 factor
should be used. Where trains are the same length, the commuter
rail capacity is simply:
(trains per hour) (seats per train) 0.90
In many cases train length is adjusted according to demand. The
longest train will be the one arriving just before the main
business start timeand vice-versa in the afternoon. Shorter
trains may be used at the extremities of the peak period. In this
case the total number of seats provided over the peak hour must
be determined and the loading diversity factor applied.
Where the commuter rail rolling stock is unknown the number
of seats per unit length of train can be used, based on the
shortest platform that the service will stop at. A number of
systems, particularly older ones, operate trains which exceed the
platform length at a number of stations. This situation is
particularly common where platforms are constrained by
physical and builtup features. Passengers must take care to be in
the correct car(s) if alighting at a station with short platforms.4
Train length on electric lines can also be limited by the amount
of current the overhead or third-rail is able to supply.
Table 9.1 shows the seats and seats per meter length of all
existing North American commuter rail cars, in descending
order. All cars have substantially the same dimensionsthe
AAR passenger car maximums of 25.2-m long (82.7 ft) and 3.2m wide (10.5 ft). A complete table of car dimensions, doors and
ADA accessibility types is provided in Appendix Three and on
the computer disk.
Passengers per meter of car range from over 7 to below 2. At
the high end are the double-deck car types, bi-levels5 and gallery
cars. 3+2 seating is needed to reach 7 passengers per meter (7/m)
length. Such seating is not popular with passengers and the
middle seats are not always occupied with some passengers preferring to stand for shorter trips. A capacity of 7/m can be used
as a maximum. A range of 5/m is the upper end for single level
cars. with 4/m preferred. These preferred and recommended
4
Another common station limitation, lack of park and ride capacity, is

considered in Chapter Six, Operating Issues.
5
Also called tri-levels on certain systems as there is an intermediate level at
each end over the trucks.
Table 9.1 Commuter rail car capacity
100
Table 9.1 Commuter rail car capacity continued
levels allow some space for toilets, wheelchairs and bicycles. If

these provisions are extensive then the car capacity should be
reduced accordingly.
Obviously the train length should exclude the length of the
locomotive(s) and any service cars, if any, and should be adjusted
for any low-density club, bar or food service cars. An allowance
for standing passengers is not recommended. However if the

nature of the service has significant short trips it may be
appropriate to add 10% to the number of seats on the train.
Heavy rail type standing densities from Chapter Five, Passenger
Loading Levels, are not appropriate for commuter rail and
should not be used.
GO TO CHAPTER 10
101
10. Automated Guideway

Transit Capacity Determination
10.1 INTRODUCTION
Automated guideway transit (AGT) generally fits into the
category of Grade Separated Rail whose capacity determination
is specified in Chapter Seven. However, there are some nuances
specific to AGT that must be considered. AGT is an almost
negligible part of urban, public, fixed guideway transitless
than 1/10th of one percent. Technology ranges widely from the
standard gauge advanced light rapid transit downtown people
mover in Detroit to small scale monorails in amusement parks.
Setting aside the possible interpretation of the Tandy shuttle
in Fort Worth as AGToperated by heavily rebuilt, manned
PCC streetcarsall AGT systems are proprietary designs. As
such their performance, acceleration, braking rate, balancing
speed and vehicle size and capacity vary greatly.1
(160 ft)based on the specific AGT values in Table 10.2, with

terms adjusted from typical rail transit values shaded Refer to
Chapter Three, Train Control and Signaling, Equation 3-15. The
results show that separation times with a simulated single-aspect
block system are two to three times longer than with the more
complex and expensive moving-block signaling system.
The moving-block results agree with those of Auer(R09), the only
10.2 TRAIN CONTROL

SEPARATION
Train control systems on AGT range from a sophisticated
moving-block signaling system to a basic manual system in
which only one train may be on a section of lineor the entire
lineat a time. Manual or radio dispatching may ensure that a
train does not leave a station until the leading train has left the
station ahead. One variant uses sectioned power supply. Power
is disconnected for a given distance behind an operating train.
These variants are not fully accommodated in the
methodology of Chapters Three and Seven. If the basic AGT
performance indices are known then the procedures of Chapter
Seven will provide an approximation of the minimum train
separation time for a range of AGT train controlsfrom a
moving-block signaling system to a simple fixed-block system.
A surrogate of this can be roughly simulated by setting the train
detection uncertainty factor (B) at four times the minimum
braking distance.
The results are shown in Table 10.1 and Figure 10.1 for trains
of typical AGT lengths12.5 m (40 ft), 25m (80 ft) and 50m
Figure 10.1 AGT train separation versus length
Table 10.2 Suggested AGT separation calculation values
Table 10.1 AGT minimum train separation times
Details of AGT system characteristics and technology are outside the scope
of this report. Details of selected systems can be found in Table 5.15 of
the ITE Transportation Planning Handbook (R42).
Default values for heavy rail. Refer to Chapter Three, Train Control and
Signaling.
102
reviewed paper specializing in AGT train control. Here, typical
short train AGT separation with moving-block control was cited
at 15 sec. The separation range is wide and highly dependent on
the train control system of the proprietary AGT system. The best
method of determining the minimum train separation is from the
system manufacturer or designer. Using the methodology of
Chapter Three should be a last resort when specific separation
information is not available.
10.3 PASSENGER FLOW

TIMES AND DWELLS
AGT systems that are part of a normal transit system can
assume flow rates and dwells as determined in Chapter Four,
Station Dwells. However, most AGT systems are classed as
institutional and the majority of passengers are unlikely to be
regular, experienced transit users. Doorways are rarely of typical
transit width or configuration. The most common arrangement is
the quadruple-flow door with associated platform doorsshown
in Figure 10.2. Doorway flow times and the associated dwells
were monitored on the three C-100 systems at SeaTac airport in
May 1995. The range of users varied greatly and included many
people with bags and a few with baggage carts. After the arrival
of a full flight with a preponderance of business passengers,
flow rates reached and exceeded transit levels. At other times,
doorway flow rates were below the transit rates documented in
Chapter Four.
Under these circumstances, calculating flow timesand from
them dwell timesis unwise. The results are unlikely to be
accurate or may reflect only a very specific subset of users.
The recommended solution for AGT systems outside the
transit sphere is simple. Accept a headway, inclusive of train
control separation, dwell time and any operating margin, that
conforms with existing operations or is suggested by the system
manufacturer. The typical headway of airport systems is 120 sec
with a few operating down to 90 sec. Claims have been made
for closer headways with some proprietary systems. Headways
shorter than 90 sec are possible but may limit dwell times and
constrain the operating margin. They should be considered with
caution unless off-line stations are adoptedsee section 10.5.

Off-line stations make closer headways possible and practical
at a price.
10.4 LOADING LEVELS

Loading levels of AGT cars tend to be atypical of normal transit
operations. Those systems such as the Detroit and Miami
downtown people movers that are integral parts of transit
networkscan use loading levels derived from Chapter Five,
Passenger Loading Levels.
Other systems range widely. At one extreme are the airport
shuttles with wide cars and no or few seats where loading can
reach 10 passengers per meter of length under pressure from
arriving business type flights. Loading diversity on airport
systems fluctuates related to flight arrival times, rather than 15
min peaks-within-the-peak. After an arriving flight, three trains
at 120-sec headways can exceed maximum loading levelsto
be followed by a number of under utilized trains.
At the other extreme are the narrow, all-seated configuration
amusement park monorails with loading as low as 2-3
passengers per meter of train length. The loading diversity factor
on the latter type systems attains unity when arrangementsand
continual passenger line-upsensure that every seat on every
train is occupiedin some cases, through all hours of operation.
The hourly achievable capacity of non-transit, AGT requires
consultation with the system supplier. The methodologies and
calculations of this report should only be used as a last resort
and then treated as a guideline.
10.5 OFF-LINE STATIONS

Off-line stations maximize system capacity. They are used on
several rail transit lines in Japan to achieve some of the highest
throughput for two-track rapid transit lines in the world. In
North America they are the exclusive preserve of one AGT
Morgantown.3
Off-line stations permit a train throughput that is partly
independent of station dwell time. Throughput is that of the train
control system plus an allowance for switch operation, lock and
clearance and a reduced operating margin.4 Morgantown and
certain other AGT systems use on-vehicle switching techniques
where even this allowancetypically 6 seccan be dispensed
with. In theory, trains or single vehicles can operate at or close
to the minimum train control separationwhich can be as low
as every 15 secrefer to Figure 10.1.
Major stations with high passenger volumes may require multiple-platform berths, otherwise partial dwell times must be
added to the train separation times to obtain the minimum headway. The achievable capacity of such specialized systems should
3
Figure 10.2 Orlando Airport people-mover doorways

Adtranz (previously Westinghouse) C-100 system.
Systems with multiple platform terminal stations could be regarded as a

sub-set of off-line stations. The Mexico City metro and PATH (New
York) are examples of such arrangements. Not coincidentally, these two
systems achieve respectively the highest passenger throughput and the
closest regular headway on the continentfor two-track rail transit
systems.
Operating margins are intended to accommodate irregularities in train
control separation and dwell times. Off-line stations remove the need to
allow for dwell time variations.
103
be determined through consultation with the system
manufacturer or design consultant.
To avoid decreasing main line capacity, the diverging moves
for off-line stations should be made at line-operating speeds
with adequate off-line station trackage for the deceleration and
acceleration distances.
Where full provision is made for these distances system

throughput becomes independent of stations and dwells
Equation 3-12 or 3-13 in Chapter Three, Train Control and
Signaling, can be used to calculate the line headway with data
values, principally length, adjusted for the specific AGT system.
104
11. Future Research

11.1 INTRODUCTION
Two issues for future research emerged from the work on this
report. The first issue was an inability to obtain meaningful
information or data on the reliability of service. The second was
the wide disparity between total station dwell time and the
actual time used for passengers boarding and alighting.
11.2 SERVICE RELIABILITY

One of the goals of this study was to develop a relationship
between closer headways and reliability of service, leading to
recommendations for how much operating margin should be
accommodated in the headway to avoid routine headway
interference and service delays. It is intuitive that as trains run
closer together the potential for service irregularities and delays
increases. A related margin, the schedule recovery provided at
each turn-back station, rarely affects achievable capacity and
was not analyzed. Schedule recovery time increases the number
of staff and cars to carry a given volume of passengers and is an
issue of economysubject to space limitations at each turnback.
The projects survey and subsequent telephone and field data
collection tasks failed to obtain any suitable material. Some
operators calculated the percentage of runs that were missed,
others had various assessments of on-time arrivalstrains that
reached their destination within five to ten minutes of schedule.
As a result, the project had to rely on the observed headway
regularity during the field data collection and on limited
headway information provided by a few operators. The results
are contained in Chapter Six, Operating Issues, Table 6.1.
Regularity was tabulated as the coefficient of variationthe
standard deviation divided by the mean. The results are shown
in Figure 11.1, in descending order of reliability.
It would be expected that light rail with on-street sections at,
or ahead, of the survey point would have less reliable headway
adherence; that automated systems should be better than
manually driven systems; and that systems with longer
headways would be better than those running trains close
together.
The results are both mixed and contrary to these intuitions.
Although Calgarys three light rail entries,1 all with on-street
sections, are at the bottom of the chart, they are mixed with
BARTs automated and longer headway entry and with the
TTCs manual subway operation (Bloor Station). BC Transit,
with its advanced automatic train supervision, meets
expectations at the top of the chart, but PATHs Journal Square
and NYCTs Grand Central listings, both manually driven and
among the closest headways in the survey, share this honor.
1
Calgarys three lines are not scheduled to interlace evenly on the downtown
trunk. This result is therefore a result of schedulingnot poor operation.
Surveys have frequently shown that reliability is a key

concern if the rail transit industry is to meet the higher customer
expectations of the future. Reliability specifically headway
adherencewas a secondary issue in this study. The data from
15 peak periods on seven systems is inadequate to draw
conclusions.
This topic merits additional research. The first two of the
TCRPs four strategic priorities for 1996 and 1997 transit
research are
Place the customer first and
Improve transit productivity.
Research into the reliability of service delivery will meet both
these goals as even headways move passengers more efficiently
with fewer trains and staff.
Figure 11.1 Headway coefficient of variation (from Table 6.1)
105
Figure 11.2 Total train operating hours lost per month

(equivalent to 0.005% of total hours operated)
Future research should summarize the many surveys of

passenger expectations; develop criteria and uniform reporting
methods for reliability; establish reliability on existing rail
transit systems through telephone and field surveys; relate
reliability to efficiency; and produce conclusions and
recommendations on the many factors that contribute to, or
reduce, system reliabilityand so efficiency.
An example of one performance criterion is shown in Figure
11.2., taken from BC Transits automated SkyTrain operation.
11.3 STATION DWELLS

The station dwell field data collection and analysis showed a
wide variation between the length of the dwell and the time
productively used for passenger flow. The bulk of the wasted
time was between flow stopping and the train starting to leave
the station. A few systems also had a significant loss between
the train stopping and the doors opening. The percentage of
productive time is shown in Figure 11.3. All data are from the
maximum load point station of lines at or close-to capacity.
Two thirds of systems with headways under 200 sec have a
flow to dwell ratio of less than 40%; five systems are at or
below 30%. Some of this unproductive time is essential. Door
opening and closing takes 4 to 6 sec. Confirmation that a train is
stopped and correctly positioned at a platform takes less than 1
sec on some automated and most manual systems, but several
seconds on others. Safety considerations require some leeway
from door closing to train leaving. There is dispute about how
much delay is required for safety. Two to 5 sec appears to be a
reasonable range used on many systems. The remaining
unproductive time, averaging 30-40% of all dwell is wasted
whether due to operational slackness or over cautious safety
concerns.
TCRP Report 4, Aids for Car Side-Door Observation, (R77) and
NCTRDP Report 13, Conversion to One-Person Operation of
Rapid-Transit Trains, (R78) concentrated on methods to permit
train operators to observe side doors as a step towards reducing
crewing from two to one on older rail rapid transit systems.
Safety and efficiency at the door interface were only reviewed
peripherally.
Two North American systems, Vancouver and Miamis Met-
Figure 11.3 Percentage of dwell time at maximum load point

stations used for peak-door passenger movements
romover; a few foreign systems; and elevators worldwide

maintain exceptional safety standard using pre-programmed
dwells without any door observation. The opportunity to tighten
up dwells and gain the associated economies is considerable.
Research into the passenger-door interface, the effects of
different door closing tones or announcements, marking
platform door positions, training passengers to wait to the side
of the door position and take more responsibility for their
actions, and reviewing interior car designs that improve flow
rates is overdue.
The benefits are considerable and consistent with the transit
industry goal to improve transit productivity. Dwell times make
up 20 to 40% of total travel time on urban rail rapid transit
systems. A modest goal of an overall 10% dwell time reduction
would reduce operating costs and car requirements by 3%. On
U.S. rail rapid transit alone that saves $120 million a year and
330 carsa future capital saving of over $600 million2 at the
estimated replacement cost of $2 million per car.
Even such modest dwell reductions would reduce overall
travel times, thus making rail rapid transit more attractive to
passengers, increasing ridership and meeting another TCRP goal
of placing the customer first.
The research brief could be expanded to examine the entire
issue of operating efficiency. This project found wide variations
across the continent. Many of the slack operating practices were
not related to restrictive labor practices but to a lack of concern
for the brisk, efficient operation that typified the better systems.
2
Based on U.S. rail rapid transit annual operating costs of $3.9 billion and a
fleet of 11,000 cars.
106
Bibliography & References

Numbered references only are summarized in Appendix One
(A.1) under the identical number. NS = not summarized
1. ABRAMOVICI, MARC, Optimization of Emergency
Crossovers and Signals for Emergency Operations in Rail
Rapid Transit Systems, APTA RT Conference, June 1982
2. ALLE, P., Improving Rail Transit Line Capacity Using
Computer Graphics, Logistics and Transportation Review,
Volume 17, Number 4, University of British Columbia,
Faculty of Commerce, December 1981
3. ALLEN, DUNCAN W., Practical Limits of Single-Track
Light Rail Transit Operation, Transportation Research
Board 1361, 1992: pp 305-311
4. AMERICAN PUBLIC TRANSIT ASSOCIATION, 1992
Transit Operating and Financial Statistics
5. AMERICAN PUBLIC TRANSIT ASSOCIATION, 1994
Membership Directory
6. AMERICAN PUBLIC TRANSIT ASSOCIATION,
Equipment Roster 1993
7. AMERICAN PUBLIC TRANSIT ASSOCIATION, Roster
of North American Rapid Transit Cars, 1993 Edition.
8. ANDERSON, J. EDWARD, Transit Systems Theory,
Lexington, 1978
9. AUER, J.H., Rail-Transit People-Mover Headway
Comparison, IEEE Transactions on Vehicular Technology,
Institute of Electrical and Electronics Engineers, 1974
10. BARDAJI, JORDI F., Regulating Headway in Barcelona,
The Urban Transport Industries Report, Campden
Publishing Limited, 1993, pp 175-176
11. BARWELL, F. T., Automation and Control in Transport.
2nd Revised Edition, Pergamon Press Limited, 1983
12. BATELLE INSTITUTE, Recommendations en vie de
lamnagement dune installation de transport compte tenu
de donnes anthropomtriques et des limites
physiologiques de lhomme, Geneva, 1973
13. BERGMANN, DIETRICH R., Generalized Expressions
for the Minimum Time Interval between Consecutive
Arrivals at an Idealized Railway Station, Transportation
Research, 1972 Vol. 6, pp. 327-341
14. BERRY, RICHARD A., CERVENKA, KENNETH J.
AND SU, CHANG-AN, Traffic and Light Rail Transit:
Methods of Analysis for DARTs North Central Corridor,
Transportation Research Record 1361, 1992: pp 224-234
15. BOORSE, JACK W., Blending LRT into Difficult Traffic
Situations on Baltimores Central Light Rail Line,
Transportation Research Record 1361, 1992: pp 197-206
NS. BOOZ ALLEN and HAMILTON INC., in association
with Abrams-Cherwony; Ammann & Whitney; George
Beetle: Merrill Stuart, Queens Transit Alternatives
Technical Appendix, Part 5a, Operations/Capacity
Analysis, NYCTA, New York, January 1981
16. BURGIN, EDWARD A., Light Rail Transit Signaling,
TRB Special Report 182, 1978, pp 119-123
17. BUSHELL, CHRIS., Janes Urban Transport Systems,

Janes Information Group Ltd, UK, 1989
18. CALLAN, DENNIS R., Toronto Transit Commissions 90
Second Headway Study: Getting More Out of Existing
Infrastructure, APTA RT Conference, Vancouver 1990
19. CANADIAN URBAN TRANSIT ASSOCIATION,
Canadian Transit Handbook, 2nd Edition, Chap. 8
Capacity, Canadian Urban Transit Association and the
Roads and Transportation Association of Canada, Toronto,
1985
20. CELNIKER, STEPHEN, and TERRY, E. WAYNE,
Trolley Priority on Signalized Arterials in Downtown San
Diego, Transportation Research Record 1361, 1992: pp
184-188
21. DELAWARE RIVER PORT AUTHORITY, 90 Seconds
Headway Feasibility Study, Lindenwold Line, Delaware
River Port Authority, January 1973
22. DEMERY, LEROY W., Jr., Supply-Side Analysis and
Verification of Ridership Forecasts for Mass Transit
Capital Projects, APA Journal, Summer 1994
23. ENVIRODYNE ENGINEERS, INC., Metro-North Speed
and Capacity Improvement Study. Tasks 1 to 5 U.S. DoT,
Urban Mass Transportation Administration, 1989
24. EUROPEAN CONFERENCE OF MINISTERS OF
TRANSPORT, Report of the 38th Round Table on
Transport Economics - Scope for the Use of Certain OldEstablished Urban Transport Techniques, Transport
Capacity, OECD, Paris, 1979, pp 24-25
25. FOX, GERALD D., Light Rail/Traffic Interface In
Portland: The First Five Years, Transportation Research
Record 1361, 1992: pp 176-183
26. GILL, D.C., and GOODMAN C.J., Computer-based
optimisation techniques for mass transit railway signalling
design, IEE Proceedings-B, Vol. 139, No.3, May 1992
27. GRAHAM, IAN R., Optimizing Headways on an
Automated Rapid Transit System: The SkyTrain
Experience, American Public Transit Association, Rapid
Transit Conference, Vancouver, B.C., 1990
28. GRAY, GEORGE E., and Hoel, Lester A., Public
Transportation Planning, Operations and Management,
Prentice-Hall Inc., 1979
NS. HIRSCHFELD. C.F., Bulletins Nos. 1-5, Electric Railway
Presidents Conference Committee (PCC), NY, 1931-1933
29. HOMBURGER, WOLFGANG S., Urban Mass Transit
Planning, The Institute of Transportation and Traffic
Engineering, University of California, 1967
30. JACOBS, MICHAEL., SKINNER, ROBERT E., and
LERNER, ANDREW C., Transit Project Planning
GuidanceEstimate of Transit Supply Parameters,
Transportation Systems Center, U.S. DOT, Oct. 1984
31. JANELLE, A., POLIS, M.P., Interactive Hybrid Computer
Design of a Signaling System for a Metro Network: IEEE
Transactions on Systems, Man and Cybernetics, 1980
107
32. KLOPOTOV, K., Improving the Capacity of Metropolitan
Railways UITP, 40th International Congress, The Hague,
1973
33. KOFFMAN, D., RHYNER, G. and TREXLER, R., Selfservice Fare Collection on the San Diego Trolley, U.S.
DOT, 1984
34. KORVE, HANS W. and WRIGHT, PATRICK M., New
Standards for Control of At-Grade Light Rail Transit
Crossings, Transportation Research Record 1361, 1992:
pp 217-223
35. KORVE, HANS W. Traffic Engineering for Light Rail
Transit, TRB Special Report 182, 1978, pp 107-114,
Transportation Research Board, Washington, DC
36. KRAFT, W. H., and BERGEN, T. F., Evaluation of
Passenger Service Times for Street Transit Systems.
Transportation Research Record 505, Transportation
Research Board, Washington D.C. 1974
37. KUAH, GEOK K. and ALLEN, JEFFREY B., Designing
At-Grade LRT Progression: Proposed Baltimore Central
Light Rail, Transportation Research Record 1361, 1992:
pp 207-216
38. KYOSAN ELECTRIC MFG. CO., LTD. Total Traffic
Control SystemTTC, Yokohama, Japan, 1986
39. LANG, A SCHEFFER, and SOBERMAN, RICHARD M.,
Urban Rail Transit: Its Economics and Technology, MIT
Press, 1964
40. LEVINSON, HERBERT S., and HOEY, WILLIAM E.
Some Reflections on Transit Capacity Proceedings of the
International Symposium on Highway Capacity,
Karlsruhe, July 1991
41. LEVINSON, HERBERT S., Capacity Concepts for StreetRunning Light Rail Transit, Australian Road Capacity
Conference 1994
42. LEVINSON, HERBERT S., Chapter 5 Urban Mass
Transit Systems, Transportation Planning Handbook,
Institute of Transportation Engineers, Prentice Hall, 1992
43. LEVINSON, HERBERT S.; ROBINSON, CARLTON, C.
and GOODMAN, LEON, Chapter 12 Capacity in
Transportation Planning, Transportation Planning
Handbook, Institute of Transportation Engineers, Prentice
Hall, 1992
44. LIN, TYH-MING and WILSON, NIGEL H.M., Dwell
Time Relationships for Light Rail Systems, Transportation
Research Record 1361, 1992: pp 287-295
45. MEDVECZKY, GEORGE. Hub-Bound Travel 1991, New
York Metropolitan Transportation Council 1992
46. MILLER, E. J. and BUNT, P. D., Simulation Model Of
Shared Right-of-Way Streetcar Operations, Transportation
Research Record 1152, 1987: pp 31-41
47. MOTZ, D., Attainable Headways using SELTRAC,
Alcatel Canada, Toronto, September 1991 (Proprietary
Reportonly non-confidential data used for the A-8
study)
48. NEW YORK CITY TRANSIT AUTHORITY, Rapid
Transit Loading Guidelines, April 1992
49. OBRIEN, W., SCHNABLEGGER, J. and TEPLY, S.,
Control of Light Rail Transit Operations in Edmonton,
TRB Special Report 182, 1978, pp 115-118
50. PARKINSON, TOM E., Passenger Transport in Canadian
Urban Areas, Canadian Transport Commission, Ottawa
1971
NS. PARKINSON, TOM, Safety Issues Associated with the

Implementation of ATCS-Type Systems, Transportation
Development Centre, Transport Canada, August 1989
51. PUSHKAREV, BORIS S., ZUPAN, JEFFREY M., and
CUMELLA, ROBERT S., Urban Rail In America: An
Exploration of Criteria for Fixed-Guideway Transit,
Indiana University Press, 1982
52. RADWAN, A. E., and HWANG, K. P., Preferential
Control Warrants of Light Rail Transit Movements, TRB
State-of-the-Art Report 2, Light Rail Transit: System
Design for Cost-Effectiveness, 1985: pp 234-240
53. RAINVILLE, WALTER S., and HOMBURGER,
WOLFGANG S., Capacity of Urban Transportation
Modes, Journal of the Highway Division, American
Society of Civil Engineers, 1963
54. RICE, P., Practical Urban Railway CapacityA World
Review, Proceedings of the 7th International Symposium
on Transportation and Traffic Theory. Kyoto
55. SCHUMANN, JOHN W., Status of North American LRT
Systems: 1992 Update, Transportation Research Record
1361, 1992: pp 3-13
56. SONE, PROFESSOR SATORU, Squeezing Capacity out
of Commuter Lines, Developing Metros, Railway Gazette
International 1990.
57. STRAUS, PETER, Light-Rail Transit: Less Can Mean
More, TRB Special Report No. 182, Light-Rail Transit:
Planning and Technology. 1978: pp 44-49
58. SULLIVAN T. J., New York City Transit News
Technology Signals Program Status Report, MTA New
York City Transit Division of Electrical Systems, APTA
Rapid Transit Conference, Sacramento, June 1994
59. TABER, JOHN and LUTIN, JEROME, Investigating the
Potential for Street Operation of Light Rail Transit, TRB
Special Report 182, 1978: pp 161-166
60. TAYLOR, P. C., LEE, L. K. and TIGHE, W. A.,
Operational Enhancements: Making the Most of Light
Rail, TRB Special Report 221, 1989: pp 578-592
61. TIGHE, W. A. and PATTERSON, L. A., Integrating LRT
into Flexible Traffic Control Systems, TRB State-of-theArt Report 2, Light Rail Transit: System Design for CostEffectiveness, 1985: pp 213-220
62. TOPP, R. M., Improving Light Rail Transit Performance
in Street Operations: Toronto Case Study, TRB State-ofthe-Art Report 2, Light Rail Transit: System Design for
Cost-Effectiveness, 1985
63. TORONTO
TRANSIT
COMMISSION,
YongeUniversity-Spadina Improved Headway Study, Final
Report Toronto Transit Commission, December 1988
64. TORONTO
TRANSIT
COMMISSION,
YongeUniversity-Spadina Improved Headway Study, Signalling
Report Toronto Transit Commission, December 1988
65. TRANSPORTATION RESEARCH BOARD, Collection
and Application of Ridership Data on Rapid Transit
Systems, Synthesis of Transit Practice, Washington D.C.,
1991
66. TRANSPORTATION RESEARCH BOARD, Gray,
Benita, editor. Urban Public Transportation Glossary,
1989
67. TRANSPORTATION RESEARCH BOARD, Highway
Capacity Manual, TRB Special Report 209, Chapter 12,
1989
108
68. US DOT FEDERAL TRANSIT ADMINISTRATION
Characteristics of Urban Transportation Systems, Revised
Edition, 1992
69. US DOT, National Transportation Statistics, Annual
Report, Sept. 1993
70. VANTUONO WILLIAM C. Signaling and Train Control,
High-Tech for High Capacity. Transit Connections,
September 1994
71. VUCHIC, VUKAN R., Urban Public Transportation
Systems and Technology, Prentice-Hall Inc., 1981
72. WALSHAW, J. R., LRT On-Street Operations: The
Calgary Experience, TRB State-of-the-Art Report 2, Light
Rail Transit: System Design for Cost-Effectiveness, 1985.
pp 221-226
73. WEISS, DAVID M., and FIALKOFF, DAVID R.,
Analytic Approach to Railway Signal Block Design,
Transportation Engineering Journal, February 1974
74. WILKINS, JOHN D., and BOSCIA, J. F., Considerations
75.
76.
77.
78.
For Effective Light Rail Street Operation, TRB State-ofthe-Art Report 2, Light Rail Transit: System Design for
Cost-Effectiveness, 1985: pp 195-202
WILSON, NIGEL H. M., MACCHI, RICHARD A.,
FELLOWS, ROBERT E. and DECKOFF, ANTHONY A.,
Improving Service on the MBTA Green Line Through
Better Operations Control, Transportation Research
Record 1361, 1992: pp 296-304
YOUNG J.A., Passenger Comfort in Urban Transit
Vehicles, Ontario Ministry of Transportation and
Communications, 1976
TELEPHONICS CORPORATION, Aids for Rail Car
Side-Door Observation, TCRP Report 4, Transportation
Research Board, Washington, DC, 1995.
HOESS, J. A. and MURPHY, P. J., Conversion of OnePerson Operation of Rapid-Transit Trains, NCTRDP
Report 13, Transportation Research Board, Washington,
DC, 1986.
109
GLOSSARY
Sources: Most of the definitions in this glossary are taken from
the Transportation Research Boards Urban Public
Transportation Glossary (1989) and from the American Public
Transit Associations A Glossary of Transit Terminology
(1984).
Caution: There is inconsistency in terminology used in the
North America transit industry. Many systems have there own
specific terminology, a motorman and guard on one system can
be an operator and conductor on another.
ABSsee control system, automatic block signal.
ABSOLUTEA block that no train may enter while the block
is occupied by another train.
ABSOLUTE PERMISSIVEA signal system for a single
track or guideway that prevents simultaneous opposing train
movements between sidings but permits following movements
at a safe distance.
ACCESSIBILITYA measure of the ability or ease of all
people to travel among various origins and destinations
AGTAutomated guideway transit; automated guided transit;
see transit system, automated guideway.
ALIGHTTo get off or out of a transportation vehicle.
AREA OCCUPANCYIn station and other facility design and
in pedestrian movement, the area provided per person.
ARTICULATED RAIL VEHICLE (articulated car)1. An
extra-long rail vehicle with two or more bodies connected by
joint mechanisms that allows bending in curves yet provide a
continuous interior. Typically, the vehicle is 56-100 ft (17-33 m)
long. It is very common on light rail transit systems but is also
found on several rail rapid transit systems. 2. Rapid transit cars
with separate bodies that share a common center truck.
ATOAutomatic train operation.
AUTOMATED GUIDEWAY TRANSIT SYSTEM (AGT)A
transportation system in which automated, driverless vehicles
operate on fixed guideways with exclusive right-of-way.
AUTOMATIC BLOCK SIGNAL (ABS)a system governing
train separation in which the signals are controlled by the trains
themselves. The presence or absence of a train in a block is
determined by a track circuit. If the circuitry fails, a restrictive
signal is displayed.
AUTOMATIC TRAIN CONTROL SYSTEM (ATC
system)1. A system for automatically controlling train
movement, enforcing train safety, and directing train operations
by computers; see also automatic train operation, automatic

train protection, and automatic train supervision. 2. A trackside
system working in conjunction with equipment installed on the
train, arranged so that its operation will automatically result in
the application of the brakes to stop or control a trains speed at
designated restrictions, should the operator not respond. The
system usually works in conjunction with cab signals.
AUTOMATIC TRAIN OPERATION (ATO)The subsystem
within automatic train control that performs such functions as
speed control, programmed stopping, and (sometimes) door
operation.
AUTOMATIC
TRAIN
PROTECTION
(ATP)The
subsystem within automatic train control that provides fail-safe
protection against collisions, excessive speed, and other
hazardous conditions.
AUTOMATIC TRAIN STOP SYSTEM (ATS SYSTEM)A
trackside system that works in conjunction with equipment
installed on the electric rail car or locomotive to apply the
brakes at designated restrictions or on a dispatchers signal,
should the operator not respond properly.
AUTOMATIC
TRAIN
SUPERVISION
(ATS)The
subsystem within automatic train control that monitors trains,
adjusts the performance of individual trains to maintain
schedules, and provides data for adjusting service to minimize
the inconveniences otherwise caused by irregularities. May also
be used for systems that merely display train status and rely on
staff intervention for any corrective action.
BARRIER-FREEContaining no obstacles that would prevent
use by a mobile physically handicapped person or any other
person.
BASIC OPERATING UNITIn rail rapid transit, the smallest
number of rapid transit vehicles that can operate independently
in revenue service, usually one to three (exceptionally more)
cars.
BI-LEVELa rail car that has two levels for passenger
accommodation. The upper level may extend through the entire
length of the car or only over a part of it; this level is sometimes
restricted to seated passengers only. Bi-level cars are used
principally on commuter rail lines. Double deck cars and gallery
cars are types of bi-level cars.
BLOCK1. A section of track or guideway of defined limits
on which the movement of trains is governed by block signals,
cab signals, or both; also known as a signal block. 2. A section
of track of defined length, the occupancy of which is regulated
110
by fixed signal(s), telephone or radio orders, or timetables; also
known as a block section.
BLOCK SIGNALa standard railroad signal system that uses
a fixed signal at the entrance of a block to govern the separation
of trains entering the block.
BOARDINGGetting on a transit vehicle.
BUNCHINGWith transit units, a situation that occurs when
passenger demand is high and dwell times at stops are longer
than scheduled. Headways become shorter than scheduled, and
platoons of transit units (vehicles or trains) develop, with longer
intervals between platoons. The same effect (one transit unit
caught by the following) can also be caused by lack of
protection from general road traffic congestion or by traffic
signal timing. Bunching can become cumulative and can result
in delay to passengers and unused capacity.
CAB1. A rail car with a driving cab. 2. A passenger carrying
car used in push-pull service and fitted with a cab at one end, to
be used to operate the train when the locomotive is pushing; see
also commuter rail.
CAB SIGNALin rail systems, a signal located in the cab,
indicating a condition affecting the movement of a train and
used in conjunction with interlocking signals and in conjunction
with or in lieu of block signals.
CAPACITY...achievableA term used in this report to avoid
the confusion whereby design capacity can mean either a
theoretical or practical maximum number of passengers that can
be transported over a given section of a transit line in one
direction during a given time period. Achievable capacity is the
design capacity factored down to reflect the uneven passenger
demand during the peak hour and the uneven loading of cars
within a train.
CAPACITY...crush (crush load)the maximum feasible
passenger capacity of a vehicle, that is, the capacity at which
one more passenger cannot enter without causing serious
discomfort to the others. Note that the crush load specification
for some rail transit vehicles does not relate to an achievable
passenger loading level but is an artificial figure representing
the additional weight for which the car structure is designed or
for which the propulsion and braking system will meet
minimum criteria.
CAPACITY...design1. For transit, the maximum number of
passengers that can be transported over a given section of a
transit line in one direction during a given time period (usually 1
hour) under prevailing traffic conditions and design comfort
standards. 2. For vehicles, the total number of spaces or people a
vehicle can accommodate.
CAPACITY...fleet (rolling stock capacity)the total
number of passenger spaces in all vehicles of a transit fleet.
CAPACITY...linethe maximum number of spaces that transit
units (vehicles or trains) on a line can transport past a fixed
point in one direction per unit of time (usually 1 hour) under

actual operating conditions; see also capacity, theoretical
line.
CAPACITY...normal vehiclesee capacity, vehicle.
CAPACITY...rolling stocksee capacity, fleet.
CAPACITY...practicalThe maximum number of passengers
that can be transported over a given section of a transit line in
one direction during a given time period (usually 1 hour) under
prevailing traffic conditions and design comfort standards.
after allowing for the uneven passenger demand during the peak
hour and the uneven loading of cars within a train. In this report
Achievable Capacity is used instead of Practical Capacity to
avoid confusion with variable definitions of this term used in
other capacity work.
CAPACITY...seating (seated capacity)the number of
passenger seats in a vehicle.
CAPACITY...standingthe number of standing passengers
that can be accommodated in a vehicle under specified comfort
standards, expressed in area per standee.
CAPACITY...theoretical linethe maximum number of
transit units (vehicles or trains) or spaces that can be carried
over a line segment during a given time period with every transit
unit operating at the minimum headway that the control system
permits. Real operating conditions may reduce this capacity. See
also capacity, line.
CAPACITY...vehicle (normal vehicle capacity, total
vehicle capacity)the maximum number of passengers that
the vehicle is designed to accommodate comfortably, seated and
standing; may sometimes refer to number of seats only.
CBDcentral business district.
CENTRAL BUSINESS DISTRICT (CBD)The downtown
retail trade and commercial area of a city or an area of very high
land valuation, traffic flow, and concentration of retail business
offices, theaters, hotels and services.
CENTRAL CITYas defined by the Bureau of the Census, the
largest city, or one of the largest cities, in a metropolitan
statistical area or urbanized area. The criteria for designating a
central city vary with the type of area and the particular census.
CENTRALIZED TRAFFIC CONTROL (CTC)in rail
systems, a traffic control system in which signals and switches
are controlled from a remotely located (centralized traffic
control) panel.
CHECKin transit operations, a record of the passenger
volume on all transit units that pass a specific location or time
point (also known as a passenger riding count or check), the
actual time the unit passes it (also known as a schedule check),
111
the number of passengers who board and alight at each stop on a
route or line (also known as an on-and-off count or check), or
any combination of these items. The checker may ride the transit
unit (an on-board check), follow it in another vehicle, or check
the transit units from a particular location (a point or corner
check).
CHECKERin transit operations, a person who observes and
records passenger counts, timing, speeds, vehicle counts,
schedule adherence, or other data useful in transit planning and
scheduling. The position may be further specified as schedule
checker, traffic checker, and so on.
CLOSE-UPin rail transit operations the process where a train
approaching a station will close-up to the train berthed in the
station to the minimum distance permitted by the signaling or
train control system. This is usually the critical line condition
that, combined with the dwell at the maximum load point
station, establishes the minimum headway.
COMMUTER RAIL CARa passenger rail car designed for
commuter rail services. It usually has many more seats than a
conventional long-distance rail passenger car. The car may be
hauled by a locomotive, have a self-contained internal
combustion engine, or be electrically propelled by power from a
third rail or overhead wire. See also cab.
COMMUTER RAILThe portion of passenger railroad
operations that carries passengers within urban areas, or
between urban areas and their suburbs, but differs from rail
rapid transit in that the passenger cars generally are heavier, the
average trip lengths are usually longer, and the operations are
carried out over tracks that are part of the railroad system in the
area.
CONDUCTOR1. In rail transit operations, the operating
employee who may control the doors on rail transit vehicles, or
who may have fare-collecting duties, or bothalso called guard
on some systems. 2. In railroad operations, the operating
employee in charge of the train and trail crew.
COUPLERa device for connecting one rail vehicle to
another. The mechanism is usually placed in a standard location
at both ends of all rail cars and locomotives.
COUPLER...automatic1. a coupler that operates
automatically. It may also be capable of uncoupling
automatically. 2. An automatic connector that joins electric or
pneumatic train lines together between rail cars.
CRITICAL LINE CONDITIONin rail transit operations the
factor that constrains headway. This is usually the close-in at the
maximum load point station or the terminal turnback process,
occasionally at junctions.
CRUSH LOADThe maximum passenger capacity of a
vehicle, in which there is little or no space between passengers
(i.e., the passengers are touching on another) and one more
passenger cannot enter without causing serious discomfort to the
others.
CTCsee centralized traffic control

DEADHEADThe movement of a transit vehicle without
passengers aboard - often to and from a garage, or from one
route to another.
DISPATCHERThe individual who is responsible for keeping
trains or other vehicles on schedule.
DOOR MOVEMENT TIMEThe time during a rail transit
station dwell that passengers are moving through train doorway
DIVERSITY loadingThe ratio between achievable (practical
capacity) and design capacity (maximum capacity) over the
peak hour, reflecting that passengers do not evenly load a car,
cars of a train or trains over the peak hour (the 3 levels).
FARE COLLECTION SYSTEMthe procedures and devices
used to collect fares and to accumulate and account for fares
paid.
FARE COLLECTION SYSTEM...automatic (AFC)the
controls and equipment that automatically admit passengers on
insertion of the correct fare in an acceptable form, which may be
coins, tokens, tickets, or farecards (stored-value farecards must
be inserted again on exit, at which point an additional fare may
be required). The system may include special equipment for
transporting and counting revenues.
FARE COLLECTION SYSTEM...fare-registering turnstile
(faregate)a turnstile that unlocks to allow a passenger to
enter the paid area after a pass or farecard or the correct amount
of money or token is inserted in it. It records the fares paid.
FARE COLLECTION SYSTEM...self-service, proof of
payment, barrier-free, honor systema fare collection
system that has no fare-registering turnstiles. This system
requires that the passenger be able to display proof of payment
(e.g., validated ticket, prepaid pass, valid transfer) while on
board the transit vehicle or in a station. Compliance is
monitored through random checking by designated transit
employees.
FAREBOXa device that accepts coins, bills, tickets, tokens,
or other fare media given by passengers as payment for rides.
FIXED-GUIDEWAY SYSTEMA system of vehicles that can
operate only on its own guideway constructed for that purpose
(e.g., rapid trail, light rail). Federal usage in funding legislation
also includes exclusive right-of-way bus operations, trolley
coaches, and ferryboats as fixed-guideway transit.
FLOW RATE (rate of flow)in transportation, the number of
units (passengers or vehicles) passing a point on a transportation facility during some period of time, usually counted or
computed in units per hour. For example, if 8 buses pass a point
in the first half hour and 15 in the second, the volume for the
hour is 23. However, the flow rate for the first half
112
hour is 16 buses/hour, and for the second half hour the flow rate
is 30 buses/hour.
or more railroads interchange cars over connecting tracks. 3. A

location at which several transit lines converge.
GALLERY CARA bilevel rail car that has seating and access
aisles on a second level along each side of an open well. Tickets
of passengers on the second level can be inspected or collected
from the lower level.
LAYOVER-TIMETime built into a schedule between arrivals

and departures, used for the recovery of delays and preparation
for the return trip.
HANDICAPPED PERSONSpeople who have physical or

mental impairments that substantially limit one or more major
life activities. In the context of transportation, the term usually
refers to people for whom the use of conventional transit
facilities would be impossible or would create a hardship. These
people are also known as transportation handicapped or as
people who have a public transportation disability.
HANDICAPPED ACCESSIBILITY (full accessibility)
The extent to which facilities are free of barriers and usable by
mobile handicapped people, including wheelchair users.
HEADWAYthe time interval between the passing of the front
ends or successive transit units (vehicles or trains) moving along
the same lane or track (or other guideway) in the same direction,
usually expressed in minutes; see also service frequency.
HEADWAY MANAGEMENTa technique for managing the
operation of transit units (vehicles or trains) that focuses on
maintaining a certain spacing between units on the same line,
instead of on adhering to a timetable. For example, if units
become bunched, corrective measures might include delaying
the units at the rear of the bunch to provide regular headways
and hence load distribution, even at the expense of reducing
timetable adherence.
HEADWAY...basethe scheduled headway between transit
unit (vehicle or train) trips during an off-peak (usually midday)
period.
HEADWAY...interferenceheadway that is so close that one
vehicle or train interferesdelaysthe next.
HEADWAY...non-interferenceheadway (usually including
an operating margin) such that in normal operations one train
does not delay another.
HEADWAY...policy1. headway prescribed by reasons other
than matching capacity to demand. 2. the maximum permissible
headway as established by the transit agency or (often) the
policy board, usually for off-peak, low demand periods.
LEVEL OF SERVICE (LOS)1. A set of characteristics that

indicate the quality and quantity of transportation service
provided, including characteristics that are quantifiable (system
performance, e.g., frequency, travel time, travel cost, number of
transfers, safety) and those that are difficult to quantify (service
quality, e.g., availability, comfort, convenience, modal image).
2. For pedestrians, sets of area occupancy classifications to
connect the design of pedestrian facilities with levels of service
(A for best through F for worst). 3. For transit rights-of way, see
right-of-way.
LIGHT RAIL CAR (LRV, LIGHT RAIL VEHICLE)a rail
vehicle similar to a streetcar. It may be larger, however, and is
often articulated. A light rail car is capable of boarding and
discharging passengers at either track or car-floor level.
LIGHT RAIL TRANSIT SYSTEM (LRT)see transit system,
light rail
LINE1. A transportation company (e.g. a bus line). 2. A
transit service operating over a specified route or combination of
routes. 3. An active (in-use) railroad track or AGT guideway. 4.
In network coding, a route and its service level, including mode
designation (type of service), line number, headway, and
sequence of transfer points (nodes). These factors describe the
lines route as an ordered set.
LINE-CLEARin rail transit, operation such that trains do not
have to stop or slow down due to the train ahead but receive a
succession of green signals. See also Headwaynoninterference.
LINE...double-track maina rail main line that has two
tracks, usually one for each direction.
LINE...single-track maina rail main line that has one track.
It requires passing sidings for bi-directional operation.
HEAVY RAILA type of electric rail transit system

characterized by exclusive rights-of-way, multi-car trains,
sophisticated signaling and high-platform loading; with the
capacity to carry a heavy volume of traffic. Also called
subways or metropolitan railways (metros). see also transit
system, rail rapid.
LOAD FACTOR1. The ratio of used capacity to offered

capacity of equipment or a facility during a specified time
period. It is usually expressed as a percentage of seats occupied
at a given point or (in continuous form) passenger miles
(kilometers) divided by seat miles (kilometers). For rail services,
the load factor is sometimes expressed as passenger miles
(kilometers) per train mile (kilometer) to account for the ability
to couple rail cars together to achieve efficiency. 2. The ratio of
passenger capacity of a vehicle; also known as a utilization
coefficient.
JUNCTION POINT1. A location at which a rail branch line

track connects with a main-line track. 2. A location at which two
LOAD FACTORThe ratio of passengers actually carried

versus the total passenger capacity of a vehicle.
113
LOADING ISLAND1. A pedestrian refuge within the rightof-way and traffic lanes of a highway or street. It is provided at
designated transit stops for the protection of passengers from
traffic while they wait for and board or alight from transit
vehicles; also known as a pedestrian island. 2. A protected spot
for the loading and unloading of passengers. It may be located
within a rail transit or bus station.
MANUAL BLOCKa system of manually governing train
movement in a block or a series of consecutive blocks by means
of signals, train orders, telephone, or radio.
MANUAL TRAIN OPERATIONa system in which train
movement is controlled by the operator (motorman) or engineer.
MAXIMUM LOAD POINT (MLP)the point on a transit line
or route at which the passenger volume is the greatest. There is
one maximum load point in each direction.
stated tolerances; for example, a transit unit (vehicle or train)

arriving, passing, or leaving a predetermined point (time point)
along its route or line within a time period that is no more than
x minutes earlier and no more than y minutes later than a
published schedule time. (Values of 0 minutes for x and 5
minutes for y are the most common).
OPERATORAn employee of a transit system who spends his
or her workday in the operation of a vehicle, e.g., bus driver,
streetcar motorman, trolley coach operator, cable car gripman,
rapid transit train motorman, conductor, etc. see also property,
operator
OPERATING MARGINAn employee of a transit system
who spends his or her workday in the operation of a vehicle,
e.g., bus driver, streetcar motorman, trolley coach operator,
cable car gripman, rapid transit train motorman, conductor, etc.
see also property, operator
MAXIMUM LOAD SECTION (MLS)the section of a transit

line or route that carries the highest total number of passengers
for that line or route and direction.
PASSENGERa person who rides a transportation vehicle,

excluding the operator or other crew members of that
transportation vehicle; see also trip, passenger; trip, linked; and
trip, unlinked.
MARRIED PAIR (MP)two semi permanently coupled rail

cars (A car and B car) that share some mechanical and electrical
equipment and must be operated together as a unit.
PASSENGER COUNTa count of the passengers on a

vehicle or who use a particular facility.
MODEa particular form of travel, for example, walking,

traveling by automobile, traveling by bus, traveling by train.
PASSENGER FLOW (passenger traffic)the number of

passengers who pass a given location in a specified direction
during a given period.
MODE...transita category of transit systems characterized

by common characteristics of technology, right-of-way, and
type of operation. Examples of different transit modes are
regular bus service, express bus service, light rail transit, rail
rapid transit and commuter rail.
PASSENGER FLOW TIME...doorway the time, in

seconds, for a single passenger to cross the threshold of a rail
transit car doorway, entering or exiting, per single stream of
doorway width.
MOTORMANTraditional term for train operator or engineer

on rapid transit systems. No longer politically correct but still in
common use.
MOVING BLOCK (dynamic block control)an automatic
train control system that spaces trains according to their location
and (sometimes) their relative velocity, stopping performance,
and a prescribed safety factor. Moving-block signaling systems
are also called transmission or communication based systems.
The latter is becoming the preferred term.
PASSENGER LOADthe number of passengers on a transit

unit (vehicle or train) at a specified point.
PASSENGER MILES (passenger kilometers)the total
number of passengers carried by a transit system for a unit of
time multiplied by the number of miles (kilometers) they travel.
A comparison of passenger miles (kilometers) and seat miles
(kilometers) provides a measure of transit system efficiency.
PASSENGER VOLUME (line volume)the total number of
passengers carried on a transit line during a given period.
MULTIPLE-UNIT (MU)a powered rail car arranged either

for independent operation or for simultaneous operation with
other similar cars, when connected to form a train of such cars.
It may be designated as DMU (diesel multiple-unit) or EMU
(electric multiple-unit), depending on the source of power.
PASSENGER...revenuea passenger who pays (or has

prepaid) a fare.
OFF-LINEnot in the main flow of traffic or not on the main

line of traffic, for example, off-line station.
PCC CAR (PCC, Presidents Conference Committee car)

a streetcar first produced in 1935. Its performance and efficiency
were significantly improved over those of any streetcar
previously built. The PCC car, characterized by (relatively)
lightweight construction, smooth and rapid acceleration and
ON-TIME PERFORMANCEthe proportion of the time that

a transit system adheres to its published schedule times within
PASSENGER...transfera passenger who changes from one

route or line to another route or line.
114
deceleration, and soft ride, became the standard for U.S.
streetcars for many years.
PEAK (peak period, rush hours)1. The period during
which the maximum amount of travel occurs. It may be specified
as the morning (a.m.) or afternoon or evening (p.m.) peak. 2. The
period when demand for transportation service is heaviest.
PEAK-HOUR FACTOR (peak-hour conversion factor)
the ratio of the volume during the peak hour to the maximum
rate of flow during a selected period within the peak hour.
PEAK/BASE RATIO (peak/off-peak ratio)1. The ratio
between the number of vehicles operating in passenger service
during the peak hours and that during the base period. 2. The
ratio between the number of passengers carried during the peak
hours and that during the base period.
PEOPLE MOVERan automated transportation system (e.g.,
continuous belt system or automated guideway transit) that
provides short-haul collection and distribution service, usually
in a major activity center. Once almost synonymous with
automated guideway transit. Now primarily used for smaller
systems such as those internal to airports.
PLATFORM (passenger platform)that portion of a transit
facility directly adjacent to the tracks or roadway at which
transit units (vehicles or trains) stop to load and unload
passengers. Within stations, it is often called a station platform.
PLATFORM....centera passenger platform located between
two tracks or guideways so that it can serve them both.
PLATFORM....higha platform at or near the floor elevation
of the transit unit (vehicle or train), eliminating the need for
steps on the transit unit.
PLATFORM....lowa platform at or near the top of the
running surface of the transit unit (vehicle or train), requiring
the passenger to use steps to board and alight.
PLATFORM....sidea passenger platform located to the
outside of the tracks or guideways, as distinguished from a
center platform located between the tracks or guideways.
PLATFORM TIMEThe time a vehicle is in revenue service.
PROPERTY (operation, operator, system)in the transit
industry, a public transit agency or a private transit company
with responsibility for transportation services such as bus, ferry,
rail; see also transit district.
RAIL DIESEL CAR (RDC, diesel rail car)a self-powered
rail car that usually has two diesel engines and can usually
operate in multiple units (diesel multiple-unit car).
RAIL RAPID TRANSITsee transit system, rail rapid
RAIL RAPID TRANSIT CAR (rapid transit car, subway

car)a rail car for rapid transit systems. It is bi-directional,
usually powered, and equipped with a control cab at one or both
ends. It may be designed to operate in single or multiple units. It
has two to five double doors per side, designed for fast boarding
and alighting from high-level platforms.
RAPID RAILA system which operates high speed, high
capacity passenger trains using exclusive fixed guideways,
grade separated and with high level station platforms for
boarding passengers.
RAPID TRANSITTransit service which is operated
completely separate from all other modes of transportation. The
term rail rapid transit frequently refers both to operation of
light rail transit vehicles over exclusive right-of-way and heavy
trail transit vehicles; the term bus rapid transit refers to
operation of motor buses over exclusive bus roads or busways.
REGIONAL RAIL SERVICEsee service, regional rail
REVENUE
MILES
(revenue
kilometers)miles
(kilometers) operated by vehicles available for passenger service.
RIGHT-OF-WAY (ROW) A general term denoting land,
property, or interest therein, usually in a strip, acquired for or
devoted to transportation purposes. For transit, rights-of-way
may be categorized by degree of their separation: A-fully
controlled without grade crossings, also known as grade
separated, exclusive, or private; B-longitudinally physically
separated from other traffic (by curbs, barriers, grade separation,
etc.) but with grade crossings; C-surface streets with mixed
traffic, although transit may have preferential treatment.
RIGHT-OF-WAY....exclusive transita right-of-way that is
fully grade separated or access controlled and is used
exclusively by transit; transit ROW category A.
ROLLING STOCKThe vehicles used in a transit system,
including buses and rail cars.
ROUTE MILES (route kilometers)various definitions
exist for this statistic: 1. One-way duplicating is total mileage
(kilometers) of routes, where the roadway or guideway
segments of each individual route are summed up in one
direction. For example, a 1 mile (kilometer) segment over which
buses operate in both directions would be reported as 2 miles
(kilometers); also known as directional route miles (kilometers)
or miles (kilometers) of roadway or route. 2. One-way nonduplicating is total mileage (kilometers) of routes, where a
particular roadway or guideway segment is only counted once
regardless of number of routes or direction of travel on that
segment; also known as line miles (kilometers) or miles
(kilometers) of directional roadway. 3. Two-way mileage
(kilometers) is total mileage (kilometers) of each route covered
from start to finish. No attention is given to direction of routes
or number of routes using any particular segment of roadway or
guideway.
115
RUNNING GEARThe wheels, axles, springs, axle boxes,
frames, and other carrying parts of a bus, truck, rail car, or
locomotive.
SECTION 15The section of the Urban Mass Transportation
Act of 1964, as amended, that authorizes the Department of
Transportation to gather statistical information about the
financing and operations of public transportation systems, based
upon a uniform system of accounts and records.
SERVICEa system or method of providing people with the
use of something, for example, transportation.
SERVICE...base periodthe level of transit operations
during the base period.
SERVICE...commutertransportation provided on a
regularly scheduled basis during peak travel periods for users
commuting to work, school and similar destinations.
SERVICE...expressservice that has fewer stops and a
higher operating speed than regular service.
SERVICE...limited1. A transit service that operates only
during a certain period of the day, or that serves only specific
stops (also known as limited stop service) or in a specified area,
or that serves only certain segments of the population. 2. Line
service with some restrictions on boarding and alighting.
SERVICE...local1. Transit service that involves frequent
stops and consequent low average speeds, the purpose of which
is to deliver and pick up transit passengers close to their
destinations or origins. 2. Transit operation in which all transit
units (vehicles or trains) stop at all stations. 3. Transit service in
a city or its immediate vicinity, as distinguished from regional
transit service or interurban lines.
SERVICE...regional rail (RGR)regional rail passenger
service, usually provided by railroad agencies, that consists of
electric or diesel-powered trains on grade-separated railroad
lines (sometimes with protected grade crossings); see also
transit system, commuter rail.
SERVICE...revenue1.
Transit
service
excluding
deadheading or layovers. 2. Any service scheduled for passenger
trips.
SERVICE...service frequencythe number of transit units
(vehicles or trains) on a given route or line, moving in the same
direction, that pass a given point within a specified interval of
time, usually 1 hour; see also headway.
SERVICE...skip-stopservice in which alternate transit units
(vehicles or trains) stop at alternate sets of stations on the same
route. Each set consists of some joint and some alternate
stations.
SHORT TURNsee turn back
SIGNAL ASPECT1. The appearance of a fixed signal

conveying an indication, as viewed from the direction of an
approaching rail unit. 2. The appearance of a cab signal
conveying an indication, as viewed by an observer in the cab of
a rail unit.
SIGNAL PRE-EMPTIONin
automatic or manual device for
phasing for the sequence of a
preferential treatment for specific
buses or trains.
highway operations, an
altering the normal signal
traffic signal to provide
types of vehicles, such as
SIGNAL...automatic blocka system in which signals are

actuated automatically by the presence of a train on the track
section. Some block signal systems can use an electric circuit to
detect the presence of any vehicle, switch positions, broken rail,
and so on.
SIGNAL...blocka fixed signal installed at the entrance of a
block to govern trains entering and using that section of track.
SIGNAL...waysidein rail operations, a fixed signal that is
located along the track right-of-way.
SINGLE UNIT (SU)a powered rail car, equipped with a
control cab at one or both ends, that operates alone.
SPACINGthe distance
measured front to front.
between
consecutive
vehicles,
SPEED see velocity

SPEED...overall trip (effective operating speed, cycle
speed)in transit operations, the average speed achieved per
round trip, including layover time but excluding deadheading
time. It is calculated by individual trips, by running time
periods, or for the entire schedule.
SPILL-BACKin on-street light rail transit operations where
trains or motor vehicles fail to clear a signalized intersection and
so prevent the following train from entering that block.
Particularly acute in downtown streets where the light rail train
can be the full length of the block.
STATION1. An off-street facility where passengers wait for,
board, alight, or transfer between transit units (vehicles or
trains). A station usually provides information and a waiting
area and may have boarding and alighting platforms, ticket or
farecard sales, fare collection, and other related facilities. It is
also know as a passenger station. 2. In railroad operations, a
place designated in the timetable by name, at which a train may
stop for traffic or to enter or leave the main track, or from which
fixed signals are operated.
STATION ACCESSIBILITYA measure of the ability of all
people within a defined area to get to a specific transit station.
STATION...all-stopin
transit
systems
with
skip-stop
116
schedule or express service, a station that is served by all
scheduled transit units (vehicles or trains).
(vehicle or train) to move from one point to another, excluding

time for stops.
STATION...maximum load pointThe busiest station on a

line where the longer dwell establishes the minimum headway.
TIME...terminal1. For passengers, the time required at the

ends of trips to unpark and park their private vehicles, including
any necessary walking time. 2. For rail vehicles, the time
allowed at a terminal between arrival and departure for turning
vehicles, recovering delays, and preparing for the return trip. 3.
The time required for a passenger to pass through a terminal
when there is a change of mode.
STATION...off-linea station at which a transit unit (vehicle

or train) stops outside of the main track or travel lane so that
other units can pass while passengers board and alight.
STATION...on-linea station in which transit units (vehicles
or trains) stop on the main track or travel lane. This is the
common design, and the term is used only to distinguish this
station from off-line stations.
STREETCARan electrically powered rail car that is operated
singly or in short trains in mixed traffic on track in city streets.
In some areas it is also know as a trolley car and, primarily in
Europe, as a tram.
SUBWAY1. That portion of a transportation system that is
constructed beneath the ground surface, regardless of its method
of construction. 2. An underground rail rapid transit system or
the tunnel through which it runs. 3. In local usage, sometimes
used for the entire rail rapid transit system, even it is not all
beneath the ground surface. 4. A pedestrian underpass.
TERMINAL1. The end station or stop on a transit line or
route, regardless of whether special facilities exist for reversing
the vehicle or handling passengers; also known as a terminus. 2.
An assemblage of facilities provided by a railroad or intercity
bus service at a terminus or at an intermediate location for the
handling of passengers and the receiving, classifying,
assembling, and dispatching of trains or dispatching of buses;
also known as a depot.
TERMINAL...stuba dead-end terminal in which the entering
rail (or other guided) transit unit must depart by the same
guideway on which it entered. Because no loop is provided, a
bi-directional transit unit (vehicle or train) is necessary.
THROUGH ROUTINGthe practice of joining the ends of
radial transit routes to travel through downtown instead of
having each route turn back in the downtown and return to its
origin.
THROUGHPUTThe volume of vehicles passing or people
transported past a point or series of points during a given period
of time.
TIME...delaythe amount of time by which a transit unit
(vehicle or train) in service is delayed from its scheduled time.
TIME...dwellthe time a transit unit (vehicle or train) spends
at a station or stop, measured as the interval between its
stopping and starting.
TIME...runningthe actual time required for a transit unit
TRACK MILES (track kilometers)the sum of the oneway

linear miles (kilometers) of all trackage in a system, including
all main track and trackage in yards, car barns, switches, and
turnouts.
TRACK MILES...revenue (revenue track kilometers)
the number of miles (kilometers) of track used in passengercarrying service.
TRACK MILES...service (service track kilometers)the
number of miles (kilometers) of track used exclusively in non
revenue service.
TRACTION SAFETY INTERLOCKin rail transit a series
circuit of electrical switches that prohibits a train from starting
unless all passenger doors are closed and locked.
TRAFFICin traffic engineering and transportation planning,
the vehicles, people or both that pass a specified point during a
given period.
TRAFFIC...annual average weekday (AAWDT)daily
traffic that is averaged over a calendar or fiscal year and that
includes only weekdays (Mondays through Fridays). It may also
exclude holidays.
TRAFFIC CONTROL DEVICEa sign, signal, marking, or
other device placed on or adjacent to a street or highway, by
authority of a public body or official that has jurisdiction, to
regulate, warn, or guide traffic.
TRAFFIC COUNTa record of the number of vehicles,
people aboard vehicles, or both, that pass a given checkpoint
during a given time period. It may be classified by type of
vehicle. See also count.
TRAILER1. An unpowered rail car operated in trains with
powered cars (rapid transit) or towed by locomotives (regional
rail). 2. In some rail rapid transit systems, a trailer may be
powered; however, it does not have operators controls and thus
can only be operated in consists with cars that do.
TRAIN1. Two or more transit vehicles physically connected
and operated as a unit; see also transit unit. 2. One or more
locomotives or self-propelled rail cars, with or without other
cars but with marker lights. 3. On a headway sheet, a single
transit unit (vehicle or train) and all the scheduled work that it
performs during the operating day.
117
TRAIN BERTHin rail operations, the space designated for a
train of given length to occupy when it is stopped at a station
platform, in a terminal, on a transfer track, or at some other
designated place.
TRANSIT SYSTEM...major activity center (MAC

system)a transit system that provides service for short trips
within small, densely populated major activity centers, such as
shopping centers and downtown areas.
TRAIN OPERATIONthe way in which a train is operated,

for example, automatic with automatic overspeed control, or
manual with either automatic or manual speed control, or skipstop.
TRANSIT SYSTEM...rail any of the family of transit

modes with rail technology. The major ones, generally in
ascending order of performance, are streetcars, light rail transit,
rail rapid transit, and commuter or regional rail.
TRAIN...push-pulla locomotive and a set of cars equipped

with one or more cab cars from which the locomotive can be
controlled. The train is either pulled and controlled from the
locomotive in the conventional manner or pushed by the
locomotive and controlled from the leading car.
TRANSIT SYSTEM...rail rapid (heavy rail transit, rapid

rail transit)a transit system that generally serves one urban
area, using high-speed, electrically powered passenger rail cars
operating in trains in exclusive rights-of-way, without grade
crossings (Chicago is an exception) and with high platforms.
The tracks may be in underground tunnels, on elevated
structures, in open cuts, at surface level, or any combination
thereof. Some local terms used for rail rapid transit are the
elevated, the metro, the metropolitan railway, the rapid, the
subway, the underground.
TRANSFER1. A passengers change from one transit unit

(vehicle or train) or mode to another transit unit or mode. 2. A
slip of paper, card, or other instrument issued to passengers
(either free or with a transfer fee) that gives the right to change
from one transit unit or mode to another according to certain
rules that may limit the direction of travel or the time in which
the change may be made.
TRANSFER PASSENGERA passenger who transfers to a
line after paying a fare on another line.
TRANSIT SYSTEMthe facilities, equipment, personnel, and
procedures needed to provide and maintain public transit
service.
TRANSIT SYSTEM...automated guideway (automated
guided transit, AGT)any guided transit mode with fully
automated operation (i.e., no crew on the transit units). The term
usually refers, however, only to guided modes with small and
medium-sized vehicles that operate on guideways with
exclusive right-of-way. The term includes the personal rapid
transit concept and group rapid transit or people mover systems.
TRANSIT SYSTEM...commuter raila passenger railroad
service that operates within metropolitan areas on trackage that
usually is part of the general railroad system. The operations,
primarily for commuters, are generally run as part of a regional
system that is publicly owned or by a railroad company as part
of its overall service. In some areas it is called regional rail.
TRANSIT SYSTEM...light rail (LRT)as defined by the
TRB Subcommittee on Light Rail Transit, a metropolitan
electric railway system characterized by its ability to operate
single cars or short trains along exclusive rights-of-way at
ground level, on aerial structures, in subways, or occasionally,
in streets, and to board and discharge passengers at track or car
floor level.
TRANSIT SYSTEM...light rail rapid (LRRT)light rail
transit with exclusive, grade-separated right-of-way for the
entire system. It may have low or high-level platforms and
visual or signal control.
TRANSIT SYSTEM...streetcar (street railway, tramway,

trolley system)a street transit system consisting of
electrically powered rail vehicles operating in one to three-car
transit units, mostly on surface streets with mixed traffic.
TRANSIT UNITone or more transit vehicles coupled and
operated together. The term includes single vehicles (bus, rail,
or other guideway) and multiple car trains (rail or other
guideway).
TRIP1. A one-way movement of a person or vehicle between
two points for a specific purpose; sometimes called a one-way
trip to distinguish it from a round trip. 2. In rail operations, a
mechanical lever or block signal that, when in the upright
position, activates a trains emergency braking system. 3. The
movement of a transit unit (vehicle or train) in one direction
from the beginning of a route to the end of it; also known as a
run.
TRIP...inbounda trip toward the central urban area, into the
central business district, or to a timed transfer point or major
activity center.
TRIP...linked (linked journey, linked passenger trip)a
trip from the point of origin to the final destination, regardless of
the number of modes or vehicles used.
TRIP...outbounda trip away from the central urban area,
out of the central business district, or away from a timed transfer
point or major activity center.
TRIP...passengerone passenger making a one-way trip
from origin to destination.
TRIP...unlinked1. A trip made in a single vehicle. 2. The
boarding of one transit vehicle in revenue service; also known as
an unlinked passenger trip. 3. Any segment of a linked trip.
118
TRIPPER1. A train inserted in the schedule to make one
peak period trip. 2. An assignment of work to an operator that is
not long enough to qualify as a full days work.
VEHICLE HOURthe operation of a vehicle for a period of 1

hour.
TURN-BACK1. In transit operations, to cut short a transit

trip (to turn back before reaching the end of the route or line),
usually to get back on schedule or to meet peak passenger
demands; also known as a short turn. 2. In rail operations, a
point along a track at which a train may reverse direction.
VEHICLE MILE (vehicle kilometer)the movement of one

vehicle over a distance of 1 mile (kilometer).
TURNOUT1. In rail transportation, the assembly of a switch

and a frog with closure rails by which rolling stock or trains can
travel from a track onto either one of two diverging tracks; also
known as a track switch. 2. A short side track or passage that
enables trains, automobiles, and similar vehicles to pass one
another.
UNIDIRECTIONAL CARa rail car (usually light rail or
streetcar) that has doors on one side and an operating cab at only
one end so that it mush be turned around by separate means at
terminals.
URBAN RAIL CARa light rail, rail rapid transit, or
commuter rail car.
VELOCITY (speed)the distance passed per unit of time, or

the rate of change in location relative to time. For transportation
vehicles it is usually measured in miles (kilometers) per hour.
WHEELCHAIR LIFTa device used to raise and lower a
platform that facilitates transit vehicle accessibility for
wheelchair users and other handicapped individuals. Wheelchair
lifts may be attached to or built into a transit vehicle or may be
located on the station platform (wayside lifts).
YARD1. In rail systems, a facility within defined limits that
has a system of tracks used for making up trains, storing rail
cars, and other purposes. 2. In transit systems, an open storage
lot for light rail vehicles, streetcars, electric trolley buses, and
motor buses.
119
A1. APPENDIX ONE

Review of North American Rail Transit
Capacity Analysis Methodologies
This appendix is the result of Task 1 of the project.
Conduct a review of North American rail transit
capacity
experience
and
capacity
analysis
methodologies.
Figures, tables and equations abstracted from the literature are
not titled, numbered or indexed, but are inserted in the text, as
reviewed. Those figures, tables and equations from this review
that are used in the report are titled, numbered and indexed
therein.
There is considerable inconsistency in use of terminology in
the transit industry. In this appendix the authors terminology is
used. Where this could be confusing an explanatory footnote is
inserted. Similarly the authors mensuration is used with
conversion to the metric units used in the report where
applicable.
Inevitably in so wide a literature survey there are
contradictions between reports. No attempt is made to reconcile
these except where specific material is used in the main body of
the report.
A1.1 INTRODUCTION
Literature searches were carried out through BC Transits and
Transport Consulting Limiteds libraries and files, and through
electronic searches of the Library of Congress; University of
British Columbia and University of Minnesota libraries; the
transportation libraries of Northwestern University and
University of California, Berkeley; and the National Technical
Information Service and the Transportation Research Boards
Transportation Research Information Systemwith listings
from British and European sources, including the International
Public Transport Union (UITP).
The electronic searches used multiple combinations and
permutations of two or three key words:
rail
transit
capacity,
rapid transit
light rail
LRT
commuter
AGT
signaling
train control
public transport
metro
local transportLibrary of Congress terminology
The electronic search was disappointing; even with broad
generic key words, such as rail transit alone, it failed to turn up
several relevant documents known to the Principal Investigator
or suggested by the Panel. In part this reveals an inadequacy in
the abstracts or summaries used. In particular, multiple-paper

documents and reports could not realistically cover a dozen or
more papers in a 200-word (or less) abstract. One important
source of rail transit information, the American Public Transit
Associations Annual Rail Transit Conference, is referenced
only by paper titleand then only for the past few years.
Similarly, many electronic databases are recent and do not
include older sources.
Mitigating these deficiencies were valuable references
obtained from the initial search reports, plus reports known to
the Principal Investigator or suggested by the Panel, which were
read and synthesized. This process doubled the number of
documents and provided some of the richest and most useful
material.
A total of 381 potential documents were identified in the
electronic searches. Abstracts were obtained on the 65 of these
that appeared useful, resulting in 33 books and reports being
obtained or ordered in hard copy. The above mentioned iterative
process increased the final total to the 67 reports listed below.
The literature search and synthesis produced considerably
more relevant material than had been envisaged. It served as a
comprehensive source to guide and steer the projects
development and evolution, and equally important, indicated
deficiencies, problems and pitfalls that the project should correct
or avoid.
A1.2 LITERATURE
SUMMARIES
More than 70 papers, books and reports were read and
synthesized with respect to Rail Transit Capacities and Capacity
Analysis Methodologies. Each item is summarized below in
alphabetic order by author.
Only material relevant to TCRP A-8 study is included. The
synthesis is not intended to be a complete prcis of any item.
Following most summaries is a brief commentary indicating the
Principal Investigators opinion of the materials strengths and
weaknesses, and expectation of the usefulness of the material to
this project.
A brief overall Summary of the literature follows as section 3
of this appendix.
ABRAMOVICI, MARC, Optimization of

Emergency Crossovers and Signals for Emergency
Operations in Rail Rapid Transit Systems, APTA
Rapid Transit Conference, June 1982
Summary: The paper presents a methodology for determining

signaling requirements and cross-over locations that will
120
minimize disruption from single-track workingwhether due to
maintenance or an emergency.
An example is given for typically spaced rapid transit
crossovers, with an intermediate running time of 4 min,
(approximately 3 km or 2 mi). Uni-directional signaling would
reduce throughput to 33% of normal. Bi-directional signaling
would permit platooning with capacity reduced to 60% of
normal.
The paper provides means to calculate the restriction of
singletrack working with and without intermediate stations. It
shows that closer cross-over spacing can provide emergency
capacity that is 80-90% of normal.
Comment: The straightforward methodology also permits
calculations of headway for light rail with single-track sections.
The report raises the issue of how much allowance capacity
calculations should contain for irregular operations.
ALLE, P., Improving Rail Transit Line Capacity

Using Computer Graphics, Logistics and
Transportation Review, Volume 17, Number 4,
University of British Columbia, Faculty of
Commerce, Dec. 1981
Summary: The study asks the following questions: How many

trains can realistically pass a point in one hour? What is the
impact of station dwell times on this throughput?
The study analyses the E and F trains on the NYCTA at
Queens Plaza Station, using actual dwell time data and statistical
probability theory to show that, by trapping 85% of the area
under the normal distribution curve, the actual dwell time will
be below 75.23 sec, 85% of the time. Using this figure it
concludes that a single track can support trains every 130 sec
almost identical to NYCTs throughput of 29 trains per hour
(124 sec), which the agency says is saturation level.
The studys dwell time methodology is:
A 95% confidence interval for the true mean is given by:
where:
X = sample mean of dwell time data

S = sample standard deviation
n = number of observations
The interval estimator for the true standard deviation makes use
of the chi-square (X2) distribution. A 95% confidence interval
for is given by:
To trap 85% of the area under the normal distribution curve,

the upper control limit becomes the mean plus one standard
deviation. Conservatively assuming the above defined and to
be the upper limits of their respective 95% confidence interval,
the upper control limit for the peak-hour station dwell becomes
( + ).
The study observed dwell times over the morning peak hour
at Queens Plaza Station from 07:30 to 08:30.
Dwell Times Used in Analysis (seconds)
These dwell times produce a sample mean of 42.7 sec and a

sample standard deviation of 18.74. (Using exact rather than the
rounded data above.) The median is 37.5 and the maximum 125
sec.
The resultant upper control limit ( + ) calculates to 75.23
sec from this data. The throughput in trains per hour (Th)
becomes:
Th = 3600/[M + ( + )]
Where M is the minimum time separation in sec provided by
the three-aspect signal system on the immediate approach to the
station. This is determined as M = 55 sec through a graphical
computing process that inserts train performance and the
physical location of signal block boundaries. The three
restrictive signal blocks approaching the station are each 200 ft
long and there are blocks at 200, 400 and 700-ft-along the
platform, the latter being the departure signal for the 700-ft-long
platform. This maximizes throughput by allowing a train (with
yellow aspects) to enter the platform before the preceding train
has completely vacated it.
The computed figure cannot be determined for other locations
without access to the study program and considerable physical
data on the signal system. However Barwell(R11) and Auer(R09)
provide simpler means to calculate this minimum signal system
time separation figure for conventional signal systems and the
55 sec can be taken as a typical figure for the common threeaspect rapid transit signal systems in North America.
Comment: This is a valuable paper with data and methods
usable in the study to show line capacity with three-aspect signal
system and variable dwell times.
The merit of this paper is that, using real life data at one of
New Yorks heavy use stations, it produces train throughput
results that are very close to actual experience without applying
any of the judgment factors used in many other calculation
methods to calibrate theory with practice. The disadvantage is
that only one station, typical as it may be, is examined.
The main lesson is that although the average peak-hour dwell
time is 43 sec, the median is 37.5 and the maximum or worst
case dwell is 125 sec, the upper control limit dwell time used to
calculate maximum train throughputon a sustainable and
reliable basisis calculated to be 75 sec. This is some 74%
higher than the mean, 100% higher than the median, 150%
higher than the often quoted typical dwell of 30 sec and 40%
lower than the maximumall figures used in methods
suggested elsewhere in this review.
ALLEN, DUNCAN W., Practical Limits of

Single-Track Light Rail Transit Operation,
Transportation Research Record 1361, 1992: pp.
305-311
Summary: The author discusses a number of assumptions applicable to light rail transit. These assumptions equate the travel
121
time in both directions, establish the fixed headway, and
optimize the signaling for the performance of the light rail
vehicles to be operated. In addition, the author assumes that the
single-track occupancy direction alternates with train meets
occurring every half headway. The paper then goes into
considerable detail to include tolerable delay factors in the
optimum design calculations.
The paper also offers some observations and opinions that a
practical application of single track to light rail operations may
take into account. The author notes that several iterations or
adjustments may be required to reach a satisfactory solution.
The specific assumptions and methodology are:
vehicle performance is uniform in both directions.

headways are fixed.
all light rail vehicles have equal priority.
signaling is optimized for vehicles used.
occupation of single-track alternates by travel direction.
meets occur every half-headway (H/2).
length of single-track is determined by design allowances
for early and late vehicles.
The amount of tolerable delay, as given in the following table,

is a key factor.
where:
TK = track kilometers
RK = route kilometers
Comment: This is an interesting paper that presents an

organized but theoretical approach to determining operational
throughput of single-track sections of light rail transit
operations. While the authors observations may be incomplete
or not apparently relevant to this projects purpose, this study
may find the conditions of tolerable delay useful.
A potential deficiency is the papers suggestion that singletrack sections less than 500-m long are unlikely to be
economicbecause of the costs of special work. This is
incomplete and possibly misleading. It is precisely short-single
track sections that can save capital costs by squeezing light rail
through an underpass or over a bridge. The high special-work
(switch) costs can be avoided by use of gauntlet track. Short
single-track sections can have little impact on capacity and
service reliability and can often be scheduled on a random
arrival, first-come first-served, basis.
AMERICAN PUBLIC TRANSIT

ASSOCIATION, 1992 Transit Operating and
Financial Statistics
Summary: Used for basic information in the study database.
5
Condition E produces a maximum occupancy time for singletrack segments. This is given by:

ASSOCIATION, 1994 Membership Directory
Summary: Used for basic information in the study database.
6
For conditions C and B, the corresponding equations are:

ASSOCIATION, Equipment Roster 1993
Summary: Used for equipment data not in the more current and
detailed rapid transit roster (R03) above. Much missing door
information has been obtained in the data collection task.
where:
= occupancy time of section

TE 1
= headway
H
TClear = signal clearance time (typical light rail
value 8 sec)
TPass = time for an entire vehicle to pass a
control point (typical value for light rail:
3 sec)
TCrit = sum of expected early and late train
times at meet point
A Condition D has been empirically derived and may give a

safer, more realistic, estimate of maximum occupancy time than
does Condition E. It is given by:
Required trackage can be determined from:

TK =
2.0(RK)(1.0 - T1/H)

ASSOCIATION, Roster of North American
Rapid Transit Cars, 1993 Edition.
Summary: Used to enter rapid transit equipment dimensions,

door widths and other data in the study database.
ANDERSON, J. EDWARD, Transit Systems

Theory, Lexington, 1978
Summary: Anderson provides a comprehensive and analytic

review of transit system theory for automated guideway transit
(AGT), including spiral transition curve and super-elevation
calculations, modal split modeling and analytic methods of
project economic evaluation.
122
Two sections pertain to rail transit capacity. Chapter Two
introduces the basic equations of motion and shows how to
calculate performance. Jerk tolerance for standing and seated
passengers is introduced showing how in initiating and ending
both acceleration and braking the rate must be tapered to control
jerk. This results in actual performance being lower than the
simplistic performance calculation common elsewhere.
The book shows how these transitions together with
accelerating performance limitations (whereby the initial
starting rate of acceleration diminishes rapidly as the train gains
speed and follows the motor curve) result in a rate of
acceleration from start to balancing (cruise) speed that will be
less than half the initial accelerationsignificantly so if the
train is heavily loaded and/or on a grade.
A critical issue in the accurate calculation of close headways
is the acceleration leaving a station and Andersons formulas
suggest that the average rate of acceleration during this period
may be 20 to 30% lower than the rate often useddepending on
grade, load and the power-to-weight ratio of the equipment.
In Chapter Four, Anderson shows formulas to calculate the
minimum separation of trains. The most restrictive headway
occurs in the approach, stop and acceleration away from the
station.
where:
Xmin
k
Vmin
ae
=
=
=
=
the minimum separation distance

a safety constant
the speed of the trailing vehicle
the braking rate of the trailing vehicle
adjusted for jerk transitions
This separation distance enables the minimum headway Hmin to

be calculated
Hmin = Tc + 2Tr + Td + 2Xmin/Vmin
where:
Tc = time for exiting train to clear platform (or

blocks), calculated in the same manner as
Xmin
Tr = control and/or train operator delay and/or
reaction time
Td = station dwell time
Comment: Transit System Theory is a misleading title because

the book deals only with AGT systems. The minimum headway
calculations use a safety multiplier in calculating braking and
clearance distances. This approach is less clear than adding a
safety distance which can be calculated from set criteria. In the
TCRP A-8 study this latter method, as outlined in Auer (R09) and
Motz (R47) following, is preferred and has been used.
Andersons Commentary on jerk limitation, transitions to
braking and acceleration rates, and the rapid fall-off of the
acceleration rate as a train gains speed is invaluable.
AUER, J.H., Rail-Transit People-Mover Headway

Comparison, IEEE Transactions on Vehicular
Technology, Institute of Electrical and Electronics
Engineers, 1974
Summary: Discusses the application of conventional block

signaling to rapid transit and AGT with details of maximum
train throughput under various conditions for both modes.

Shows how the WMATA signaling system is designed for 75sec minimum headways with trains of maximum length. This
can be reduced to 18 sec on AGT systemswith the same
brick-wall safety standards.
The author describes time delays that limit signaling
throughput:
train operator reaction time varies, 0 with ATO

cab signal communication delay, 2.0 sec
overspeed detection delay, 0.75 sec
switch lock-to-lock time, 3.0 sec
MLH = 0.682K(TL + SBD + TDUD + SCBD)/CS
where:
MLH = Minimum Line Headway (sec)

K
= Safety Factor, must be 1 for brick-wall
standard
TL
= Train Length
SBD = Safe Braking Distance based on runaway
propulsion failure plus reduced braking
factor
TDUD = Train Detection Uncertainty Distance
SCBD = Service Control Buffer Distance (AGT
only)
CS
= Command Speed
Other equations are developed to calculate the headway on a

conventional three-aspect block signaling system under a variety
of conditions and assumptions, including the impact of
Automatic Train Operation (ATO) and cab signals over manual
operation. Cab signals can improve minimum headway by a
calculated 1.7 sec at an approach speed of 50 km/h while ATO
can effect a further reduction of some four sec at the same
speed.
Auer shows the components in the minimum headway, at a
command speed of 50 km/h, for a conventional three-aspect
signaling system. The total headway of 73 sec includes a 20-sec
dwell. The minimum line headways can be expressed as 53 sec
plus the controlling dwell. This corresponds closely to Alles
work (R02) which suggests a three-aspect signal system with
ATO can sustain a headway of 55 sec plus the upper control
limit dwell time.
The variation of this minimum headway (73 sec) with train

length is shown in the following figure.
The variation of headway with train command speed is shown
below. The minimum headway is 71.2 sec at 44 km/h
including a nominal 20-sec station dwell.
123
coded circuits in 1933, he suggests that moving-block systems
may take over many high speed inter-city applications where the
signal system must accommodate trains of differing lengths,
performance and speeds.
Barwell discusses rail junction optimization techniques and
the simulation of train following behaviorparticularly relevant
when train spacing is perturbed. He develops the minimum train
separation Se as:
Se = TL + BL + 0.75V2/aK
where:
Note that the command speed is the speed restriction imposed
by the signal system approaching and leaving a stationnot the
cruise or maximum speed between stations. Typical command
speeds will be in the 30 to 40 km/h range allowing a 75-sec
headwayclose to the optimal minimum of 71.2 sec. However
where there are restrictions, approaching or leaving a station,
due to special work or curves, the minimum headway can
increase significantly. At a more restrictive command speed of
20 km/h, the headway increases to 100 sec. Discounting the 20sec station dwell, this is an increase from 55 to 80 sec45%.
Comment: Auers paper provides one of the best, concise

summaries of a conventional three-aspect signaling system
throughput for both rapid transit and AGT. The results
correspond closely to actual field data. When combined with the
upper control limit dwell time calculations of Alle(R02) it
suggests both simple and complete methods for the study to
determine line throughput. It has been used in the study as the
best representation of three-aspect signaling systems.
10 BARDAJI, JORDI F., Regulating Headway in

Barcelona, The Urban Transport Industries
Report, Campden Publishing Limited, 1993, pp.
175-176
Bardaji describes how automatic regulation increased the
practical capacity of the Barcelona subway by 5%.
11 BARWELL, F. T., Automation and Control in

Transport. 2nd Revised Edition Pergamon Press
Limited, 1983
In this standard text, the late Professor Barwell covers many
aspects of rapid transit operation and control. Among his many
Comments are that transport problems generally reduce to the
consideration of headway at a bottleneck and admonitions that
some of the mathematical theory presented does not correspond
to actual field experience without practical adjustments.
In discussing multiple-aspect signaling systems he points out
that the law of diminishing returns operates very powerfully.
It is rarely economic to move beyond the typical three-aspect
signaling system although four aspects have been used to
increase capacity on some European high-speed inter-city
railroads. In noting that track circuits were first used in 1872 and
TL
BL
V
a
K
= train length
= block length plus safety distance or
block overlap plus sighting distance
= train speed
= braking rate
= a safety constant
Minimum headway (Hmin) is shown as:

+ recovery
Hmin = Se/V + maximum
station dwell
time
+ reaction
times
Comment: Barwell provides a useful way to calculate the

minimum train spacing for a moving block systemwhere
theory corresponds closely with practice. However both here
and in the train separation equation above, the introduction of
safety factors makes the calculation subjective. Barwells work
provides methods to calculate junction constraints on capacity.
12 BATELLE INSTITUTE, Recommendations en

vue de lamnagement dune installation de
transport compte tenu de donnes
anthropomtriques et des limites physiologiques
de lhomme, Geneva, 1973
Summary: The relevant parts of this report deal with
recommended comfort levels for many aspects of public
transport vehicles, including temperature, ventilation, noise,
floor slope, acceleration, rate of change of acceleration (jerk)
and passenger standing density. Information is provided for
three conditions, comfortable, uncomfortable and unacceptable.
The passenger standing density recommendations are
comfortable
uncomfortable
unacceptable
2-3 passengers per m2

5 passengers per m2
>8 passengers per m2
Details are provided on the projected body space of passengers

in various situations. The most useful of these for rail transit
capacity are tabulated for males.
124
13 BERGMANN, DIETRICH R., Generalized

Expressions for the Minimum Time Interval
between Consecutive Arrivals at an Idealized
Railway Station, Transportation Research, 1972
Vol. 6, pp. 327-341
Summary: Bergmanns mathematical treatise explores the

principal determinant in rail transit throughputthe minimum
time between successive arrivals at a station.
He expands on the basic equations of motion, examining in
particular limitations and effects of train approach speed, train
length, and the emergency braking rate. Four expressions are
developed for differing limits of these three variables.
The basic expression for minimum headway TA;i/i+1 without
limits is:
where
td = station dwell time

tr = emergency braking response time of
following train
Li = length of leading train
Vm = constant speed station approach
De = emergency deceleration rate
Do = operational deceleration rate
A = acceleration rate
Three other expressions are derived for variations or limits to

approach speed, train length, and the emergency braking rate.
These are plotted against approach speed to show a minimum
headway of 31 sec plus station dwell at an approach speed of
approximately 35 km/h (22 mph). Higher approach speeds show
a linear relationship to headway when operational and
emergency deceleration are equal. When emergency
deceleration is higher than operational deceleration the
minimum headway remains constant with approach speed.
Bergmann then compares his results with other authors before
concluding that increasing the emergency deceleration rate will
decrease minimum headwayswith the caution that the
approach is theoretical and does not take into account the effect
of finite signal blocks on train separation.
14 BERRY, RICHARD A., CERVENKA,

KENNETH J. AND SU, CHANG-AN, Traffic
and Light Rail Transit: Methods of Analysis for
DARTs North Central Corridor, Transportation
Research Record 1361, 1992: pp. 224-234
Summary: This paper presents a detailed summary of the
application of computer modeling of delays to automobile traffic
caused by DARTs (Dallas, Texas) North Central light rail line.
Grade crossing methods, at-grade or grade separated, are
proposed based on the effect of light rail on traffic.
Comment: Berry et al. are concerned exclusively with the
effect of the light rail on general traffic flow and do not address
the capacity of the light rail line itself. As such, this work is of
little use to the project except to confirm other similar work that
grade crossings have little, if any, impact on light rail capacity
compared with the constraint of signaled sections.
15 BOORSE, JACK W., Blending LRT into

Difficult Traffic Situations on Baltimores Central
Light Rail Line, Transportation Research Record
1361, 1992: pp. 197-206
Summary: Discusses light rail and traffic signal control for
intersections with long (110-sec) cycle time. Travel time
improvements are possible with sequencing. Confirms other
work that suggests on-street segments with traffic control are
generally less restrictive of capacity than the signal system used
on segregated track sections. For example, two trains platooned
through each traffic light cycle provide a throughput of 65 trains
an hourversus the 30 trains per hour of the signaling system
and 8 trains an hour limit of the single-track sections.
16 BURGIN, EDWARD A., Light Rail Transit

Signaling, Transportation Research Board Special
Report 182, 1978, pp. 119-123
Summary: Overview of light rail signaling, cab controls and
interlockings with breakdown of safety critical areas and
nonvital areas. Details Munis original Market Street light rail
subway cab control signaling with the three codes for 16, 43 and
80 km/h and automatic overspeed braking that occurs at 3.2
km/h over the set limit. (This signaling is now being replaced by
a moving-block system to increase capacity.)
17 BUSHELL, CHRIS., Janes Urban Transport

Systems, Janes Information Group Ltd., UK,
1989
Comment: The papers extensive analysis is interesting in
introducing the difference in minimum headways due to
operational and emergency deceleration rates and showing the
optimum approach speeds under various conditions. The
analysis does not take into account practical limits on
acceleration and deceleration with respect to passenger comfort,
not does it allow for performance variations, grades or
operational allowances. His calculated minimum headway of 31
sec plus dwell is applicable only to moving-block signaling
systems but provides an interesting lower limit for such systems.
Summary: A comprehensive reference to rail transit systems.
18 CALLAN, DENNIS R., Toronto Transit

Commissions 90 Second Headway Study: Getting
More Out of Existing Infrastructure, APTA Rapid
Transit Conference, Vancouver 1990
Summary: See reviews
based.
(R61)
and
(R68)
on which this paper is
125
19 CANADIAN URBAN TRANSIT

ASSOCIATION, Canadian Transit Handbook,
2nd Edition, Chap. 8Capacity, Canadian Urban
Transit Association. and the Roads and
Transportation Association of Canada, Toronto,
1985
Summary: This work gives a broad-ranging introduction to the
subject of transit capacity. Capacity is cited as being an elusive
figure. Both rail and bus modes are covered in easily
comprehensible language.
The chapter deals with the following determinants of
capacity: loading standards, headways and signaling, dwell
times and vehicle performance. The paper closes with a table of
design flows for selected transit modes.
Full utilization of capacity is limited to short periods of time.
Capacity is an elusive figure which is determined partly by the
level of service (speed, degree of crowding) desired. The
handbook defines three terms relating to capacity:
volume
demand
capacity
=
=
=
actual flow
potential flow
possible flow
A basic equation for determining capacity is given. Units are

passengers per hour.
(1)
(2)
where: h = headway (minutes)
f = frequency (units per hour)
n = number of vehicles per transit unit
p = passengers per vehicle
The author states that capacity is determined by a number of
factors which can be readily grouped into categories as follows:
1. Vehicle Characteristics
fleet size
maximum number of vehicles per transit unit
vehicle dimensions
seating configuration
number and location of doors
maximum speed
acceleration and deceleration rates
2. Right of Way Characteristics
cross-section design
degree of separation from other traffic
intersection design (at-grade or separated)
horizontal and vertical alignment
3. Stop Characteristics
stop spacing
on-line or off-line (latter allows passing stopped
vehicles)
fare collection method

high- or low-platform loading
length of platforms
turnaround facilities at terminals
4. Traffic Characteristics
volume and nature of other traffic (for shared right-ofway)
cross-traffic at intersections (at-grade)
5. Method of Headway Control
type of control separation standards for safety.
The report states that the permissible level of passenger
crowding on transit vehicles is an important determinant of
capacity. Standing densities of 0.1 m2 per passenger have been
observed in some cities but a value of between 0.2 and 0.7 m2
per passenger is more typical in North America.
Passenger behavior is also important in determining loading
standards as loading in cars and trains tends to be uneven.
Allowing passengers to travel between cars through end doors
can help even loading on a train. Irregular densities in cars can
be caused by passengers congregating around doors, stanchions
and the like.
Minimum headway is determined by the degree of separation
from other traffic, the method of headway control, and by dwell
time effects. Most rail transit modes other than streetcars have
controlled headways. For streetcars, the maximum frequency
is around 120 units per hour in mixed traffic. However, at this
frequency the service quality is reduced due to poor service
reliability. At such frequencies the traffic lane used by transit
essentially becomes a transit-only lane by default. A more
realistic maximum frequency would be 60 units per hour in
mixed traffic or 75 units per hour on an exclusive right-of-way.
Headways are governed by the type of signaling system. With
automatic train operation (ATO), door control and initiation of
acceleration remain under manual control. Automatic train
control (ATC) fully automates all aspects of train operation and
allows full driverless operation with possible throughput
increases.
The minimum headway for a block-signaled line can be
determined from:
where:
h
T
L
V
d
a
t
K
=
=
=
=
=
=
=
=
minimum headway(s)
station dwell time (s)
train length (m)
operating speed (m/s)
deceleration (m/s2)
acceleration (m/s2)
reaction time (s)
safety factor
A common value of h for lines signaled for minimum

headway is 90 sec. Sustained headways of 120 to 130 sec are
more common and allow for peak station delays. Headways of
dieselelectric commuter rail service tend to be on the order of 10
to 12 min to allow for grade crossings, longer braking distances
and mixed use of the rail line.
126
Average speed depends upon vehicle characteristics, traffic,
stop separation and dwell times. It is given by the following
equations:
transit modes. The chapter closes with a comprehensive view of

expected capacities for urban transit.
20
where:
S = stop spacing in meters
CELNIKER, STEPHEN, and TERRY, E.

WAYNE, Trolley Priority on Signalized
Arterials in Downtown San Diego,
184-188
If S is not constant then:
where:
k = number of stops
r = terminal time to turnaround (sec)
L = route length (m)
The main vehicle independent factors governing average

speed are dwell times and stop separation. For wide stop
separation, maximum speed is the most important vehicle
characteristic, while acceleration and deceleration rates are more
important with narrow stop separation. Acceleration and
deceleration values also partly regulate safe following distances.
A value of 1.25 m/s2 is reasonable for these characteristics.
Dwell times are controlled by the following factors:
number of passengers boarding and alighting

method of fare collection
number of loading positions
high/low level car entry and exit
door arrangement and number
seating arrangement
Typical ranges for on-board fare collection, low loading

equipment are 23 sec per boarding passenger and 1.52.5 sec
for each alighting passenger.
The chapter closes with a discussion of comparative design
flows and a table of capacities for various transit modes. Two
key points are that service quality and reliability are
compromised at the upper limits of design capacity, and that
design flows are generally only reached for short periods. Some
sample hourly capacities for the various modes are:
Streetcars, mixed traffic..............................6,060 ppphd
Streetcars (2 cars) exclusive lanes ............15,150 ppphd
Commuter rail, bi-level.............................13,750 ppphd
LRT, articulated cars ................................24,300 ppphd
Rail Rapid Transit.....................................43,000 ppphd
Comment: The chapter gives a broad ranging introduction to

the subject of transit capacity. It discusses full capacity as
limited to short periods of time because of practical peak service
considerations. These are covered in considerable detail with
regard to vehicle characteristics, right-of-way characteristics,
and methods of operational control. In addition, the chapter
covers the effect on capacity of passenger loading standards, and
on the physical and control limitations on headway for various
Summary: The trolley priority signaling system in downtown

San Diego was altered in 1990 as a result of the original
systems inadequacy following increases in San Diego Trolley
services. The original pre-emption mechanism gave light rail
vehicles full priority at all intersections. With more frequent
light rail service, this resulted in excessive delays to pedestrian
and vehicular traffic crossing streets used by the light rail. A
lack of an allowance for light rail trains traveling in the opposite
direction and a tendency to fail further high-lighted the need for
improved signaling.
The replacement system integrates light rail operations into
the downtown traffic signal progression system. This allows the
light rail trains to travel unimpeded from one station to the next
in a green window. In theory, all waiting for signals is done as
dwell time at stations. In the morning peak, some trains must
leave stations after the green window has passed but while the
nearest signal is still permissive. This allows the following train
to enter the station but results in the first train waiting between
stations.
Comment: The installation of an improved trolley priority

signaling system has improved light rail travel times in Centre
City San Diego. A capacity limitation created by the unidirectional nature of the earlier signal system has been removed.
21
DELAWARE RIVER PORT AUTHORITY,

90 Sec Headway Feasibility Study, Lindenwold
Line, Delaware River Port Authority, January
1973
Summary: To accommodate proposed new branches, methods

were examined to decrease headways on the inner (Camden Philadelphia) portion of the PATCO line to 90 sec. The analysis
shows that this headway can be achieved by a combination of
adjusting block boundaries and both increasing and reducing
speedsincreasing speeds where speed limits (curves) increase
the critical station close-in time, reducing those speeds that
produce limiting safe braking distances.
To mitigate the cost and inconvenience of increased travel
time, options were presented to increase the assured braking rate
and reduce the braking system reaction time within the
capabilities of adhesion and the existing slip-slide detection and
controlspecifically the overspeed control reaction time
which applies the brakes if the non-vital speed governor fails.
The report states Braking distance is one of the most important factors in the calculation of minimum headways because it
determines minimum safe train separation. A train must always
127
be separated from the train ahead, or end-of-line bumping block,
by at least the worst case stopping distance plus safety
marginstermed the safe braking distancea function of
speeds, curves, grades, braking rate, available adhesion and the
reaction times of on-board and wayside train control equipment.
PATCO uses automatic train operation with full automatic
driving. On this equipment the worst case reaction time occurs
when the speed governor fails just before receiving a lower
speed code with the train already close to the overspeed limit.
This worst case failure assumes the train is under full power
until the vital overspeed protection system intercedes and
applies braking. In a worst case, such emergency braking
assumes the failure of one set of braking equipment
(independent for each truck) on the shortest consist.1
A separate analysis examined changes necessary to
accommodate 90 sec headways in the downtown turnback. To
achieve this involved a combination of reducing the terminal
approach speed, relocating the terminal scissors cross-over from
behind to in front of the station and extending the tail tracks
behind the station.2 This had the added benefit of decreasing
turn-around time, in part compensating for increased running
times elsewhere.
The analysis in this report was based, in part, on the
separately summarized paper: Weiss, David M., and Fialkoff,
David R., Analytic Approach to Railway Signal Block Design,
Transportation Engineering Journal, February 1974.
Four supply-side parameters are defined:

Peak traffic share (PTS) (passengers per peak-hour
direction as a percentage of weekday ridership)
Vehicles per hour
Passengers per vehicle
Average weekday ridership (AWR)
The relationship between these four parameters is expressed as:
(AWR) (PTS) = (Veh/hr) (Pass/veh).
Following further discussions of the supply-side, the report
details the relationship between average weekday and peak-hour
ridership, citing data from many cities to show a North
American range of 9 to 24 percent with a mean of 15 percent.
The maximum service that a fixed-guideway transit facility
can supply or field is stated to be a function of maximum train
length and maximum frequency of service, with the former
determined by platform length and the latter by the train control
system. Other factors are stated to be vehicle performance,
maximum speed between stations, average dwell times at
stations and other operating considerations.
The report tabulates average peak-hour occupancy derived
from data between 1976 and 1990.
Comment: This report provides useful information on

providing higher capacity by reducing headways with track
circuitbased automatic train operation. The thorough, yet
concise, description of safe braking distance, and its constituent
components, is applicable to many rapid transit systems.
22
DEMERY, LEROY W., Jr., Supply-Side

Analysis and Verification of Ridership Forecasts
for Mass Transit Capital Projects, American
Planning Association Journal, Summer 1994
The report discusses capacity limitations on recent light rail

lines as they relate to the signaling system, single track sections
and maximum train length.
Summary: Demerys paper deals extensively with the

differenceand often confusionbetween the demand for and
the supply of service on rail transit.
... peak-period capacity is not an issue in most United
States and Canadian cities ... ... observed peakpoint
loads outside New York, Montreal and Toronto are well
below the theoretical capacity of the heavyrail and lightrail modes.
1
Reviewers Note: The worst case safe braking distance (sometimes called
the safety distance) is calculated from the worst case reaction time
assuming the heaviest passenger load, plus any possible snow and ice
load, tail wind (if any), steepest applicable down grade, adhesion limits,
and partial brake system failure. Note that the terminology worst case is
misleading. The truly worst case would be a total braking failure. In these
analyses worst case means reasonable failure situations. Total brake
failure is not regarded as a realistic scenario on modern rail transit.
2
Reviewers Note: The report recommended extending the underground tail
tracks by 125 ft. The possible alternate of energy absorbing train arrestors
was not discussed.
Demery states the maximum train length in San Diegos Centre City is 3
cars and that the four-car trains have a car added or removed at the 12th
and Imperial station. Other sources state that four-car trains are broken
into 2 two-car trains to move through city streets.
Demery discusses three reasons why vehicle loadings fall far

short of the theoretical levels.
Maximum peak-period demand occurs over intervals of 15
to 20 min (quoted as the sub-peak rather than peakwithin-the-peak).
As the number of standing passengers increases, loading
and unloading times also increase, extending dwells and
reducing schedule adherence.
128
Outside the largest, most congested urban areas, the level
of crowding that transit passengers appear willing to
tolerate falls well short of theoretical design or
maximum vehicle capacity.
After brief reference to different vehicle lengths and widths,
Demery suggests that, for the purpose of capacity calculations,
an upper plausible limit for vehicle occupancy is 150 passengers
per car with occupancy higher than 100 unlikely to occur
outside, New York, Boston, Montreal and Toronto. Long
before crowding levels ... reached New York levels, prospective
passengers would choose to travel by a different route, by a
different mode, at a different time, or not at all.
The report tabulates and compares daily and peak-hour
ridership and passengers per vehicle for 19 New York CBD
trunks for 1976 and 1991, as abbreviated below:
such as 30,000 to 50,000 passengers per peak hour direction for

heavy rail and 10,000 to 30,000 for light rail. It also says rail
transit ridership stabilizes when peak-period vehicle occupancy
reaches the point where prospective riders are no longer willing
to boardoften at a point well below that implied by the phrase
full standing load.
Comment: The gist of Leroy Demerys recent report deals with

the relationship between the demand for and the supply of
service on rail transit and is not relevant to this study. However
there are numerous useful insights on the issue of capacity. One
is the caution with respect to ridership data from the four light
rail systems that have CBD free zones. Another is the relatively
low average loading density in the peak hour on all US and
Canadian systems in the range of 0.9 to 3.7 passengers per m2
with the highest outside New York, Boston, Toronto and
Montreal being 2.3 passengers/m2.
The tabulation of average peak hour loadings per vehicle in
New York in 1976 and 1991 shows an 11 per cent decline in the
median over 15 years. Despite a few lines showing increases,
many othersdeemed, now and then, to be saturated or at
capacityhave lower loading densities. This would suggest an
expectation of better standards and Demery comments clearly
that new rail transit systems in cities with palatable transport
alternatives will not achieve these densitiesand if they reduce
service levels to increase vehicle loadings, as appears to be the
case on several systems, then riders will go elsewhere.
This suggestion has significant implications for a study of
future capacity based on existing and past ridership.
23
The report then makes a case that ridership forecasts are

prepared with little or no reference to supply-side parameters
and that ridership will be below forecast when the delivered
service frequency is below initial plansoften because too few
cars were purchased or there are inadequate operating fundsor
the line was not designed or signaled to accommodate the
frequencies used in initial forecasts.
Ridership figures can be misleading in cities with free
downtown zones. In Pittsburgh 20% of use is short trips in the
free zone.4
Ridership can increase without additional peak-hour supply
due to spreading periods of peak demand, a rise in off-peak and
reverse-peak use, and/or a willingness of passengers (and
prospective passengers) to tolerate higher levels of crowding.
The report states that effective capacityor likely maximum
ridershipfalls well below the routinely quoted capacity figures
4
Reviewers Note: The percentage is lower in Pittsburgh and Portland and

higher in Calgary (30%).
ENVIRODYNE ENGINEERS, INC., MetroNorth Speed and Capacity Improvement Study.

Tasks 1 to 5 US Department of Transportation,
Urban Mass Transportation Administration,
1989
Summary: The volumes in this series summarize the major

locations on the Metro-North rail system where capacity is
constrained. The key limitations include the following:
interlocking locations and layout, lack of grade-separation at
junctions, inadequate number of tracks, and short platform
lengths. The Port Jervis line faces an additional problem of
competition with freight trains for track access.
The capacity at each of the locations studied is given in
combinations of the number of express and local trains which
could be operated given current and future conditions.
Generally, express operations allow a higher throughput of
trains since there are no station stops during which time the
track is occupied. A particular problem is in finding pathways
for local trains which stop in more than one of the express zones
as the current track configuration is often not designed for this.
The provision of more local trains between zones is necessitated
by the growing suburb to suburb travel market.
Comments These reports outline specific instances of many of

the general capacity constraints faced by commuter rail
operators. An emphasis on conclusions reached, rather than the
129
simulation methodology used, restricts the usefulness of these
studies for general application.
24
EUROPEAN CONFERENCE OF
MINISTERS OF TRANSPORT, Report of the
38th Round Table on Transport Economics Scope for the Use of Certain Old-Established
Urban Trans. Techniques, Transport Capacity,
OECD, Paris, 1979, pp. 24-25
operation. It addresses a number of ways of reducing light rail

travel time through traffic signal pre-emption and shows that
with careful traffic engineering neither road not light rail
capacity is reduced by the grade crossingsat the headways and
specific circumstances of the Portland system.
26
Summary: Provides a European aspect to capacity with the

following list for maximum capacity by mode:
Comment: The passenger capacity figures can be misleading

because they do not indicate consistent length or loading
density. The maximum number of trains per hour reflects
European practice and is higher than similar North American
data.
25
FOX, GERALD D., Light Rail/Traffic Interface

In Portland: The First Five Years, Transportation
Summary: Fox summarizes the use of railway crossing gates,

traffic signals and stop signs to control grade-crossings on
TriMets 24.3 km light rail line in Portland. Signaling of the line
is also dealt with peripherally. The majority of the line is
operated on sight with 11.3 km of private right-of-way being
governed by automatic block signals (ABS).
A description of traffic signal pre-emption techniques is
given, ranging from wayside push-buttons to the Philips Vetag
inductive loop system. Installation of the latter system allowed
the addition of two intermediate stations to the line while
maintaining the same overall travel time.
The principal traffic control lessons learned from Tri-Mets
initial light rail line are:
Use conventional traffic signal equipment for public
familiarity and ease of maintenance.
Do not give motorists more information than they need
because it only causes them confusion.
Controlling traffic movements is generally more effective
than prohibiting them. Motorists tend to ignore
prohibitions but are more receptive to controls.
Light rail construction often involves lengthy street
closures which alter traffic flows. Such adjustments in
traffic flow can continue after construction is complete, so
reducing conflicts between light rail and vehicular traffic.
Comment: This paper updates previous work by the author on

this subject with practical experience gained from five years of
GILL, D.C., and GOODMAN C.J., Computerbased optimisation techniques for mass transit
railway signalling design, IEE Proceedings-B,
Vol. 139, No. 3, May 1992
Summary: This recent British paper compares and presents

analytic treatment of the capacity of fixed-block, multi-aspect
cab control and transmission based train control systems before
suggesting that the many nuances are beyond analytic methods
and require computer simulation.
The authors state that, in addition to the major headway
components of station close-in time plus station dwell, a margin
must be added to allow for small delays and variations in train
performance. They suggest an allowance of 15-20 sec and use
the term minimum service headway when this margin is
included, or signal headway5 when it is not.
Increases in line capacity require either increases in train
length or increases in positional resolution by creating shorter
block sections, possibly with an increase in either the number of
visual aspects or the number of automatic train protection
codesor the introduction of a moving-block signaling system.
Theoretical minimum headway (between stations), H, is
defined as:
where
v = velocity
b = deceleration
l = train length
Maximum train frequency, F, is 3600/H, setting dF/dv to zero

produces the speed (vop) for the closest headway)
As capacity is proportional to train length and inversely

proportional to headwayitself a function of train length, the
above equations can be merged to show that capacity is a
function of the square root of train length.
At the optimal running speeds in the above figure, train
frequencies range from 150 to 260 per hour without station stops
or with off-line stations. With station stops the typical maximum
practical train throughput is a much reduced 20 to 30 trains per
hour. As these high throughputs (between stations) are not
required, block lengths can be extended away from stations with
considerable cost saving and no impact on throughput.
5
Reviewers Note: CAUTION. Other authors rarely include station dwell in

the definition of signal headway.
130
On conventional rail transit systems with stations theoretical

headway calculations must take into account the time a train
takes to decelerate from line speed, stop at the platform and
accelerate out. With simplifying assumptions, Bergmann6 shows
that the theoretical minimum headway Hmin is given by:
where
vm
tr
tw
a
b
l
=
=
=
=
=
=
maximum velocity
ATO equipment response delay
station dwell time
acceleration
deceleration
train length
Under typical rail transit conditions, with a 140 m (460 ft) train
and a 30-sec dwell, this equation gives a minimum headway of
70-sec plus any operational margin.
Bergmann also derives the optimal line speed for maximum
throughput as:
where
bs = minimum service deceleration

be = inimum emergency deceleration
Under typical rail transit conditions this equation gives an

optimal line speed of 37 km/h. The authors specifically note that
this is the station approach speed and does not preclude higher
inter-station speeds.
The paper then analyzes the improvements in headways
which are possible by increasing the number of visual signalling
aspects or the number of automatic train protection codes. The
results, shown below, support their conclusion of diminishing
returns and indicate the optimum line speed approaching a
station of approximately 40 km/h.
6
BERGMANN, D.R.: Generalized expressions for the minimum time

interval between consecutive arrivals at an idealized railway station.
Transportation Research 1972, Vol. 6, pp. 327-341.
The paper then adjusts Bergmanns work to add allowances for

jerk limits into and out of acceleration and deceleration, grades
and curvesspecifically in station approaches and safety
distance adjustments. The equations are complex and still
require assumptions for train control and vehicle equipment
response or reaction time, driver reaction time (if any) station
dwell time, an operations allowance or margin, reduction in the
nominal acceleration rate as speed increases and fluctuations in
traction power voltage (and hence train performance) as trains
accelerate in each specific supply section.
Recommendations are made that computer simulation is the
preferred approach, combining a train performance program
with a signal layout design program. To compensate for such
refinements as traction voltage fluctuation and jerk, such
programs should be run at increments of 0.1 sec. The paper
points out that programs do not necessarily take coasting into
account.7
The results of such computer simulations are provided for the
following typical rail transit conditions:
train length
maximum speed
aspects
reference speeds
service braking
emergency braking
minimum jerk rate
140 m (460 ft)

80 km/h (50 mph)
4
80.0, 69.5, 53.3, 0.0 km/h
1.0 m/s/s (m/s2)
1.3 m/s/s (m/s2)
0.75 m/s/s/s (m/s3)
The resulting minimum headway was 74.8 sec plus dwell time
and an operational allowance. A 30-sec dwell and a 15-sec
operational allowance would produce a headway of 120 sec. The
programs were run for a moving-block signaling system under
the same conditions. The close-in headway was reduced to 43.9
sec, producing a minimum headway of 89 secleading to the
7
Reviewers Note. Coasting is a period when neither power nor braking is

applied. It is required by some operators as an energy conserving measure
and is often omitted in peak periods when the maximum system
performance is required. While coasting increases running time between
stationsand hence decreases system capacity with a given vehicle fleet
sizeit should not affect the minimum service headway (other than by
causing minor increases in supply voltage).
131
conclusion that, under typical conditions, a moving-block
signaling system can increase line capacity by 33%.
Comment: Gill and Goodmans lengthy paper provides the

most detailed analysis of train control system throughput in the
reviewed literature. Despite the analysis accommodating
nuances such as jerk and multiple equipment and driver reaction
times, ignored in most other work, there are still many variables
that are best accommodated by computer simulation.
The initial analyses of throughput without station stops appear
somewhat academic but allow the determination of any inter
station speed controls or speed limits. Such restrictions may
reduce throughput with stations only if they reduce the station
approach speed below the optimum 37 km/h (23 mph).
Otherwise inter station speed controls or speed limits only
reduce running times and impose the economic penalty of
requiring additional vehicles to serve a given passenger demand.
The results show minimum service headways that are longer
than most other work reviewed, even with station dwells only
estimated. The comparison of conventional multiple aspect
signaling systems and moving-block signaling systems is
valuable.
27
GRAHAM, IAN R., Optimizing Headways on

an Automated Rapid Transit System: The
SkyTrain Experience, American Public Transit
Association, Rapid Transit Conference,
Vancouver, B.C., 1990
The NYCTA sets crush loading standards at 255 passenger per

18 m cara density of 5 per m2. This makes the maximum
capacity of a ten-car train on a single trackwith a signaling
throughput of 30 trains an hoursome 76,500 passengers per
peak hour direction. Such a capacity is not realistic, however, as
it is based on the crush capacity of the cars.
John Fruin in Chapter 10 shows that the shoulders of the 95th
percentile male occupy 0.14 m2, and that unavoidable contact
between standees occurs at a space occupancy of 0.26 m2. Space
requirements in free standing lines or platform waiting areas are
0.5 to 1.0 m2 per person.
Comment: Fruins work is valuable in discussing the preferred

and minimum space per standing and per waiting passenger.
29
HOMBURGER, WOLFGANG S., Urban Mass

Transit Planning, The Institute of Transportation
and Traffic Engineering, Univ. of California,
1967
Summary: A comprehensive course text with examples of

actual rail system capacities. Useful table, albeit with out-dated
data, of peak-within-the-peak relationships (data from various
sources).
Summary: Describes how the use of moving block train control

with sophisticated Automatic Train Supervision allows close
matching of supply to demand by varying headways second by
second through each peak period.
Comment: Provides useful information on the relationship of
peak-within-the peak to average peak-hour demand. Data show
the loading standard difference between normal operation and
after delays where standing passenger density increases from a
mean of 2.8 per m2 to 5 per m2.
28
GRAY, GEORGE E., and HOEL, LESTER A.,

Public Transportation Planning, Operations and
Management, Prentice-Hall Inc., 1979
Summary: Comprehensive transit textbook with chapters by

individual authors.
Professor Vukan Vuchics Chapter 4 defines transit modes
and various terms, offering the following capacity ranges.
William Vigrass in Chapter 5 cites planned and actual

maximum capacities for selected examples. Munis light rail
metro is designed for 9,000 passengers per peak-hour direction.
Comment: The ratio of peak hour to peak-within-the-peak

capacity is an important part of TCRP A-8s approach to Rail
Transit Capacity. The above table has been extended,
recompiled with current data, and disaggregated by mode in the
studywhich designates this ratio as the first level of diversity.
30
JACOBS, MICHAEL., SKINNER, ROBERT

E., and LERNER, ANDREW C., Transit
Project Planning GuidanceEstimate of Transit
Supply Parameters, Transportation Systems
Center, US Department of Transportation, Oct.
1984
Summary: Chapter 4 deals with the estimation of capacities,

with many data from sources referenced elsewhere in this study.
The chapter cites level-loading doorway flow at 1.5 to 2.0 sec
per person per door lane with low-loading light rail increasing to
1.5 to 2.5 sec per person per door lane unloading and 2.0 to 8.0
sec per person per door lane boardingthe higher figures
relating to train operator fare collection.
132
The following factors in train headway are listed:
braking rate (with adjustment for any grade)

train control delays
maximum speed
type of signaling
train length
dwell times
block length
The basic modeling equation is:

Tj = T(xj+3 - TLmax) - T(xj)
with the constraint
Tj TD for all j
where:
North American platform lengths ranged from 70 to 213 m. The

closest European light rail headways at low speeds is quoted in
the range of 37 to 58 sec, North American range is quoted at 90
to 120 sec with the possibility of down to 40 sec.
A comprehensive section on vehicle space per passenger
suggests that gross vehicle area is the most practical data to use.
While 53% of U.S. rapid transit lines enjoyed rush hour loadings
of 0.5 m2 per passenger or better, the following data were
offered. (compiled from two separate tables from different
sources, average of 58 routes):
j
Ti
xi
xi+3
Tmax
TD
=
=
=
=
block number
block cycle time
block length of controlling joint
block length of controlled joint (3
blocks downstream from the
controlling joint)
= maximum train length
= desired headway (less dwell)
The model showed that the block lengths could be defined for
nine car trains (162 m) to permit a headway of 83 sec, plus
station dwell of 37 sec, for the design total of 120 sec, this is
down from the initial Montreal design standard of 150 sec.
Comment: An interesting and comprehensive approach to

optimizing the throughput of a conventional three-aspect
signaling system without overlays.
32
The report is one of the few to discuss the diversity of
standing densities within a carhigher in doorways/vestibules,
lower in aisles and at car ends (unless the car has end doors).
The report includes extensive references, tables of data and a
glossary.
Comment: As one of the most comprehensive compilations of

loading standards, this has been useful to the project.
31
KLOPOTOV, K., Improving the Capacity of

Metropolitan Railways UITP, 40th International
Congress, The Hague, 1973
Summary: Klopotovs report is derived from questionnaires

sent to 38 international rapid transit systems, three-quarters of
which stated they were working to increase capacity.
The percentage of peak-hour passengers that are seated ranges
from 12.5% in Tokyo to 70% in Liverpool (PATH 30%, SEPTA
55%). (Systems with a 100% seated policy are excluded.)
Average peak-hour loading density varies widely:
JANELLE, A., POLIS, M.P., Interactive

Hybrid Computer Design of a Signaling System
for a Metro Network: IEEE Transactions on
Systems, Man and Cybernetics, 1980
This comprehensive computer modeling approach looks at

how to obtain the maximum train throughput in designing a
three-aspect signaling system for rapid transit. Although specific
to Montreals rubber tired Metro, the approach is adaptable to
any three-aspect signaling system.
The model makes use of the repetitive nature of rapid transit
operations and assumes Automatic Train Operation that
regulates speed and controls station stopping. Block ends are
assumed fixed at station platform ends and interlockings, and a
train separation of two blocks is maintained at all times. The
model adjusts other block lengths to maximize throughput using
the following relationship among travel time, block length and
capacity.
Controlling station dwell to increase capacity shows that 54%

of the systems surveyed have four double doors per car side,
each in the range of 1.2 to 1.4-m wide with the great majority
close to 1.4 m. Door opening and closing times range from 1.0
to 4.5 sec with most in the 2- to 3-sec range. Brief mention is
made of the Paris Metros dwell control method of closing off
platform entry as a train approaches and Copenhagens method
which is to start opening the doors before a train has come to a
full stop.
A common dwell reduction feature is doorway setbacks so
that standing passengers do not block the flow. 71% of surveyed
systems had setbacks of 200 mm or more.
Most systems had sustained peak-hour headways at or greater
than 120 sec with the exception of Tokyo (110 sec), Leningrad
and Philadelphia Market-Frankford (105 sec), Paris (95 sec),
133
PATH (90 sec) and Moscow (80 sec) The latter required an
expensive move from a two- or three- aspect to four-aspect
signaling system.
Methods employed or planned to increase capacity ranged
from decreasing seating space to removing cabs from all but the
end cars, with the most common approach being new or
improved signaling to reduce headways.
Signaling changes including adding automatic train operation
and automatic train supervision, using more realistic safety
distances, adjusting block lengths or adding blocks. Where
station capacity was a limitation, improvements were suggested
to increase passenger flow to and from platforms. These
included separating entry and exit flows and operating escalators
at higher speeds. While most escalators in the United States run
at 0.46 m/s, 0.6 to 0.75 m/s is used occasionally in Canada and
frequently in Europe with certain former Soviet bloc cities
doubling flows with speeds of 0.75 to 0.9 m/s.
Comment: Although outdated, this report presents

comprehensive information on rail transit capacity,
unfortunately diminished by the poor translation and editing
from Russian. Russian and Japanese rapid transit systems
achieve the highest capacity in the world by a combination of
very close headways and high densities of standing passengers.
Several countries show that close headways can be operated
reliably and (when adjusted for North American loading levels)
provide an upper limit to rapid transit capacity.
33
KOFFMAN, D., RHYNER, G. and

TREXLER, R., Self-service Fare Collection on
the San Diego Trolley, US Department of
Transportation, 1984
Summary: Chapter 3Transit Operations provides a

comparison of dwell times between light rail in San Diego and
Boston. In both cities observers with stop watches rode the light
rail lines counting and timing passengers entering and leaving
each car, along with the number of passengers remaining onboard. Data was collected at all stations in San Diego and in
three sets for Boston: fare free (station collection) zones (two
routes) and inbound cars with train operator fare collection.
The model used multiple regression analysis with loading
time (dwell time) as the dependent variable and total on, total
off, total on-board as the independent variables. (The San Diego
model included zero-one variables to represent whether there
was any boarding or alighting activity at a stop.) The
coefficients of these variables include the extra time needed in
San Diego for the first boarding or alighting passenger who
operate the doors themselves. (Similar variables were tested in
Boston but, as could be expected, performed poorly).
After testing a variety of variables, including various powers,
exponentials, logarithms and interaction terms, a linear model
produced the best results. The only non-linear terms which
improved any models were the squares of ons and offs in San
Diego. These made only a minor improvement and were not
used as they have no physical interpretation, may be due to
errors in the data collection process and make the comparison

between the two cities difficult.
The result was a two part loading model with one linear
relationship for passengers movements from zero to one and
another linear relationship for all passengers movements above
one.
Only the San Diego model results are shown below.
Note that data is excluded from terminal stations and train operator relief
points, dwell times are from first door open to last door shut and excludes
time when the door is open without any passenger activity.
Finally, a composite model was developed using the constant

and zero-one coefficients from the San Diego model and the
remaining coefficients from the Boston inbound model.
Loading Time = 7.76 + (1.91) (Any Ons):
+ (1.12) (Any Offs):
+ (3.12) (Cash Ons):
+ (1.94) (Non-cash Ons):
+ (1.61) (Offs):
+ (0.87) (Passengers on-board):
The 95% confidence interval is 2 sec, computed from the
estimated variances and co-variances in each component
model. The report shows that without self-service fare collection
the San Diego running times would increase from the then 42
min, to 47 to 48 min.
Comment: Chapter 3 provides a comparison of loading times

between San Diego MTDBs self-service fare collection system
and that part of the MBTAs Green Line where on-board train
operator collection is used. The methodology and data provide
useful information for use in estimating light rail station dwells
with low loading.
34
KORVE, HANS W. and WRIGHT,

PATRICK M., New Standards for Control of
At-Grade Light Rail Transit Crossings,
Transportation Research Record 1361, 1992:
pp. 217-223
Summary: There is very little consistency of traffic control

devices used on American light rail lines. Variation can be
found not only between, but also within systems. Korve and
Wright outline the need for an American standard system of
traffic control devices and the efforts of an Institute of
Transportation Engineers committee to create such a standard.
134
35
Boarding and Alighting Time = -4.0 + 2.0(ons)

1 (offs) 8 and 6 (ons) 13
Data sets = 5
Coefficient of Determination = 0.94
Standard error of estimate = 1.50
KORVE, HANS W. Traffic Engineering for

Light Rail Transit, Transportation Research
Board Special Report 182, 1978, pp. 107-114
Summary: Korve provides an alternative definition of light rail

and shows that light rail road crossings can be separated in
space or time, detailing control options for the latter.
Stop signs are acceptable for grade crossings with traffic <
5,000 vehicles/day and light rail > every 5 min. Total
preemption is feasible down to 2-min light rail headways with
multiphase traffic signals and cross-traffic as high as 25,000
vehicles/day. On inter-connected traffic signals, progression can
be adjusted to favor light rail. Where possible, light rail stop
placement can be arranged to enhance progression speed.
The report contains acceleration and braking curves for
modern light rail vehicles and shows various methods to
accommodate traffic turning left across median light rail tracks.
Comment: The wide variation in results from city to city,

vehicle to vehicle and mode to mode suggest caution in
developing a general equation for dwell times. Kraft comments
that platform congestion could increase alighting times but
provides no data to substantiate this.
There are several deficiencies in this report which has been
quoted in several other papers. As such, the report is of little
value to the study other than to suggest caution in system to
system comparisons.
37
Comment: The report shows that light rail grade crossings
should rarely impact line capacity as good engineering can
ensure that a train can move through a grade crossing on each
light cycleand, in certain circumstances (limited train length),
a platoon of two trains per cycle. This condition permits a
throughput of 60 to 120 trains per hour, well beyond the
capacity of any signaling system on other sections of a typical
light rail line.
However, such throughput will impose delays which can be
minimized (or eliminated) with properly timed progression and
coordinated station placementbut only in one direction.
Progression timing can be adjusted to favor the peak direction.
36
KRAFT, W. H., and BERGEN, T. F.,

Evaluation of Passenger Service Times for Street
Transit Systems. Transportation Research
Record 505, Transportation Research Board,
Washington DC 1974
Summary: Kraft analyzed 1500 entry and exit observations to

derive an expression for passenger loading times, using the
method of least squares. All were on surface vehicles,
disaggregated by type of fare payment, time of day and by the
following types of flow.
all passengers boarding
all passengers alighting
mixed flows
The results show linear relationships with distinct differences
for elderly, handicapped and commuter passengers. Off-peak
passenger times were more leisurely. The applicable results for
low-loading streetcar (light rail) with exact fare, and doubledoors were:
Boarding Only Time = 3.4 + 0.9(ons)
Data sets = 7
Coefficient of Determination = 0.64
Standard error of estimate = 0.90
KUAH, GEOK K. and ALLEN, JEFFREY B.,

Designing At-Grade LRT Progression: Proposed
Baltimore Central Light Rail, Transportation
Summary: Kuah and Allen analyze the effect of modifying the

traffic signal progression in downtown Baltimore to allow the
light rail service on Howard Street to benefit from progressive
signaling. Computer modeling of the proposed changes shows a
10% increase in downtown traffic flows.
Comment: The paper does not directly address capacity but
does provide information on the related issue of light rail
signaling on city streets. It is interesting that the current
signaling is not mentioned as a capacity limitation.
38
KYOSAN ELECTRIC MFG. CO., LTD.

Total Traffic Control SystemTTC,
Yokohama, Japan, 1986
Summary: Many Japanese electric railways, typically a cross

between rapid transit and commuter rail, operate intensive
service. This report describes the track layout, signaling system
and operations of one of the busy two-track lines in the suburbs
of Tokyo.
The Keio Teoto Electric Railway has a two-track main line
between Keio-Hachioji and Shunjuku. Four branches merge into
this line and many trains continue through into central Tokyo
via joint running with the subway system. There are 49 stations
and a total of 63 route kilometers.
The Keio Teoto Electric Railway operates 30 trains in the
peak hour over a single track, combining four levels of express,
semi-express and local service. This frequency is made possible
by four platform and off-line platform stations, where faster
trains pass local trains, and an Automatic Train Supervision
system. The signaling system is a relatively conventional threeaspect block system.
Comment: This manufacturers description shows how commuter rail capacity can be increased with multi-track stations,
135
precision operation and the assistance of a computerized
automatic train supervision system.
39
LANG, A SCHEFFER, and SOBERMAN,

RICHARD M., Urban Rail Transit: Its
Economics and Technology, MIT Press, 1964
Summary: Lang and Sobermans book on rail transit economics

and technology is reportedly the first since Dooliitles treatise of
1916. Three sections relate to the A-8 rail transit capacity
project.
Parts of Chapter Three, Stations, deal with the interaction of
train and station design with dwell times. Loading time is
dependent on the distribution of passengers along the platform,
the ratio of total door width to car length and the number of
boarding passengers. Obtaining a uniform distribution of
passengers along the platform is desirable but difficult,
particularly so when crowded platforms impede flows in the
rush hour.
Sufficient entries and exits to adequately sized platforms are
necessary and must be evenly spaced for best distribution.
Passageway flow rates of up to 100 passengers per minutes per
meter of width are quoted (30 per minute per foot). Downward
stairs reduce this flow by some 25%, upward stairs by 40%.
These flow rates diminish when crowding exceeds 4 persons per
square meter (0.4 square feet per person).
In Chapter Four, Rail Transit Vehicles, Section 4.5 Car
Capacity and Dimensions discusses seating provision relative to
compromises between capacity and comfort. Suggesting that all
rapid transit cars are substantially similar in width, the report
equates passengers per square foot versus the percentage seated.
This ranges from 0.3 passengers per square foot with 50%
seated to 0.6 passengers per square foot with 15% seated. This is
then translated into passengers per Linear Foot of Train, as
shown below. The maximum vehicle capacity is 4 passengers
per linear footapproximately 2.5 square feet per passenger.
to attract passengers, higher comfort levels, i.e., less crowding,

are desirable.
Chapter Five of this text deals entirely with capacity. Capacity
is calculated as the number of trains per hour multiplied by train
length and the passengers per linear foot from the above graph.
Using the mathematics of Appendix A the minimum headway is
expressed as:
where
h
L
T
V
a
d
=
=
=
=
=
=
headway (s)
total train length (ft)
station stop time (s)
maximum train speed8 (ft/s)
rate of acceleration (ft/s2)
rate of deceleration (ft/s2)
Applying this equation at a maximum approach (close-in) speed

of 32 km/h (20 mph) and a dwell of 40 sec produces the
following optimum headways for different train lengths, and
capacity in passengers per peak-hour direction. These use a
vehicle loading of 3.1 passengers per linear foot with average
acceleration (a) of 3.0 mph/s (1.33 m/s2) or 2.0 mph/s (0.89
m/s2)
Appendix A, Some Considerations of Minimum Headway,

develops the above minimum headways with equations for
wayside signals and theoretical minimum headways and
minimum headways with automation.
The theoretical minimum headway is expressed as
The minimum headway with cab signals, assuming the

following train stops behind the preceding train before entering
the station is
For a completely automated system
where
The authors also discuss the importance of ease of ingress and

egress, recommending minimum distances between seats and
doorways and discouraging three abreast seating. Comfort levels
are discussed relative to smoothness of operation and the issue
of supply and demand. Where systems are oversubscribed and
few attractive alternate forms of transportation are available,
high levels of crowding will be tolerated. Where systems wish
h
L
T
a
s
r
c
=
=
=
=
=
=
=
headway (s)
total train length (ft)
station stop time (s)
rate of acceleration (ft/s2)
safety distance (ft)
operator reaction time (s)
communication time (s)
Reviewers Note: The maximum train speed, in feet per second, is the
maximum speed in the final approach to the stationnot the maximum
speed between stations.
136
Comment: In one of the earliest modern texts on rail transit
Lang and Soberman have provided a succinct yet thorough
outline of capacity issues. Their calculations, regarded by the
authors as conservative, tend to show passenger volumes higher
than would be regarded as practicaldue to their use of dwell
times of 30 to 40 secwhich do not take into account an
allowance for irregular running.
40
LEVINSON, HERBERT S., and HOEY,

WILLIAM E. Some Reflections on Transit
Capacity Proceedings of the International
Symposium on Highway Capacity, Karlsruhe,
July 1991
Summary: The authors comment that transit capacity is far

more complex than highway capacity. They show that train
headway is the sum of dwell time plus the reaction time, braking
time, acceleration time and time to clear the station.
They caution that because actual capacities may vary in a
way that cannot actually be described in a formula ... capacities
obtained by analytical methods must be cross-checked against
operating experience...
The study cites the historic high train throughput on the
Chicago Loop with visual rules (70 trains per hour) versus the
maximum NYCTA throughput on a three-aspect signaling
system of 35 trains per hour, achieved by use of key-by
procedures. Similar historic experience has shown streetcar
throughput on a single track of up to 145 cars per hour.
The same general (transit, all modes) capacity formula is
shown as in the ITE Transportation Planning Handbook below.
Comment: A useful general paper which repeats the cautions

necessary in an analytic approach.
41
where:
LEVINSON, HERBERT S., Capacity Concepts

for Street-Running Light Rail Transit, Australian
Road Capacity Conference 1994
Summary: The report compares historic streetcar service

capacities of up to 150 cars per track per hour with current
services that reach 96 cars per track per hour (Hong Kong) and a
passenger volume up to 8,500 passengers per peak hour
direction (Calgary).
On-street light rail capacity is related to the loading and
unloading times at the busiest stop, train length and traffic signal
cycles. Train length is limited to the shortest city block. Dwell
time is related to the loading level (platform height) and fare
collection system.
Basic capacity is defined with a simplified version of the
formula in the same authors Chapter 12 Capacity in
Transportation Planning, of the Transportation Planning
Handbook.
The formula for trains per hour per direction with signalized
intersections is given as:
Cp = trains per hour per track

tc = clearance between trains is defined as
the sum of the minimum clear spacing
between trains plus the time for a train
to clear a station, with typical values of
25 to 35 sec. (Some transit agencies use
the signal cycle length as the minimum
clearance time).
D = dwell time at stop under consideration,
typically ranging from 30 to 40 sec,
sometimes to 60 sec.
time variations and/or uncontrolled
variables associated with transit
operations. R values are tabulated from
1.0 in perfect conditions with level of
service E to 0.634 with level of
service A, assuming a 25%
coefficient of variation in dwell times.
Maximum capacity under actual
operating conditions would be about
89% of that under ideal conditions
resulting in about 3,200 effective sec of
green per hour.
g = effective green time, sec, reflecting the
reductive effects of on-street parking
and pedestrian movements as well as
any impacts of pre-emption
= cycle length, sec
c
Passenger spaces per car, needed in this equation to determine

capacity, are suggested at an occupancy level of two passengers
per m2, compared with a crush load of 4 per m2.
The results quote a maximum capacity of 40 to 45 trains per
track per hour at level of service E,9 reducing to 36 to 40
trains when variations in arrival and dwell times are
consideredequivalent to 10,000 to 13,500 passengers per
peak-hour direction with trains 46 to 69 m long.
System planning based on level of service D is
recommended. The following table extract shows light rail
capacities in trains per track per hour at level of service D,
with 23-m long cars, a typical 50% green cycle ratio, an R of
0.80, a station clearance time of 5 sec per 25 m of train, and a
further reduction to 80% of maximum capacity.
Mention is made of the possibility that two single-car trains may

be able to berth in a station simultaneouslydoubling the
capacity. Train spill back10 is discussed and two
recommendations made:
9
10
Levinson, Herbert S, Chapter 12 Urban Mass Transit Systems,

Transportation Planning Handbook,
Failure of a train to clear an intersection within the green cycle.
137
The length of trains should not exceed the street block
length.
There should be no more that one train every other block
to reflect variations in arrival and dwell times, suggesting
that there should not be more than one train every other
signal cycle where blocks are less than 122 m.
These recommendations result in a design capacity of 30 trains
per hour for 60-sec cycles, reducing to 20 for 90-sec cycles and
15 for 120-sec cycles. The equivalent capacity, based on a 30sec dwell time, ranges from 4,500 to 10,000 passengers per
peak-hour direction for two-car trains to 6,000 to 13,500 for
three-car trains.
FACTORS INFLUENCING TRANSIT CAPACITY

(* non-rail factors removed or adjusted)
1. Vehicle Characteristics
a.
b.
c.
d.
2. Rights of Way Characteristics

a.
b.
c.
d.
e.
f.
Number of tracks
Degree of separation from other traffic
Intersection design
Horizontal and vertical alignment
Route branching and junctions
Turnaround conditions at terminals
3. Stop Characteristics
a.
b.
c.
d.
e.
f.
g.
h.
i.
The report concludes with a list of useful planning guidelines.

dwell times should be minimized by using cars with high
platforms or low floors, multiple doors and fare
prepayment.
Green time for trains should be maximized.
Exclusive lanes should be provided
Routing patterns should minimize the number of on-street
turns
Central area junctions should be kept to a minimum.
Number of cars in train

Car dimensions
Number and configuration of seats
Number, location, width and actuation of doors
Spacing
Dwell Time
Design (on-line or off-line)
Platform height (high or low level boarding)
Number and length of loading positions
Method of fare collection
If on-board fares, type of fare
Common or separate areas for boarding or alighting passengers.
Passenger accessibility to stop
4. Operating Characteristics
a.
b.
c.
d.
Service types* (express, local)

Layover and schedule adjustment practices
Time losses to obtain clock headways or crew reliefs
Regularity of arrivals at a given stop
5. Passenger Traffic Characteristics

Comment: This recent paper adds to the substantial transit
capacity work by author Levinson with information on light rail
on-street operation. It contributes useful information to the
study.
The one train every two light cycles provides a basis for the
simple capacity calculations and conveniently coincides with
the typical maximum frequency on signaled segregated track of
30 trains per houralthough several new U.S. light rail lines are
only signaled for 17 trains per hour. (3.5-min headways). The
spill back situation has been investigated in the study and the
more detailed calculations can be used to help determine dwell
times and to analyze junction clearance and turnback times.
42
LEVINSON, HERBERT S., Chapter 5 Urban

Mass Transit Systems, Transportation Planning
Handbook, Institute of Transportation Engineers,
Prentice Hall, 1992
Summary: The author provides a comprehensive outline of

transit services with definitions and extensive data tables.
Characteristics and capacities are shown for numerous transit
vehicles, including some performance curves and formulas to
calculate performance.
Figures 5.10 and 5.11 show the relationship between
maximum speed, station spacing and average speed is
documented and a tabulation11 shows one second per passenger
per lane for level boarding and alighting and 1.7 sec for lowlevel (light rail) alighting.
A table of Factors Influencing Transit Capacity, derived from
the Highway Capacity Manual and Canadian Transit Handbook
is the most comprehensive in the literature.
11
Table 5.16 Average Boarding and Alighting Intervals for Transit Vehicles
a. Passenger distribution among major stops

b. Passenger concentration and interchange at major stops
c. Peaking of traffic (peak-hour factors)
6. Street Traffic Characteristics

a. Volume and nature of traffic (on shared right-of-way)
b. Cross traffic at intersections (where at grade)
c. Curb parking practices
7. Method of Headway Control

a. Automatic or by train operator
b. Policy spacing between trains (* or safety distance)
Comment: A wealth of information. The above table and

certain performance information has been used in developing
the Analytic Framework.
43
LEVINSON, HERBERT S.; ROBINSON,

CARLTON, C. and GOODMAN, LEON,
Chapter 12 Capacity in Transportation Planning,
Transportation Planning Handbook, Institute of
Transportation Engineers, Prentice Hall, 1992
Summary: Chapter 12 follows the more general information of

Chapter 5 to present a wide range of capacity information with
material synthesized from many sources.
The general equation for capacity of a transit line is given as:
138
and for light rail with controlled intersections:
where:
Cp = passengers per hour per track

tc = clearance between successive cars or
trains, in sec
D = dwell time at the major stop on the line
under consideration, in sec
n = number of cars in train
time and arrival time variations (0.833
suggested in text for maximum theoretical
capacity for buses, 0.89 in later railspecific references)
g = traffic light green time, in sec
c = traffic light cycle length, in sec
Various passenger load factors are shown based on a

percentage of seats. Loading standards A through F (crush) are
tabulated. The suggested schedule design capacity is 2.8 to 3.3
passengers per m2, 25% below the crush capacity. The peakhour factor is discussed for 15-min peak-within-the-peak. A
range of 0.70 to 0.95 is suggested, approaching 1.0 in large
metropolitan areas. Diversity of loading between cars of a train
is mentioned but only limited data is provided.
Specific capacity for rapid transit is shown as:
Numerous examples are given of actual capacity with rapid

transit maximums ranging from Hong Kongs 81,000 passengers
per peak hour direction to NYCTs 53rd Street tunnel at 54,500
in 1982, down from 61,400 in 1960. The calculated maximum
attainable for 10 car trains every 120 sec is shown as 57,300
passengers per peak hour direction after a 15% reduction for
unequal passenger distribution.
Historic streetcar or light rail volumes are shown reaching
10,000 passengers per peak hour direction in North America.
Three articulated light rail vehicles are calculated to handle up
to 17,000 passengers per peak hour direction, with 35 trains per
hour and a density of 3.25 passengers per m2.
Commuter rail in North America is shown as achieving
15,500 passengers per peak hour direction with 15 trains per
hour per track (LIRR). Comparable European capacities can
reach 28,520 passengers per peak hour direction with 30 trains
per hour. As a result, several European cities signal and operate
commuter rail in a manner equivalent to rapid transit. (The
lower volume is due to the common commuter rail policy of a
seat per passenger.)
Comment: An outstanding and comprehensive report.
44
LIN, TYH-MING and WILSON, NIGEL

H.M., Dwell Time Relationships for Light Rail
Systems, Transportation Research Record 1361,
1992: pp. 287-295
Summary: Lin and Wilson make a detailed analysis of dwell

time determinants at two stations on the subway portion of the
Massachusetts Bay Transportation Authoritys Green Line light
rail. Both linear and non-linear models are used to explain the
dwell time data with the latter being only slightly more
effective. Data for one- and two-car trains were analyzed
separately so exposing a considerable difference in contributing
factors according to train length.
The linear equations giving the best fit to the data as a whole
are reproduced below.
For one-car trains:
DT = 9.24 + 0.71 * TONS + 0.52 * TOFFS + 0.16 * LS
This gives an R2 value of 0.62.
For two-car trains:
DT = 13.93 + 0.27 * TONS + 0.36 * TOFFS + 0.0008 * SUMASLS
This gives an R2 value of 0.70.
where:
DT
TONS
TOFFS
LS
SUMASLS
=
=
=
=
=
Dwell time(s)
Total boarding passengers
Total alighting passengers
Number of departing standees
Sum of (TOFFS)*(arriving
standees) +(TONS)*(departing
standees)
The constant term for two-car trains in the equations is larger

but the lower multipliers give a lower marginal dwell time for
boarding compared with one-car trains. Note that the effect of
crowding on the cars (the last term in the equations) is much
lower for two-car trains. There is also evidence that the effect of
crowding may cause a non-linear increase in dwell time during
congested periods.
The paper closes with a brief discussion of the service
implications of variable dwell times. Uneven dwell times cause
uneven loading in a self-perpetuating cycle. Mixing different
train lengths on the same service is likely to cause uneven
loading.
Comment: While the information given by Lin and Wilson is

specific to Bostons Green Line, the basic form of their
equations and conclusions is likely applicable elsewhere. As
such, this paper is a valuable reference in discussions of dwell
times and their effects on capacity.
45
MEDVECZKY, GEORGE. Hub-Bound Travel

1991, New York Metropolitan Transportation
Council 1992
Summary: Comprehensive statistics on transit and vehicular

movements in Manhattan. Cordon counts provide peak-point
passengers on trunk lines. Additional New York cordon counts
139
for the study were acquired directly from MTA - New York City
Transit.
46
MILLER, E. J. and BUNT, P. D., Simulation

Model Of Shared Right-of-Way Streetcar
Operations, Transportation Research Record
1152 1987: pp. 31-41
Miller and Bunt introduce the reader to a computer program

designed to simulate streetcar operation on the 501 Queen line
in Toronto, Ontario. The number of inputs to the model is
exhaustive and includes a directly proportional relationship
between standee numbers and boarding passenger service times.
Comment: Much of the effort expended in the program is in

creating a routine for explaining the short-turning of cars to
assist in determining the best way to increase service regularity
and capacity on the Queen line. As traditional streetcar service
is only a small part of the capacity study, this report is of limited
value.
47
MOTZ, D., Attainable Headways using

SELTRAC, Alcatel Canada, Toronto, September
1991 (Proprietary Reportonly non-confidential
data used for the A-8 study)
Summary: Seltrac was one of the first transmission-based

moving-block signaling systems. It is now in its fifth generation
and is used in five North American locations. It is currently
being installed on the Muni Metro light rail subway to increase
throughput.
The system is based on the brick-wall stop safety criteria
and allows trains to operate at the closest possible spacing with
separation defined as the normal braking distance plus a safety
distance. Braking distance is a readily determined or calculated
figure for any system. The safety distance is less tangible, being
comprised of a calculated component adjusted by agency policy.
In certain systems this safety distance is a fixed quantity;
however, the maximum throughput is obtained by varying the
safety distance with speed and location.and where different
types of equipment are operated, by equipment type.
In theory, the safety distance is the maximum distance a train
can travel after it has failed to act on a brake command before
automatic override (or overspeed) systems implement
emergency braking. Factors in this calculation include:
system reaction time

brake actuation time
speed
train load (mass)
grade
emergency braking rate
normal braking rate
train to track adhesion
an allowance for partial failure of the braking system
The paper shows safety distances in worst case failure

situation for the London Undergrounds new Jubilee Line that
could be as long as 190 m for a fully loaded train at maximum
speed (90 km/h) on a maximum down grade (5%).
In contrast, the constant safety distance used on the Seltrac
equipped rapid transit system in Vancouver is 50 m, in part due
to the better assured emergency braking provided by magnetic
track brakes.
The paper describes the simulation of other capacity
constraints at junctions, turnback and terminal stations,
including situations with late trains, to show that a throughput of
36 trains per hour can be sustained with a train irregularity
(behind schedule) of up to 60 sec.
Comment: It is not usually appropriate to reference a

proprietary paper that is not available in the public domain.
However, this is the only known source that explains and
derives the safety distance for a moving-block-signaling system
with conventional rapid transit equipment.
As such it sets the upper limit of throughput that could be
achieved on any existing or new rapid transit whether that
system uses Seltrac or one of the other moving-block signaling
systems that have recently entered the market, including French
and British systems and the recently announced BART/Hughes
Aircraft development.
In principle, a moving-block signaling systems allows
headways to decrease from the optimum with a three- aspect
signaling system of 55 sec plus dwell to 25 to 35 sec plus dwell.
However, at such closer headways, constraints at junctions and
terminals and the issue of irregular operation become
increasingly critical. (Note that the worst case braking rate
used in the report is relative and does not assume total braking
failure but rather no electric braking and partial air brake
failureretaining 75% of normal braking ability.)
The paper does not comment on the safety distance selected
by the London Transport management or the regulatory
authority (the United Kingdom Railway Inspectorate) as a result
of this study, but it is possible that it is less than the 190 m
calculated.
Moving-block signaling systems, constraints and recovery
issues are fully discussed in the study. The data in the paper
have been used to set a range of safety distances that, in
conjunction with the maximum dwell time, establishes the
minimum headway on both moving-block signaling systems and
conventional multiple-aspect signaling systems.
48
NEW YORK CITY TRANSIT

AUTHORITY, Rapid Transit Loading
Guidelines, April 1992
Summary: This policy paper gives the loading and service

standards which have been applied, with minor modifications, to
the New York subway system since 1987. The guidelines
provide for slightly more space per passenger than those in
effect until 1986. Modifications have allowed for a relaxation in
the nonrush hour passenger loading guideline to allow for the
operation of short trains.
The loading guidelines were established from test loadings of
different car types, loading surveys of revenue service at the
140
peak load point and comparisons with the policies of other rail
transit operators. Additional concerns such as passenger
comfort, dwell time effects, uneven loading within trains, and an
allowance for slack capacity in the event of service
irregularities and fluctuations in passenger demand were also
considered. A rush hour standard of 3 sq ft per standing
passenger (3.6 passengers per m2) was generated from this work.
The policy recognizes that this condition is only to be met at the
maximum load point on a route and so is effective for only a
short time and small portion of the overall route. For
comparison, the agencys calculations of the maximum capacity
of each car type are based on 6.6 - 6.8 passengers per m2.
The graph below compares the loading standards of a number
of systems.
The application of these guidelines resulted in a 6.4% increase

in weekday train miles, a minor increase (0.3%) on Saturdays
and a 1.0% decrease on Sundays.
Comment: This useful paper gives a look at how loading

standards are developed and their effects. It confirms the
importance of considering the effects on dwell times when
creating loading standards. The need to give passengers with
access to alternative transportation a comfortable ride is also
given importance with the variable loading standards applied to
less frequent rush hour and non-rush hour services.
49
Standards for loading in the non-rush hours are substantially

more generous with a seated load at the maximum load point
being the general standard. If this would require headways of
four minutes or less, or preclude operation of short trains, a
standard of 125% of seated capacity applies. This consideration
of passenger comfort also extends to rush hour service on lines
where the headway is longer than 4 min. In these cases a sliding
scale is used to ensure lower standing densities on routes with
longer headways, as shown in the following graph.
OBRIEN, W., SCHNABLEGGER, J. and

TEPLY, S., Control of Light Rail Transit
Operations in Edmonton, Transportation
Research Board Special Report 182, 1978, pp.
115-118
Summary: This paper gives a pre-opening report on the control

of light rail and traffic on the Edmonton, Alberta Northeast light
rail line. The signal blocks on the light rail are stated to be 1-km
long which places a severe constraint on capacity. The authors
place emphasis on the need to maintain consistent service on the
outlying portion of the light rail line in order to ensure proper
utilization of the downtown tunnel. This is achieved with light
rail pre-emption of the nine grade crossings on the line. Grade
crossing signals and gates are integrated into the signal
controllers of adjacent intersections to ensure smooth traffic
flow and prevent queuing on the rail tracks.
Comment: This report supports other literature information that
signaling systems, not grade crossings, are generally the
capacity constraints on light rail systems.
50
Minimum headways for each day and service period were also
developed with the results shown in the following table:
PARKINSON, TOM E., Passenger Transport in

Canadian Urban Areas, Canadian Transport
Commission, Ottawa 1971
Summary: The Principal Investigators 1971 report quotes

maximum rapid transit volumes from sources referenced
elsewhere in this review. The ratio for the peak-within-the-peak
is discussed for both a 5-min and 20-min flow level. The report
looks briefly at the difference between theoretical and practical
maximum capacities for rail transit and the headway reductions
possible with automatic train operation.
141
Space per passenger (square meters)
Comment: This study is one of a small number that suggest

higher throughput with automatic train operation compared to
manually driven systems.
51
PUSHKAREV, BORIS S., ZUPAN,

JEFFREY M., and CUMELLA, ROBERT S.,
Urban Rail In America: An Exploration of
Criteria for Fixed-Guideway Transit, Indiana
University Press, 1982
Summary: Pushkarev et al. use a unique approach to rail transit,

discussing the number of rail transit tracks (65) that enter CBDs
in the USA and Canada; of which 38 operate in the peak hour
with the luxury of more than 0.5 m2 of space per passenger.
Only 6 tracks operate at system capacity, 5 in New York and
one in Montreal. The authors point out that in the United States
outside New York, no rail system operates at more than 33% of
nominal system capacity.
Data compilations and presentations are numerous and have
been cited and reproduced elsewhere. The relationship between
peak hour volumes, space per passenger and theoretical capacity
of lines in the United States is shown. The first two data sets are
illustrated below:
The report suggests using gross vehicle floor area as a readily
available measure of car occupancy and applies the following
quality of service standards:
ADEQUATE0.5 m2 provides comfortable capacity per

passenger space
TOLERABLE WITH DIFFICULTY0.35 m2 lower
limit in North America with some touching
TOTALLY INTOLERABLE0.2 m2 least amount of
space that is occasionally accepted
The report discusses two of the three types of occupancy
diversitypeak-within-the-peak and uneven loading between

cars of a train.
The book states that the physical capacity of a rapid transit
line is frequently misunderstood but is basically controlled by:
1. Policy determination
2. Car Width
3. Platform Length
4. Minimum operational headway
Car width can be assigned to two groups: narrow-2.5 to 2.8 m
(generally old systems IRT, PATH, SEPTA MarketFrankford, Montreal, Chicago and Boston) and wide-3.05 to
3.20 m (IND, SIRT, SEPTA Broad Street, Cleveland, Toronto
and all newer systems). Platform length ranges from 70 m
(Boston, currently being extended) to 213 m (BART). The
authors comment that minimum operational headway must be
sustainable reliably and has three major components:
type of signaling
complexity of route
dwell times
They cite the common limit of 30 trains per hour with the
typical three-aspect signaling system and state that in practice
this is lower if there are merges but can be increased with
careful and precise operation, as for example, with the NYCTs
33 trains an hour on the Flushing Line or the 38 on PATHs
World Trade Center linemade possible only by the multiple
track terminal. The highest routine frequencies in the world (on
a two-track system with on-line stations and no junctions) are
the 40 trains an hour of the Moscow Metro. However, AGT can
operate at closer headways using off-line station as shown in the
15-sec and 18-sec headways in Morgantown and Dallas-Fort
Worth.
The report has only minor content on light rail quoting Pittsburgh PCC car headways of 23.5 sec with on-sight operation
142
and SEPTAs 29 sec with block signals on the Market street
subway at a reasonable schedule speed of 20 km/h. This is
achieved by allowing train operators to pass red signals,
operating on-sight, and with multiple station berths (4).
The authors discuss performance in terms of installed power
per tonne, suggest 80 km/h as a suitable maximum speed which
should be achieved in 25 secbut takes 60 sec in a few cases
where old, under-powered equipment is still in service. They
address some confusion in defining average speeds and use the
terms:
Schedule Speed is the net average operating speed without
terminal layover time. Gross Average Operating Speed adds
terminal layover time.
Comment: Pushkarev, Zupan and Cumellas book is one of the

most comprehensive, readable and complete treatises on North
American rail transit. It uses principally new data, specifically
acquired for the book, presented with outstanding clarity and
exceptional graphics.
The section on headways is perceptive, introducing one of the
factors not mentioned elsewhere in the literaturethat capacity
is heavily dependent on policyranging from New York with
290 passengers per car (crush load) through Washington with
service specified for an average of 170 passengers per car to
BART with a policy of 90 in a larger carbut not currently
achieved. The authors clearly indicate that passenger loading
densities of the older subway systems will not be accepted on
new North American systems.
52
RADWAN, A. E., and HWANG, K. P.,

Preferential Control Warrants of Light Rail
Transit Movements, Transportation Research
Board State-of-the-Art Report 2, Light Rail
Transit: System Design for Cost-Effectiveness,
1985: pp. 234-240
Summary: Radwan and Hwang attempt to quantify the delay

caused to light rail and general traffic by the use of light rail
traffic signal pre-emption. The following version of Websters
delay model was used in their research:
where:
d = average delay per vehicle on the particular

intersection approach
c = cycle time
= proportion of the cycle that is effectively
green for the phase under consideration
(g/c)
q = flow
s = saturation flow
x = degree of saturation
The authors have endeavored to create a model that does not

discriminate against the transit mode; as most comparisons
based on intersection level of service do. As a result, their model
assesses both the delay and savings experienced by road
vehicles and the light rail trains.
Their findings showed that, for a two-phase intersection with

no left turns, the overall intersection gain due to signal
preemption is linearly proportional to light rail volume. For a
three-phase intersection with an exclusive light rail phase almost
no intersection gain was observed. In the case of a three-phase
intersection with an exclusive left-turn phase it was found that
there is an optimum main-arterial volume at which the overall
intersection gain is maximum for a given constant left-turn
volume.
Comment: While providing some interesting results, the model
used in the study has some faults which may have biased the
results. The most important of these are assuming an overly
optimistic car occupancy of 1.4 and light rail volumes of 40-50
trains per hour. This level of light rail service is far beyond that
operated on North American lines running at-grade with signal
pre-emption.
The study does not mention that at-grade light rail capacity is
limited by grade-crossings.
53
RAINVILLE, WALTER S., and

HOMBURGER, WOLFGANG S., Capacity of
Urban Transportation Modes, Journal of the
Highway Division, American Society of Civil
Engineers, 1963
Summary: The late Walter Rainville was Chief of Research for

the then American Transit Association (now APTA) and was
noted for his no-nonsense approach. In the transit section of this
paper he defines:
Effective transit capacity =
Vehicles per hour Passengers per vehicle
The paper then lists typical fully loaded capacities of rapid
transit cars 35 years ago, the number of trains per hour (20 to
32) and a range of actual capacities. The paper points out that in
theory a two-track rapid transit system could be built to handle
90,000 passengers per peak-hour direction whereas in practice
the maximum in the country, then, was the NYCTA IND 6th
and 8th Avenue expresses at 71,790.
Rainville discusses peak-hour loading diversity and shows an
average for heavy volume lines in New York and Toronto of
87.6 (peak hour/peak-within-the-peak rate), an exceptional 95.6
for the NYCTA 7th Avenue line and the lower figure of 72.9 for
less heavily loaded lines in other cities.
Comment: The now historic data in this, and other references,
provides an insight into the maximum capacity of rail transit in
an era when ridership and loading levels were higher.
54
RICE, P., Practical Urban Railway CapacityA

World Review, Proceedings of the 7th
International Symposium on Transportation and
Traffic Theory. Kyoto
Summary: Rices long paper combines two diverse areas. The

first is a survey of the headways, capacities and commercial
143
speeds of 53 urban railway systems throughout the world based
on available published data. The second is an analysis of
minimum headways that expands on the work of Lang and
Soberman (R39) and Bergmann (R13) to compensate for reduced
acceleration as a train increases speed
coasting between stations
closely spaced stations that result in a station approach
below the optimal speed for minimum headways12
the distance a train must move out of a station before the
following train receives signal clearance to enter
The survey shows three13 systems that operate 40 trains per
hour, thirteen systems that operate 30 to 36 trains per hour and
twelve that operate 24 to 27 trains per hour on a single track.
The highest quoted capacity is 72,000 passengers per peak-hour
direction per track, three systems quote capacity between 60,000
and 70,000. All other systems (49) show capacities below
50,000 passengers per peak-hour direction per track. The data
shows that the 53 rail transit systems have a mean route length
of 14.6 km and a mean overall station spacing of 1.1 km.
Rice analyzes a typical station to station run of 1.6 km (1 mi)
with modern rail transit equipment. Constant acceleration to the
point where station braking must commence produces a
theoretical run time of 89 sec. However as the speed of a train
increases acceleration tapers offultimately to zeroas the
train moves along the motor performance curve. Using a typical
performance curve results in a practical station to station time to
111 sec25% higher.
Adding the maximum realistic level of coasting increases
travel time by a further 9 sec to 120 secan 8% increase with
an estimated energy saving of 23%.14
Rice also tabulates performance and capacity data for the 53
systems. The overall mean normal service braking rate is 1.14
m/s2, the mean emergency braking rate is 1.51 m/s2 and the
mean initial acceleration rate is 1.12 m/s2. The overall mean
design maximum speed is 79.4 km/h (50 mph). The overall
mean packing density is 3.61 passengers per square meter.
The headway equations that are developed contain constraints
for conditions where the optimal approach speed cannot be
obtained due to coasting practices (or to speed control), due to
tapering of the initial acceleration, and due to any run out
distance from a stationa distance that a train must cover
before the following train receives.
Rice acknowledges the importance of dwell time in
determining the minimum practical headwayand the difficulty
in estimating the dwell time. He quotes a dwell time in sec for a
heavy departure load at
Dwell = 17.5 + 0.55(number of passengers per double door)

and for a medium departure load at
Dwell = 13 + 0.49(number of passengers per double door)
Only limited results of applying the numerous equations derived
in the report are shown. The most significant are (for a 1.6 km
station to station run):
the optimum approach speed for typical rolling stock is 32
km/h (19 mph) which produces a headway of 80.4 sec with
a nominal 30-sec dwell
headways increase at approach speeds above and below
this optimum. For example at 50 km/h (31 mph) the
headway increases to 86 sec, a 7.5% decrease in capacity;
at 20 km/h (12.5 mph) the headway increases an identical
amount to 86 sec
removing the adjustments due to the tapering of the
acceleration curve results in a linear acceleration
decreasing the headway to 80.1 sec, a 0.4% improvement.
(0.6% improvement without considering the dwell)
Comment: Rice, in attempting to accommodate performance

and station spacing nuances in train performance has added
considerable complexity to the calculations and imposed several
conditions. The results do not seem to justify the complexity.
There are few conditions of station spacing, (or speed control)
and train performance where an optimal approach speed of 32
km/h (19 mph) cannot be achieved. Using actual motor
characteristics rather than assuming linear acceleration only
changes the calculated headway by 0.4%. The calculation of the
actual impact of this improvement is valuable.
Despite the added complexity several assumptions have still
to be made, for example driver and equipment reaction time, and
the use of the very variable emergency braking rate, rather than
the service braking rate, to determine minimum separation times
is unusual. This higher braking rate and a lower estimate of
reaction time than other workers (2 sec) may explain why Rices
calculated minimum headway of 50 sec plus dwellfor a threeaspect fixed-block system with typical train lengths and
performanceis lower than the 55 sec typical of other work.
The largest deficiency, considering the elaborate analysis, is
that no allowance is made for schedule recovery to avoid any
headway interference.
55
12
This constraint would be the same is speed controls were used that limit
the optimal approach speed.
The three closest headway systems (40 trains per hour) are quoted as
Moscow, PATH and NYCT. As NYCT operates no more than 30-32
trains per hour on its heaviest trunk routes the data are suspect. It may be
that the information relates to the theoretical throughput of the signaling
system rather than actual trains operated.
14
The calculation of energy consumption is not specified and probably does
not take into account power other than traction use. i.e. hotel load power,
the bulk of which is for heating or air conditioning. On systems with
weather extremesmost East coast systemsthe hotel load can be as
high as the traction load cutting the coasting savings in half.
13
SCHUMANN, JOHN W., Status of North

American LRT Systems: 1992 Update,
3-13
Summary: This paper provides a concise overview of recent

North American light rail developments. Future plans of
systems are also outlined. Seven tables are used to gather
together many of the basic statistics for U.S. and Canadian light
rail systems operating in 1992. A brief section also discusses the
interest in low-floor light rail cars shown by many transit
agencies.
144
Comment: Schumann provides a useful but brief summary of
some of the aspects of light rail which are relevant to this
project. The information in the tables may be directly useful or
form a base to seek more current data. The introduction of lowfloor cars will have effects on capacity as a result of reduced
dwell times through faster passenger movements and better
accessibility to the mobility impaired.
56
SONE, SATORU, Squeezing Capacity out of

Commuter Lines, Developing Metros, Railway
Gazette International 1990.
Summary: Professor Satoru of the University of Tokyo gives a

broad outline of many of the factors limiting rail transit
capacity. Station dwells are introduced as the key factor in
determining capacity. The minimum practical headway on an
uncomplicated line is around 40 sec plus dwell time at the
busiest station plus the time needed for a train to move its own
length from a standing start. Even with an infinitesimally short
dwell time, the minimum headway is thus at least 50 to 60 sec.
One method of reducing overall line headway is to have some
trains by-pass lightly used stations or to use an A/B stopping
pattern where lighter stations are served by either the A or B
services with heavier locations and transfer points being served
by both.15 Dwell times at AB stations are still a major limitation
on headway. Commuter rail services with complex stopping
patterns are often able to be more flexible than rail rapid transit
and so trains can be scheduled to pass through relatively busy
outlying stations when other services are provided.
An even passenger distribution on board the trains is
important to ensure that maximum use is made of the rolling
stock. Station design can be used to create an even distribution
of passengers throughout the train. This can be achieved by
designing cross-platform transfers, and distributing platform
entrances and exits along the length of the platform and varying
their locations at different stations. Stub-ended termini are a
particular problem which can, at least, be partially improved by
adding platform access at the outlying ends of the platforms.
Additional platforms can be used to reduce dwell times by
allowing boarding on one side of a train and alighting on the
other. Throughput can also be increased with additional track by
converting side platforms to island platforms and running
alternate trains on either side of the platform. This is a much
more economical solution than adding a parallel main line.
Junctions can be improved with grade separation or by
shifting the interchange function to a major station nearby with
excess platform capacity.
The city terminus is a common limiting station on rail transit
lines. Creating run-through stations by linking terminus stations
is an excellent, albeit expensive, solution. Building a loop giving
direct access to all platform tracks is another successful way of
increasing station throughput. Allowing higher speed
approaches to stub-end stations by extending the station tracks a
short safety distance beyond the platforms is also possible in
some cases. Double-decking is another effective but expensive
15
The AB skip stop system was used extensively on the Chicago Transit
Authoritys rail lines until 1995.
station improvement. Train schedules can also be adjusted to

increase throughput. Three main categories can be defined:
All trains stop at each station.
Fast trains over-take slower ones at four-track stations.
Each train serves all stations in a zone then runs express to
the city terminus.
The first pattern works best with less than 10 stations of
similar traffic generation. The second is effective with a large
homogeneous system but does not give the higher number of
fast trains near the central hub which is desirable on a radial
system.
The last pattern (3) is ideal for branching, radial commuter
lines since it gives high capacity and fast journeys. Passengers
traveling between intermediate stations may be inconvenienced
by the need to change trains but their numbers are small.
A number of Japanese examples of capacity increases are
given. Several of these are of running trains of similar service
characteristics in succession in a practice commonly known as
platooning. In one case the first train leaves 130 sec before the
second, stops at one additional station and arrives at the
terminus 90 sec before its slightly faster counter-part. The
double-track Seibu Railway, which operates such patterns, runs
30 trains into its Tokyo terminus in the morning peak hour and
has plans to add three additional trains. This is despite the
terminus being stub-ended with only three full length tracks.
Care must be taken when increasing capacity to ensure that
additional ridership does not simply create another choke-point
at stairs or passageways.
Future capacity increases will likely require the use of offline
stations, on-board switching, and train-to-train safety control or
collision avoidance technologies.16 Off-line stations can only be
practical where the platform loop track is long enough to allow
acceleration and deceleration to take place off the main line.
Headway improvements may, however, be marginal since a
train approaching a facing points switch must be able to stop
short if the switch has failed in mid-position. On-train switching
equipment could remove this restriction with more development.
Comment: This paper presents a comprehensive overview of

the factors restricting the upper limits of rail transit capacity. It
gives useful examples of capacity increases obtained on several
Japanese rail transit servicesseveral of which have both the
highest train, and highest passenger, densities in the world
57
STRAUS, PETER., Light-Rail Transit: Less

Can Mean More, TRB Special Report No. 182,
Light-Rail Tran: Planning and Technology.
1978: pp. 44-49
Summary: Mr. Straus makes a strong argument for keeping the

light in light rail transit and resisting the temptation to build
light rail lines to rapid transit standards. A particularly interesting
16
Reviewers Note: Essentially a moving-block signaling system.
145
table showing capacities of various light rail alignment options
is reproduced here.
Another relevant point made in the article is that higher

speeds can lead to reduced capacity because of the need for
longer following distances. Mention is also made of the faster
boarding possible with high-level platforms, as found in the
Muni Metro subway17. The use of low level platforms (with
moveable steps on the cars) on the surface maintains the
flexibility and simplicity of light rail operation elsewhere on the
system.
58
SULLIVAN T. J., New York City Transit

New Technology Signals Program Status
Report, MTA New York City Transit Division
of Electrical Systems, APTA Rapid Transit
Conference, Sacramento, June 1994
Summary: NYC Transits existing train control system is an

automatic fixed block wayside signal system. Virtually all track
circuits are single rail. Much equipment dates to the original
installation and has a high failure rate and maintenance costs.
Following the 1992 14th Street derailment, a $14 million speed
protection system is being installed at 31 priority locations.
A 5-year, $1 million study of train control systems has
concluded with broad support for Communications-based
signalingalso referred to as transmission-based or movingblock signaling. The principal attribute is continuous two-way
communication and control, increased safety, increased
functionality, and lower life cycle costs.
A survey of signaling technology around the world showed
numerous advantages for moving-block signaling systems,
including increases in capacity. Other advantages include
improved schedule adherence, reduced power consumption and
the inherent ability to operate in both direction on any track with
full automatic control.
The report discusses the issue of adapting the traditional
failsafe signaling concept to the equivalent, but different, safety
standards of computer based controls. Despite concerns, and
resistance to the introduction of new technology in train control,
many rail transit operators have selected moving-block signaling
systems, including London Transport and Stockholm Transit.
An overlay track-circuit system, SACEM, with some movingblock attributes, has increased train throughput on the Paris RER
line A. The report describes the selection and successful
operation of moving-block signaling systems by eight other rail
transit operators in North America and Europe.
17
Reviewers Note. In the subway the double front door is not used due to
the large gap created by the tapered car end.
The report expresses concern with acquisition of a single

proprietary system. The market for moving-block signaling
systems has been dominated by one company for two decades.
However all the signaling suppliers contacted had or were
developing moving-block systems. NYC Transit intends to work
with the international signal industry, to seek an engineering
partner and to use a development-driven approach to test and
select moving-block signaling systems. Sullivan expresses hope
that this process will develop a standard for such systems and
that NYC Transit will have more than two suppliers to ensure
competition and the lower costs that moving-block signaling
systems have the potential to deliver.
Comment: The strong endorsement of moving-block signaling

systems by two of the worlds largest rail transit operators,
MTA New York City Transit and London Transport indicates
that this technology and its multiple advantages has become
acceptable.
59
TABER, JOHN and LUTIN, JEROME,

Investigating the Potential for Street Operation of
Light Rail Transit, Transportation Research
Board Special Report 182, 1978: pp. 161-166
Summary: While using data collected in 1973, this paper has

some interesting figures of delay for streetcars in Toronto.
Traffic signals were found to cause 50% of the delays to
streetcars while passenger service (boarding) times accounted
for 40%.18 Delays caused by traffic congestion were only 3.3%
of the total. On the St. Clair line boarding delays accounted for
only 27% of total delay. This is believed to be a benefit of the
extensive use of island stops on this route.
60
TAYLOR, P. C., LEE, L. K. and TIGHE, W.

A., Operational Enhancements: Making the
Most of Light Rail, Transportation Research
Board Special Report 221, 1989: pp. 578-592
Summary: This mis-titled work summarizes the efforts to

minimize the effects of the Los Angeles Blue Line light rail on
roadway capacity. This is achieved by varying the priority given
to the light rail trains according to road traffic volumes. During
peak traffic periods, the light rail is accorded a lower signaling
priority to prevent disruption of motor traffic. At off-peak hours,
the light rail can be allowed greater priority with minimal
impact on motor traffic.
Comment: At the 6-min headways under consideration, the
light rail is seen as limiting road capacity and not the reverse.
Reducing priority for light rail at peak hourswhen it is most
neededis negative. It reflects badly on the traffic engineering
process whereby the number of vehicles, rather than the number
of people, moved is prioritized.
18
Reviewers Note. The latter figure has no doubt dropped since the
adoption of an exact fare policy.
146
61
TIGHE, W. A. and PATTERSON, L. A.,

Integrating LRT into Flexible Traffic Control
Systems, Transportation Research Board Stateof-the-Art Report 2, Light Rail Transit: System
Design for Cost-Effectiveness, 1985: pp. 213220
Summary: Tighe and Patterson offer a general discussion of

integrating light rail into vehicular traffic signaling. Their ideas
are then applied to the Woodward corridor in Detroit, and the
Guadalupe corridor in Santa Clara County, Different solutions
are offered in each case to reflect the specific alignment
characteristics.
For the Woodward Corridor, the light rail is proposed to run
in the exceptionally wide median of Woodward Avenue. This
allows the use of two-phase traffic signals (i.e. no left turns
permitted) at all intersections since the median can be used to
create U-turn bays between intersections. Cars wishing to turn
left are able to use a combination of right-turns and U-turns to
achieve the same result. Intersection spacing is such that the
light rail can easily run with the progressive signaling at crossstreets while pre-empting the U-turns when required.
In the Guadalupe corridor example, the medians of North
First Street and Tasman Drive are of a more conventional width
making the U-turn arrangement impractical. Instead, multiple
phase traffic signal controllers with a total of up to 16 phases
(some of which can run concurrently) will be used to
accommodate heavy volumes of turning traffic. The degree of
light rail pre-emption will be variable so as not to unduly hinder
automobile flows at peak times. During off-peak periods a
greater degree of pre-emption will be permitted.
Comment: The omission of any mention of a reduction of light

rail capacity due to less than full signal pre-emption in this
paper indicates that, at the headways under consideration (4 - 6
minutes), pre-emption is not necessary for providing sufficient
light rail capacity.
62
TOPP, R. M., Improving Light Rail Transit

Performance in Street Operations: Toronto Case
Study, Transportation Research Board State-ofthe-Art Report 2, Light Rail Transit: System
Design for Cost-Effectiveness, 1985
Summary: The Queen streetcar line in Toronto, ON

experiences service irregularities due to extremely heavy use
(75,000 passengers per day) and a lack of transit priority. This
paper summarizes some of the operational problems of the route
and details the results of two studies aiming to solve them. Key
to improving the service on the route is a reduction in the
number of unscheduled short-turns required to maintain
headways and capacity on the central portion of the line.
One approach was solely to look at operational adjustments
which would improve service reliability. Passenger service time
was found to take 12 - 18 percent of total travel time. Signal and
queue delays accounted for 13 - 15 percent of total travel time.
Suggestions included extending running times, increasing the
service gap required to initiate a short-turn, adding scheduled

short-turns, and using larger, articulated vehicles.
The second approach was to study ways of improving service
through the use of transit priority measures such as pre-emptive
signaling. In some cases this could simply mean re-timing the
traffic signals to improve general traffic flow.
Comment: This is one of few papers to address the operational

problems of a traditional streetcar service in mixed traffic with
no priority measures.19 The speed of the service and number of
cars required is heavily affected by the current conditions.
63
TORONTO TRANSIT COMMISSION,

Yonge-University-Spadina Improved Headway
Study, Final Report Toronto Transit
Commission, December 1988
Summary: This staff report, based on studies by consultants

Trans mode and Gibbs and Hill, examines a range of options to
increase capacity on the TTCs Yonge-University-Spadina
(YUS) subway.
In 1988 the Yonge subway south of Bloor was close to its
rated capacity of 34,000 passengers per peak hour direction.20
This capacity is based on maximum length six-car trains, 140-m
(450-ft) long, operating at headways of 130 sec (28 trains per
hour).
The Yonge subway, opened in 1953, was the first new
postwar subway in North America. It uses a conventional three
aspect color light signaling system based on track circuits
designed for 120-sec headways (30 trains per hour), on the basis
of station dwells of no more than 30 sec. Actual dwells at the
major Bloor-Yonge interchange station of 45 sec prevent
undisturbed operation of more than the 28 trains per hour.
Analysis of downtown developments had indicated a future
demand, on this critical section of the subway, increasing by
33% to 45,000 passengers per peak hour by the year 2011.
A detailed analysis of the signaling system confirmed that the
Bloor station dwell was the only bottleneck preventing 120-sec
headways. However if the signaling system was upgraded for
closer headways other bottlenecks would appear, particularly the
Finch turnback used by all trains. (At the other end of the line a
short-turn divided the turnbacks between two stations, so
avoiding any restrictions.)
The study examined three signaling improvements that would
progressively reduce headway. The first option made minor
signal adjustments in the vicinity of Bloor station to permit 122sec headways. The second set of improvements to signaling
reduced the headway to 112 sec but required a major
reconstruction of the Bloor station to ensure dwell times within
30 sec, and changes to the terminal at Finch.21
The third improvement was to replace the signaling system
19
The Toronto Transit Commission has recently managed to obtain priority

for streetcars on sections of its network.
20
Ridership has decreased in the last few years.
21
A fourth option that would permit a 105-sec headway required extensive
modifications to the existing signaling system and was discarded as
impractical.
147
with automatic train operation that would permit 90-sec
headwaysagain with a major reconstruction of the Bloor
station and both terminals.
Each option required additional vehicles and the yard

expansions to accommodate them. The results are summarized
above with cost estimates in millions of 1988 Canadian dollars.
The study showed that the most cost effective way of
reducing dwells at Bloor station was to widen the station and
add a Centre platform so that trains could simultaneously open
doors on both sides.
Terminal changes involved extending the tail tracks and
adding a second pocket track so that peak period trains could
reverse behind rather than in front of the station. The improved
headways could not be accommodated by using the scissors
crossover ahead of the station due to the wide track separation
dictated by the center platform and the resulting high traverse
time.
PROPOSED TERMINAL CHANGES (Not to Scale)
Original
Crossovers
Original
Pocket
New pocket
& tail tracks
The study did not evaluate the considerable operating cost

repercussions. All options required additional crews to permit a
set-back operation at the terminals while the first two options
imposed speed controls that reduced the average system speed,
increasing vehicle and crew requirements. Option Threes
automatic train operation offered the significant potential saving
of reducing train crews from two to one.
Implementation of Options Two and Three was lengthy and
difficult as changes had to be made while the subway was
operating, work being restricted to limited hours, even with
proposed early closing each night.
The study also reviewed alternate methods of increasing
capacity. Widening vehicle doors was suggested as a way to
reduce dwells. An increase of 22 cm (9 in.) to a total width of
1.37 m (4.5 ft) was proposed. This was not practical on existing
cars but may be implemented on future car orders.
Adding a short (50 ft) car to each train would be possible
within the existing platform length of 152 m (500 ft). This would
increase capacity by 11% while concurrently reducing dwell
time by an estimated 12%. The costs was estimated at $47 million.
Comment: The TTCs capacity problem stems from a failure to

operate the Bloor-Danforth subway as originally planned. A
wye junction at Bay/St. George was designed so that each
alternate Bloor-Danforth train ran downtown via the University
subwayavoiding the need for passengers to physically transfer
to downtown trains. This operation was abandoned after a six
month trial in 1966 as uneven train arrivals made the merge
difficult. The uneven arrivals were primarily due to the lack of
any intermediate timing points on the long cross-town BloorDanforth subway.
Twenty years after the subway opened, intermediate timing
points (dispatch signals) were added. By this time the University
subway had been extended along Spadina and the wye operation
was no longer feasible. 30% of Yonge subways peak-point
passengers and 48% of the University subways peak-point
passengers transfer from the Bloor-Danforth subway.
The study offers valuable information on capacity limitation
and upgrade alternatives. The possibility of operating 7 car
trains of existing cars does not seem to have been considered.
Such a consist would extend beyond the station platform but all
doors would be (just) within the platformautomatic train
operation would be desirable or necessary to achieve the
required berthing accuracy.22 There is no supporting evidence
that widening doors would reduce dwells. Information
elsewhere suggests that the 1.15 m (3.75 feet) wide door, while
narrower than normal for heavy rail vehicles, supports two
streams of passengers and that little gain would be achieved
until the doorway is sufficiently wide for three streams.
The addition of automatic train operation and rebuilding
Bloor station appear to be the only way to meet future passenger
demand. This would be easier, cheaper and faster using a
transmission based signaling system (moving-block), avoiding
the difficult, potentially service disruptive, changes to the
existing signaling equipment. Transmission based signaling
systems have been selected by MTA-NYCT and London
Transport as the most practical way to upgrade or replace
existing conventional signaling systems. This omission from the
study is all the more surprising considering that the TTC already
operates a transmission based signaling system on the
Scarborough line,an extension to the Bloor-Danforth subway.
64
TORONTO TRANSIT COMMISSION,

Yonge-University-Spadina Improved Headway
Study, Signaling Report Toronto Transit
Commission, December 1988
Summary: This staff report, based on studies by consultants

Trans mode and Gibbs and Hill, expands on the signaling system
options required to increase the capacity of the TTCs YongeUniversity-Spadina subway described in Yonge-UniversitySpadina Improved Headway Study, Final Report (above).
65
TRANSPORTATION RESEARCH BOARD,

Collection and Application of Ridership Data on
Rapid Transit Systems, Synthesis of Transit
Practice, Washington DC, 1991
Summary: A comprehensive account of rapid transit data

collection practices. The report comments on the generally low
22
The transmission based automatic train control on the TTCs Scarborough
line achieves stopping accuracy 8 cm (3 inches).
148
technology approach that is mainly devoid of any field survey
design or sampling techniques. Toronto is an exception using
optical readers to enter field data into the computer. Several
systems are starting to use electronic registers in the field.
Indications of accuracy are not quantified but the report infers
that most operators achieve the FTA Section 15 requirements in
passenger counts of accuracy within 10% at the 95% confidence
level. Toronto and Atlanta claim accuracy to within 5%. NYCT
states its checkers cannot monitor heavily loaded trains and at a
certain (unspecified) level of crowding just mark such cars as
crush loaded. NYCT also estimates that its exit counts are light
by 15%.
On-board counts vary widely with the NYCTs Rapid Ridecheck being among the most comprehensive, measuring: actual
arrival time; alighting passengers; boarding passengers;
passenger load leaving; actual departure time and scheduled
departure time.
cles is 11,000 to 13,000 passengers per peak hour direction. (30

to 35 trains per hour) at level of service D. This range
increases to 17,500 to 20,000 passengers per peak hour direction
at passenger service level E with 0.3 m2/passenger.
Comment: The simple set of capacity information and
calculations derived in this study are expected to be a suitable
replacement for the Highway Capacity Manuals rail transit
capacity section.
68
US DEPARTMENT OF
TRANSPORTATION FEDERAL TRANSIT
ADMINISTRATION Characteristics of Urban
Transportation Systems, Revised Edition, 1992
Summary: Contains many tabulations of rail transit information

including a compiled range of rapid transit passenger space
occupancies.
Comment: Provides a useful indication of the data collection
process and probably accuracy level. NYCT offers possibility
for a detailed dwell time analysis from the large quantity of
Rapid Ridecheck data but actual NYCT peak counts and any
loading diversity within a train is tainted by the lack of actual
checker counts on crush loaded cars.
66
seated passengers
standing passengers
crush loading
0.28 - 0.46 m2
0.22 - 0.26 m2
0.17 m2
A list of AGT car capacities has been used in the AGT data
table. Examples of dwell times are higher than used by many
other examples in this literature survey.

Gray, Benita, editor. Urban Public
Transportation Glossary, 1989
Summary: A comprehensive glossary used with the APTA

glossary and definitions from several summarized reports, to
compile the rail transit capacity specific glossary in this report.
67

Highway Capacity Manual, Special Report 209,
1989
Summary: This much referenced report devotes a modest space

to rail transit capacity. It tabulates observed peak hour capacities
in the United States and Canada, suggesting that peak 15- to 20min volumes are about 15% higher. Typical maximum train
throughput is suggested at 30 with reference to higher levels
PATHs 38 trains per track per hour and the CTAs 78 (prior to
the use of a cab control signaling system on the elevated loop.)
The formula for rapid transit capacity is the same as shown
above in Levinson. Suggested loading levels for capacity
calculations are, level D, an average of 5 sq ft per passenger
(0.46m2).23 The resulting suggested maximum capacity for twotrack rapid transit lines is 18,000 to 30,000 passengers per peakhour direction.
The formula for light rail capacity is also shown above in
Levinson(R42). The resulting suggested maximum capacity for
two-track light rail lines with three-car articulated light rail vehi-
Comment: The values for seated floor occupancy and for

commuter rail dwell appear low.
69
Summary: A comprehensive tabulation of transportation

statistics with limited general information on urban transit.
70
Reviewers Note. Much the same as the 0.5 m2 of Pushkarev (reviewed

above)particularly when Pushkarevs use of gross vehicle floor area is
taken into account.
VANTUONO WILLIAM C. Signaling and

Train Control, High-Tech for High Capacity.
Transit Connections, September 1994
Summary: This magazine article discusses advanced train

control systems relative to a need to increase capacity.
Communication based24 signaling systems in use and under
development are summarized.
24
23
US DEPARTMENT OF
TRANSPORTATION, National Transportation
Statistics, Annual Report, Sept., 1993
AUTHORS TERMINOLOGY. Communication based signaling systems

are also called transmission based or moving-block systems. As not all
communication or transmission based signaling systems are moving block
the A-8 report will use moving-block signaling system to avoid confusion.
149
Pointing out that moving-block signaling systems have been
in use in Europe and Vancouver, Canada for several years the
author discusses the selection of the Seltrac system for San
Franciscos MUNI resignaling and an unspecified similar
system for the modernization on New Yorks subway lines. It
comments that other US rail systems are expected to follow
New Yorks lead, quoting NYCT after an intensive study and
international peer review, communications based technology is
the best, most cost-effective system for our purposes.
The article describes moving-block signaling systems from
nine suppliers25:
General Railway SignalATLAS

Union Switch & SignalMicroBlok TM
AEG Transportation SystemsFlexiblok TM
Alcatel CanadaSELTRAC TM
Harmon IndustriesUltraBlock TM
Siemens Transportation Systems
Matra TransportMETEOR TM, SACEM TM,
MAGGALY TM
CMW (Odebretch Group Brazil)
Morrison Knudsen (with Hughes and BART)
TM
Comment: One of the most comprehensive and current

descriptions of moving-block signaling system. The only known
system omitted is that of Westinghouse Brake and Signal (UK)
currently being installed on a portion of London Transports
Underground.
The article is somewhat optimistic, claiming possible
headway reductions to 60 sec. It also steers around the
considerable industry controversy related to moving-block
systems in which the hardware based fail safe features of
conventional signaling are replaced by a software equivalency.
Until NYCT announced the selection of a transmission based
system, several of the above manufacturers were vociferously
opposed to the software based train control systems (despite
some of them offering software controlled interlockings).
71
where:
= numbers of vehicle per Train Unit

n
Cv = Passenger spaces per vehicle
hs min = minimum headway (station)
The vehicle capacity (passenger spaces per vehicle) is shown as:

Reviewers Note. Several of these moving-block signaling systems are
under development and it will be some years before they are proven in
service.
= vehicle floor area loss factor for walls

Ag = gross vehicle floor area
Al = vehicle floor area used for cabs, stairwells
and equipment
m = number of seats
= floor area per seat
= floor area per standing passenger
Suggested values for space per seat are 0.30 to 0.55 m2, for
space per standee 0.15 to 0.25 m2. Operating capacity, Co, is
defined as:
Co = Cone hour < C
The scheduled line capacity utilization factor, , is defined as:
The capacity utilization coefficient26 is defined as:
where:
P = number of passengers transported past a

point in one hour
Professor Vuchic develops the concept of Linear Vehicle

Capacity
= Cv/l'
where:
l = length of vehicle
Suggested values of are 7.08.5 for light rail vehicles and

8.010.0 for heavy rail cars. The maximum way capacity Cw is
developed as:
VUCHIC, VUKAN R., Urban Public

Transportation Systems and Technology,
Prentice-Hall Inc., 1981
Summary: Professor Vuchics comprehensive text devotes 70

pages to capacity, introducing some unique definitions and
taking an approach that defines two capacities: Cwway
capacity and Csstation capacity. Maximum offered line
capacity C is defined as the minimum of way or station capacity.
25
where:
where:
so
tr
K
v
b
= safety separation
= reaction time27
= safety factor
= train speed
= braking rate
Ten different safety regimes from Friedrich Lehner (1950) are

introduced.28 Using the above equation for way capacity and the
brick-wall scenario, Vuchic calculates the way capacity for
BART at 185 trains per hour and 350 trains per hour for 2 car
articulated light rail vehicles.
The book then develops a station capacity equation incorporating dwell times. Station capacity is shown to be 1/4 to 1/7th of
26
Reviewers Note The capacity utilization coefficient is more commonly

called the load factor.)
Reviewers Note: Operator and equipment reaction time for manual trains,
equipment (brake) time only for automatic train operation.
28
Reviewers Note: Only one is accepted in North American rapid transit,
namely the brick wall scenario with a service braking rate
27
150
way capacity. The theoretical throughput and optimum speed is
shown as:
Suggested practical capacities are 15,000 to 20,000

passengers per peak-hour direction for light rail and 55,000 to
65,000 passengers per peak hour direction for rapid transit.
The Yamanote Line in Tokyo is referenced as possible the
highest capacity line with 165,000 passengers per peak-hour
direction on four tracks. Actual examples of minimum
headways and capacities are tabulated. Streetcars are shown to
have operated historically at headways down to 23 sec on street
and 30 sec on segregated tracks. Signaled light rail has
demonstrated headways down to 27 sec.29
Rail rapid transit headways as low as 70 sec are shown in the
Soviet Union with 90 sec the closest operated elsewhere.
Vuchics mathematical analysis of capacity concludes with
extensive comments on the relation ship between theoretical and
practical capacities of transit modes:
Capacity is not a single fixed number but is closely related
to system performance and level of service.
Operation at capacity tends to strain the system to its
maximum abilities and does not represent a desirable
condition.
There is a significant difference between design capacity
and the number of persons transported during one hour.
Theoretical capacities are often quite different from
practical capacities.
Way capacity is a different concept from station capacity,
station capacity always governs line capacity.
There can be friction between boarding and alighting
passengers that impacts dwell time calculations.
Comment: Professor Vuchic develops by far the most

comprehensive mathematical treatise of rail transit operation
and capacity. As with other mathematical treatments, the
difference between theory and practice is difficult to reconcile
or quantify.
The concept of passenger capacity per linear unit of a train
has merit and is developed in the study.
Except possibly for automated guideway transit with off-line
stations, the use of way capacity has little relevance and
produces dubious results. It is difficult to see the value of a line
without stations and questionable whether such a line could
throughput the calculated 185 BART trains per hour or 350 light
rail trains per hour.
The book acknowledges this and states that station stops are
the capacity constraint on rail transit systems. In calculating the
clearance times for these station stops dwell times are poorly
dealt with and several factors are omittedparticularly issues
29
Reviewers Note: With multiple berth stations and without automatic train
stops to allow operators to proceed through red signals on a line of sight
basis.
of a trains initial acceleration diminishing rapidly, speed limits

and/or grades entering and leaving stations, braking transition
times (jerk limitation) and worst case braking conditions due to
either equipment failures or adhesion limitations.
Other sources (Alle(R02) on dwell times, Auer(R09) on minimum
headways and Motz(R47) on safety distances) provide methods to
calculate minimum headways that include better treatments of
dwell time and incorporate factors not considered in this book.
72
WALSHAW, J. R., LRT On-Street Operations:

The Calgary Experience, Transportation
Research Board State-of-the-Art Report 2, Light
Rail Transit: System Design for CostEffectiveness, 1985. pp. 221-226
Summary: This paper describes the operation of the 7th

Avenue transit mall in Calgary, AB. In the peak hour, 176 trains
and buses use the mall. Light rail headways were expected to be
reduced from 5 min to 2.5 min with the opening of the Northeast
Line in 1985. Light rail operation benefits from a progressive
signaling system that keeps signal delays down to 7-8 % of mall
travel time.
Comment: The paper provides useful information with respect

to buses and light rail sharing a right-of-way.
73
WEISS, DAVID M., and FIALKOFF, DAVID

R., Analytic Approach to Railway Signal Block
Design, Transportation Engineering Journal,
February 1974
Summary: This paper describes computer based methods to

design a fixed-block signaling system for high capacity rail
lines. Five programs were developed.
1. A passenger station dwell time program using information
on passenger traffic, number and size of doors, distribution
of passengers on the platform and train and the ratio of
boardings to alightings.
2. 1A train performance simulator that produces train speed,
time and location based on a lines grades and curves and
on the trains traction performance.
3. A braking distance program that utilizes braking rates, jerk
limitation and reaction times. This program calculates the
worst case stopping distance plus safety marginstermed
the safe braking distancea function of speeds, curves,
grades, braking rate, jerk rates, available adhesion and the
reaction times of car-borne and wayside train control
equipment. The exactness of the safe braking distance
calculation contributes to higher capacity and eliminates
the need for additional margins to be addedtermed
ignorance factors.
4. A minimum headway program utilizing the outputs from
the above three programs.
5. 1A graphical plotting program.
151
A composite schematic of the final output is shown below:
Braking distances cause large headways at high speeds

where between station maximum speeds may become the
limiting factor in minimum headway. However the time to travel
a train lengthcritical to the close-in timewill blow up
hyperbolically at low speeds. In between lies a speed or profile
that will optimize headway.
The paper tantalizingly offers a method to equate passenger
volume with dwell times but offers no details.
74
NOTE Recovery margin is operationally desirable but not essential.
The paper describes the selection of cab signal speed

commands, locating signal block boundaries and the
development of the optimum train design profile.
An appendix calculates the value of train speed which
minimizes headway as:
where
MT
V
B
N
L
= minimum headway in sec

= constant train velocity
= constant braking rate
= brake application reaction time
= train length
Differentiating this equation relative to V shows that for

minimum headway:
Substituting this optimum value of velocity back into the

minimum headway equation results, relative to two trains
traveling at a constant speed, in an expression for minimum
headway that is independent of velocity:
The authors warn that trains do not usually maintain constant

velocity and that the factors influencing braking distance are
continually changing, making the calculation of minimum
headways more complex. In most situations it is the station stop
times that determine the minimum headwaynot the speed
between stations.
Comment: This paper provides a useful and concise outline of

signaling system optimization. In most cases the minimum
headway is the station stop time, comprising the sum of the
close-in time, dwell time and recovery margin. The paper shows
that the braking distances that establish the close-in time can be
approximated by quadratic functions of train velocity.
WILKINS, JOHN D., and BOSCIA, J. F.,

Considerations For Effective Light Rail Street
Operation, Transportation Research Board
State-of-the-Art Report 2, Light Rail Transit:
System Design for Cost-Effectiveness, 1985: pp.
195-202
Summary: Wilkins and Boscia outline their views on designing

light rail for on-street operation. Some portions are relevant to
capacity issues.
Throughput is lower but this can be partially offset by train
operation.
Dwell times are longer with low platforms unless selfservice fare collection and safety islands are used.
Average speed is reduced because of pedestrian and
vehicle interference.
Comment: The paper provides indications of capacity

limitations with on-street light rail operation.
75
WILSON, NIGEL H. M., MACCHI,

RICHARD A., FELLOWS, ROBERT E. and
DECKOFF, ANTHONY A., Improving Service
on the MBTA Green Line Through Better
Operations Control, Transportation Research
Record 1361, 1992: pp. 296-304
Summary: Wilson et al. examine the operational control system

of the MBTA Green Line light-rail system in Boston in this
paper. Particular attention is paid to methods of maintaining
even headways, such as short-turning, express running and
deadheading, in order to maintain as even a service as possible.
The existing operating practice relies on the intuition of
inspectors stationed in the subway stations to decide the action
to be taken to maintain service. Interestingly, all the correctional
methods described are applied in the downtown portion of the
line, not the outlying branches. The actions of the inspectors
were examined by the authors and found to be generally
beneficial to reducing passenger travel time. The researchers
also created a correctional decision making routine for each line
which is based on the preceding and following headways for
each train. A different routine is required for each line given the
discrete riding patterns on the individual branches. This
framework would take much of the guess work out of
dispatching and further reduce the number of deleterious
dispatching decisions.
152
Determining the following headway is not possible with the
current manual train supervision methods but this problem will
be more readily corrected with utilization of the recently
installed Automatic Vehicle Identification (AVI) system for
field dispatching. While the AVI system does not automatically
calculate preceding and following headways, the authors argue
that modification of to the AVI system could enable automatic
headway calculation and so make correctional actions still more
effective.
Comment: This paper examines the operational control of the
busiest light rail system in the United States. The discussions of
maintaining even headways are highly relevant to the provision
of capacity on any rail transit line. As the authors point out, their
work is especially applicable to the light rail systems in
Philadelphia and San Francisco which, like Boston, have
multiple surface lines funneling into a downtown tunnel trunk
line.
76
YOUNG J.A., Passenger Comfort in Urban

Transit Vehicles, Ontario Ministry of
Transportation and Communications, 1976
Summary: Contains useful tables:
transit seat dimensions for several rail systems

detailed car dimensions
chart of ratio of door openings to car length
transit vehicle entry step heights
transit vehicle door flow rates
Useful recommendations on optimal door widths, aisle widths

and interior designs. Data on car lighting, noise and vibration
levels are not relevant to the TCRP A-8 study.
Comment: The seat data should allow the development of a
North American rapid transit average which could avoid the
complexity of determining floor space used by seats on a system
by system basis. Equating the total door width along the side of
a car as a percentage of the cars length and relating this
percentage to boarding and alighting flows has merit.
A1.3 REVIEW SUMMARY

The literature review of North American Rail Transit Capacities
and Capacity Analysis Methodologies has produced a wealth of
information, data and methodologies.
A1.3.1 BASICS AND CAUTIONS

Several authors caution that there is no absolute determination
of rail transit capacity, that capacity is subject to many variants
which can change from mode to mode and system to system.
There are several cautions concerning the accuracy of ridership
information, particularly with respect to individual car counts
under crowded conditions.
There is general agreement that the definition of rail transit

capacity is the number of passengers that can be carried past a
single point, in a single direction, in a single hour. Many authors
discuss the relationship between peak hour and peak-within-thepeak capacity, others concentrate on the latter short term
capacity. This results in an overstatement of a full hours
capacity.
One author argues that a case can be made that the peakwithin-the-peak is the actual maximum capacity of the system
and, if there were an adequate supply of passengers, that rate
could be sustained for a full hour.30 Several authors discuss this
issue of supply versus demand, both with respect to capacity and
in two cases with respect to the quality of service. Here the
argument is that if service is provided that exceeds demand, the
level of crowding will decrease and more passengers may be
attracted.
A valuable input on this topic is the suggestion that new rail
transit systems must move away from providing service based
on the loading levels of older systems. If their goal is to attract
riders then the quality of service must be improved. Three
papers peripherally mention that this was the original goal of
BARTthat all passengers have a seatsubsequently lost to
the realities of operating economics.
A1.3.2. INFLUENCING FACTORS

The literature clearly indicates the two major factors that,
multiplied together, determine rail transit capacity. The first is
line capacity, the throughput of trains per hour, the second is
train capacity.
Line capacity is a function of two major factors, each of
approximately equal weight. One is the time between a train
starting from a station and the next train berthing at that station.
This is a function of the train control system, both the type of
system and the design of that type. For example the
conventional three aspect signaling system can be designed for a
minimum station separation of 55 sec, but is often, particularly
for light rail, designed for longer separation times which require
fewer blocks and lower capital and maintenance costs.
The literature introduces several minor factors that influence
line capacity. These include speed limits at station approaches
and exits and the rapid fall off of the acceleration rate as a train
gains speed. Three authors state that automatic train operation
can increase throughput within a range of 5 to 15%. None
provide data to support this proposition. Many of the discussions
on line capacity fail to consider constraints due to junctions or
turnbacks. Where such limitations are discussed it is invariably
without the detailed analysis that has been applied to the
headway limitation at stations. Several papers indicated that the
maximum or average speed of trains between stations is a factor
in capacity. This is only true when a finite quantity of rolling
stock is taken into account.
The second major factor pertaining to line capacity is the
station dwell time. This is extensively dealt with in 26 papers,
listed in the framework chapter, section 3.6.5. Suggestions range
30
Reviewers Note: This argument glosses over the practice of several

operators who insert one or more trains to handle the peak-within-the-peak
demand, then remove them at the end of their run as the system cannot
reliably sustain that number of trains over a longer period.
153
from using average or typical dwells in the 20- to 30-sec range,
to a detailed methodology to calculate an upper control limit
based on measured dwells over a peak hour at the busiest
station.
The relationship between passenger movements and dwell
times is a component of most dwell discussions. Those that
included analysis concluded, without exception, that linear
regression provided the most suitable fit for both rapid transit
and light rail with high and with low loading. Three references
improved the data fit by including the number of passengers
onboard a car as a variable. One study used multiple regression
and showed a small improvement in data fit with the variable of
on-board passengers to the power of 2.0 or 2.5. One paper
evaluated a variable to account for passenger actuated doors on
the San Diego Trolley.
The literature contained many references to train or car
capacity, methods of calculation based on net floor area, gross
floor area and length of train, and examples of loading levels
throughout North America. One paper contains useful
information on capacity variations with different door and
interior arrangements.
Although the literature had an abundance of information on
these three major factors, train control throughput, dwell times
and train or car capacity with one exception it was mainly silent
on the fourth major capacity issuepolicy. While this is a
difficult area to analyze it can have a massive impact on
capacity. Suggestions that new rail lines should be based on all
passengers with a seat can reduce capacity, as normally defined,
by a factor of three or four. In effect such policy issues are the
most important of the four main rail transit capacity factors.
of capacity calculations the modes were better grouped by the

types of operation. These groups are defined and presented in
the framework chapter, section 3.3.
A1.3.4 LIGHT RAIL SPECIFICS

No fewer than 37 of the reviewed papers dealt specifically
with light rail. In particular the issue of traffic engineering for
shared right-of-way and grade crossings was extensively
covered. Capacity issues on lines without full grade separation
broke the literature into two groups. One group indicated that
capacity was rarely an issue as the demand for service under
such situations was far below the train headways that could be
provided.
Other work suggested that capacity on lines with grade
crossings was effectively limited to one train per traffic signal
cycle.31 Another suggested that where train length approached
the street block length, one train every second traffic signal
cycle was more realistic.
A1.3.5 STATION CONSTRAINTS

Beyond two unsuccessful attempts to equate dwell times with
the level of crowding on station platforms there was little
discussion in the literature on the impact of station constraints
on capacity. This is not unreasonable as most of the station
constraints impact the number of people using that station, that
is the demand, not the capacity of the rail transit line.
A1.3.3 GROUPING
A1.3.6 CONCLUSIONS
The literature generally dealt clearly and specifically with the

different modes, rapid transit, light rail, commuter rail and
automated guideway transit. It became clear that for the purpose
The literature has produced a wealth of information,

methodologies and data so aiding this project to maximize its
use of existing information and data.
31
Reviewers Note: Papers that dealt with traditional streetcar operation

suggested much higher throughputsreaching as high as two or
occasionally three single cars per cycleor over 100 cars per hour.
154
A2. APPENDIX TWO

Rail Transit Survey
This appendix is the result of Task 3 of the study.
Survey rail transit services in North America to
determine system characteristics and factors that
influence and constrain capacity.
The survey was carried out in June and July 1994. Data have
been updated using 1993 FTA Section 15 reporting contained in
the 1993 National Transit Database, published in 1994.
A2.1 INTRODUCTION
A2.1.1 PURPOSE OF SURVEY
A telephone survey of North American rail transit systems was
conducted to determine the availability of existing ridership
data, capacity and capacity constraints from each system. The
opportunity was also taken to ask other relevant questions
regarding line and station constraints, dwell times, signaling
systems, and other issues of relevance to the A-8 study. Table A
2.1 through Table A 2.4 show the systems surveyed by mode.
The Vancouver SkyTrain and Toronto Scarborough RT lines are
included in the rail rapid transit category as they are not typical
of automated guideway transit in ridership and route
characteristics.
A2.1.2 SURVEY METHODOLOGY

Letters were sent to the CEO or General Manager of each
agency in mid April, 1994 requesting the designation of a
contact person. 22 responses were received from 43 letters.
Contact persons from non-responding agencies have been
obtained by telephone query. Multiple mode systems often
required separate contacts for each mode or division.
As a result of the principal investigators work on a light rail
system in Mexico City, English speaking contacts were obtained
for four of the five Mexican rail systems and complete data
acquired for two systems. Limited data was obtained for a third
system. The remaining two systems were dropped after three
telephone calls failed to get responses. Basic information and
annual ridership was obtained from other sources to enable
complete survey listings.
A questionnaire was developed from a relational database
derived from APTA data and the initial analytic framework,
showing each system and mode. System and vehicle data,
including car dimensions has been incorporated in this database.
The questionnaire was tested with a series of initial telephone
interviews. It was not satisfactory and numerous changes
resulted. A sample of the final questionnaire, completed for the
Washington Metropolitan Area Transit Authority, is attached.
The survey itself was conducted in June and July 1994 with
each system answering the same 24 questions. The same one
page survey was used for all modes to ensure consistency in the
study. For multi-modal systems, a separate questionnaire was
completed for each mode. A few mode-specific questions were
included to deal with unique aspects of particular modes, such
as passenger actuation of light rail transit doors. Emphasis was
also placed on determining the accessibility of each system to
the mobility impaired and the resulting effects on service quality
and capacity. When possible, ridership reports, car details and
timetables were obtained. Information gathered from this survey
was used to update and expand the database in preparation for
the remainder of the study.
A variable in the survey was the level of interest and
knowledge shown by the contacts. Many were enthusiastic to
talk about their system and volunteered additional useful
information. Other staff members were more restrained and only
dealt with questions asked directly. In numerous cases the
contact requested that the questionnaire be faxed to allow
additional staff people to assist in answering the questions.
Others wished to answer the questionnaire in written form to
ensure accuracy. Project staff met these requests with some
reservations as voice communication can convey nuances and
useful asides which are not readily given in short written
answers.
155
Sample Telephone Survey
1. Do you have individual route peak point ridership

data by hour ! by trains ! by short time periods "
how many ___ mins?
Do you have riding counts (ride-checks) ! 24/month
On systems with 4-car or longer trains. Do you have
individual car counts for peak hours at the peak
points on one or two representative days?
!____________________________________
2. Do you issue ridership statistics or a summary? !
Can you send us this as a starting point? ! Offer
address, fax number.
3. No of cars in trains? 2-4-6 ____________________
4. Are there any stations on the system which
regularly experience pass-ups? " Which route
and station(s) __________________________
sometimes at Union after commuter train arrives
_________________________________________
5. Do you serve stadiums? ! have any event
ridership? " notice higher densities? ! _______
6. Do you have any station constraints that reduce
ridership?
Full parking lots " __________ Ticketing line ups "
Long walks " ___________Congested platforms "
Other congestion " _____ Safety/security issues "
No transfers " ________________ Poor access "
Transfer cost " ______________ Other reasons "
Not really, some may apply i.e. walks _________
7. Do you calculate the maximum capacity of the
system in passengers per peak hour direction?
How? 170 no. of cars
8. What is the full peak-hour capacity of your cars?

seats 68 / 80 standing total 170 Use end of form if
different car types.
Is this determined by a formula? " by
experience _______________________________
Is this an agency policy? " ___________________
Do you have any folding seats " _______________
9. Do you have any published standards or policies
you can send us? ! policy headway of 6 mins in
peak _____________________________________
10.Do you measure the ratio of ridership to
capacity? " _______________________________
11.What type of signaling system is used? __________
3 aspect cab signals with ATO _______________
12.What is the closest headway scheduled? 2 mins
__________________________________ 00 secs
13.What is the theoretical closest? _____ 1 mins30
secs
14.What limits the closest headway? Station Dwells
!
Turnbacks " ____________________ Signaling !
Station Approach " _____________ Single track "
Junctions " ________________________ Other "
occasional turnback problems _______________
15.Is driving manual " or Automatic Train
Operation ! If ATO is manual driving allowed or
practiced !? once a week/driver ______________
16. If not tabulated above Type of fare collection
system?
Cubic stored value, being upgraded __________
17. If not tabulated above Wheelchair accessible
"? Type? _________________________________
18.We are trying to relate stations dwells with
passenger volume and door width to passenger flow
per second. Do you have information on maximum
station dwells, number of passengers entering and
exiting a train versus stopped time.
" _______________________________________
Contact rail superintendent Tom Ferer 962-2760
may have some dwell data __________________
19. Only for systems at or close to minimum headway.
We are interested in schedule adherence at close
headways. Do you have peak hour, peak point
information "? ______
Not to level you seem to want _______________
20. Only for heavy volume systems if there is no dwell
data. Later this year we may want to time dwells in
the peak period at peak-point stations. Would this
be possible? How should we set it up?
Probably _________________________________
21. Only for LRT Are car doors passenger actuated
"? ______________________________________
Does this cause any delays " _________________
Do you have any data on such delays " _________
22. Only if accessible. How many wheelchairs are
carried each day ___, each month _____? Line
156
by line _____? Is there data on any delays so
caused? " contact Avon Mackel 962-1083 for
use data (Task 5) _________________________
23. Only where no APTA data (not CR) Do you have
dimensioned floor plans of major car types in order
to determine number of seats, area for standing
passengers and door widths? " ____
24.Further Notes and Comments
Both Rohr (80 seats) and Breda (68 seats) are
deemed to have same peak capacity of 170
Table A 2.1 Light rail systems surveyed
and counts support this compares with

manufacturers rated crush capacity of 220-230
respectively 2 min. headway from 2 6 min.
services plus inserted extra train(s) _________
Possible dwell time survey location _________
Follow up wheelchair data in Task 5 _________
Very helpful & informative __________________
_________________________________________
Use other side of form for more comments or
information
157
Table A 2.2 Rail rapid transit systems surveyed
Table A 2.3 Commuter rail systems surveyed
Table A 2.4 AGT systems surveyed
158
A2.2 RIDERSHIP
INFORMATION
Table A 2.5 Summary of available ridership information (all

modes)
A2.2.1 COLLECTION AND AVAILABILITY

OF RIDERSHIP INFORMATION
Ridership data collected from agencies is presented in Appendix

Three Data Tabulations. Not all information categories
requested are included in the appendix and reference may be
made to the files on the computer disk for categories not
appearing in the tables in the appendix.
Ridership information for systems using the proof of payment
fare system (most light rail transit, some commuter rail and one
rail rapid transit system) is generally derived from ticket
machine revenue. Data from ride checks is used to give a ratio
between fare revenue and the number of passengers riding the
system. This ratio is then used to calculate ridership on a more
regular basis than would be affordable with ride checks alone. A
contact at BC Transit, which uses this technique, emphasized its
inaccuracy.
In several cases the mailed ridership count material has
contained more information than the contact indicated was
available. Some contacts have also discussed their data with
considerable skepticism regarding its accuracy In discussions
with contacts of systems operating at or near minimum
headway, the importance of station dwell times in governing
headway was apparent.
Only a few systems had data for loading of individual cars in
a train. Sufficient information for assessing the second level of
diversityuneven loading between cars in a trainwas
available for a number of rapid transit systems.
Commuter rail systems generally had the most exhaustive
collections of ridership data. This is made possible by the use of
conductors to collect fares and count passengers. Some
agencies, however, remarked that conductor counts tended to
overstate ridership in comparison with the results from
dedicated ridechecking staff. Efforts to improve the accuracy of
the conductor counts were being made to remedy this situation.
Most commuter rail operators were able to provide line-byline ridership summaries along with station on/off data for all
trains operated. Peak hour and peak 15-min ridership for
commuter rail was generally calculated from train-by-train data.
Given the limited number of Automated Guideway Transit
systems and their small size, little information could be
collected regarding this mode. To supplement the information
on AGT gathered during the survey, Chapter One Rail Transit in
North America includes a table of AGT ridership data compiled
from Trans 21 data.
In summary, for the 52 systems surveyed, the ridership
information indicated in Table A 2.5 is available. The commuter
rail systems account for the bulk of the systems providing
station on/off data.
It should also be noted that, where counts by train are
available, hourly ridership and ridership by short time periods
can be derived from that information if not presented separately.
A2.3 CAPACITY AND

POLICIES
Much car capacity information was compiled from APTA data
before the telephone survey commenced. Where possible the
information was checked with other sources and agency contacts
to confirm its accuracy. This data can be found in Appendix
Three Data Tabulations to this report. Train lengths were
determined from agency contacts.
Some contacts were able to provide floor plans of their cars
while others indicated that these would be available if required.
A2.3.1 LOADING STANDARDS

Acceptable loading standards varied between modes and
systems. Light Rail cars are generally designed to seat most
passengers in the off-peak. Loading standards for rail rapid
transit systems were found to vary considerably between
agencies. An example of this contrast can be seen by comparing
load factors between San Franciscos BART and New Yorks
PATH. In this example, load factors are the number of
passengers on the car divided by the number of seats. BART
passengers are reported as accepting load factors up to 1.5 on a
regular basis, although 2.5 was reached following the 1988
Loma Prieta earthquake. PATH, on the other hand, uses a load
factor of 4.1 as its standard car capacity index.
Commuter rail carriers attempt to provide one seat per
passenger and standing is rare although it is generally
considered acceptable near the downtown terminals. The sole
exception is on the Long Island Rail Road between Jamaica and
Penn stations where standing loads are common in the peak
hours. Agencies whose cars have 2+3 seating observed that
passengers will often stand voluntarily rather than sit three to a
bench.
Automated guideway transit offers an extreme alternative to
the all-seated policy of most commuter rail agencies. Miamis
Metro-Mover supplies only 8 seats for a car with a total capacity
of 100 passengers. Such a situation is made acceptable by the
short trips typical of circulator systems. While these loading
levels are also common on airport AGT systems, leisure systems
generally offer a seat per passenger.
A2.3.2 TRAIN LENGTH

Train length for light rail transit systems is limited by the length
of street blocks in sections of street running. This is a problem
159
not faced by the other modes with the occasional exception
where commuter trains could interfere with grade crossings
when stopped.
Systems handle the light rail transit block length problem in
different ways. In Portland, Tri-Met is limited to running twocar trains by the short blocks in that citys downtown. SRTD in
Sacramento runs four-car trains at peak hours resulting in
blockage of cross-streets during station stops downtown. This is
evidently made possible by a relaxed attitude on the part of the
city street department. The San Diego Trolley takes still another
approach and splits four-car trains in half before they enter the
downtown street-running portion of the line.
Table A 2.6 Summary of additional ridership and service

information (all modes)
A2.3.3 PASS-UPS
the telephone. This ratio was more commonly available

immediately from those agencies with a policy load factor.
Calculation of maximum system capacity was also often
handled in the same way. Unfortunately such calculations
frequently produced the current capacity of the system with the
existing fleet rather than the ultimate capacity by failing to take
into account increased train frequency and other possible service
enhancements.
A summary of data collected on the subjects above is
presented in Table A 2.6.
Commuter rail is strongly represented in the measurement of
ridership/capacity ratios and schedule adherence. Both of these
indicators are monitored closely by most commuter rail
operators, especially when service or track usage is provided on
a contract basis.
Few systems reported regular pass-up situations on their lines.

Conditions caused by unplanned service irregularity are not
included in this tabulation. New York City Transit was an
exception with pass-ups reported on a regular basis. Further
inquiries suggest that three of 11 NYCT trunks in Queens and
Manhattan are overloaded. Pass-ups are also routine in Mexico
City and occur to a lesser extent on systems in Montreal,
Toronto and Vancouver.
Pass-ups were reported on four light rail transit and two
commuter rail systems with none on AGT systems. However,
the light rail transit and commuter rail pass-ups are atypical for
these modes and the study team doubts that they are routine.
For other systems, pass-ups were often voluntary as a result of
passengers waiting for less crowded vehicles. This was
particularly the case at rail rapid transit stations adjacent to
downtown commuter rail terminals. Washingtons WMATA
reported pass-ups to be a problem when commuter trains arrived
at Union Station during peak hours.
In general, pass-ups were limited to stops near the edge of
downtown during narrow time periods. This was the case in
Edmonton where the recent light rail transit extension south to
the University has boosted ridership by 50% and caused trains to
become full before leaving the north edge of downtown. This
may be a temporary situation.
A2.4 HEADWAY
LIMITATIONS
As shown in Table A 2.7, headway constraints varied by mode.
One difficulty found with the answers to this question is that
staff of systems not running at maximum capacity were not
familiar with the ultimate constraints faced by their system. This
concern would be particularly marked for dwell time, turnback
and junction effects which would not be as evident with low
service frequencies.
A2.3.4 EVENT RIDERSHIP

In response to the panels request, systems serving sports
stadiums were identified and asked whether they had specific
ridership figures for special events. Many agencies do keep
some track of the patronage gained from such service. Most of
this information is in the form of estimates of ridership and
travel market share. Little information about high car loading
was available although BC Transit reported loads 25% in excess
of standard peak-hour maximum car capacity.
A2.3.5 RIDERSHIP/CAPACITY RATIO

Remarkably few agencies aside from commuter rail operators
indicated that they regularly calculated a ridership/capacity
ratio. Many calculations of this information were made while on
A2.4.1 SIGNALING
A majority of contacts (67%) reported signaling to be a major
constraint on their systems. In many cases the signaling system
was designed to accommodate a level of service below the
maximum that could be provided given right-of-way and vehicle
characteristics. Reported actual and theoretical minimum
headways are shown in Table A 2.8. This allowed systems with
relatively long headways to report signaling as a constraint. This
is illustrated by the Edmonton light rail transit line which has
already reached its minimum theoretical headway of five
minutes despite operating on a largely grade-separated line with
full grade crossing protection. The Calgary light rail transit
system, which uses the same vehicles and has less right-of-way
protection, operates every three minutes on signaled sections
with higher frequencies possible on the downtown transit mall.
160
Table A 2.7 Headway constraints by mode (excludes those systems for which responses were not obtained)
Table A 2.8 Minimum headways for systems surveyed
Notes:
N/A indicates not available and/or applicable.
Minimum headways for many commuter rail systems are a
result of the contract with the host railroad and are not due to
practical headway constraints.
161
On some rail rapid transit lines and light rail transit trunks
headways have reached the minimum possible with the current
signaling system. In these cases efforts are being made to
upgrade the signaling to allow more frequent service. Even
relatively recent and advanced signal systems such as those on
BART and the MUNI Metro subway have reached their limits
and are being replaced with more capable technology.
The shortest theoretical headways given represented the
extreme ends of the spectrum. New Jersey Transits Newark
City Subway, operating PCC cars with wayside automatic block
signals, was quoted as having a minimum headway of 15 sec.
Such frequencies are made possible by manual operation at
relatively low speeds, possibly with red signals taken as
advisories, and multiple station berths. Similar conditions permit
SEPTA to operate light rail vehicles 30 sec apart.
For fully signaled systems, Metro-Dades MetroMover AGT
has a minimum theoretical headway of one minute and 10 sec. A
large number of rail rapid transit systems reported minimum
theoretical headways of one minute, 30 sec but such frequencies
are only regularly operated on BC Transits SkyTrain. Here,
trains currently operate as close as every minute and 35 sec.
This is made possible with the Seltrac moving block signal
system. The Morgantown PRT can operate at exceptionally
close 15-sec headways thanks to the use of small vehicles and
offline stations.
The issue of light rail transit street running is related to
signaling in its effects on limiting headways. The only light rail
transit operation to cite street running as a headway limitation
was Baltimore. Given that the current headway on the line is 15
min, it is unlikely that this is a practical problem. Traffic
congestion was reported as a problem for the Toronto streetcar
system but this is not a typical contemporary light rail transit
operation. Also of relevance is the practice of the San Diego
Trolley of splitting long trains when they enter downtown. This
increases the number of trains operating on street but apparently
has not caused an operational problem on the line segment
governed by traffic signals.
Signaling of commuter rail systems is a very complex area
given the wide variety of signal types which can be found on
some of the systems surveyed. Complicating this are factors
such as ownership of track by other than the operating agency
and discrepancies between signaling practices between
railroads. Even the two large New York commuter rail
operations, Metro-North and the Long Island Railroad, reported
signaling ranging from centralized traffic control (CTC) to
manual block system (MBS) despite controlling almost all of
their lines. In many commuter rail operations, headways are
limited by the contract with the host railroad and not by the
signaling system.
A2.4.2 TURNBACKS
Turnbacks were cited as a problem on five rail rapid transit
systems, three light rail transit systems and two commuter rail
services. Turnbacks are a common limitation when line capacity
is neared or where a rapid turnaround is required to maintain
schedules. The latter is the case on the Los Angeles light rail
Blue line where the train operators drop back one train in order
to minimize terminal time. The other light rail transit operator
facing turnback difficulties is SEPTA which operates a number

of high frequency routes converging on a central terminus.
However, as this terminus is a loop, the delays may be more
properly attributed to long dwell times resulting from passengers
boarding and alighting.
Rail rapid transit agencies with intense service, New York,
Boston and Vancouver, indicated turnbacks as a constraint. Staff
in Los Angeles claim that the Red Line subway also faces this
constraint despite long (6 min) headways.
Commuter rail contacts rarely indicated turnbacks as being a
problem. This is understandable since in many cases equipment
is only able to make one peak direction trip in each peak period.
Agencies identifying this factor as a problem were GO Transit
and New Jersey Transit. The latter stated that trains required a
minimum 30-min turnaround time at New Yorks Penn Station
before returning to service.
A2.4.3 JUNCTIONS
Junctions are a minor constraint with only five of 57 systems
reporting them as limitations. The relevant rail rapid transit
systems are the CTA and SEPTA. Commuter rail operators
facing this difficulty are Chicagos Metra and New Yorks
Metro-North. Other busy systems avoid this problem through
the use of flying junctions which obviate the need for at-grade
crossings. A recently installed rail/rail underpass west of
Torontos Union Station provides a relatively simple example of
this technique.
A2.4.4 STATION APPROACH

This limitation was cited even less often than junctions by
agency contacts with only three systems indicating difficulty.
BC Transit was the only rail rapid transit system to give station
approach as a problem. In this case, the station approach
difficulty is a result of turnback limitations at the downtown
terminus and may perhaps be better seen as a turnback problem.
The two New York commuter rail operators, Long Island
Railroad and Metro-North, both encounter this constraint at their
large, congested Manhattan terminals.
A2.4.5 SINGLE TRACK

Single Track operation was a difficulty primarily encountered
by light rail transit (32%) and commuter rail (19%) operators.
Light rail transit single track operation has been reduced in
Portland, Sacramento and San Diego, through double tracking
projects. San Diego has eliminated single track from its current
network but the Santee extension which is under construction
will feature a single track section limiting headways to 15
minutes. The new light rail transit line in Baltimore also features
considerable single track operation but this route has been
designed to allow double tracking in the future. Older light rail
transit lines with single track running include SEPTAs Media
and Sharon Hill routes.
Single track is also a problem on some of the newer commuter
rail lines where passenger train service has brought substantial
162
increases in the number of trains operated. This is the case on
the Los Angeles Metrolink and San Diego Coaster services, and
on the Tri-Rail line in southern Florida. A number of other
commuter rail operations also reported double track as being a
limitation. Such was the case for Marylands MARC service on
portions of Amtraks busy Northeast Corridor Line.
A2.4.6 STATION DWELLS

Station dwells were found to be an important limitation on
capacity with 28% of agencies indicating them as a headway
limitation. Station dwells and related topics are discussed in
section.
and is unique to those few light rail operations with on-board

fare collection.
A2.5 STATION
LIMITATIONS
Table A 2.9 indicates the constraints that limit capacity at rail
transit stations.
A2.5.1 FULL PARKING LOTS

A2.4.7 OTHER HEADWAY CONSTRAINTS
While only 18% of all systems gave other headway constraints,
44% of commuter rail operators responded to this category. The
principal reason for this is that most commuter rail systems
operate on tracks owned by other railroads and so must rely on
the track owner to provide pathways for commuter trains. This
constraint seemed to be the strongest for the MARC and
Virginia Railway Express services in the Washington DC region
which have faced great resistance from the owning railroads to
the operation of additional trains.
While in most cases commuter trains operate on tracks owned
by freight railroads or Amtrak, Philadelphias SEPTA also owns
some track used by the freight companies. This gives SEPTA a
better bargaining position for those commuter routes which
operate over freight trackage. In other areas, such as Chicago
and southern California, the commuter rail agencies are
acquiring lightly used track from the freight railways. While this
imposes a maintenance burden on the commuter rail agency, it
does allow a greater priority to be accorded to the passenger
service.
Two light rail transit systems reported other constraints,
Toronto for extensive street running and SEPTA for delays due
to electronic fare boxes. The latter factor extends dwell times
By far the largest station constraint reported by systems was that

of full park and ride lots. 56% of all systems noted a shortage of
parking space but the response was even stronger from
commuter rail systems with 81% indicating full lots. The
importance of parking to ridership can often be linked to the
orientation of the system towards suburban or urban customers
with the former requiring more parking.
Some commuter rail systems, such as Chicagos Metra, have
taken to establishing cornfield stations whose main purpose is
to allow the construction of park and ride lots outside population
centers.
A2.5.2 TICKETING LINE-UPS

Ticket purchase line-ups were generally not a problem except
near month-end when monthly pass purchases are made. Pass
purchase queues were especially pronounced for commuter rail
systems with a number of agencies offering ticket by mail
programs to reduce line-ups at stations. The San Francisco
CalTrain peninsula commuter rail service offers an incentive of
a 2% discount on passes sold by mail, in comparison to the
service charge made by other operators.
Table A 2.9 Station constraints by mode (excludes those systems for which responses were not obtained)
163
A2.5.3 CONGESTED PLATFORMS

Platform congestion was a relatively small problem confined to
the two most heavily used rail rapid transit systems (New York
and Mexico City) and the two largest commuter rail systems
(Long Island and Metra.) The only light rail transit system
reporting congested platforms is the STC in Mexico City,
however, their light rail transit line is light rail in name only and
has most of the characteristics of a rail rapid transit system.
A2.5.4 OTHER CONGESTION

Only Mexico City and New York experienced trouble with
congestion at additional locations. In the case of New York,
entry and exit turnstiles create congested conditions for
passengers.
A2.5.5 NO TRANSFERSTRANSFER COST

Most responses here were due to a lack of fare integration
between modes and the practice of levying a surcharge for
transfers. Most systems without fare integration indicated that
work was being done to remedy the situation. New Yorks MTA
is working to permit bus-subway transfers.
Another factor, particularly for commuter rail and some rail
rapid transit lines, is the convenience of the downtown terminals
to workplaces since a well located terminal can obviate the need
for many transfers. Such is the case with PATCOs route in
Center City Philadelphia.
A2.5.6 SAFETY AND SECURITY

A quarter of the systems surveyed indicated that concerns over
safety and security could have an effect on ridership. These
concerns were greatest on large, urban systems but were also
apparent on smaller light rail transit lines (Sacramento,
Edmonton) during the evening.
While most commuter rail trains were viewed as being safe,
parking lot security was a major concern at many systems. This
problem is also experienced on some rail rapid transit lines with
one parking lot on the BART line in Oakland not filling largely
as a result of security issues.
A.5.7 LONG WALKSPOOR ACCESS

A quarter of all systems reported access problems with there
being very little differentiation between each mode. Some of the
factors which influenced these answers included poor station
location, poor station design and large park and ride lots.
A2.5.8 OTHER STATION CONSTRAINTS

Two systems reported short platforms as being limitations, the
GCRTA rail rapid transit line in Cleveland and the Long Island
Railroad. In the former case the platforms on the affected line
segment are being lengthened to eliminate the constraint. On the

LIRR station length is limited by the presence of adjacent grade
crossings.
Another difficulty reported on commuter rail systems,
particularly in low density areas, is a lack of feeder buses to and
from stations. This is being remedied in some areas by the use
of dedicated feeder buses from rail stations to important work
sites.
A beneficial station effect has occurred at Trenton, NJ where
SEPTA and New Jersey Transit service connect to offer travel
between Philadelphia and New York. This has increased the
ridership on both systems.
A2.6 DWELL TIMES

A number of factors affect rail transit dwell times. Two of the
most important are platform height and method of fare
collection. Wheelchair access is also of importance and this is
dealt with in detail in section below. Appendix Three contains
tables of these factors for each system surveyed.
As noted in Table A 2.6, only 20% of systems have dwell
time data, some of which was noted as being outdated and of
questionable value. Passenger flow through doors was not
immediately available from any system. Car door widths have
also been determined for many systems, these are included in
the tabulations of Appendix Three.
Some systems were able to supply policy dwell time
information. The San Diego Trolley has a policy minimum
dwell of 20 seconds while Bostons MBTA has policy dwells at
each of its rail rapid transit stations ranging from 15 to 30 sec.
Some systems, such as Calgarys light rail transit and the TTC
subway, have an enforced safety delay of a few seconds once
the doors have closed before the train can move.
Dwell times on commuter rail lines ranged widely depending
on car design and station usage. New Jersey Transit gave a
range of 20 sec to 8 min depending on the line and station. The
Long Island Railroad also operates a variety of equipment
resulting in dwell times being more of a problem for
conventional, lowloading, diesel-hauled trains than on electric
multiple unit trains designed for rapid, high-level boarding and
alighting.
A2.6.1 FARE COLLECTION

Fare collection effects on dwell times are principally a light rail
transit issue although fares are collected on-board by the
operator on exceptional rail rapid transit lines, as on Clevelands
Red Line. Fare collection by a conductor, as used in Chicago
and on many commuter rail lines, does not affect dwell times.
Fare collection by the light rail transit operator is exclusive to
the older light rail transit lines. Even here, fare collection in the
Central Business District (CBD) is usually handled by station
agents to ensure high passenger flow capacities. SEPTA
reported that the addition of electronic fareboxes to its light rail
transit fleet has also resulted in extended dwell times outside the
CBD.
All new light rail transit lines, some commuter rail operators
and one new rail rapid transit line have opted for the proof of
payment (PoP) system which eliminates any effect of fare
164
collection on dwells. This system is also used on the heaviest
streetcar line in Toronto to allow all car doors to be used for
boarding and alighting and so reduce dwells.
Seven of the rail rapid transit systems surveyed use turnstiles
which accept magnetically encoded tickets and passes to speed
passenger movements. The use of automated ticket vending
machines (TVMs) is also becoming widespread, both in
conjunction with proof of payment fare systems and as a way to
speed ticket purchase for other fare systems.
commuter rail systems use a wide variety of wheelchair access

methods ranging from level loading to car and platform
mounted lifts. The only light rail operation to use car-mounted
lifts is the San Diego trolley, all later systems use platform lifts
or special mini-high platforms which provide access to the
accessible location on each train. The low-floor car, which
overcomes much of the accessibility problem, is not yet in use in
North America; however, Portland and Boston have ordered
cars of this type.
A2.6.2 PLATFORM HEIGHT
A2.7 SCHEDULE
ADHERENCE
Platform elevation has a considerable effect on dwells since

lowlevel platforms necessitate the use of steps on the car to
reach the passenger areas. Rail rapid transit and AGT systems
universally feature high platform loading with its inherent speed
advantages. Light rail transit and commuter rail systems
featured both high and low level boarding, with some lines
allowing both through the use of dedicated doors and/or
moveable steps. The latter solution is used on the MUNI Metro
network in San Francisco to allow high-capacity operation in the
downtown subway and traditional street running on the surface.
In the subway one door cannot be used due to the cars end
taper.
A2.6.3 WHEELCHAIR EFFECTS

Wheelchair boarding and alighting can have major effects on
dwell times, particularly when some form of boarding aid, such
as a lift, is required. The accessibility of the systems in this
study is summarized in Table A 2.10. Light rail transit and
Table A 2.10 Summary of wheelchair accessibility
36% of the systems surveyed (see Table A 2.6) indicated that

they measure on time performance on a regular basis and it is
likely that all of the commuter rail systems measure this
variable, whether this was reported or not. Schedule adherence
for commuter rail is important in the case of contracted service
where this data can be a determinant of the fees paid to the
contractor.
A2.8 COMMENTS AND

RESULTS
With the survey complete, an adequate range of data; by peak
hour, peak-within-the peak, individual trains and individual cars
selected operators) was obtained for use in Task 5.
Although several agencies have and can make riding counts
(ride-checks) available, these do not always clearly show dwell
times relative to the passengers boarding and alighting, nor do
they show levels of crowding in the cars and on the platforms.
Such dwell time data and wheelchair boarding and alighting
times were the principal areas for the field data collection
requirements of Task 5.
The telephone survey achieved its goals with only a few
systems not responding satisfactorily. Most agency contacts
have proved to be quite helpful and accommodating.
Information gathering to supplement the survey continued
during the remainder of the project. The valuable contacts made
during the survey were invaluable for the field data collection
component of the project.
165
A3. APPENDIX THREE

Data Tabulations
A3.1 INTRODUCTION
This appendix compiles much of the basic information
contained in the database assembled for this project. The
complete database, in Microsoft Access 2.0 format, is included
on the companion computer diskette to this report.
Not all the information categories listed could be determined
for each operator, route, car type, etc. as a result of inadequacies
in the data supplied by transit agencies and existing
compilations. Particular emphasis was placed on getting as
much information as possible from the major U.S. and Canadian
operators.
The data used are as up-to-date as possible and were collected
from agencies between April 1994 and May 1995. Data were
updated where appropriate during the course of the study. As
many of the basic statistics (primarily in Table A 3.1) were
determined from the FTAs 1993 National Transit Database,
systems which commenced operation in 1993 and later may not
have complete operating statistics. This applies, for example, to
Bi-States Metrolink light rail transit line in St. Louis and the
Metrolink and Coaster commuter rail services in Southern
California.
A3.2 NOTES ON THE TABLES

Route ridership information was generally compiled directly
from agency data. Peak-hour and peak 15-min flows were often
Table A 3.1 Rail transit annual operating statistics
calculated from individual train counts, particularly for

commuter rail. In these cases, calculations were based on strict
definitions of the time periods considered. For example, if trains
were scheduled to arrive at 7:30, 8:00 and 8:30, the peak hour
would be determined by the sum of the loads on either the first
pair or second pair of trains, not all three. This ensures that the
time interval examined does not exceed the stated interval as
including all three trains would give an interval of up to 1 hour,
59 sec.
The table of trunk lines (Table A 3.4) does not include the
Mexican systems; they can be found in Table A 3.3. The figures
in the minimum operated headway column should be used with
care, particularly for commuter rail where multiple-track lines
and station approaches can allow simultaneous movements on
parallel tracks.
The total car capacity figures in Table A 3.5 should also be
used cautiously as each agency has its own standards for
determining this value. Scheduled loading levels were used
wherever possible, often based on a standing density of 4/m2. In
some cases transit agencies provided the manufacturers
maximum stated load; often at a crush loading level not
acceptable in regular service. Chapter Five, Passenger Loading
Levels, discusses loading levels and provides recommendations.
166
Table A 3.1 Rail transit annual operating statistics (continued)
167
Table A 3.2 Train length, loading and fare collection characteristics
168
Table A 3.2 Train length, loading and fare collection characteristics (continued)
Table A 3.3 Route characteristics and ridership1
SEPTA commuter rail ridership data was determined from SEPTA: Regional Rail Ridership Census 1993-94, SEPTA 1994.
169
Table A 3.3 Route characteristics and ridership (continued)
170
171
172
Table A 3.4 Trunk characteristics and ridership2
Ibid.
173
Table A 3.4 Trunk characteristics and ridership
174
Table A 3.5 Rail transit car specifications
175
Table A 3.5 Rail transit car specifications (continued)
176
Includes double-stream high-level center door.
177
178
STC (Mexico City) and STCUM (Montreal) are coincidentally adjacent listingsand the only operators of the French metro pneumatique
system. The cars on both systems are substantially identical in dimensions, number of doors and seatings. Montreal rates total capacity at 160.
Mexico City offered no such rating but loadings on the busiest lineline 3reach 260 passengers per car. This is almost 6 passengers per
m2by far the highest in North America. A more palatable total capacity of 220 passengers has been assigned to the Mexican fleet, less to the
dimensionally identical steel-wheeled versions which experience less intense loading.
THE TRANSPORTATION RESEARCH BOARD is a unit of the National Research Council, which
serves the National Academy of Sciences and the National Academy of Engineering. It evolved in
1974 from the Highway Research Board which was established in 1920. The TRB incorporates all
former HRB activities and also performs additional functions under a broader scope involving all
modes of transportation and the interactions of transportation with society. The Boards purpose is to
stimulate research concerning the nature and performance of transportation systems, to disseminate
information that the research produces, and to encourage the application of appropriate research
findings. The Boards program is carried out by more than 400 committees, task forces, and panels
composed of more than 4,000 administrators, engineers, social scientists, attorneys, educators, and
others concerned with transportation; they serve without compensation. The program is supported by
state transportation and highway departments, the modal administrations of the U.S. Department of
Transportation, and other organizations and individuals interested in the development of transportation.
The National Academy of Sciences is a private, nonprofit, self-perpetuating society of distinguished
scholars engaged in scientific and engineering research, dedicated to the furtherance of science and
technology and to their use for the general welfare. Upon the authority of the charter granted to it by
the Congress in 1863, the Academy has a mandate that requires it to advise the federal government on
scientific and technical matters. Dr. Bruce M. Alberts is president of the National Academy of
Sciences.
The National Academy of Engineering was established in 1964, under the charter of the National
Academy of Sciences, as a parallel organization of outstanding engineers. It is autonomous in its
administration and in the selection of its members, sharing with the National Academy of Sciences the
responsibility for advising the federal government. The National Academy of Engineering also
sponsors engineering programs aimed at meeting national needs, encourages education and research
and recognizes the superior achievements of engineers. Dr. Harold Liebowitz is president of the
National Academy of Engineering.
The Institute of Medicine was established in 1970 by the National Academy of Sciences to secure
the services of eminent members of appropriate professions in the examination of policy matters
pertaining to the health of the public. The Institute acts under the responsibility given to the National
Academy of Sciences by its congressional charter to be an adviser to the federal government and, upon
its own initiative, to identify issues of medical care, research, and education. Dr. Kenneth I. Shine is
president of the Institute of Medicine.
The National Research Council was organized by the National Academy of Sciences in 1916 to
associate the broad community of science and technology with the Academys purpose of furthering
knowledge and advising the federal government. Functioning in accordance with general policies
determined by the Academy, the Council has become the principal operating agency of both the
National Academy of Sciences and the National Academy of Engineering in providing services to the
government, the public, and the scientific and engineering communities. The Council is administered
jointly by both Academies and the Institute of Medicine. Dr. Bruce M. Alberts and Dr. Harold
Liebowitz are chairman and vice chairman, respectively, of the National Research Council.

TCRP Report 13

Uploaded by

Copyright:

Available Formats

TCRP Report 13

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

TCRP Report 13

Uploaded by

Copyright:

Available Formats

What organizations are discussed in the document?

What organizations are discussed in the document?

What is the purpose of the Transportation Research Board?

What is the purpose of the Transportation Research Board?

T

The Federal Transit Administration

Rail Transit Capacity

Transportation Research Board

TCRP OVERSIGHT AND PROJECT

TRANSPORTATION RESEARCH BOARD EXECUTIVE COMMITTEE 1996

TRANSIT COOPERATIVE RESEARCH PROGRAM

Research Sponsored by the Federal Transit Administration in

NATIONAL ACADEMY PRESS

TRANSIT COOPERATIVE RESEARCH PROGRAM

Published reports of the

Printed in the United States of America

This report will be of interest to transportation and rail-transit planners,

Under TCRP Project A-8, research was undertaken by Transport Consulting

2.2.2 Footnotes and References ......................................8

8.3 SIGNALED SECTIONS................................................. 89

A2.5 STATION LIMITATIONS ........................................ 162

COOPERATIVE RESEARCH PROGRAMS STAFF

PROJECT PANEL A-8

Berkowitz the projects statistical advisor. Murray McMillan of

evaluating transit line performance,

Inside the Report

FTAFederal Transit Administration. Section 15 of the Urban Mass

has been prepared containing spreadsheets into which system

Footnotes and References

Table S.1 North American rail ridership by mode

Table S.2 Transit ridership summary (million)

The basics of rail transit capacity are very simplethe product of

... long before crowding levels...... reached New York

Determining how many passengers will fit on a train is a policy

related to passenger flow with reasonable accuracy. It is,

The Morgantown Automated Guideway Transit system, with off-line

The minimum train separation is based on systems designed

Figure S.3 Headway components for cab-control signaling

Figure S.4 Toronto Transit Commission King Station S/B

added one second per passenger on average. Light rail with

Figure S.5 Summary of rail transit doorway average flow

Table S.4 Linear load summarypassengers per meter

Figure S.6 Linear passenger loading of articulated LRVs.

the calculation of maximum achievable capacity. On-board fare

Three levels of loading diversity were reviewed. The diversity

Capacity determination was broken down into the four modes

Figure S.8 Headway components of selected North American

mended range to be applied in capacity determination is 15 to 25

Figure S.10 shows the capability of various train control systems

Figure S.9 Light rail travel time over single-track section.

Figure S.11 Achievable capacity with a multiple-command

States and Canada, no lines exceed 50,000. NYCTs two-track

S11 INCREASING CAPACITY

Where higher capacity is required there are the obvious steps of

S12 ECONOMIC ISSUES

THE DISK IS NOT REQUIRED TO CALCULATE

All project spreadsheet work has been carried out in Microsoft

USING THE SPREADSHEETS Instructions,

CAUTION Reasonable care has been taken in obtaining

ORDERING The disk is available on request to