400180001A
transportation
air quality
analysis
sketch planning methods
Volume
-------
NOTICE
This information document is isssued by the Office of
Transportation and Land Use Policy, U.S. Environmental Protection
Agency, in response to Section 108(f) of the Clean Air Act. This
document provides a range of quantitative analytical techniques to
evaluate transportation measures and packages of alternative measures,
and is designed to assist State and local air pollution control
agencies and Section 174 lead planning agencies to perform
transportation-air quality planning. Although examples are provided,
this document is not intended as a substitute for the necessary
case-by-case analysis by approporiate local planning organizations.
A limited number of copies of this document are available from EPA
Regional Offices. Additional copies may be obtained, for a nominal
cost, from the National Technical Information Services, 5285 Port Royal
Road, Springfield, Virginia 22151.
This technical report was furnished to the Environmental
Protection Agency by Cambridge Systematics, Inc., Cambridge,
Massachusetts 02142, in fulfillment of Contract No. 68-01-4977. The
opinions, findings, and conclusions expressed are those of the authors
and not necessarily those of the Environmental Protection Agency or of
cooperating agencies. Mention of company or product names is not to be
considered as an endorsement by the Environmental Protection Agency.
-------
Technical Report Documentation Page
EPA 400/1-sW-OOla
TRANSPORTATION AIR QUALITY ANALYSIS -
SKETCH PLANNING METHODS
Volume I; Analysis Methods
7. Author'.)
9. Performing Organization Name and Address
Cambridge Systematics, Inc.
238 Main Street
Cambridge, Massachusetts Uzlnz
12. Spo-Mwing Agency Nome ond Address
Office of Transportation & Land Use Policy
Environmental Protection Agency
401 M Street S.W.
Washington, B.C. 20460
3. Recipient's Cotolog No.
5. Report Date
December, 1979
6. Performing Organization Code
78015
8. Performing Organization Report No.
10. Work Unit No. (TRAIS)
11. Contract or Grant No.
68-01-4977
13. Type of Report and Period Covered
Final Report
14. Sponsoring Agency Code
IS. Supplementory Notes
Part of a two-volume final report describing sketch planning methods for trans-
portation and air quality planning.
16. Abstract
Analytical methodologies are described (in Volume 1) and illustrated (in Volume II
for use by metropolitan planning organizations and other state and local transportation
agencies in analyzing the air quality potential of candidate urban transportation
measures. As sketch planning techniques, the methods are designed to produce first-
cut estimates of a proposed transportation measure's impact for a relatively small in-
vestment of time and effort. Quantitative methods oriented to auto restricted zones,
high occupancy vehicle priorities, transit improvements, parking programs, carpool/
vanpool incentives, and staggered work hours are provided. The methods use worksheet,
programmable calculator, and computerized approaches to apply disaggregate behavioral
models. They can be used to predict traveller demand as a function of transportation
system characteristics, transportation facility operations as a function of their usage
and their physical characteristics, and special impacts including vehicular emissions,
fuel consumption, and operating costs. Guidelines are provided both to those respon-
sible for designing the transportation-air quality analysis approaches in specific
local areas, and to those who will carry out these analyses. In addition, references
are provided to documents which provide additional detail on the methods.
17. Key Words
Air Quality Planning
Urban Transportation Planning
Sketch Planning
Transportation Systems Management
18. Distribution Stotement
19. Security Clossif. (of this report)
unclassified
20. Security Clossif. (of this poge)
unclassified
21. No. of Poges
22. Pnce
Form DOT F 1700.7 (8-72)
Reproduction of completed poge authorized
,,r*»i protection Agoncy
r. g. Envirr-n.-r-.'rioaJ. *;.. , ^
^-;ri ^J"Room 1670
60G04
-------
EPA 40071-80-001 a EPA Contract No. 68-01-4971
TRANSPORTATION AIR QUALITY ANALYSIS
SKETCH PLANNING METHODS
VOLUME I
DECEMBER 1979
Prepared for:
Environmental Protection Agency
Office of Transportation and Land Use Policy
in Cooperation with the
Urban Mass Transportation Administration
Office of Planning Methods and Support
-------
PREFACE
The sketch planning methods described in this handbook have been
assembled and illustrated under contract to the U.S. Environmental
Protection Agency and the Urban Mass Transportation Administration in
order to provide assistance to local transportation agencies as they
conduct the planning and analyses required to develop the transportation
portions of State Implementation Plans. The handbook is presented in two
volumes:
Volume I: Analysis Methods
Volume II: Case Studies
The project was performed by Cambridge Systematics, Inc. Earl R.
Ruiter, Project Manager, and John H. Suhrbier, Principal Responsible,
provided the overall direction and management of the work performed. The
development of the handbook benefited greatly from the advice provided by
Marvin L. Manheim in the sketch planning applications of transportation
analysis and programmable calculator methods areas, and by Adolph D. May
(University of California, Berkeley), and Frederick A. Wagner
(Wagner-McGee, Inc.) in the area of highway facility operations. The
development or enhancement of specific analysis methods was carried out
by Ellyn S. Eder and Melissa M. Laube. Additional major contributors to
the handbook were Elizabeth A. Deakin, Lance A. Neumann, Daniel S. Nagin,
Terry J Atherton, Scott D. Nason, William D. Byrne, and Greig Harvey.
-------
Important contributions have been made by EPA staff members Chris
Shaver, David Levinsohn, Gary Hawthorn, and Joseph Ossi. Their support
and individual inputs have been very much appreciated. The contents of
this report, however, reflect the views of Cambridge Systematics, Inc.,
and they are fully responsible for the facts, the accuracy of the data,
and the conclusions expressed herein. The contents should not be
interpreted as necessarily representing the views, opinions, or policies
»»
of the Environmental Protection Agency, the Urban Mass Transportation
Administration, or the United States Government.
-------
iii
TABLE OF CONTENTS
1. SKETCH PLANNING AND AIR QUALITY ANALYSIS
1.1 Purpose of this Handbook 1-1
1.2 Basic Concepts in Sketch Planning 1-6
1.3 The Range of Sketch Planning Techniques Available 1-9
l.M Choosing the Appropriate Technique 1-1M
2. TECHNIQUES FOR ESTIMATING TRAVEL DEMAND IMPACTS
2.1 Manual Methods 2-2
2.1.1 Worksheet Mode Choice Methods—Overview 2-4
2.1.2 Nomograph/Formula Techniques—Overview 2-16
2.1.3 Systematic Data Analysis—Overview 2-30
2.2 Programmable Calculator Methods 2-38
2.2.1 Program HHGEN—Developing a Sample of Households 2-40
2.2.2 Program 2MODE-AGG—Synthetic Mode Choice 2-44
2.2.3 Program 3MODE(VAN)-AGG—A Pivot-Point Mode Choice
Model 2-47
2.3 Computer Methods 2-52
2.3.1 CAPM—Community Aggregate Planning Model 2-59
2.3.2 SRGP—Short Range Generalized Policy Analysis 2-64
2.3.3 Transit Sketch Planning Procedure 2-73
3. TECHNIQUES FOR ANALYZING FACILITY OPERATIONS
3.1 Manual Techniques 3-4
3.1.1 Traffic Flow Relationships 3-5
3.1.2 Graphical Techniques 3-10
3.1.3 Manual Areawide Traffic Engineering Analysis Method 3-13
3.1.4 Transfer of Experience 3-15
-------
IV
TABLE OF CONTENTS (Continued)
3.2 Calculator Methods 3-20
3.2.1 Program BUS 3-21
3.3 Computer Methods 3-23
3.3.1 TRANSYT 3-30
3.3.2 FREQ 3-39
4. IMPACT ANALYSIS TECHNIQUES
U.I Automotive Emissions Estimation Procedure 4-3
4.2 Automotive Fuel Consumption and Operating Cost Estimation
Procedure H-H
4.3 BUSPOL and ENERGY—Environmental and Energy Impacts of
Bus Operations 4-6
4.4 MOBILEl--Motor Vehicle Emissions 4-7
5. DESIGNING AN ANALYSIS APPROACH
5.1 Developing an Analysis Strategy and Selecting Techniques 5-3
5.2 Representing Transportation System Changes in Sketch
Planning Methods 5-15
5.3 Market Segmentation Guidelines 5-19
5.4 Factors Influencing the Accuracy of Sketch Planning
Techniques 5-23
5.4.1 Input Data Accuracy 5-24
5.4.2 Level of Analysis Detail 5-26
5.4.3 The Reliability of Travel Models 5-28
5.4.4 Sensitivity Testing 5-29
5.4.5 Transferability 5-31
5.5 Data Sources for Air Quality Analysis 5-41
APPENDIX A: Worksheets for the Manual Pivot-Point Mode Choice Method
APPENDIX B: Worksheets for the Manual Synthetic Mode Choice Method
APPENDIX C: Incremental Work Trip Mode Choice Program Documentation
APPENDIX D: An Automotive Emissions Estimation Procedure
APPENDIX E: An Automotive Fuel Consumption and Operating Cost
Estimation Procedure
APPENDIX F: Bibliography
-------
LIST OF TABLES
Pace
1.1 Reasonably Available Transportation-Air Quality 1- 2
Measures to be Analyzed for SIP Revisions
1.2 Typology of Sketch Planning Techniques 1-10
1.3 Cross-Reference Table for Analysis Method and Case Studies 1-12
2.1 Average Vehicle Trip Rates and Other Characteristics 2-19
of Generators
2.2 Detailed Trip-Generation Characteristics 2-22
2.3 Trip-Generation Parameters 2-25
2.M Daily Home-Based Vehicular Trips that Might 2-36
Be Attracted to the Bicycle
3.1 Comparison of Field Measurements with Davidson Equation 3- 9
Computations
3.2 Traffic Signal Timing Optimization Impacts 3-16
3.3 Freeway Ramp Control System Impacts on Average Speed 3-17
3.4 Impact Changes Due to Traffic Management Strategies for 3-37
Base Conditions
M.I Inputs to the Automobile Fuel Consumption and Operating 4- 5
Cost Procedure
5.1 Applicability of Demand and Facility Operations Analysis 5-11
Methods
5.2 Transferability of Work Mode Choice Model to Different Cities 5-35
5.3 Comparison of Elasticities for Three Mode Choice Models 5-37
5.U Transportation-Air Quality Data Requirements and Sources 5-43
-------
VI
LIST OF FIGURES
Page
2.1 Sequence of Basic and Supplemental Worksheets 2- 9
2.2 Data Flows in Program 3MODE(VAN)-AGG 2-50
2.3 Interrelationships of Travel Demand Models 2-66
2.M Major Steps in Sketch Planning Procedure 2-75
3.1 Basic Intersection Diagram 3-H
3.2 Traffic Signal Progression 3-11
3.3 TRANSYT Program Logic 3-33
3.1 Revised TRANSYT Program Logic 3-35
3.5 Display of Platoon Formation at a Traffic Signal 3-36
Adapted from TRANSYT Output
3.6 Freeway Geometry Definitions for FREQ 3-U1
3.7 Schematic View of FREQ Simulation Procedure 3-U2
3.8 FREQ Priority Entry Optimization Procedure 3-^3
3.9 An Idealized Speed Contour Map from FREQ 3-^5
5.1 Representative Actions Within Eight Classes of
Transportation-Air Quality Measures 5-6
5.2 Work Mode Choice Model: Definition of Variables 5-34
-------
CHAPTER i. SKETCH PLANNING AND AIR QUALITY ANALYSIS
1.1 Purpose of this Handbook
This handbook presents a selection of techniques for transportation-
air quality planning. It describes each technique and suggests the
purposes and conditions for which it would be appropriate. The emphasis is
on "sketch planning" techniques—ones which can produce a first-cut
estimate of a proposed transportation measure's impact for a relatively
small investment of time and effort.
The analysis approaches described here should be useful in developing
the transportation portions of State Implementation Plans (SIP's), as
required to meet national ambient air quality standards under the Clean Air
Act, as amended (42 U.S.C. 1857 et seq.). The Act, and the joint Environ-
mental Protection Agency/Department of Transportation guidelines issued
pursuant to it, call for the analysis of a number of transportation
measures which potentially could improve air quality (Table 1.1). The
effects these measures would have—on travel; the transportation system;
energy conservation; and a host of other social, environmental, economic,
and financial concerns; as well as on air quality—must be evaluated within
a broadly participatory, interactive planning process. Furthermore, the
analyses must be completed expeditiously, in order to meet legislative
deadlines for SIP adoptions and submittals. These combined requirements
necessitate analytic capabilities which produce results quickly and yet
provide accurate information on a wide range of impacts.
-------
1-2
TABLE 1.1
Reasonably Available Transportation-Air Quality Measures
to be Analyzed for SIP Revisions
Transportation System Measures;
• improved public transit (short- and long-range)
• exclusive bus and carpool lanes
• area-wide carpool programs
• private car restrictions
• on-street parking controls
• pedestrian malls
• park-and-ride and fringe parking lots
• employer programs to encourage carpooling and vanpooling, mass
transit, bicycling and walking
• bicycle lanes and storage facilities
• staggered work hours (flexitime)
• road pricing to discourage single-occupancy auto trips
• traffic flow improvements
Vehicle and Equipment Measures:
• inspection and maintenance programs
• alternative fuels or engines and other fleet vehicle controls
• other than light duty vehicle retrofit
• extreme cold start emission reduction programs
• controls on extended vehicle idling
• vapor recovery
1Note: This report focuses on the first category of measures, those
affecting the transportation system. The second catgory of measures are
considered in some of the emissions estimation techniques, however.
-------
1-3
Metropolitan Planning Organizations (MPO's) for the most part are
«
taking the lead in transportation-air quality analysis. Often, however,
the MPOs' analytical techniques center around large computer modelling
systems which originally were designed to investigate the effects of long-
term activity shifts and/or major capital investments in transportation.
Such model systems tend to be too costly, too data-demanding, and too time
consuming for use in analyzing the numerous alternatives to be considered
in transportation-air quality planning. Because the models were not
intended for analysis of transportation operations, management policies, or
small changes in facilities, they often omit variables necessary to study
such measures. Some of the variables which are included enter the models
in ways that cannot respond to the influence of proposed actions, or that
can do so only by receding or reprogramming. Because the models were meant
for regional studies, they frequently are so "aggregate" (i.e., coarse-
scaled) that small or localized change cannot be discerned. The models
usually do not have the capability to focus on a subarea or corridor except
after considerable modification, and then only with a great deal of work.
They rarely distinguish among various socioeconomic groups or other popu-
lation subsamples. Thus, the conventional model systems are not well-
suited for transportation-air quality planning—nor for most transportation
system management (TSM) or transportation-energy conservation planning
efforts, which likewise emphasize quick response analysis of management and
operations policies and small capital investment projects. Methods which
are cheaper, quicker, more flexible, and more responsive are needed.
-------
In recent years, a number of sketch planning tools have been devel-
oped. They range from specialized techniques designed to address a parti-
cular type of measure or impact to general-purpose procedures and method-
ologies, and they cover a smililarly broad range of sophistication and
complexity. In the following chapters, a selection of these sketch plan-
ning tools is reviewed, their applicability to various analysis problems is
evaluated, and the resource requirements for each technique are assessed.
Special emphasis is given to techniques suitable for transportation-air
quality analysis, although most of the methods are more generally appli-
cable to transportation system management planning and transportation-
energy conservation planning. The techniques are grouped into three
categories:
• travel demand analysis methods; methods which focus on user
response to changes in transportation systems;
• facility operations analysis methods; method which focus on capa-
city, speeds, flows, and other operating characteristics; and
• special impacts analysis methods; methods for analyzing emissions
and energy conservation;
However, a case could be made that several of the techniques cover two or
even all three classifications. Within each category, the techniques are
further grouped by their computational approach--manual, calculator pro-
gram, and computer methods. Some relatively complex and costly methods are
included where their potential usefulness or interest warrants attention;
however, the bulk of the handbook is devoted to simpler and cheaper
techniques.
-------
1-5
In the remainder of this chapter, issues in transportation sketch
planning analysis are reviewed, and considerations in designing an analysis
approach are discussed. Chapters 2 through H then present the selected
techniques. In Chapter 5, the discussion returns to technical issues in
the design of an analysis approach. Case studies are provided in a sepa-
rate Volume II.
This handbook is designed with two major types of readers in mind:
• Those who are designing transportation-air quality analysis
approaches, and in doing so must select an appropriate set of
analysis techniques.
• Those who are conducting transportation-air quality analyses,
and require reference material on particular analysis methods.
Both types of readers can obtain the general introduction and required
background to the terminology used in this handbook by heading the remain-
der of Chapter 1. Then, those requiring reference material can use the
relevant parts of Chapters 2 through U to obtain an overview of various
kinds of methods, descriptions of specific methods, and pointers to further
reference material. Also, the Case Studies in Volume II can be used as
extended examples of the use of these methods. Section l.U provides a
guide to the methods discussed in Chapters 2 through U.
Readers who must design specific local analysis approaches should skip
directly from Chapter 1 to Chapter 5, which goes beyond the introductory
material in the remainder of this Chapter, discussing a number of analysis
design issues which must be considered. Then, further information on the
methods found to be potentially useful can be obtained by going back to
Chapters 2 through U and the Case Studies, again using Section l.U as a
guide.
-------
1-6
1.2 Basic Concepts in Sketch Planning
Sketch planning methods span a wide range of sophistication and com-
plexity. The term covers a spectrum of approaches, from the drawing of
qualitative and order-of-magnitude inferences based on experience in
various settings (a case study approach) to use of computerized analysis
packages. The unifying characteristic is that all are quick-response,
expense-saving ways of evaluating the impacts of transportation proposals.
The economies associated with sketch planning can be achieved in any
of several ways:
• by examining only those impacts which typically occur in the
short run (e.g., five years or less)
• by reducing the geographic scale of the analysis, either by
focusing on a particular subarea or corridor or by using large
(i.e. , aggregated) analysis zones
• by focusing on the effects on travel demand (i.e., user
response to transportation system changes), treating facility
operations as given (exogenous)
• by focusing on the effects on facility operations, treating
travel demand as given (exogenous)
• by making other simplifications, such as using average values
for variables, using coefficients or data developed in another
area, or relying on inference and extrapolation.
Such simplifications are justifiable, but they must be treated with
caution. The caution is based on the recognition that transportation
interacts in complex ways with land use and socioeconomic factors. The
characteristics of the transportation system—physical and operating—help
to determine the amount of travel which takes place, which in turn affects
/
the performance of the transportation system. Over the longer run, this
interaction influences choices about household location, the location of
-------
1-7
firms, personal employment, auto ownership, and so on. Nonetheless, the
first-order effects of a transportation measure can be viewed as limited
both spatially and temporally.
Because transportation-air quality planning focuses primarily on the
period between now and 1987, analysis can focus on near-term effects,
checking only to make sure that detrimental longer-term effects are
unlikely. Similarly, when a proposed measure applies to a limited geogra-
phic area, analysis can focus on that area, ignoring (for the most part)
the "spin-off" effects which might occur in distant parts of the region.
Thee simplifications should cause only minor losses of accuracy.
Interactions between demand for transportation and facility character-
istics and operations are more complicated, as they occur continuously and
iteratively. Thus, a complete, detailed analysis would represent the
effects changes in demand would have on facilities' operations, and the
effects changes in physical facilities or their operations would have on
demand, using techniques termed "network equilibrium" or "capacity
restraint". Nevertheless, a proposed measure can be classified as falling
generally into one of two categories--it can affect demand primarily or it
can affect facility characteristics and operations primarily. A number of
analysis tools have been developed for each of these categories: some
emphasize demand, some facilities. The analysis methods which emphasize
demand represent facility characteristics as fixed or exogenous.
Similarly, methods which emphasize facility characteristics generally
represent demand in a simplified way.
-------
1-8
The measures most frequently considered in transportation-air quality
analysis are oriented toward:
• improving the flow of travel, reducing emissions per mile; or
• reducing the amount of travel by auto, by inducing travellers
to walk, bicycle, use transit, or share-a-ride instead of
driving by themselves.
The first type of measure tends to be facility/operations-oriented; the
second type, demand-oriented. Either, however, may entail a physical
change to the transportation system—adding a lane for high-occupancy
vehicles, adding additional buses to a route, or may simply be a policy
change--a11owing right turns on red, eliminating tolls for carpools.
Impact analyses—of emissions; energy conservation; and other social,
economic and environmental concerns—follow from the basic traveller beha-
vior and facilities analyses. Once the changes in travel choices, average
speed, and so on are known, many impact calculations are straightforward.
However, from the perspective of decision-makers, the most pressing concern
may be the incidence of impacts—which areas, or which groups within the
population, are most affected by proposed measures. In order to examine
the incidence of impacts, it is necessary to distinguish among such areas
or population groups. This differentiation has been termed "market
segmentation"; it can be based on a variety of factors, including income,
auto ownership levels, availability of transit or other ridesharing modes,
or trip destination (CBD/suburban). When data are sufficient to permit
such market segmentation, the analysis can be more detailed and accurate
and can reveal the distribution of travel responses and impacts among
different groups or geographic areas.
-------
1-9
1.3 The Range of Sketch Planning Techniques Available
Numerous sketch planning techniques have been developed in recent
years. Three categories of such techniques are presented in this handbook:
• Travel demand analysis methods are those which predict travel-
ler behavior in response to change in the transportation
system. Techniques for trip generation, destination choice,
mode share, and route choice are all reviewed.
• Facility operations analysis methods predict the operating
characteristics of transportation facilities as a function of
changes in capacity and operating policy. Intersection
improvements, signal timing, capacity changes, and flow meter-
ing all are addressed by the techniques discussed.
• Special impact analysis methods focus on particular effects of
transportation changes. Methods for assessing changes in vehi-
cular emissions and fuel consumption are presented.
Within these categories, the techniques are further classified by the
technology used in applying them:
• Manual methods are techniques which utilize worksheets, formu-
las, nomographs and the like to carry out hand calculations, as
well as approaches for making use of data or study results from
other urban areas.
• Programmable calculator methods are adaptations of manual
methods which, by capitalizing on recent developments in inex-
pensive calculating equipment, allow for more detailed and pre-
cise analysis at no significant increase in effort.
• Computer-based methods are model systems for which time and
expense are minimized by making simplifying assumptions or
otherwise limiting the scope of the analysis.
Table 1.2 illustrates how the classification is used in this hand-
book. The methods were selected as representatives of the range of
approaches which have been developed. The various techniques can accom-
modate different amounts and types of data, can be used at different levels
of detail, and require various levels of staff expertise or experience,
-------
TABLE 1.2
Typology of Sketch Planning Techniques
Technology
Demand
Facility Operations
Special Impacts
Manual
worksheet mode choice
quick response urban
travel estimation
techniques
systematic data
analysis
traffic flow formulae
graphical techniques
areawide traffic
engineering method
transfer of experience
emissions worksheets
auto fuel consumption
and operating costs
Programmable
Calculator
HHGEN
2MODE-AGG
3MODE(VAN)-AGG
BUS
BUSPOL
ENERGY
Computer
CAPM
SRGP
transit sketch planning
procedure
TRANSYT
FREQ
MOBILE1
-------
1-11
labor, and other resources. Furthermore, techniques which are applicable
to the full range of transportation-air quality measures have been
included. The handbook thus should be useful in disparate settings and
under disparate conditions.
An index to the methods listed in Table 1.2 is provided in Table 1.3.
For each general class of method, and for each individual method, the
appropriate section is indicated. This table also shows the Case Studies
in Volume II in which the various methods are demonstrated, as well as the
policies illustrated in each Case Study.
To illustrate the use of Table 1.3, consider first the analyst who
wishes to see examples of the analysis of parking programs. Under this
column, both Case Studies I and IV appear. Case Study I illustrates a
number of manual demand and impact methods, and Case Stlidy IV illustrates
SRGP, a computer demand method. Secondly, consider an analyst who wishes
to obtain information on calculator demand methods. The table provides
Section references to the general overview of this class of methods (2.2),
and to the three methods described in detail in this handbook (2.2.1,
2.2.2, 2.2.3). The table also shows that two Case Studies (II and III)
illustrate the use of these methods.
-------
TABLE 1.3
Cross-Reference Table for Analysis Methods and Case Studies
VOLUME II: CASE STUDIES BY POLICY CLASS
METHOD
Manual Demand
Pivot Point Mode Choice
Synthetic Mode Choice
Quick Response Urban Travel
Estimation
Systematic Data Analysis
Calculator Demand
HHGEN - Household Samples
2MODE-AGG - Synthetic Mode Choice
3MODE(VAN)-AGG - Pivot Point Mode
Choice
Computer Demand
CAPM - Community Aggregate
Planning Model
SRGP - Short-Range Generalized
Policy Analysis
Transit Sketch Planning
Manual Facility Operations
Traffic Flow Formulae
Graphical Techniques
Areawide Traffic Engineering
Transfer of Experience
VOL I
SEC-
TION
2.1
2.1.1
2.1.1
2.1.2
2.1.3
2.2
2.2.1
2.2.2
2.2.3
2.3
2.3.1
2.3.2
2.3.3
3.1
3.1.1
3.1.2
3.1.3
3.1.4
Auto
Restric-
ted
Zones
II
II
HOV
Prior-
ities
I
I
I
III
III
III
Traffic
Flow
Improve-
ments
II
II
VII
Transit
System
Improve-
ments
I
I
I, II
III
III
II, III
IV
V
Park-
ing
Pro-
grams
I
I
I
IV
Pric-
ing
Poli-
cies
Carpool/
Vanpool
Incent-
ives
I
I
I
IV
Stag-
gered
Work
Hours
I
I—•
NJ
-------
TABLE 1.3 (Con't)
Cross-Reference Table for Analysis Methods and Case Studies
VOLUME II: CASE STUDIES BY POLICY CLASS
METHOD
Calculator Facility Operations
BUS - Bus Operations
Computer Facility Operations
TRANSYT - Arterial Street Systems
Analyzer
FREQ - Freeway System Analyzer
Manual Impacts
Emissions Worksheets
Auto Fuel Consumption and Operating
Costs
Calculator Impacts
BUSPOL/ENERGY - Bus Emissions and
Fuel Consumption
Computer Impacts
MOBILE1 - Auto Emissions
VOL I
SEC-
TION
3.2
3.2.1
3.3
3.3.1
3.3.2
4.1
4.1
4.2
4.3
4.3
4.4
4.4
_
Auto
Restric-
ted
Zones
II
II
Mi^HI
HOV
Prior-
ities
III
VI
I, III
III
Traffic
Flow
Improve-
ments
VI
II, VII
II
Transit
System
Improve-
ments
III
1,11,111
II
III
Park-
ing
Pro-
grams
I
Pric-
ing
Poli-
cies
Carpool/
Vanpool
Incent-
ives
I
Stag-
gered
Work
Hours
I
f—l
LO
-------
1-14
l.H Choosing the Appropriate Technique
Factors which should be considered in selecting a sketch planning
technique or method include:
• the types of effects the method can represent;
• the adequacy with which various kinds of impacts are addressed;
• the data processing capacity of the method;
• time and budget constraints;
• the skill level of available personnel;
• data availability and quality.
The first test of a method's suitability for a particular analysis
application is whether the method can represent the effects of the
measure(s) under study. For example, a method designed for analyzing how
transit improvements affect transit mode share may be unable to represent
the effects of carpool incentives on transit ridership. The user must
hypotheize about the effects a measure might have, and then select an
analysis technique, or a set of techniques, which can address those effects.
The adequacy with which various kind of impacts are addressed is
another important consideration in selecting a technique. As discussed in
the preceeding section, sketch planning analysis techniques generally focus
on either the travel impacts or the facility operation impacts of a
measure, treating the other impacts in a simplified fashion. The user will
need to determine whether the primary effects of the measure(s) to be
analyzed are on travel demand or on facility operations, and choose an
analysis method accordingly. Note, however, that in many instances
measures will have important effects on both demand and operations. The
-------
1-15
user should consider whether the method provides sufficient representation
of all key impacts; in some cases, it may be necessary to partition the
analysis into separate steps or to use several techniques to obtain reason-
able results.
The suitability of a technique also is a function of its data proces-
sing capacity, in comparison to the amount of data which must be analyzed
to provide the desired level of detail or number of results. From a prac-
tical point of view, for instance, a manual procedure might be a good
choice for examining the areawide implications of improved transit service,
but would be too cumbersome to use in analyzing the improvement on a route-
by-route basis for a large transit system. Conversely, if only one or two
routes are to be affected, it may be inappropriate to use a technique which
can handle, or requires, large amounts of data; a simpler approach may
suffice.
The appropriateness of a methodology to a particular analysis problem
cannot be determined in isolation from such practical considerations as
time, funding, and labor constraints and data availability. Time and
budget limitations often will dictate that the user select a technique
which provides less accurate and reliable results than may be desired in
the ideal. The level of expenditure of time and money on an analysis also
should be compatible with the scale and importance of the measure being
considered. A first look at a proposed measure might best be done with a
relatively fast, general technique, for instance.
The skill level of available personnel constitutes another important
consideration, since there are substantial differences among various
analysis methods with regard to their degree of difficulty. Technical
-------
1-16
skills are needed to apply the computer methodologies: if such skills are
in short supply, one of the calculator or manual methods may be a better
choice. On the other hand, many of the techniques requiring the fewest
calculations do require the most judgment and experience in problem formu-
lation and assessment; these requirements imply more senior-level involve-
ment in the analysis process.
Limitations on the availability of required data often will prove to
be a binding constraint on the amount of analysis which can be performed
and its complexity. The sophistication of any methodology chosen should be
commensurate with the quality of available data. Since the costs of data
collection can amount to a substantial share of total analysis cost, data
requirements should be taken into account in evaluating the cost-
effectiveness of alternative methodologies.
In the following chapters, the selected techniques are reviewed. The
presentation includes:
• a description of the technique's key features
• an assessment of its strengths and limitations
• suggestions on applicability
Among the factors considered are the following:
• measures addressed by the method
• nature of the analysis method—its theoretical or methodological
features
• analysis time frame—short-, medium-, or long-range
• geographic scale—region, metropolitan area, corridor, localized
• behavioral representation—does the model attempt to replicate
travel behavior or facility operation?
-------
1-17
• critical assumptions--what assumptions are made which may have a
significant effect on the results?
• ease of use and understanding—staff requirements to execute
and/or interpret the method
• cost--personnel, computer or other costs required to execute the
method (both initially and for repeated appliations)
• input requirements—data which must be supplied by user
• output--information provided by the method, especially outputs
with specific relevance to air quality analyses
• previous use—is the method well tested and well documented; has
it performed well in the past?
-------
CHAPTER 2 TECHNIQUES FOR ESTIMATING TRAVEL DEMAND IMPACTS
Travel demand analysis asks the question: How will travel behavior
change in response to changes in the transportation system, land use and
development patterns, and population attributes? Thus it examines where
and how much travel will occur, at what times, by what mode, and over which
rbutes.
Transportation-air quality planning focuses on incremental changes in
the transportation system and their effects on travel. Such changes, cumu-
latively and over time, also may contribute to changes in land use and
location and thus to the distribution of economic, social, and demographic
characteristics. Nonetheless, over the time frame of primary interest in
air quality, energy conservation, and transportation system management
planning—a horizon of less than ten years—such secondary, "induced"
changes are likely to be modest. (Note that other forces may change both
land use and population attributes significantly, however.)
The recognition that analysis can focus on travel impacts, with less
attention given to shifts in land use and population, gives rise to a num-
ber of sketch planning techniques for demand analysis. A selection of
manual, calculator program, and computer program techniques is described
in the following pages.
-------
2-2
2.1 Manual Methods
Manual methods of forecasting travel demand can provide quick and in-
expensive estimates of the impacts of transportation measures. Manual
methods require less time, a lower level of staff specialization and less
computing equipment than do many other techniques. On the other hand, the
results they produce almost inevitably are less detailed and precise than
those obtained with a more sophisticated methodology. The manual travel
demand estimation procedures described here are representative of the broad
range of methods which are available and are judged to be the most useful.
The techniques which are included are:
• Worksheet mode choice methods:
Pivot-point mode choice method
- Synthetic mode choice method
• Nomograph/formula techniques:
- Quick response urban travel estimation package
• Systematic data analysis:
- Transfer of data on travel characteristics and elasticities
Inference from previous study results or applications
- Analysis of area-specific data
This variety of methods itself spans a considerable range of sophis-
tication, detail and accuracy. A method could be used by itself, or in
combination with others; for instance, an analysis of area-specific data
could be used to evaluate the potential applicability of results from other
studies, or to assess the reasonableness of transferring an elasticity.
Furthermore, manual methods could be used to produce a preliminary estimate
of a proposed measure's impacts, followed by the application of more com-
plex calculator or computer program techniques if warranted.
-------
2-3
Some caution must be exercised in the application of manual
techniques—particularly the simpler ones. There is a distinct tradeoff
between the simplicity of the method and the amount of judgment it takes to
apply it appropriately. For example, transfer of data or results from one
location to another necessitates a decision on the "similarity" of the two
areas, which usually will require a fairly high level of experience and
knowledge. In some ways, then, the more rigorous methods can be used more
readily, simply because they are more generally applicable.
-------
2-4
2.1.1 Worksheet Mode Choice Methods - Overview
Worksheet modejChoice methods forecast choices of travel mode as
functions of transportation level-of-service and transportation costs. The
models can therefore be used to predict the impact of transportation poli-
cies which affect the level-of-service or cost of one or more transpor-
tation alternatives.
The first worksheet method described in this handbook is a "pivot-
point" method—it predicts incremental changes in mode choice compared to a
"base case" level. The "base case" may reflect existing conditions (e.g.,
current mode shares) or it may represent the conditions expected to be in
existence as a result of other factors such as population growth, new
development, and implementation of transportation measures other than the
one being studied. (For example, the "base case" for 1982 might encompass
a 10 percent increase in population over the 1979 level and a 20 percent
increase in transit mode share as a result of system expansion in the 1979
to 1981 period. Changes in addition to these, such as additional bus
service in 1982, would then cause an incremental change from the 1982 base
case. )
The second worksheet method is a "synthetic" method—it synthesizes
mode shares independent of any observed or base case conditions. It can be
used to generate base case work trip mode shares as well as the impacts of
specific policies.
Both of these worksheet methods are based on a specific mathematical
formulation frequently used in mode choice models—the "multinomial logit"
formulation. The basic multinomial logit form is used in the synthetic
mode choice method. This form is:
-------
2-5
wh ere:
m
U
m
m
m
z
all modes, i
is the probability of choosing a given mode, m
is the utility of mode m
is the base of natural logarithms, 2.718
The incremental form of this model provided by the pivot-point method is:
AU.
P1
m
P e
m
m
L
all modes, i
Pi e
AU.
wh ere:
'm
m
AU,
m
is the revised probability for mode m
is the base probability for mode m
is the change in utility for mode m
In both of these forms, the mode-specific utilities, Uffl, are typically
defined as:
where:
Si
a. and b
im
im
Um = f aim Si + f bim Xim
is the ith socioeconomic variable; e.g., the
traveller's income level or the employment density in
the zone of attraction
is the ith level-of-service variable for mode m;
e.g., the in-vehicle travel time by transit
are estimated coefficients
-------
2-6
When the coefficients of the logit model (or any other travel demand model)
are estimated using observations of individual tripmakers, then the models
are termed "disaggregate" models. This is to distinguish these models from
"aggregate" models—those estimated using the average characteristics of a
number of tripmakers and their trips. Most conventional travel demand
models are aggregate models. In recent years, however, the increased
information available from individual observations has been used to obtain
more detailed models using disaggregate estimation methods.
-------
2-7
a. Pivot-Point Mode Choice Method (1H)
The pivot-point transportation demand estimation method produces esti-
mates of incremental changes in modal shares for drive alone, shared ride,
and transit modes of travel; changes in carpool size and auto occupancy;
and changes in daily vehicle miles of travel in response to individual
transportation measures or combinations of measures. Modal shares are
predicted as a function of the relative "utility" of one or more modes.
The changes in utility are determined on the basis of changes in the
out-of-vehicle and in-vehicle travel time and cost resulting from each
measure. The method is versatile with respect to the degree of market
segmentation (number of population groups), number of zones, and geographic
scale to which it can be applied. Once the analyst is familiar with the
basic calculations required by the method and the base data is collected,
the worksheets normally require only four hours to analyze each measure,
assuming three or four market segments are defined.
Calculation Procedure—The methodology consists of the five basic and
five supplemental worksheets included in Appendix A. Several steps are
required prior to using the worksheets:
• First, the user must determine the appropriate geographic scale
of analysis: regional, corridor, specific facility, etc.
• Second, the user must decide on the number and characteristics of
"market segments" or population subgroups for which separate
analyses will be performed. The analysis population can be
differentiated into a number of discrete, relatively homoge-
neous subgroups, on the basis of such criteria as geographic
location, transit availability, or trip orientation, to more
accurately reflect differences in behavior among specific groups
of travellers.
-------
2-8
• Third, base case modal shares must be estimated for each market
segment, taking into account the specific level-of-service or
costs experienced by each of these subgroups. Methods of
obtaining these base case modal shares are discussed in Section
5.4. Obviously, the method is much easier to use when these data
are readily available.
After these initial three tasks have been performed, the user can begin to
apply the worksheets. The worksheet sequence is represented graphically in
Figure 2.1.
Base data on socioeconomic characteristics and travel patterns are
entered on Worksheet I. Required data items consist of the following:
• Percent of total population analyzed
• Average annual household income
• Average number of non-work auto trips daily
• Base work trip modal shares (drive alone, shared ride and transit)
• Average carpool size
• Average trip length (for work and non-work trips)
• Average daily vehicle miles of travel (VMT)
For policies differentiating between 2-person and 3 or more-person
carpools, additional data items required are:
• Base modal shares for 2-person and 3 or more-person carpools
• Average size of 3 or more-person carpools
Worksheet II summarizes changes in transportation level-of-service and
costs resulting from transportation measures under study. For transit,
shared ride and drive alone alternatives, the following data are required:
• In-vehicle travel time
• Out-of-vehicle travel time (e.g. walk, wait, transfer, pickup,
drop-off)
-------
2-9
FIGURE 2.1
SEQUENCE OF BASIC AND SUPPLEMENTAL WORKSHEETS
I-A. BASE VMT
I. BASE DATA
II-A. CHANGES IN
TRANSPORTATION
LEVEL OF SERV-
ICE BY CARPOOL
SIZE
II. CHANGES IN
TRANSPORTATION
LEVEL-OF-
SERVICE
III-A. ESTIMATION OF
CHANGES IN
CARPOOL SIZE
III-B. REVISED
MODAL SHARES
RESULTING
FROM VANPOOLS
ESTIMATION OF
REVISED WORK
TRIP MODAL
SHARES
IV. ESTIMATION OF
CHANGES IN VMT
REPEAT FOR EACH
POPULATION SUBGROUP
V. SUMMARY OF
CHANGES IN VMT
V-A. SUMMARY OF
CHANGES IN
MODAL SHARES
-------
2-10
• Out-of-pocket travel cost
• For the shared-ride mode only, whether or not its usage is
subject to employer incentives
Revised work trip modal shares for each market segment are estimated
on Worksheet III, based on the calculation of a "change in utility"
function for each mode, which is used to obtain the change from existing
modal shares. The change in utility function is calculated by multiplying
the changes in travel time and cost from Worksheet II by appropriate
coefficients and then exponentiating the product. Default coefficients
are included in the worksheet for this calculation; these coefficients are
transferred from a model calibrated on Washington, D.C. data (5).
User-supplied coefficients may be substituted in their place.
Worksheet IV translates the revised modal shares into changes in
vehicle miles of travel, using base data, including numbers of workers per
household, average vehicle occupancy, and average trip length, from
Worksheet I. Worksheet V summarizes the changes in VMT calculated in
Worksheet IV by market segment.
Note that while Worksheets I, II, and V are used only once for each
policy analysis, Worksheets III and IV are repeated for each market segment
included in the analysis.
Five supplemental worksheets are provided for use in certain special
situations. Each form is numbered so that it corresponds with the basic
worksheet it supplements. Thus, Worksheet I-A can be used to calculate the
base VMT data required on Worksheet I. Worksheet II-A allows the user to
•'•The mathematical form of the mode choice model underlying this
method is described in Section 2.1.1.
-------
2-11
differentiate changes in transportation levels of service by carpool size,
specifically data for 3 or more-person versus 2-person carpools. Revised
modal shares by carpool size are calculated on Worksheet III-A. Worksheet
III-B can be used to estimate the changes in modal shares which would
result from introduction of a vanpool program. The average areawide
changes in the use of each transportation mode can be estimated using
Worksheet V-A.
Suggested Applications—The pivot-point travel demand method can be
used to analyze the effects of virtually any measure which could induce
changes in work trip mode choice as a result of changes it makes in the
travel time or cost associated with a transportation alternative. It can
be applied in the analysis of such measures as priority treatment of high-
occupancy vehicles, traffic flow improvements, transit fare changes,
pricing policies to discourage low occupancy vehicles, and carpool/vanpool
incentives. Thus, the method is applicable for at least some aspects of
the analysis of all major transportation-air quality measures, given
sufficient information on transportation supply conditions, e.g. changes in
travel time.
To apply the worksheets, it is necessary to represent the measure
under study in terms of changes in travel time and/or cost. For some
measures, e.g. transit fare policies, this is straightforward. However,
for some measures (e.g. the designation of an exclusive bus lane), the user
will be required to estimate the changes in travel time or cost through use
of a highway facility operations model or some other procedure. Case
Study I in Volume II demonstrates the use of the manual worksheets to
-------
2-12
analyze the effects of a high-occupancy vehicle lane; this example
illustrates how the impacts of a program measure are translated into terms
which can be used in the worksheets.
The method is not appropriate for some analyses, particularly those
where travel time or cost changes are not major impacts. For instance:
• Auto restricted zones (ARZ's), in which the impact on trip
distribution must be determined externally from the model
• Parking programs, in which supply constraints can be of major
importance
• Variable work hours, in which the major focus of analysis
concerns the number of employees who will be affected and their
change in work hours.
-------
2-13
b. Synthetic Mode Choice Method
The synthetic mode choice method, unlike the pivot-point methodology,
can be used to estimate base case or existing work trip modal shares, as
well as specific policy impacts. However, the synthetic model is more
difficult to apply. In most cases, it is necessary to collect socio-
economic and service level data by zone or by market segment, which is not
required in the case of the pivot-point method.
The .method entails manual application of the logit model described in
Section 2.1.1. Modal shares are estimated as a function of the travel time
and cost assocated with each mode. The worksheets in their present form
can be used to forecast the split between two alternative modes. However,
the technique can be adapted to include other modes.
Calculation Procedure—The method involves the use of the three
worksheets included in Appendix B. Worksheet I, which contains the actual
model, is used to estimate the passenger volumes for each of the two modes
considered. A separate worksheet is required for each market segment
studied. A utility function is calculated as the product of empirically
estimated coefficients and in-vehicle and out-of-vehicle travel time and
user cost variables, and exponentiating the resulting utility function.
The worksheet includes default coefficients, which can be replaced by
values supplied by the user. Similarly, the general form of the worksheets
can be used with alternative logit models, especially if these have been
developed for the metropolitan area being studied. The utility function
yields the fractional shares for each mode, which are then multiplied by
the total market segment volume to yield the actual volume of passengers in
the market segment which use each mode.
-------
2-14
Worksheet II is for analysis of variations in transit fare. The
utilities of auto and transit alternatives are calculated as in
Worksheet I, except that the transit fare term is separated from the other
components of the calculations, which are subtotalled such that the value
of the transit fare term can easily be varied and the total utility
adjusted with a minimum of recalculation. A similar worksheet could be
developed for analyzing the impact of travel time variations.
Worksheet III can be used to organize the information necessary to
determine the level-of-service data required as input to Worksheets I and
II. The various components of in-vehicle and out-of-vehicle time and cost
for each alternative are entered on the worksheet and the necessary calcu-
lations performed.
Suggested Applications—The synthetic manual mode choice worksheets
provide an efficient and accurate manual method for estimating base case
modal shares, given adequate household and service level data by zone and/
or other market segment. However, a procedure with reduced data require-
ments may be more appropriate for many applications.
If the synthetic manual mode choice method is used for estimation of
base case modal shares, it is then relatively easy to apply the model to
forecast the impacts of measures resulting in variations in fare or service
levels. However, application of the method generally is more difficult
than the use of the pivot-point methodology because the worksheets must be
modified in many cases, and because its data requirements are greater. It
may frequently be easier to use the pivot-point methodology to forecast
policy impacts even if base case conditions are estimated with the
synthetic model.
-------
2-15
Because policies are represented in the synthetic model as revised
travel costs and in-vehicle and out-of-vehicle travel times, in some
instances it will be necessary to use facility operation methods, such as
those discussed in Chapter 3, to translate the effects of a particular
policy into measurable revisions in travel time which can be used as input
to the synthetic demand model.
-------
2-16
2.1.2 Nomograph/Formula Techniques—Overview
Nomograph/formula techniques merit consideration as sketch planning
analysis tools for several reasons. The worksheet methods presented in
this handbook directly estimate only work trip mode choice; when estimates
of trip generation, distribution, or route assignment also are needed,
nomographs or formulas can provide quick, if approximate, results. Also,
although the nomograph/formula techniques are relatively imprecise for
forecasting policy impacts, they are a simple way of estimating "base case"
conditions, e.g. existing modal shares. A user may prefer them over the
synthetic modal split worksheet when base case estimates can be approxi-
mate. Using nomograph/formula techniques to develop base case data along
with pivot-point worksheet models to forecast the impacts of policy changes
can be very effective from the standpoint of both forecasting accuracy and
ease of application.
-------
2-17
a. Quick Response Urban Travel Estimation Techniques (18)
These techniques, developed as part of the National Cooperative
Highway Research Program, provide manual methods for carrying out the four
steps of the conventional urban transportation modelling process (trip
generation, trip distribution, mode choice, and traffic assignment), as
well as guidelines for determining auto occupancy and time of day
distributions for different trip purposes. Each of the estimation
procedures may be used alone; or they may be used in combination as an
integrated travel forecasting system.
The procedures are applicable to areawide, corridor, or site analysis,
with varying levels of computational detail provided for each of the
steps. They utilize data and parameters drawn from typical urban settings,
in tabular and nomograph form, and in the case of the mode choice
procedure, travel demand models estimated on Washington, D.C., and Atlanta
data.
The time required to use the techniques depends on the particular
policy being analyzed, the impacts of interest, and the level of detail and
accuracy desired in the analysis. An illustrative evaluation of residen-
tial development incorporating trip generation, distribution, assignment,
mode split, and auto occupancy calculations required 60 hours of analyst
time, not including data collection and preparation.
Like the computer-based travel demand modelling approaches on which
they are based, the nomograph techniques are not capable of analyzing some
transportation measures. Specifically, the mode choice procedure predicts
only transit and auto shares and thus is unable to capture the effects of
-------
2-18
ridesharing incentive programs. Auto occupancy is predicted independent of
travel level of service based on city size, trip purpose, time of day,
income, and parking cost; income and parking cost are expressed only in
relative terms (low, medium, high).
Although the data requirements for the models are flexible, in
general, more data will be required to support an equivalent level of
analysis with the nomograph technique than the pivot-point models. However,
the nomograph method offers somewhat different analysis capabilities than
the pivot-point model.
The following paragraphs briefly outline the features of the principal
component of the nomograph techniques. Detailed descriptions of the tech-
niques, sets of tables and nomographs, and sample calculations for each
procedure and the model system as a whole are available (18).
Calculation Procedure--The estimation of trip generation involves the
use of one of three tables, the choice depending on the scope of the
analysis, availability of data, and level of detail and accuracy desired
for the analysis:
• Table 2.1 provides trip generation rates for specific trip
generators. Vehicle trip rates per day, peaking factors, typical
auto occupancy, and typical transit share of person trips are
reported for a large number of specific generator types ranging
from single-family housing to hospitals.
• Table 2.2 contains estimated household person-trips classified by
urban area population, household income, auto ownership, and trip
purpose. This table is useful for estimating total travel in an
urban area, but may also be used to develop estimates of total
tripmaking to and from analysis zones given the distribu- tion of
population and income or auto ownership in each zone.
-------
2-19
TABLE 2.1
Average Vehicle Trip Rates and Other Characteristics of Generators
>ar**at «J«*3i8S«i.
CEHERATOR*
Residential
Single F<""4ly
1 Du/acre
2 Du/acre
3 Du/acre
1* Du/acre
5 Du/acre
Medium Density
(Duplex,
Townhouse*
etc.)
5 Du/acre
10 Du/acre
15 Du/acre
Apartments
15 Du/acre
25 Du/acre
35 Du/acre
50 Du/acre
60 Du/acre
Mobile Home
Park
5 Du/acre
10 Du/acre
15 Du/acre
Retirement
Community
10 Du/acre
15 Du/acre
20 Du/acre
Condominiums
10 Du/acre
20 Du/acre
30 Du/acre
Planned Unit
Develop .
5 Du/acre
15 Du/acre
25 Du/acre
Miscellaneous
Service
Station
Race Track
Pro-Baseball
Military Ease
Bratfaifc; JSsCSMBTSa!
VEHICLE TRIPS0 TO t FROM
PER DAT PER
WELLING
irarr ACRE
9.3 9.3
9.3 IS. 6
10.2 30.6
10.2 fcO. 8
9.1 "»5-5
7.0 35.0
7.0 70.0
7.0 105.0
6.0 90.0
6.0 150.0
6.0 210.0
6.0 300.0
6.0 360.0
5-5 27.5
5-5 55.0
5-5 82.5
3.5 35.0
3.5 52.5
3.5 70.0
5.9 59.0
5.9 118.0
5.9 177.0
7.9 39.5
7.9 118.5
7.9 197.5
.SEE INDIVIDUAL GENERATOR BELOW
Station Pump
~W~ 133
Seat Attendee
0.61 1.08
0.16 1.18
Military Civilian Total
Personnel Employees Employeei
2.2 7.1 1-8
lOOOsq.ft.
GFA ifoployee Acre
PERCERT TRIPS IH
HOUR SHOWH
L.M. P.M. PEAK HR.
'EAR PEAK OF QEH.
8.0 10.8 10.8
8.0 10.8 10.8
8.0 10.8 10.8
8.0 10.8 10.8
8.0 10.8 10.8
8.0 10.8 10.8
8.0 10.8 10.8
8.0 10.8 10.8
7.9 10.8 10.8
7.9 10.8 10.8
7.9 10.8 10.8
7.9 10.8 10.8
7.9 10.8 10.8
8.3 10.8 12.5
8.3 10.8 12.5
8.3 10.8 12.5
.2.1 12.1 12.1
.2.1 12.1 12.1
.2.1 12.1 12.1
7.1 7.1 7.1
7.1 7.1 7.1
7.1 7.1 7.1
.0.1 10.1 10.1
.0.1 10.1 10.1
.0.1 10.1 10.1
1.5 3.0 U.O
TYPICAL
AUTO
3CCUPAHCY
1.62
1.62
1.67
1.67
1.62
1.57
1.57
1.57
1.56
1.56
1.56
1.56
1.56
1.5*
1.5*
1.5*
1.U8
1.U8
1.U8
1.56
1.56
1.56
1.58
1.58
1.58
1.55
2.05
2.05
1.U2
TTPICAL
% TRANSIT
OF TOTAL
PERSOJ
TRIPS*
3.2
3.2
3.2
3.2
3.2
5.6
5.6
5.6
12.*
12.*
12.1.
12.*
12.lt
1.0
1.0
1.0
6.0
6.0
6.0
9-0
9-0
9-C
7.1
7.1
7.1
Source: (18)
-------
2-20
Table 2.1 (Continued)
D
GENERATOR
Free Standing
Supermarket
Discount Store
Discount Store
vith Super Mkt
Department
Store
Auto Supply
Hew Car
Dealer
Convenience
(21* hrs.)
(15-16 hrs.)
Shopping Center
Regional (over
1 million
sq. ft.)
(1/2-1
million sq.ft
Community
(100,000-500,001
sq. ft.)
Neighborhood
(under 100,000
sq. ft.)
Central Area
(High Dens.)
Industrial/
Manufacturing
General Manu-
facturing
Warehouse
Research/
Develop.
Industrial
Park
General Light
Ind.
All Industry
Avg.
Offices
General
Medical
Governmental
Engineering
Civic Center
Office Park
Research Cente
Restaurants
Quality
Restaurant
Other Sit-
Dovn
Fast Food
Banks
Parks & Recre-
ation
Marina
Golf Course
Bowling
Participant
sports
City Park
County Park
State Park
Wilderness
ParV
VEHICLE TRIPS6 TO & FROM
PER DAY PER
SEE INDIVIDUAL GENERATOR BELOW
135.3 - 1000
50.2 57.2
81.2 30.3
36.1 32.8 900
88.8
Wt.3
577.0
322.0
33.5 30.9 580
3lt.7 20.1* 370
1*5-9 20.6 330
97.0
1*0.0 - 900
It. 2 2.3 1*0.5
5-3 U.l* 67.5
5.1 2.U 60.6
8.8 3.9 71.9
5.5 3-2 52.U
5-5 3.0 59.9
11.7 3.5 Ili5
63-5 25.0 1|26
It8.3 12.0 66
23.0 3.5 282
25.0 6.1 33
21.0 3.3 277
9-3 3.1 37
56.3 - 200
L98.5 - 932
533.0 - 1825
.388 75
259.0 18.5
3U. 2 " 7- It
296.3
26.5
60.0
26.5 5-1
6l.l 0.6
0.01
PERCENT TRIPS IS
HOUR SHOWN
A.M. P.M. PEAK HR.
PEAK PEAK OF GEH.
0 8.7 12.6
0 5-1 9-7
0 6.9 11.1
_ _ _
...
_ _
• _ .
_
1.9 9-7 11.5
2.8 9.6
11.2 11.3
3.3 11.5 12. U
_ _
I8.lt 19-3 32.2
12.7 32.2
21.1. 20. U
13.2 lit. 7
21.1 20. U 21.2
15.8 19.1*
20.7 19-1
8.5
8.5 16.0
16.9 Hi. 6
9.0 ll.l*
16.9 Ik. 6
16.0 18.5 20.2
1.8 6.0 12.5
29.0 6.1*
16.0 5.7
-
- -
- -
-
-
.
. • .
.
-
TOPICAL
AUTO
OCCUPASCT
1.6U
1.6k
1.61.
1.6U
1.6U
1.6k
1.6k
1.6U
1.6U
1.61.
1.61*
1.61*
1.61*
1.33
1.25
1.33
1.33
1.33
l.UO
1.35
1.U5
1.35
1.35
1.35
1.35
1.35
1.93
1.93
1.93
1.U5
2.05
2.05
2.05
2.05
2.05
2.05
2.05
2.05
TTPICAL
% TRANSn
OF TOTAI
PERSON
TRIPS4
1
1
1
2
1
1
1
1
3
3
3
3
12
5
5
5
5
5
5
5
5
5
5
5
5
5
3
3
1
-
_
_
•
_
_
_
_
-------
2-21
Table 2.1 (Continued)
flTTH L'U A'PA 7J
\JLtu CtTlA IV n
Parks & Kecre-
•tlon (cont'd)
National Monu-
ment
Ocean Front
Lake/Boating
An1m«1 Attrac-
tions
Hospitals
All Categories
General
Childrens
Convalescent
University
Veterans
Nursing Home
Clinics
Educational
All Categories
Four Year Univ
Jr . College
Secondary
School
Elementary
School
Combined Elem
Sec.
Libraries
Airports
General Avia-
tion
Commercial
Betel/Motel
Hotel
Motel
Resort Hotel
VEHICLE TRIPS0 TO & FROM
PER DAT PER
OOOsq.ft.
FA EMPLOYEE ACRE
11.9
21.6
3.6
72.2
Staff Bed Acre
6.1 lit. 8 1*0
5.9 !"*:§ -
10.1 25.2
1«.5 3.2
7.8 37.0
2.2 3.8
2.7
5.9
Student Staff
1.8 13.6
2.5 9-8
1.5 28.2
1.1* 19-9
0.6 11.7
0.8 11.8
in. 8 51.0
Take-Off/ Employee Acre
Landing
2.5 6.5 3.6
11.8 16.8
Room Employee
10.5 11.3
9.6 10.6
10.2 10.3
PERCENT TRIPS IN
HOUR SHOWN
A.M. P.M. PEAK HR
PEAK PEAK OF GEN.
— _ _
— • -
_ _ _
- - _
_ _
11.7
18.0 9.0
-
_
12.5 10.5
11.0 16.5
5.2 7.8 13.3
.
_ - -
11.0 9.0
11.5 7.5 11.9
11.5 >».9
31.1* 2.0
_ _ _
16.0
11.8 10.5 15-7
9-7 17.3
7.9 5-7 8.3
6.7 5-9 9.0
2.6 6.8 7.8
TYPICAL
AUTO
OCCUFANCT
2.0:
2.05
2.05
2.05
1.1*0
1.1*2
1.1*2
1.1*2
1.1*1
1.32
1.1*0
1.1*0
1.1*0
1.1*0
1.55
1.55
1.55
1.55
1.55
1.52
1.52
1.56
1.56
1.93
TYPICAL
% TRANSIT
OF TOTAL
PERSON
TRIPS'1
_
_
_
_
-
17
17
10
10
10
10
10
-
13
13
__
It
1* -e
iJ
6
1
3
2
0
0
The trip rates given are based on a limited number of studies and thus must be
used with caution. The ITE Trip Generation Report(6) provides current data which
is also periodically updated. The vehicle trip rates include external-internal and
internal-external trip ends at generators as well as trucks, taxis and bus.
Most of the generators examined are located outside the central business districts
of cities. The trip rates may thus be inapplicable to sites located within the
dense urban core, particularly in large cities. Variations in generation rates
may also exist because of the location of the generator either within a metropolitan
area or outside that area.
The vehicle trip rates presented are actually volumes into and out of the
site. As such, they may include some trips that would be passing the
site on the adjacent street system, in any case, while making a trip for
another reason, and they are induced to stop for impulse or convenience
shopping, personal business or to drop off or pick up a passenger. The
proportion of these trips has not been identified. Note also that ranges
in trip rates can be expected and these can vary depending upon local
conditions.
The typical transit f shown has a wide range of variation based on loca-
tion within an urban area, level of service provided, etc., and as such,
should be used only to provide gross approximations.
Does not include school bus transit.
-------
2-22
TABLE 2.2
Detailed Trip-Generation Characteristics
URBANIZED AREA POPULATION: 50,000-100,000
Cncone Range
f 1970 *
1 (000'*)
0,3
M
--5
5-6
6-7
7-8
8-9
9-10
10-12.5
12.5-15
15-20
20-25
25+
Average
Arg Autos
Per HH*
0.56
0.81
0.88
0.99
1.07
1.17
1.25
1.31
1.U7
1.69
1.85
2.03
2.07
1.55
Average
Daily Person
Trips Per HHe
k.5
6.8
8.1*
10.2
11.9
13.2
ait.it
15.1
16.1.
17-7
18.0
19.0
19-2
llt.l
% HH by Autos Owned
0
53
32
26
20
15
11
8
6
3
2
2
1
1
12
1
39
58
61
62
6U
6U
62
60
k9
38
28
21
19
1»7
2
7
10
12
17
20
23
28
32
It It
52
57
58
59
35
3+
1
1
1
1
1
2
2
2
3
8
13
20
21
6
Average Daily Perse
Per HH by Bo of Aut
0
2,0
2.2
2.6
3.0
3.0
3.5
U. 8
5.5
6.2
6.1
6.0
6.0
6.0
1.. 6
1
6.5
8.0
9.5
11.0
12.5
13.3
1U.O
lit. 3
15.0
15.0
13.5
13.0
12.5
12.6
2
11.5
13.0
llt.5
15-5
16.5
17.0
17.5
17.5
18.5
19.0
19.5
20.0
20.0
17.2
n Trips
os/HH
3+
12.5
15.0
16.5
18.0
19-5
21.5
22.5
2U.O
25.5
25.5
23.0
23.0
23.0
21.lt
% Average Daily
Tries bv rurnoq
21
21
21
IB
18
16
16
16
15
1U
13
13
13
16
HBNW
57
57
57
59
59
61
61
61
62
62
62
62
62 '
61
Person
HHB
22
22
22
23
23
23
23
23
23
2k
25
25
25
23
1970 $
(OOO's)
0-3
3-U
lt-5
5-6
6-7
7-8
8-9
9-10
10-12.5
12.5-15
15-20
20-25
25+
Wt. Avg.
Avg Autos
Per HHd
0.1*9
0.72
0.81
0.9lt
1.01
l.llt
1.25
1.3U
1.50
1.65
1.85
2.01
2.07
1.55
URBA
Daily Person
Trips Per HI?
lt.0
6.8
8. It
10.2
11.7
13.6
15.3
16.2
17.3
18.7
19.6
20. It
20.6
lit. 5
WIZED AREA POPULATION:
% HH bj
0
57
36
29
21
17
12
9
6
It
2
2
1
1
lit
r Autc
n
37
56
61
65
66
65
61
58
50
ItO
28
20
19
US
>s Ovi
6
8
10
13
16
21
28
33
ItO
51
57
61
59
33
>edb
0
0
0
1
1
2
2
'3
6
7
13
18
21
6
100,000-250,000
Averag
Per HH
1.0
1.7
2.5
3.5
It. 5
5.U,
5.8
6.3
6.8
7.0
7.2
7.5
7.5
5.1t
by No
7-5
9.2
10.2
ll.U
12.5
13.8
15.0
15.8
16.0
16.0
15.0
15.0
15.0
13.7
Perso
, °f *^
10.5
13.3
1U.5
1U.5
15.6
17.0
17.5
18.0
19.0
20.U
21.0
21.0
21.0
18.1.
n Tri-os
tos/me
13.8
16. It
17.6
19.0
20.5
22.2
23.0
23.5
2U.5
25.0
25-5
25.5
25.2
22.lt
% Average
Trips t
20
22
22
22
20
20
20
19
19
19
18
18
18
20
Daily I
v Purpo
63
60
58
58
58
57
57
57
57
56
56
55
55
57
'erson
[£
17
18
20
20
22
23
23
2lt
2lt
25
26
27
27
23
Source: (18)
-------
2-23
Table 2.2 (Continued)
llRRANTZm AREA POPULATION: 250.000-750,000
IncomeRange
1970 $
( 000 ' s )
0-3
3-1.
»t-5
5-6
6-7
7-8
8-9
9-10
10-12.5
12.5-15
15-20
20-25
25+
Wt. Avg.
Avg Autos
Per HHd
O.U7
0.77
0.88
1.01
1.10
1.2U
1.33
l.Uo
1.58
1.72
1.88
2.0U
2.08
1.1*1
Average
Dally Person
Trips per HHe
3.3
5.8
6.9
8.1*
9.5
10.9
11.7
12. U
13.5
1U.6
15-5
16.0
16.2
11-8
f HH by Autos Owned*
0
58
38
29
20
lit
8
6
b
2
2
2
1
1
12
1
37
50
57
62
65
6U
60
57
k6
36
26
20
17
1*1*
2
5
10
12
16
19
25
30
35
1*6
53
58
59
61
37
3+
0
2
2
2
2
3
1*
k
6
9
lit
20
21
7
Per HH by Ho. of Autos /HHC
0
1.1*
2.0
2.5
2.9
3.5
it.O
it. 6
5.2
5.5
5.5
5.2
5.0
5.0
U. 3
1
5.6
T.k
8.0
9.0
9-7
10.6
11.0
11.2
11.3
11. l»
11.5
12.0
12.0
10.0
2
9.3
10.8
11.5
12.1
12.7
13.5
lit. 2
1U.8
15.6
16. U
16.7
16.7
16.7
ll».U
ft,.,.
9.3
11.1
11.9
12.7
13.5
ll».l»
15.3
16.2
17.6
18.8
19.1
18.6
18.6
15.8
Jf Average
Tri-Ds V
HBW
10
13
21
22
22
20
20
20
20
20
' 20
20
20
20
Daily !
IV PuT-no
HBNW
67
6U
57
56
55
55
55
55
55
53
52
50
50
55
"erson
r£
1JHB
23
23
22
22
23
25
25
25
25
27
28
30
30
25
URBANIZED AREA POPULATION: 750,000-2,000,000
;ncomeRang<
1970 $
(OOO's)
0-3
3->»
U-5
5-6
6-7
7-8
8-9
9-10
10-12.5
12.5-15
15-20
20-25
25+
Wt. Avg
Avg Autos
Per TST
O.U7
0.68
0.78
0.81*
0.95
1.06
1.16
1-25
1.1*1
1.60
1.77
1.95
2.02
1.31
Average
Daily Person
Trips per HH
1.9
3.7
U.5
5.1
5.8
6.5
7.2
7.7
8.5
9.1*
9-9
10.6
11.0
7.6
% HH by Autos Owned*1
0
58
UO
32
28
22
16
12
9
5
2
2
2
2
15
1
37
52
58
60
62
63
63
61
56
1*5
35
21*
20
1*8
2
5
8
10
12
15
20
23
27
3U
1*6
51
56
58
32
3+
0
0
0
0
1
1
2
3
5
7
12
18
20
6
Average Daily Person Trips
Per HH by No. of Autos/HHc
0
0.8
2.0
2.5
2.7
2.9
3.0
3.2
3.3
3. U
3.6
3.6
3.7
li.O
3.1
1
3.2
It. 5
5.0
5.6
6.1
6.5
6.9
7.1
7.5
7.5
7.6
7.0
7.0
6.5
2
5.7
7.0
7.5
8.0
8.6
9-2
9-5
9.9
10.1
10.7
10.7
11.1
11.3
9-5
3+
7.3
9.2
10.0
11.0
n.6
12.2
12.6
13.0
13.6
lit. 2
1U.6
1U.8
11*. 8
12.6
% Average Daily Person
Trips by Purpose*
HBW
1>*
22
28
28
28
27
27
27
26
25
2U
21*
23
25
HBNW
66
59
53
53
53
53
53
5lt
53
53
53
53
5>t
5lt
NHB
20
19
19
19
19
20
20
19
21
22
23
23
23
21
-------
2-214
• Table 2.3 contains trip production and' attraction estimates by
urban area population; i.e., for urban areas of different popula-
tion categories, factors for determining external tripmaking
levels, and relationships for determining trip attractions by
purpose for analysis areas based on employment, retail activity,
and population. The trip production and attraction estimates may
be combined to produce a production-attraction matrix for use in
the second step of the travel demand analysis procedure.
Trip distribution produces zone-to-zone tripmaking (trip interchange)
estimates. A production-attraction matrix developed using Table 2.3 of the
trip generation phase, or supplied by another source (such as previous
travel demand estimation efforts), forms an estimate of the "size" of each
zone. Interchange estimates are then based on this "zone size" and the
ease of travelling between the zones, measured by a "friction factor" or
"impedance." The approach developed for manual estimation has been
streamlined by:
• shortcut calculation of zone-to-zone impedance based on airline
distance.
• a set of friction factor nomographs for four urban area sizes and
three trip purposes (home-based work, home-based non-work, and
non-home-based).
• simplified worksheets for calculating trip interchanges.
The only required inputs are a production/attraction matrix and a map of
the study area for determining airline distances between zones. Manual
calculations are expected to be manageable using this technique with up to
80 zones--a 34-zone distribution was conducted in under 26 person-hours.
(5) for an in-depth discussion of the Gravity Model which
provides the basis for this trip distribution procedure.
-------
2-25
TABLE 2.3
Trip-Generation Parameters
I PART A - TRIP PRODUCTION ESTIMATES 1
Urbanized
Area
Population
$0,000- 100,000
100,000- 250,000
250,000- 750,000
750,000-2,000,000
Average
Daily
Person
Trips
Per HIT
lU
1U
12
8
% Average Daily
Person Trips
by Modeb
Public
Transit
2
6
8
13
Auto
Passenger
UO
30
31
30
Auto
Driver
58
6U
61
57
% Average Daily
Person Trips
by Purpose*
HEW
16
20
20
25
HBNU
61
57
55
5U
NHB
23
23
25
21
Auto Person Trips
as a t of Total
Person Trips0
HEW
96
88
8U
7U
HBNW
99
97
96
93
NHB
98
9U
92
86
Auto Driver Trips
. as a % of Total
Person Tripsd
HEW
70
64
62
56
HBNW
54
54
54
53
NHB
68
66
64
60
|PART B - USEFUL CHARACTERISTICS FOR TRIP ESTIMATION |
Urbanized
Area
Population
50,000- 100.000
100.000- 250,000
250,000- 750,000
750,000-2,000,000
External Travel Characteristics
% of Total External
Trips Passing
Through Area
21
15
10
It
% of Total External
Trips to
the CBD*
22
22
18
12
Total Areavide
Truck Trips
as a % of
Areavide Auto
Driver Trips'
27
17
16
16
PART C - TRIP ATTRACTION ESTIMATING RELATIONSHIPS
(All Population Groupings for either Vehicle or Person Trips)
TO ESTIMATE TRIP ATTRACTIONS FOR AN ANALYSIS AREA
HBW Trip Attractions • Fx Q..7 (Analysis Area Total
/Analysis Area\
HBHW Trip Attractions • P, 10.0 1 Retail I +
\ Employment /
/Analysis Area\
NHB Trip Attractions • Pj 2.0 1 Retail 1 +
\ Employment /
Where: Fj_, P2 and F3 are areavide control factors.
TO DEVELOP AREAWIDE CONTROL FACTORS, USE:
. m Areavide Productions for HBW Trips
1 1.7 (Areavide Total Employment)
- Areavide Productions for HBNW Tries
2 T~ 1 Areavide \ / Areavide \
llO.Ol Retail 1 + 0.5 [Non-Retail] +
[_ \Eaploynent/ \Employment/
- m Areavide Productions for NHB Trips
3 |~ / Areavide \ / Areavide \
2.0 I Retail } + 2.5 1 Non-Retail j +
|_ \pnploymenty ytaploymeny
. USE:
Employment }J
/Analysis AreaX
0.5l Non-Retail 1 +
\ Employment /
/Analysis Area\
2.5 1 Non-Retail 1 +
\Eaployment J
/AreavideC")
1.0 (Dvelling)
\ Units /J
/AreavldeVl
0.5 (Dvelling)
\ Units /J
1
0
/Analysis Area
1.0 I Dvelling
\ Units
Analysis Area
0.5 I Dvelling
\ Units
Source: (18)
-------
2-26
Mode shares can be predicted for two modes: auto and transit. No
provision for estimating carpooling activity is incorporated into the
procedure. For each origin-destination pair, three steps are required
to determine the share of trips made by transit and auto:
• Determine auto impedance—auto impedance is treated as a function
of airline distance, average operating speed, parking cost and
auto operating cost per mile. A set of nomographs is provided
for determining zone-to-zone auto impedance.
• Determine transit impedance—transit impedance is a function of
fare, airline distance (which must be adjusted if a transfer is
involved in making the trip), and access mode (auto or walk).
Two nomographs (for the auto and walk access modes) are provided
for determining transit impedance.
• Determine transit mode share—based on the impedances for trips
between each zone pair by both transit and auto, determine the
transit mode share. Nomographs are provided for home-based work,
home-based non-work, and non-home-based trips.
Several rules of thumb (based on current experience in the transit
industry) for determining transit patronage are also included for quick
estimation purposes.
In order to conduct the mode choice analysis, the following items are
required:
• An area map for determining airline distance between zones
• A map of the existing or proposed transit system for use in
determining transit airline distance (taking into account
necessary transfers)
• Transit fares
• Estimated auto operating costs per mile
• Average zone-to-zone highway speeds, which may be determined
using nomographs (based on airline distance, urban area popula-
tion, and percent of freeway and arterial mileage travelled
between zones) or knowledge of the study area.
-------
2-27
Because this procedure uses distances rather than travel time, it
cannot be used to evaluate the impact of transit headway changes. However,
an alternative approach for determining impedance for the transit mode
which does incorporate travel time is provided; impedance may be determined
by the following equation:
FARF
Im = IVTT +2.5 OVTT +
where:
I™ = Transit impedance
IVTT = In-vehicle travel time
OVTT = Out of vehicle travel time (wait time)
FARE = Transit fare
HHI = Household income
This impedance measure then could be used to examine transit mode shares
under various operating policies.
The traffic assignment procedure involves forecasting passenger and
vehicle volumes on the transportation system given the results of the
previous modelling steps. Three levels of traffic assignment capability
are provided in the procedure:
• Traditional traffic assignment
• Traffic generation and decay
• Traffic diversion and shift
The traditional traffic assignment methodology is the common "all or
nothing assignment" process: all traffic between two zones is assigned to
the shortest path available. In order to keep the analysis manageable, the
-------
2-28
number of zone-to-zone interchanges must be kept relatively small. A tech-
nique for determining turning movements also is incorporated into the
traditional methodology.
Traffic generation and decay is used to analyze a specific traffic
generator and its effects on the nearby street network. Average street
volumes are read from one chart, and the volumes due to the proposed
traffic generation (e.g., a shopping center) are added to the base volume
according to the following characteristics of the generator:
Density and size
Level of service
Auto ownership
Transit availability
Residential/non-residential mix
Freeway diversion factors
Traffic diversion and shift is a simple and quick methodology which
enables the user to determine the effects of changes in highway level of
service on highway segment volumes. Traffic diversion due to improvements
on a facility is an example of the type of impact analysis capability
provided by the diversion and shift graphs.
Two additional information sources related to travel demand estimation
are provided in the procedures--current average automobile occupancy
characteristics and current average time of day characteristics. Sets of
tables are provided which relate auto occupancy and the time distribution
of travel to urban area size, trip purpose, land use characteristics and
other factors. These averages, however, are not responsive to policy
measures designed to affect ridesharing activity or the hourly distribution
of travel.
-------
2-29
Suggested Applications—The set of quick response techniques would be
most useful for:
• the development of "base case" data if this information is not
available from other sources for given urban area
• adjustment of available base data to represent a future analysis
year, taking into account planned development projects and urban
growth patterns
• traffic volume impact analysis of specific site developments and
expected residential growth pattens.
The method is not fully satisfactory for:
• analysis of measures designed to encourage shifts from single-
occupant autos to the transit and shared ride modes.
If the route choice and traffic volume impacts of a particular measure are
of interest, the manual pivot-point analysis could be used in place of the
mode choice analysis procedure provided in this set of procedures.
Case Study III in Volume II shows the use of these procedures as part
of the process of specifying a base case for an urban corridor analysis.
-------
2-30
2.1.3 Systematic Data Analysis—Overview
Several procedures in common use for estimating both base case
transportation conditions and probable impacts are described below. These
procedures do not forecast with the accuracy of the methodologies discussed
previously, and they rely to a greater degree on the judgment of the
analyst. However, they usually require fewer data than the methods
described above; and some of them can provide the kind of information that
many decision-makers request—evidence of what happened elsewhere or what
local conditions are, for example. The procedures should be used with
care, since they rely on the accuracy of assumptions of similarity or
correlation. They may be best suited to preliminary feasibility studies and
other situations where a great degree of precision is not required. There
also are instances when coi^traints on time, data, labor and/or funds will
dictate the use of these simple approaches.
Approaches of this type are discussed in this section under the
following headings:
• Transfer of travel characteristics and elasticity data
• Inference from previous studies or applications
• Analysis of urban area data
-------
2-31
a. Transfer of Travel Characteristics and Elasticity Data
Analysts often find that data needed to study a proposed
transportation measure are unavailable or are so old that they are
unreliable. When new data collection is impractical, a common solution is
to use data compiled in another area or averaged over several areas. Such
data as average auto occupancy, household auto ownership levels for various
socioeconomic groups, and trip frequency by trip type and household
characteristics are typical of those that might be "borrowed" in this
fashion (See Section 3-1.3). Elasticities of travel demand, e.g., the
percent change in transit ridership in response to a 1 percent change in
fare or the percent change in VMT in response to gas price, are a special
type of data whose transfer frequency is considered; the elasticities may
have been determined through observation or may have been estimated in a
model.
The reliability of such transferred data depends on how similar
certain socioeconomic conditions and other factors are from one location to
the other. Factors which normally affect travel behavior, such as popula-
tion, income, auto ownership, and land use patterns, can cause travel char-
acteristics and elasticities to vary, and the validity of any transfer of
data between sites will depend on the degree to which variations in these
conditions can be controlled. In general, transferred data are more likely
to be valid if the locations have similar population, land use, and auto
ownership characteristics.
Chan suggests that data should be transferred only between cities of
the same "category," where cities are categorized in a fashion which should
minimize within-category variation in travel behavior (7). The proposed
classification divides cities into four categories by population and
-------
2-32
spatial structure, characteristics which are associated with travel demand
patterns and the kinds of transportation systems which are available. The
categories are:
• Large (population over 800,000)/multi-nucleated
• Large (population over 800,000)/core-concentrated
• Medium (50,000-800,000)/multi-nucleated
• Medium (50,000-800,000)/core-concentrated
Research indicates that trip frequencies and trip lengths tend to be
greater in large cities. However, auto ownership per household and the
share of total trips by auto tend to be greater in small cities. Empirical
evidence suggests that multi-nucleated cities generally have lower levels
of total travel activity, as measured in vehicle miles of travel, than
core-concentrated cities of equivalent size, since the dispersion of
activity centers tends to result in shorter average trip lengths between
residential, employment, and other locations.
Suggested Applications—The transferred data or elasticities generally
should be used only at a fairly large geographical scale of analysis (i.e.,
city or areawide) because they are not accurate enough for use in the
assessment of a small-scale transportation improvement. For example, a
transferred elasticity could be used to produce a rough estimate of the
systemwide change in transit patronage resulting from a fare increase, but
it would be inappropriate in most cases for forecasting the effects of
individual transit route changes. Transferred data may be of particular use
in "first-cut" or "initial screening" analyses. Case Studies I and II in
Volume II show the use of transferred data to define base conditions for
studies of an urban highway corridor and an auto-restricted zone.
-------
2-33
b. Inference from Previous Studies or Applications
Most of the measures which are being proposed in transportation/air
quality planning already have been studied and/or implemented somewhere.
One way of evaluating such measures is to look into the results of such
previous studies or applications and extrapolate their findings to the
situation currently under study.
The validity of drawing such inferences will depend on the similarity
of conditions between the two sites. Differences in conditions or charac-
teristics which influence transportation behavior, particularly socio-
economic and spatial characteristics, could result in the same measure
having markedly different results in different locations. Since it is
extremely difficult, if not impossible, to fully control for the difference
among areas, one should be cautious in drawing conclusions based on extra-
polated findings. Nevertheless, careful study of the literature can prove
insights into the probable impacts of a measure, particularly if a wealth
of experience exists or if less extensive experience has been remarkably
consistent. It generally should be possible to base qualitative statements
about likely impacts on a review of earlier findings, and sometimes order-
of-magnitude/quantitative conclusions can be drawn.
Suggested Applications—This approach might be used at a preliminary
stage of analysis to obtain a general indication of the appropriateness of
a measure. However, depending on the quality and detail of the literature
available on the measure under consideration, the information obtained from
A useful reference summarizing the results of previous
applications is provided by (55). A guide to further references by
measures is also provided.
-------
2-34
review of previous studies or applications could serve as a guide for some
quantitative analysis using available local data. The detail and accuracy
of the resulting estimates will not equal those obtainable using some of
the methods previously described, which suggests that this type of analysis
should be confined to preliminary analyses where precision is not critical.
This approach may also be useful for evaluating programs which are
difficult to analyze with other manual techniques, e.g., bicycle policies.
-------
2-35
c. Analysis of Urban Area Data
Data specific to the area under study can be of great value,
particularly when home interview surveys are outdated or insufficient. The
U.S. Census publishes data on modal shares, median distance, and median
trip time for work trips for 60 SMSA's, 20 of which are surveyed each year
on a rotating basis. Most MPO's have at least some data on hand which
could be used to estimate both work and non-work trips—ranging from modal
forecasts to cordon counts, small-area surveys, and other limited-scope
data.
Such data can be used to estimate current and future base case con-
ditions, and analysis of urban area data can help to identify appropriate
target areas or "markets" for various measures. When used in conjunction
with some knowledge or reasonable assumptions about the effects a proposed
measure might have, data analysis can provide a quick first-cut policy
analysis.
Suggested Applications—Data analysis is useful:
• for preliminary feasibility analyses
• when available data are insufficient to use worksheets
• when worksheets cannot easily be adapted for analysis of the
particular measures under consideration
Case Studies I and II in Volume II show the use of available urban area
data to define base conditions for the analysis of an urban highway
corridor and an auto-restricted zone. Also, Section 5.5 discusses the
sources of urban area data in greater detail.
Table 2.4 illustrates how trip generation data, in combination with
assumptions on mode share potential, can be used to assess the potential
impacts of new bicycle facilities (3). The analysis illustrated in the
-------
2-36
TABLE 2.4
Daily Home-Based Vehicular Trips that Might
Be Attracted to the
Trip Total Nuaber of
Purpose Daily Home Based
Vehicular Trips
School 160,000
Recreation 917,000
Personal 666,000
Business
Shopping 566,000
Work 829,000
Medical 48.000
Total 3,086,000
Source: "A Sicuury Report of
Metropolitan Council
Percentage
of Vehicular
Trips Less
Thin 6 Kin.
in Duration
(2)
21. 1Z
35. OZ
40. SZ
48. 6Z
18. 9Z
14. OZ
Travel in the
' •'•/Based on 1970 Minneapolis/St. Paul trip
\^-/Low percentages in this category are due
V. -3 /Less than 1000 tri?s.
Assuaed Percentage
of Vehicular Trips
Less Than 2 miles
in Duration That Might
So Hade by Bicycle
50.0Z
35. OX
30. OZ
20. OZ
10. OZ
S.OZ
Bicycle
Kesultanc Per- Nusber of Vehlc-
centage of All ular Trips That
Vehicular Trip* Would Be Attracted
Attracted to to Bicycle Use If
the Bicycle Proper Facilities
Were Provided
10. 0/. 16,000
12. OZ 100,000
12. or. si, ooo
9.71 55,000
2.0Z 16,000
0.7* —(3)
S.7Z 263,000
Percent
of Total
Bicycle
Trips
61
37Z
30:
21Z
6:
OZ
Twin Cities Metropolitan Area" (Draft Report), Twin Cities
data.
to the high percentage of
pedestrian trips vhlchve not counted
as vehicular
Source: (3)
-------
2-37
table forecasts the number of daily bicycle trips which might occur in the
Minneapolis/St. Paul metropolitan area if proper facilities were provided.
Actual data on trip frequencies and distances are used to determine the
number of daily trips shorter than two miles. Of these trips, it is
assumed that a certain percentage, specified by purpose, may be attracted
to the bicycle mode.
-------
2-38
2.2 Programmable Calculator Methods
Calculator programs are an adaption of manual methods of analysis.
While calculator methods do not represent a methodological advancement,
there are advantages in the use of calculator programs which make them a
powerful innovation:
• Data collection and preparation will be essentially the same
whether a manual approach or a program is used; but use of a
program does away with the time to set up and repeat the
calculation steps. This time savings can be spent in collecting
more data, testing variations of the proposed measures, and
analyzing impacts on additional "market segments" or population
subgroups. The programs thus allow the analysis to be more
detailed, without increasing costs or effort.
• Calculator methods can be less expensive than either manual
methods (by saving staff time) or computer methods (by avoiding
computer use). Progammable calculators without a "carry-over
memory", i.e., a capability of storing a program indefinitely,
cost as little as $100 for 100 program steps and 20 memory
locations. In the intermediate price range ($200-$500) storage
capabilities are sometimes available within the calculator
itself, and capacities range up to 960 program steps or 100
memory locations (with a tradeoff of eight program steps for each
memory location). Magnetic card programmable calculators will
store a program in "hard copy" indefinitely, without dedication
of program space in the calculator; prices currently start at
about $250. Whichever programmable calculator is chosen, its
cost should be recovered in a short time by the value of time
saved in performing transportation measure analyses. In addition,
the calculator can be used in other work.
• Calculator methods provide results with greater precision, since
none of the intermediate calculations are rounded. This is a
small consideration, since the input data are likely to be
estimates, but it could make a difference when similar policies
are compared.
• It is not necessary to have special training or in-depth
understanding of the methdology to use the calculator programs.
(This is also true of some manual methods.) Worksheets with
detailed instructions allow junior staff members to complete
analyses after they have been set up by analysts familiar with
the selected technique. Thus, planning agencies which elect to
use calculator program methods will not need to invest in
extensive staff training.
-------
2-39
• Compared to computer methods, calculator programs require less
data collection and preparation, and they obviously avoid the
need for computer facilities, a programmer's time, and a computer
time budget. However, their capabilities are limited in
comparison to computer-based techniques, and they may not provide
enough detail for analyses of certain policies or market
segments. It nonetheless may be appropriate to use the faster,
cheaper calculator programs to analyze a large number of
measures, to focus on a particular market segment, and to refrain
from computer use until a smaller number of alternatives
deserving more detailed analysis are selected.
The calculator programs described here cover a range of transportation
and air quality analysis requirements but the list is by no means
comprehensive. New programs, or modifications to available programs, can
be developed quite easily in response to available input data, analysis
needs, or output requirements.
The types of calculator programs available for demand analysis include:
• mode choice models; a number of mode choice models are available
(48); some provide estimates of base case mode shares while
others are simply calculator versions of the worksheet
pivot-point methods. Descriptions of two programs are included
in this section.
• trip table estimation: programs for estimating base case trip
tables either by mode or for total person-trips also are
available. Programs have been developed for three commonly used
trip distribution models (gravity, intervening opportunities, and
competing opportunities) though the methods are limited to about
15 zones (52). Programs based on the Fratar growth factor method
for adjusting trips between zones based on growth rates in each
zone have also been developed (52). A calculator program for
developing a transit trip table from boarding and alighting
counts is also available (48).
• generation of household samples; a program is available to
generate a sample of households for use in the mode choice models
described earlier (MM). The program uses published tract and
block level reports from the U.S. Census to provide information
on such household characteristics as average household size,
income, and auto ownership.
Three calculator programs—two mode choice models and a method for
generating household samples—are described in the sections which follow.
-------
2-MO
2.2.1 Program HHGEN—Developing a Sample of Households (M8)
Socioeconomic statistics for a residential area usually do not
illuminate the variations from the average which may be of particular
importance in impact prediction. It is sometimes desirable to forecast the
impacts of transportation system changes on a sample of households so that
variations in response can be examined. This sample may be provided by a
typical home interview survey; however, an up-to-date survey may not be
available, or available survey results may not adequately represent the
particular group which would be impacted by a proposed measure (e.g.,
households living near a proposed new bus route).
HHGEN may be used to develop a "synthetic" sample of households from
tract and block data provided by the decennial U.S. Census. An adjustment
procedure allows the household characteristics to be updated if necessary,
using data more recent than the Census. The household sample generation
methods used in HHGEN represent initial, illustrative approaches applicable
using programmale calculators. Research is continuing at MIT, where HHGEN
was developed, to devise improved methods of using programmable calculators
in this way.
Related manual techniques for synthesizing work locations and the
number of licensed drivers for each household also are documented. This
set of households may then be used in conjunction with the manual
disaggregate demand forecasting techniques to determine base condition
travel behavior within the study area, as well as future behavior in
response to transportation system or other changes. The results developed
using the household sample are then scaled up to represent the entire study
-------
2-11
area population. Case Study III in Volume II illustrates a typical
analysis using HHGEN in conjunction with calculator travel demand
forecasting techniques.
Calculation Procedure—The household generation process within HHGEN
is sequential, assigning to each household the following characteristics:
household type (family/unrelated individual)
housing type (single family/multi-family)
housing tenure (rent/own; families only)
number of workers (0/1/2/3+)
income (dollars per year)
family size (2,3,U...)
number of autos (0,1,2+)
The input data required by HHGEN are available from the following tables
provided by the Census:
• Characteristics of Housing Units and Population by Blocks (County)
• Income Characteristics of the Population—Tract Level (SMSA)
• Structural, Equipment, and Financial Characteristics of Housing
Units—Tract Level (SMSA)
Both tract and block data may be used, but data at the block level of
detail may be omitted if this level is not required. The following input
data items are required for Census tracts:
Total population
Percent living in group quarters
Number of single family houses
Average value of owner-occupied units
Number of rental units
Average rent
Number of single-person households
Number of households with roomers, boarders, or lodgers
Mean family income
Median unrelated individual income
Number of households owning 1, 2, and 3+ autos
Block level data (if used) includes:
-------
2-42
• Population
• Percent in group quarters
• Number of single family houses
• Number of owner-occupied units
• Average value of owner-occupied units
• Number of rental units
• Average rent
• Number of single-person households
• Number of households with roomers, boarders, or lodgers
The only other input data item required is the number of households
which are to be generated, which is specified by the user based on the
total number of census blocks to be represented, the desired accuracy of
the forecasts developed using the household sample, and the budget and
analyst time available for the analysis. In general, the greater the
number of households generated, the more accurate the forecasts developed
will be. However, manual or calculator forecasting on a very large sample
of households can be quite time consuming and tedious. For most
applications, a sample of 1000 should be sufficient, and a sample of a few
hundred is enough for many problems.
The program HHGEN, using the above data items, develops a set of
households, each with specific auto ownership levels, income, housing type
and tenure, etc., reflective of the average and frequency statistics
reported by the Census. Many of the characteristics are determined using
Monte Carlo methods in which the value of a given characteristic is
selected at random from a distribution which reflects the probabilities of
each value occurring. Some characteristics are determined contingent on
the values of previous characteristics. For example, auto ownership is a
function of income and housing. Should the number of licensed drivers or
workplace be of interest, simple manual techniques employing similar Census
or other information have also been designed.
-------
2-U3
Runs of HHGEN are repeated for each Census block or tract within the
study area to develop a set of households representing that area.
Suggested Applications—Because of its orientation to Census blocks
and tracts, HHGEN is best suited to the analysis of actions within a rela-
tively limited study area; HHGEN is particularly useful in the analysis of
alternative transit services and routings within a specific area or
corridor, as illustrated in Case Study III. In conjunction with a synthe-
tic disaggregate demand model, HHGEN may be used to develop base data when
current travel behavior is not known.
HHGEN is not well suited to areawide analysis, for which a home inter-
view survey would be a more appropriate source of a sample of households.
The number of calculation steps required can be quite cumbersome when the
technique is used in areawide analysis or when a wide variety of measures
are being explored. When using manual or calculator approaches to do such
areawide or wide-ranging analyses, average statistics for more aggregate
groupings of the population are more practical than those which employ a
random sample of households; these averages can be used to obtain compara-
tively accurate forecasting with much less effort.
-------
2.2.2 Program 2MODE-AGG—Synthetic Mode Choice
This program uses the multinomial logit model to predict, for sample
individuals representative of market segments (e.g., income levels, auto
ownership levels, etc.), the probability of choosing one of two modes (auto
and transit) for the work trip. The choice probabilities are then multi-
plied by the population for each market segment to yield aggregate modal
volumes. The sample individuals' socioeconomic (auto ownership level and
income) and trip level-of-service characteristics (trip distance, and
travel time and cost for each mode) are used to predict mode choice proba-
bilities. Any number of market segments may be used in the analysis, the
number depends on the level of data detail available and the level of
accuracy and detail desired in the analysis. The model within 2MODE-AGG is
of the synthetic type and therefore may be used to develop estimates of
existing travel behavior if this information is not available.
Calculation Procedure—Program 2MODE-AGG is designed for use with a
magnetic card programmable calculator. Base data for each representative
individual or market segment is recorded on magnetic cards, along with the
coefficients of the travel demand model to be employed. Default model
coefficients are also supplied.
The required data for each individual or market segment includes:
• household income
• household auto ownership
• number of licensed drivers in household
• work trip origin and destination.
Trip data, recorded for origin-destination pairs, includes:
• in-vehicle travel time for auto and transit
• out-of-vehicle travel time for auto and transit
• out-of-pocket travel cost for auto and transit
• trip distance.
-------
2-45
All data items are recorded on magnetic cards and then run through the
calculator program which reports total volumes for auto and transit. A set
of worksheets for organizing the necessary input data also has been devel-
oped. Any number of market segments or representative individuals may be
used in the analysis. A potential source of a set of representative house-
holds is the use of Program HHGEN described previously.
Suggested Applications—The most effective use for 2MODE-AGG is in
developing base case mode split data for auto and transit. Its usefulness
for air quality planning is limited by the restriction to two modes, how-
ever. In order for the output of the program to be useful for policy
analysis, an independent estimate of ridesharing activity would have to be
developed. A particularly attractive application of 2MODE-AGG is the
analysis of specific transit service proposals within a limited corridor or
service area, as shown in Case Study III, Volume II. In combination with
program HHGEN, the program can be used to develop estimates of transit
service patronage in areas where no transit service currently exists.
When base conditions have been determined using 2MODE-AGG, it is
fairly easy to apply the technique for forecasting the impact of program
measures designed to encourage transit use. Carpool promotion, high-
occupancy vehicle lanes and other ridesharing incentives cannot be analyzed
using 2MODE-AGG, however. Despite the relative ease of using the program
for policy analysis once base conditions have been determined, it may be
even easier to employ a pivot-point model such as 3MODE(AGG)-VAN or the
manual pivot-point worksheets, because both have more modest input data
requirements.
-------
2-H6
Because of the model's limitation to the analysis of two modes and its
requirement for socioeconomic data for each market segment or representa-
tive household, if base modal shares are known, policy analysis may be more
effectively carried out using the incremental pivot-point worksheets or
SMODE-AGG(VAN). These techniques afford the analyst greater flexibility
and acuracy in program measure impact analysis while requiring less data.
-------
2-47
2.2.3 Program 3MODE(VAN)-AGG—A Pivot-Point Mode Choice Model1
This program is designed to calculate revised work trip mode shares
for up to four modes, including vanpool. It can be used to analyze an
unlimited number of market segments and to aggregate the results for all
segments to produce areawide average mode share and vehicle miles of travel
impacts resulting from the policy being analyzed.
The program is based on the generalized multinomial logit model of
work trip mode choice identical to that used in the Manual Pivot-Point
Worksheets. The program incorporates the specification and coefficients of
a model developed using 1968 Washington, D.C. data (5). It is also possible
for the user to insert coefficients from a locally estimated model and the
program can adapt to models with a variety of specifica- tions. Extra
storage capacity for model coefficients allows this flexibi- lity in the
form of the specific model used in the program.
3MODE(VAN)-AGG extends the demand estimation programs found in (48) to
include procedures for analyzing the demand for ridesharing modes. Proce-
dures are available within the program to predict the mode shifts resulting
from policies with differential impacts on carpools of varying size. Most
frequently, these procedures would be used to determine the carpool size
impact of policies including special incentives for the formulation of
large carpools, such as high-occupancy vehicle lanes limited to vehicles
with a specified minimum number of occupants.
Appendix C for a complete program listing and detailed
documentation for 3MODE(VAN)-AGG.
-------
2-U8
The calculator program is based on the pivot-point or incremental form
of the logit model. It predicts new mode shares and vehicle-miles of
travel based on existing shares and estimated changes in transportation
level of service. For this reason, the data requirements for
3MODE(VAN)-AGG are modest compared to those of the 2MODE-AGG program.
Calculation Procedure—3MODE(VAN)-AGG as listed in Appendix C is
designed for implementation on the Texas Instruments TI-59 calculator and
its companion printer. The program may be stored on magnetic cards once it
has been entered into a calculator.
The analyst identifies any number of market segments for analysis and
the program determines the impacts on each segment separately. The results
for each market segment are printed and then combined with previous market
segment results. At any point in the analysis, the aggregate results up to
that point can be printed for later review. Through careful ordering of
the market segments to be analyzed, the impact of the measure under
analysis can be examined for varying levels of aggregation of the popula-
tion. Thus, for a regionwide policy aimed at increasing ridesharing, the
impacts on CBD workers with and without access to transit, all CBD workers
as a group, all central city workers, and all workers in the region can
each be printed by 3MODE(VAN)-AGG.
The input data required by the program varies, depending on the policy
under analysis. In the following list, all items not classified as option-
al are always required. The optional items are required for some calcu-
lator options:
-------
2-49
• average round trip length
• annual household income (optional)
• average carpool size
• market segment size (population)
• average vanpool size (optional)
• vanpool circuity factor (extra distance for picking up
passengers—optional)
• base work trip mode shares
- drive alone
- carpool
- transit
- vanpool
- other
• changes in transportation 1eve1-of-service for each mode
- in-vehicle travel time
- out-of-vehicle travel time
- out-of-pocket cost
Figure 2.2 illustrates the data flows within the program.
Default values based on national averages are provided for some of the
input data items for use when local values are unavailable. A set of
worksheets (included in Appendix C) has been developed to aid in the
organization of the input data for each market segment, and in the execu-
tion of the program.
Suggested Applications—The program is applicable to the analysis of
any transportation-air quality measure intended to affect the demand for
alternative transportation modes. Because of the flexibility in analysis
i
detail which it provides, the program can be used for broad assessments of
the relative attractiveness of a. large number of measures, or quite
detailed evaluations of the mode choice and VMT impacts of a more limited
set of measures. The program is particularly useful for the analysis of
-------
2-50
Traveller
Characteristics
Travel
Distance
Base Case
Mode Shares
Total Person
Travel
Changes in Level of
Service:
In-vehicle Time
Travel Cost
Out-of-Vehicle Time
Incremental Work
Trip Mode Choice
Program
Person-Trips
by Mode
Vehicle-Miles
of Travel
FIGURE 2.2
Data Flows in Program 3MODE(VAN)-AGG
-------
2-51
measures to encourage carpooling and vanpooling, since many other available
techniques cannot accurately represent the impacts of such measures. Case
Studies II and III in Volume II show the use of the program to analyze
auto-restricted zone and transit corridor alternatives.
With minor modifications, involving the application of user-supplied
coefficients to the demand model, the program can be used to analyze non-
work as well as work travel impcts. Case Study III in Volume II provides
an example of the use of 3MODE(VAN)-AGG in analyzing non-work travel
impacts.
The use of the program is limited only by the availability of base
case data to support an analysis using the pivot-point technique. The
travel behavior (particularly mode shares) as well as the socioeconomic
characteristics of each market segment must be known or estimated. The
quality of the base case data can have an important effect on the accuracy
of the resulting analysis.
-------
2-52
2.3 Computer Methods
Like all sketch planning methods, computerized sketch planning tech-
niques involve a certain degree of simplification in some aspects of the
analysis. However, their use allows more detailed and complex analysis
than would be feasible with manual or calculator techniques. The price of
these improvements is a larger investment in trained staff, data, and
computer resources.
For the most part, detailed urban transportation planning models are
based on a four-step process for predicting travel demand:
• Trip generation models based on regressions or cross-
classifications of trips by purpose against zonal characteristics
such a number of dwelling units, household income, and number of
employees.
• Trip distribution models, generally of the gravity type, in which
the attractiveness of each potential destination is based on a
log-linear function of the zone's attraction characteristics and
its accessibility. These models predict trip destinations, given
the trip ends predicted in the trip generation process.
• Mode split models which predict the fractions of trips by mode.
These models are generally share models, used to estimate the
fraction of total trips using a particular mode based on the
"utility" of that mode and all alternative modes.
• Traffic assignment methods which assign trips to links in the
transportation network based on all-or-nothing assignment to
minimum travel time paths. Often the ability to apply iterative
capacity restraint (to attempt to balance facility volumes and
travel times) also is provided.
A variety of computer-based sketch planning approaches are available which
address all or some of the steps of the process outlined above. Aggregate
approaches use average characteristics of the population residing in a
geographic area (i.e., a "zone"), such as average income per household,
average household size, etc., to predict the travel behavior of a group
-------
2-53
(e.g., total trips per zone). Disaggregate approaches use individual
observations and predict individual behavior—i.e. , they predict the
behavior of an individual (a person, a household) based on the character-
istics of that particular individual. Such individual predictions then
must be "factored up" to obtain a prediction of areawide, or population-
wide, travel behavior. A number of approaches represent a combination of
aggregate and disaggregate techniques. For example, model coefficients may
be developed using disaggregate data but the model may then be applied
using aggregate, or zonal average data.
A number of aggregate sketch planning systems have been developed
which replicate this four-step process, but at a lower level of detail.
These systems qualify as sketch planning tools either because their
packaging and structuring makes them relatively easy to use, or because
they can be used for large-zone (or district-level) analyses, or in most
cases, both. Many of them are based on UMTA's Urban Transportation
Planning System (UTPS) (71,72) which is not itself a sketch planning tool.
SNAP (35), TAP (33), TASSIM (37), and TRIMS (50) fall into this group.
COMPACT (U6) and IMPACT (65) are similar systems developed in England
which, while not UTPS-based, do incorporate the four-step prediction
process.
Another set of methods, including CRISTAL (62) and TRANS (75), is
based on aggregate travel demand models, but uses idealized city structures
rather than detailed zonal systems and networks. This simplification
greatly reduces analysis preparation costs and computer running times, but
it also limits the method's ability to be sensitive to localized
transportation system changes.
-------
2-5M
Most (if not all) of the aggregate sketch planning model systems are
applied at a higher level of geographic aggregation than their more
detailed counterparts. TRIMS (50), for example, is used to predict travel
for the Washington, D.C. metropolitan area based on 150 "districts," as
compared with the 1,250 "zone" system used for more detailed analysis.
Savings in computer costs with higher levels of aggregation can be signifi-
cant, particularly for calculations involving zone-to-zone matrices: A
data matrix at the 150 district level would have 150 x 150, or 22,500
elements; zonal level matrices would have 1,250 x 1,250, or 1,562,500
elements.
The loss of detail inherent in highly aggregate approaches sometimes
makes them inappropriate for analyzing certain transportation measures,
because the aggregation may make it difficut or impossible to represent the
differences in the transportation system before and after a measure is
implemented. As an example, consider changes in parking charges in a small
area within the CBD. If the zone in which the changes occur is very large,
these changes will have an insignificant effect on zonal average charges.
The model system then cannot predict the impact of the changes, unless its
zone system can be tailored to the specific problem being analyzed.
Some aggregate model systems allow the simultaneous use of different
levels of aggregation over different areas. It is possible to "window in"
on a particular sub-area by using a detailed zone system to represent it,
while using more aggregate zones (districts) for the remainder of the
area. For example, TAP (33), as it is currently set up in the Dallas/
Fort Worth area, can use HO districts, more than 6,000 zones, or any com-
bination of districts in some areas and zones elsewhere. By allowing
-------
2-55
detailed analysis of selected areas, these windowing techniques retain much
of the accuracy associated with more detailed model systems where such
accuracy is most important, while achieving cost savings compared to those
realized by other sketch planning models.
Another major simplification found in most sketch planning procedures
based on aggregate model systems is a lower level of detail in network
representation of the transportation system. In TRIMS (50), for example,
the district level highway network is composed of principal arterials only,
and a simplified transit network is used which is coded essentially as a
highway network, with separate wait times at each origin district for trips
destined to each of three separate areas (i.e., CBD, fringe, suburbs). In
TASSIM (37), a "spider" network is used which represents the highway system
by aggregating links between district pairs into a composite link. With
this procedure, only 582 links are used to represent a network which at the
most detailed level consists of 18,000 links. Similar problems to those
associated with large zone systems can arise when coarse networks are used
to analyze very localized transportation measures. However, those proce-
dures which have the "windowing" capacity described earlier usually allow
network components, specified at different levels of detail, to be used in
a single run.
Computerized sketch planning techniques often place more emphasis on
transportation system evaluation than the model systems designed for more
detailed analysis. SNAP (35), for example, incorporates modules which
generate a range of cost, consumer surplus, accident, air and noise
pollution, accessibility, and community impact measures in easily-obtained
matrices based on the groups impacted and the mode of travel.
-------
2-56
The omission of a number of important variables from the traditional
aggregate models may limit their usefulness in analyzing policies affecting
certain kinds of transportation policies. For example:
• Trip generation models rarely include variables representing the
amount or quality of transportation serving the various zones.
Thus, the number of trips predicted to originate in each zone is
unaffected by changes in the transportation system character-
istics. This usually is acceptable for work trips, which (at
least in the short term) mostly must be made regardless of any
changes in the transportation system. But for the more
discretionary trip purposes such as shopping or personal
business, the number of trips made is_ likely to change in
response to transportation level-of-service changes. A measure
such as an increase or decrease in gasoline price could affect
the total number of discretionary trips made, but a traditional
model system could not predict that.
• Trip distribution models often use distance or travel time by
auto as the only measure of the level of transportation
service--cost is omitted, as are transit service variables. The
models thus cannot predict the effects of such policies as park-
ing price increases or transit travel time reductions might have
on destination choice.
• Mode split models traditionally have included only auto and
transit. Such models cannot differentiate shared-ride (carpools,
vanpools) from drive-alone, although such a differentiation is
increasingly a factor in transportation operations planning.
Other computerized sketch planning models are available to deal with
aspects of transit or paratransit forecasting. Paratransit modes, which
generally are excluded from mode split models because of the difficulty in
characterizing their level-of-service variables, may be studied with
specially designed analysis tools. SAIM (73) is designed to identify
potential areas for successful paratransit operation, while FORCAST (12) is
designed to equilibrate supply and demand conditions for the analysis of
dial-a-ride systems. Special techniques also exist for analyzing more
conventional transit systems alternatives. One such method, the Transit
Sketch Planning Procedure (11), is described later in this section.
-------
2-57
In addition to these limitations which apply in general to most aggre-
gate model systems, certain characteristics of individual sketch planning
procedures also may limit their applicability. For example, several proce-
dures treat transit in a very general fashion or in some cases ignore it
completely In TRIMS (50), for example, transit trips are considered for
only one trip purpose (usually work) in any given run. Other models deal
only with automobile use and may be used to estimate the amount of auto
travel by highway system (SSDA), by highway system and time of day (RHEM),
by highway system, time of day, and subarea (CAPM), or link-by-link inde-
pendent estimates of auto volumes DTEM) .
Finally, in recent years a number of MPOs have integrated disaggregate
travel demand models—usually for mode choice—into their transportation
modelling process. In almost all cases, these models have coefficients
which were estimated using observations of individual travel behavior, but
they are being applied using aggregate zonal-based data. They thus
produce predictions of how "average" travellers would behave, and much of
the variation in travel behavior is lost. In some recently-developed
application procedures, individual households rather than traffic ones
serve as the basic prediction unit. In this procedure, termed household
sample enumeration, a representative sample of households is selected from
the study area. The "study area" need not encompass an entire urban area;
if major corridor or activity center-oriented measures are considered, it
1These and other computerized methods are described in (70). CAPM
is also described in Section 2.3.1.
biases inherent in such application procedures are discussed in
(43).
-------
2-58
is possible to draw samples from these areas only. The disaggregate model
then is applied to each household in the sample. Aggregate forecasts are
obtained by expanding these travel predictions for individual households to
represent areawide travel patterns.
One example of the application of this procedure to the analysis of
transportation measures is the Short Range Generalized P_olicy Analysis
method (SRGP,7). A description of SRGP is included as one of the three
computerized sketch planning methods described in greater detail in the
subsections which follow. The other methods are CAPM and the transit
sketch planning procedure.
-------
2-59
2.3.1 CAPM--Community Aggregate Planning Model1
CAPM is a highway sketch planning tool included in UTPS which is
designed for the preliminary evaluation of large numbers of broadly defined
alternatives. CAPM has the ability to directly address the following kinds
of issues:
• Decisions as to the location, magnitude, and functional type of
urban highway investments.
• Formulation of highway operating strategies useful in obtaining
environmental and system, performance objectives, such as pollu-
tion abatement or fuel conservation.
• Examination of the highway-related implications of future land
development policies.
While CAPM does not treat transit systems explicitly, its utility is
not limited to the evaluation of highway alternatives. CAPM can be used in
conjunction with complementary transit sketch planning tools (e.g., transit
corridor analysis) or can be used to assess the highway-related impacts of
transit policies (e.g., the doubling of total patronage or the increase of
transit's share of trips to the CBD by "X" percent).
An important feature of CAPM is that it presents side-by-side displays
of the various performance measures for base year conditions and a future
alternative set of conditions. This is accomplished by a two-step
process. Initially, a program run analyzes base year conditions and
creates an output data set. All subsequent runs with alternative data sets
read the base year output and write the side-by-side comparisons. The
impacts of any system improvements or policy change are, therefore, readily
apparent.
references are available, including 21, 23, 42, 58, 59.
-------
2-60
A second feature is ease of input preparation. The basic geographic
analysis units are communities, ranging in size from 8 to 30 square miles.
The only inputs necessary for the representation of an alternative are for
each community:
• Exogenously estimated vehicle trip ends,
• Total lane-miles of arterials,
• Route numbers, average numbers of lanes and center-line miles of
each freeway.
Other d.ata necessary for the estimation of the large number of evaluation
criteria output assume internal default values, although these may be
easily changed if desired.
Another major feature of CAPM is that its outputs include a large
number of performance measures which can be directly provided to decision-
makers. These include measures of travel, system performance,
environmental/social impacts, and system supply costs. These are output
for the base and test alternative in juxtaposition and appear in three
separate summary tables:
• for each individual community,
• for groups of communities having similar characteristics; e.g.,
all suburbs, and
• for the region as a whole.
The CAPM program consists of three interrelated components:
• Travel generator; Assuming the stability of an input base year
regional work trip time, the generator uses a modified version of
a direct assignment model to compute a system-sensitive estimate
of the daily regional average trip distance (length). This, in
turn, is used to estimate the average trip distance for each
community by examining its geographical position relative to all
other communities. These trip distances are passed to the trip
distributor.
-------
2-61
• Trip Distributor; The average distance from the generator module
is used in an exponential trip distance distribution function to
estimate total travel in each community as the sum of three
components: (1) trips with both ends in the given community; (2)
trips with one end in the community; and (3) trips with neither
end in the community. A special assignment logic is used to
assign each component to the proper highway system. This assign-
ment logic assumes that freeway speeds are twice those on surface
arterials. Under congested conditions this tends to overload
freeways. A capacity restraint mechanism has thus been provided
which adjusts the volumes on a given community freeway facility
to ensure that its speed in the peak hour will be at least 15
percent higher than the competing surface arterial system in the
same community. VMT in each community on freeways, surface
arterials, and locals is passed to the performance module for use
in criteria estimation.
• Performance Measurer: Using factors which are input by community
type, the VMT's passed from the distributor are split direction-
ally and temporally (peak, off-peak, other hours). By comparing
volumes to capacities, speeds are calculated and in turn travel-
related performance measures and impacts are computed for each
direction/time period. These are summed (and weighted where
appropriate) to yield daily totals. In addition, the direct
impacts and costs of going from the base to the alternative test
system are estimated. The output reports are produced by this
section of the model. A list of the reports is as follows (all
are optional):
- Community Type Statistics
- Performance Factors
- Community Location Data
- Base Community Attributes
- Base Freeway Route Data
- Base Freeway Route Intersections
- Alternative Community Attributes
- Alternative Freeway Route Data
- Alternative Freeway Route Intersections
- Metropolitan Area Specific Data
- ADT on Freeway Segments
- Community Travel Breakdown by System
- Metropolitan Area Summary
- Community Type Summary
- Community Specific Reports.
As an option, the impacts and costs can be reported for autos and
trucks separately, as well as totalled, in the last three reports.
-------
2-62
Suggested Applications—As a tool for the analysis of air quality-
oriented transportation measures, CAPM is significantly limited by its lack
of explicit treatment of transit systems and of the mode choice aspect of
travel behavior. These missing components limit its applicability to urban
areas (generally small) with no existing or planned public transit service;
and for cities with transit service, to transportation measures which are
strictly highway-oriented and are not expected to impact transit usage in
any significant way. CAPM is also limited to the analysis of alternatives
which can be realistically represented by changes in the following highway
system characteristics:
• by freeway facility: maximum speed, capacity, ramp spacing,
number of lanes, and connectivity with other freeway facilities;
• averaged over entire communities (8 to 30 square miles): surface
arterial maximum speed, capacity per lane, traffic signals per
mile, and number of lane-miles; local street average speed.
The following types of air quality-oriented highway system measures
are not amenable to analysis using CAPM:
parking supply or pricing changes
high-occupancy vehicle lanes and roadways
very localized traffic flow improvements
bridge or roadway toll policies
carpool and vanpool incentives
CAPM can predict average trip distances and trip routing, but all other
aspects of travel demand must be provided as input data, either in the form
of auto trip ends by community, or as person trips, transit trips and auto
occupancy by community.
To summarize, given CAPM's limitations discussed above, it can be
considered for use in the following ways as part of the analysis of air
quality-oriented transportation measures:
-------
2-63
• to analyze policies such as auto-restricted zones, traffic flow
improvements, and staggered work hours in cities with no public
transit service;
• to develop base data on highway system usage and the resulting
cost, energy consumption, air quality, safety, an travel speed
impacts; and
• after the prediction of the demand impacts of any transportation
measure using other techniques, to develop estimates of revisions
in highway system usage and the resulting impacts.
-------
2-64
2.3.2 SRGP—Short Range Generalized Policy Analysis1
Short Range Generalized Policy Analysis (SRGP) is a computerized
sketch planning methodology which employs a set of interrelated
disaggre- gate travel demand models to predict the near-term impacts of
transporta- tion measures—e.g., changes in household auto ownership
level, work trip mode choice, and non-work trip frequency, destination,
and mode choice. SRGP was designed to combine theoretical innovations
with a number of soft- ware features to produce a user-oriented,
practical planning tool respon- sive to the transportation, air quality
and energy conservation analysis needs of metropolitan planning
organizations. Case study applications of SRGP were carried out in
cooperation with the Denver, Dallas/Fort Worth, and San Francisco
MPO's. Since then, SRGP has been used in a number of applications in
each of these urban areas, and has produced analysis results quickly and
efficiently. Examples of these applications appear in Case Study IV in
Volume II.
Calculation Procedure—SRGP links together eight disaggregate
travel demand models to predict auto ownership, work trip mode choice,
and non- work travel, all conditional on fixed household and employment
locations. This conditionality implies that while SRGP captures the
near-term impact transportation measures would have, impacts which might
occur over a longer term, such as the number and distribution of work
trips and location and land use patterns, are assumed to remain the same
during the analysis
Complete user documentation for SRGP may be found in (8).
In-depth discussion of SRGP's capabilities is contained in (6) and (7).
-------
2-65
period. SRGP shares this assumption with most traditional model systems
in which land-use patterns are generated exogenously and work trip
generation is not predicted as a function of transport service levels.
The models included in SRGP are:
• an auto ownership model for households with one or more workers;
• an auto ownership model for households with no workers;
• a work mode choice model;
• a carpool size model for work trips;
• a shopping trip generation model;
• a social-recreational trip generation model;
• a simultaneous destination and mode choice model for shopping
trips; and
• a simultaneous destination and mode choice model for social-
recreational trips.
With three exceptions, each of these models is of the multinomial logit
form. The two non-work trip generation models are specified as linear
regression models.
As illustrated in Figure 2.3, these models are linked together in two
ways:
• Shorter-term or lower-level choices are conditional on the out-
come of longer-term or higher-level choices as represented by the
solid lines in Figure 2.3. For example, decisions concerning
work trip mode choice are conditional on predetermined household
auto ownership levels. This type of linkage corresponds to that
typically found in traditional aggregate model systems.
^Note, however, that it is possible to predict the near-term
impacts of a transportation measure implemented during some future
analysis year. In Denver, for example, 1985 was used as the base year;
household and employment locations were assumed to be unchanged by the
policies tested (9).
-------
2-66
Predetermined Household
Locational Variables
Auto Ownership
Work Mode Choice
i
Non-Work Trip Generation
T
Non-Work Trip Joint
Destination/Mode Choice
I
-1
I
1
I
I
FIGURE 2.3
Interrelationships of Travel Demand Models
-------
2-67
• Higher-level decisions are based on expectations of lower-level
choices. For example, the choice of household auto ownership
level is influenced, at least in part, by the characteristics of
alternative means of travel for both work and non-work travel.
Similarly, non-work trip generation is sensitive to transporta-
tion levels of servic.e for each potential destination. These
linkages from lower level choices to higher level choices are
illustrated by the dotted arrows in Figure 2.3 and typically are
not found in aggregate model systems.
Software Capabilities—SRGP was designed and developed within the UTPS
framework and was written as a user-coded subroutine of UMODEL (71). This
UMODEL setting imposes a user-oriented structure on SRGP and provides
standardized input and output facilities for communications with other UTPS
programs. The input data are specified in such a way that they can be
produced easily from a typical household interview survey and normal net-
work analysis techniques.
SRGP has been designed to allow the user to tailor any model run to
the specific conditions being analyzed. SRGP has an "update" feature which
permits the user to modify any input data on a zonal or interchange basis.
Multiplicative and/or additive changes to the data are specified by making
simple changes to the job control parameters at the time the run is sub-
mitted. Thus, rather than developing new input data to represent a parti-
cular transportation measure, the base data are modified within the program
to reflect any changes in level of service for those areas or trip inter-
changes affected.
The user also can specify the demand models to be used for a
particular run. This feature results in significant savings when analy-
zing measures which have direct impacts only on home-to-work travel and for
which any secondary impacts on non-work travel can be ignored. Predicting
-------
2-68
work trip mode choice, which involves at most three (or possibly four)
modal alternatives per observation, is considerably less expensive than
estimating choice probabilities and related impacts for all mode/destina-
tion combinations available to a household for non-work trips, which typi-
cally involves many more alternatives.^"
It is also possible to alter SRGP model coefficients. However, this
option has been used primarily for adjusting model coefficients when
initially setting up SRGP in an urban area to ensure that SRGP predictions
match base year observed data.
Adaptation to a Specific Urban Area—The work required for an MPO to
integrate SRGP within its planning process can be organized into three
tasks:
• Develop base year data files;
• Adjust model coefficients;
• Update base year data as required;
Three types of base year data are required by SRGP: a sample of
households, transportation levels of service (both peak and off-peak), and
land-use characteristics on a zonal basis. The ideal source of required
1In the Denver study, the computer cost for the analysis of work trip
mode choice only for a sample of 2,027 households was approximately $4.00 per
run. For the analysis of both work and non-work travel for 506 households
using separate samples of 50 destinations each for shopping travel and
social/recreational travel, though, the run cost was $37. In order to reduce
run costs, an optional destination sampling procedure has been incorporated
within SRGP. Using this option, it is possible to select a sample of
destinations for which choice probabilities will be calculated. Experience has
shown that sampling 50 out of 276 possible destinations results in no
significant decrease in accuracy.
-------
2-69
household data is a recent home interview survey or an older survey updated
to reflect current population characteristics. These surveys typically
provide information on a much greater number of households than necessary
and are generally quite complete in terms of fulfilling the data needs of
the model system.
If such a survey does not exist, there are two alternatives:
• Conduct a new, small sample survey;
• Synthesize artificial household-level data from Census data.
The medium used for conducting a new, small sample survey need not be
restricted to the home interview. In the Dallas/Fort Worth area, for
example, telephone interviews with mail-out/mail-back follow-ups were used
to develop the required household sample.
Most urban areas have transit and highway networks coded in standard-
ized formats from which the required transportation level-of-service data
can be obtained. Similarly, land-use studies have been conducted in most
urban areas and can provide the required population, employment, and other
land use-related data.
Experience in adapting SRGP to specific urban areas has shown that
this task, development of base year data, usually requires the greatest
amount of effort. Depending on the availability and status of required
data, this task typically requires six to ten person-weeks.
Some of the model coefficients incorporated within SRGP generally have
to be adjusted in order to replicate base case conditions. There is a
theoretical basis, supported by empirical evidence, for transferring the
estimated relationships among travel time, cost, income, etc.; however, no
-------
2-70
such basis exists for transferring alternative-specific constant terms,
which essentially capture the effects of all other factors influencing the
choice process not explicitly modelled (5*0. Thus the strategy for
transferring models among urban areas is to accept the estimated
coefficients for time, cost, income, etc., and adjust only the constant
terms. This is done by applying the model in the same way it is applied in
forecasting. The predicted results are compared with those observed, and
the constant terms are adjusted to compensate for any differences. This
task is relatively straightforward, typically requiring one person-week to
accomplish. This process is described in Section 5.5.5.
For many applications, some data updating may be necessary if the
available data is not representative of the time period for which the
analysis is to be performed. For example, in some cases it may be neces-
sary to utilize an out-of-date interview survey as the source of household
data; other applications may require that a future year be used as a base
for the analysis. In either case, it is desirable to update from one point
in time to reflect actual or projected changes in population and employment
characteristics at a later point in time. Similarly, level-of-service and
zonal characteristics may require updating. The procedures most
appropriate for updating these data vary, depending on the type of data
being updated (i.e., zonal, level of service, or household) and the level
of detail desired.1
1Several such procedures are discussed in (7).
-------
2-71
Suggested Applications—SRGP can be used as a sketch planning tool by
MPO's in the analysis of a wide range of transportation policies. It has
been used in varied applications in several urban areas. Unlike many
traditional travel demand models, SRGP explicitly considers carpooling as
a distinct alternative in the choice of mode for work travel. Vanpooling,
too, can be included as a separate mode.
In terms of applicability to specific transportation measures, two
steps are required:
• the measure must be translated into changes in variables included
in the model system;
• those individuals actually affected by the measure must be
identified within the model system.
The majority of transportation measures being considered can be expressed
in terms of changes in auto or transit travel time or costs and therefore
are easily represented by SRGP. In the case of carpool matching and promo-
tion, an "incentives" variable is included in SRGP, which is based on
earlier modelling work.
The second requirement is that the group affected by a measure be
identified. For those transportation measures defined on a geographic
basis (i.e., areawide ridesharing programs, CBD parking restrictions,
etc.), this is straightforward, although for very small areas, care must be
taken to ensure that the sample size is adequate. For those transportation
measures that are generally restricted to major employers (such as carpool
1See Case Study IV, Volume II, and reference (6) for more detail on the
carpooling incentives variable.
-------
2-72
and vanpool programs), SRGP is capable of distinguishing impacts by
employer size. Facility or route-specific transportation measures (i.e.,
HOV lane, tolls, etc.) are more difficult to analyze. Although it is
possible to judgmentally assign trip interchanges to a specific facility,
unless the facility is isolated (such as a bridge crossing), such a proce-
dure may not be realistic. Further, unless the facility is heavily used,
the number of trips in the household sample that would use the facility may
be too small to allow any meaningful analysis. A more appropriate applica-
tion of SRGP to measures such as these would be to analyze their potential
on a system-wide basis (i.e., HOV lanes on all major facilities serving the
CBD, areawide program of signal timing improvements, etc.).
SRGP has trip-based emissions and fuel consumption submodels which
interface directly with the individual travel demand models and are
sensitive to changes in average speed, trip length, cold-start conditions,
and vehicle weight. Thus, SRGP can analyze measures which improve traffic
flow leading to increased VMT, but which result in a net decrease in
emissions by improving operating conditions.
-------
2-73
2.3.3 Transit Sketch Planning Procedure (11)
This technique was developed primarily as a sketch planning model for
transit systems analysis and design; the highway system is treated at a
very general level. The procedure links together a series of computer and
manual calculations into an analysis methodology similar to that used in
detailed urban transportation planning models, but introduces a number of
features which significantly reduce time and cost requirements while main-
taining a relatively high level of accuracy.
Given a set of analysis year trip tables, the procedure predicts mode
shares for a representative sample of origin and destination (0/D) pairs,
and then expands these to all 0/D pairs to obtain transit and auto flows
for the entire area. These transit flows are then assigned to a simplified
transit network. Manual worksheets can be used in the steps for which less
detail is needed or which involve calculations for the limited number of
sampled 0/D pairs; a computer is used in those steps which involve exten-
sive calculations for all 0/D paris. This worksheet/computer combination
minimizes much of the time and expense associated with an "all computer"
approach, while eliminating much of the tedium that can be associated with
sketch planning procedures.
This procedure has been applied in studies of a wide range of transit
alternatives in several urban areas. In a transit study for the City of
Regina, Saskatchewan, Canada, for example, about 30 transit system designs
ranging from fixed-route buses to detailed personal rapid transit networks
were examined in less than five person-weeks of analyst's time (53). The
procedure also has been used to study the feasibility of automated guideway
-------
2-74
transit systems in Dallas, Chicago and Atlanta and has been applied in
Mexico City in an evaluation of several regional transit alternatives. It
is suited for Phase I of UMTA's transit alternatives analysis process.
Calculation Procedure--Figure 2.4 shows the sequence of steps used in
applying the model system. Initially, a zonal system is developed by
combining existing traffic zones to form a maximum of 100 analysis zones.
Then, a series of transit alternatives are formulated. The characteristics
of each alternative are specified in system description log sheets, which
include information such as headways, fares, station spacing, vehicle size,
etc. (An example of such a log sheet is presented in Case Study V in
Volume II.) In addition, a simple transit network, representing only net-
work structure and link speeds, is coded, using a highway (non-
line-oriented) network building program (UTPS progam HR).
The next step is selection of zone pairs which are representative of
the types of transit service available in the urban area. Typically, the
zone pair sample size ranges from 6 to 30. Because detailed demand
analysis is performed only for these representative zone pairs, the
sampling concept is one of the major efficiencies of this procedure. Each
zone pair is divided into six "market segments" (or groups) based on the
group's mode of access to transit (walk, auto, feeder) at both origin and
destination (e.g., walk/walk, walk/feeder, feeder/walk, etc.). Transit
service levels then are developed manually for each market segment, using
the transit network (or a system map) and the system description log
sheets. Similarly, auto service levels are developed manually. In this
case, though, one service level is used for all market segments for a
particular zone pair.
-------
2-75
Develop Zone System
Formulate Transit
Alternative(s)
System Description
(Log Sheet, Map)
Develop Service Levels for
Representative Zone Pairs
Estimate Mode Shares for
Representative Zone Pairs
Map Representative Mode
Shares onto All Trips
(UMODEL)
Assign Transit Trip
to Network (UROAD)
Cost/Revenue Estimation
Environmental Impact
Estimation
Code Transit Network (HR)
Choose Representative Zone
Pairs,
Mapping Matrix
Demand Model(s)
Socio-Economic Data
Market Segments
Trip Tables
Mapping Matrix
Peaking Factors
Market Segments
FIGURE 2.4
Major Steps in Sketch Planning Procedure
Source: (11)
-------
2-76
With service levels developed for each representative zone pair,
user-supplied demand models (e.g., work and non-work mode choice models)
are applied manually to obtain mode shares of work and non-work trips for
each market segment within each representative zone pair. Assuming 18 zone
pairs have been selected, this requires 2 trip types X 6 market shares X 18
zones pairs, or 216 calculations. At this point it is possible to
introduce further market segmentation based on socioeconomic characteris-
tics (i.e., auto ownership level, income, etc.). The increase in accuracy
resulting from this finer market segmentation, of course, is obtained at
the expense of additional calculations. However, the computation of the
demand models for these segments typically is quite easy, especially if a
programmable calculator is available.
Next, these representative mode shares are mapped into all origin/
destination pairs using a user-coded version of the UTPS program UMODEL
(72). The input data required in this step are peak and off-peak person-
trip tables, a "mapping" matrix relating each 0/D pair in the system to its
corresponding representative 0/D pair, and, for each zone, an estimate of
the proportion of total trips for each market segment. The outputs of
UMODEL are transit and auto trip tables for peak (work) and off-peak
(non-work) periods. These transit trip tables then are assigned to the
simplified transit network using the UTPS program UROAD. Finally, a series
of worksheets is completed to obtain estimates of transit operating and
capital costs, and changes in auto emissions and fuel consumption.
Examples of these worksheets are presented in Case Study V in Volume II.
-------
2-77
Adaptation to a Specific Urban Area—Because of the flexibility
inherent in this procedure, its implementation in an urban area is
straightforward. Initial data requirements are relatively small: a
24-hour trip table, "peaking factors" (i.e., the percent of total trips
made during the peak hour) to obtain separate peak and off-peak trip
tables, and depending on the degree of market segmentation used and the
variables in the travel demand models selected, certain socioeconomic
characteristics for each analysis zone. The only other-data required are
those which are specific to each transit alternative. Many of the latter
data are developed manually, and their accuracy depends on the analyst's
familiarity with the urban area as well as the judgment and care exercised
in developing these data. Guidelines for developing these data are
available (11).
Two of the three computer programs used in this procedure are standard
UTPS programs (HR and UROAD). The third program is a user-coded version of
the UTPS program UMODEL. This program, which serves as the demand applica-
tion program, accepts as input person-trip table data, and from it (and a
number of other inputs) generates analysis year modal trip tables. All
program options and control parameters are entered at run time by means of
user control parameters; no modifications to the program itself would be
required for use in other urban areas.
Suggested Applications—This procedure is applicable to transit
improvement measures, including both capital improvement policies and a
wide range of operating policies. Case Study V in Volume II shows an
example of the use of the procedure to study a mixture of demand-
responsive, express, and standard bus transit services.
-------
2-78
Because auto travel has been treated only at a very general level of
detail, vehicular emissions calculations are based on VMT only. If zone-
to-zone auto distances and travel times were available, though, a more
sophisticated emissions analysis procedure could be used with this sketch
planning procedure. Using standards UTPS software, the auto person-trip
table, already available with this procedure, could be merged with auto
time and cost matrices to produce VMT summaries by trip length and speed
class. These summaries then could be used in conjunction with the manual
emissions worksheet described in Section 4.1.
-------
CHAPTER 3. TECHNIQUES FOR ANALYZING FACILITY OPERATIONS
Transportation facility operating characteristics are an important
factor in transportation-air quality planning. Characteristics such as
highway capacity and operating conditions and transit system capacity,
frequency of service, and route coverage are critical determinants of
traveller response to transportation measures. Facility operations
potentially are affected, either directly or indirectly, by all
transportation-air quality measures. Physical and operating
characteristics may be directly altered by certain measures: by adding a
highway lane or a turning channel; by adding new bus routes. Highway
flows and transit level of service are also influenced indirectly by
nearly all transportation measures, through changes in the demand for
auto or transit travel. For example, higher auto costs may reduce the
number of cars on the highways and thereby improve travel times, but
increase the load on certain transit routes.
Because of the importance of highway operations analysis in
transportation-air quality planning, this chapter focuses primarily on
techniques for highway operations analysis. However, some techniques for
analyzing bus system operations also are discussed.
Operations analysis needs vary significantly with the stage of
planning. In the early stages of planning, there is a need for order of
-------
3-2
magnitude estimates of the relative implications of a large number of
proposals to see which offer the greatest promise or have the highest
probability of negative side effects. "Back-of-the-envelope" calculator
or worksheet based methods, and transfer of relevant experience from
other areas, would be appropriate at this level of analysis. In the
later stages of planning, however, very detailed analysis methods are
needed to determine the best design characteristics of facility and
operations measures. These contrasting analytical requirements are
exemplified by ramp metering, which can be screened for applicability by
drawing on evidence from other areas with implementation experience, but
must be planned with tools which can select the optimal set of metering
rates given exact highway characteristics and traffic volumes.
At each level of analysis, the concern is with certain measures of
transportation system performance, including times and costs to
travellers, and assorted impacts of traffic, such as pollutant emissions,
noise, and fuel consumption. All of these characteristics can be
determined from two basic factors—traffic volume and speed. For highway
operations, varying the level of analysis consists primarily of obtaining
volume and speed estimates at different levels of detail, as in average
speed over an entire freeway corridor, versus average speed over each
freeway link, versus continuous speed profiles along the freeway.
Analysis methods for highway operations are restricted by
computational difficulties. Traffic engineering is one of the most
intensively studied aspects of transportation. There is a growing body
of evidence from previous experience with such measures which can be used
-------
3-3
to screen an initial list of strategies. Further, traffic engineers have
developed a good understanding of traffic flow theory—i.e., the
relationships among traffic volumes, speeds, and vehicle density on
facilities—and of queuing and delay characteristics at intersections and
when volumes approach capacities. Equations which embody this
understanding can be used to study the traffic characteristics of
individual highway and street links and intersections. But realistic
application of traffic flow principles over a larger segment of the
highway system requires a level of analytical complexity which is handled
most appropriately by a computer. Computer models have been developed
for many applications, including preferential freeway lanes, preferential
treatment on local streets, ramp metering, and network signal
optimization, as well as more mundane roadway improvements and changes in
traffic volume which result from other transportation measures. Because
of their expense and level of effort, many of these computer models are
impractical for transportation-air quality planning, but others can and
should be employed when the resources are available.
-------
3-4
3.1 Manual Techniques
In view of the data requirements of computerized highway models,
there is a strong incentive to fall back on simpler, more approximate,
non-computerized methods. Many such techniques are available. They fall
roughly into three categories:
• empirical and theoretical traffic flow relationships;
• graphical techniques; and
• transfer of experience, locally and from other areas.
The sections which follow discuss each of these categories in turn.
-------
3-5
3.1.1 Traffic Flow Relationships
Flow/density/travel time and queuing relationships initially were
developed to describe the behavior of traffic on individual components of
the highway system. Many of these equations depict isolated highway or
street traffic phenomena, such as delay at signalized intersections and
volume-travel time relationships on uninterrupted links.
Intersection delay/signal timing methods are perhaps the most
familiar. In 1966, Webster and Cobbe published an intersection analysis
approach based on queuing theory and on observations of traffic in the
United Kingdom (76). They provided equations for determining:
• saturation flow (i.e, capacity) of each approach to an
intersection, given roadway geometry;
• optimal cycle length, given traffic and pedestrian volumes
and pedestrian crossing times;
• optimal green time; and
• average delay (or time spent at the intersection) for each
approach.
While other intersection analysis methods are available, Webster's
has received the widest field application. His equations are ideally
suited for programmable calculator and worksheet approaches. Program
libraries which include the basic Webster equations are available
(18,U8). These can be extremely useful for transportation-air quality
planning in a limited but important way, by helping to identify
intersections which may be mis-timed and by providing a first-cut
estimate of the potential time savings from a better signal timing
scheme. Unfortunately, Webster's approach is inappropriate for actual
-------
3-6
signal re-timing in an urban network, because it assumes an
isolated intersection and random arrivals of vehicles from all
directions. Courage has taken some initial steps in his work to account
for non-random, platooned arrivals which characterize most urban arterial
and network intersections (19). Additional refinements to translate
intersection programs to provide useful arterial analysis tools are
considered feasible, through the use of iterative (stepwise) calculations
of intersection-by-intersection performance. Howevber, the time and cost
of following such a procedure could rival the expense of setting up and
running computer programs such as FREQ and TRANSYT.
Isolated links in the highway system also have been treated
extensively in the traffic engineering literature. There is both an
empirical and a theoretical strain in this work, but the two viewpoints
turn on the same fundamental concept: that speed (miles per hour),
traffic density (vehicles per mile) and traffic volumes (vehicles per
hour) are related in obvious ways.
Theoretical researchers have sought to infer these relationships
directly from physical and probabilistic principles, while empiricists
have collected data (particularly density/speed observations) and
estimated parameters for various equations. Some data are presented in
2
raw form, as tables, charts, or nomographs.
FREQ and TRANSYT programs are described in Section 3.3.1 (57).
the Highway Capacity Manual (36).
-------
3-7
Both approaches have produced reasonably accurate tools for
predicting travel times on uninterrupted stretches of roadway, such as
long freeway links with constant physical characteristics and no ramps.
Even freeway segments with on- and off-ramps are modelled adequately by
these equations, at least for volumes well below capacity. However,
there is a significant problem: the relationships do not hold as volumes
approach capacity, yet congestion is the most important aspect of a
traffic flow analysis. Special calculations are necessary to represent
the consequences of queue spillback at "oversaturated" locations.
For non-congested conditions, there still may be some value in a
straightforward volume/speed relationship. In particular, some
researchers have developed approximate equations for entire arterial
segments including links and intersections, both signalized and
unsignalized. An example is the model for arterial travel time
originally proposed by Davidson (22). This model, developed using
elements of queuing theory, is an equation which relates traffic flow and
average route travel time. The equation is:
t s t
o 1 - q/c
where:
t = travel time per unit distance
tQ = minimum travel time per unit distance, e.g. at zero flow
J = the Davidson parameter
q = traffic flow
c = road capacity
-------
3-8
J is intended to represent physical/environmental factors, although
attempts to classify roads by environment have not been successful. The
tQ parameter approximates what could be considered travel time at free
flow speed, including some service delay at signals. Thus, t could be
measured in early morning conditions on a route, but substituting the
travel time at free-flow speed is sufficient. The value of c can be
computed easily, and J can be calculated through regression or some other
ad hoc estimation method for an existing arterial data base, or adapted
from evidence on prior values.
Jovanis has applied this model to Bay Area data with a great deal of
success (40). Observed versus predicted travel times for a major
arterial are shown in Table 3.1. This method could provide a useful
approximation to the travel time effects of change in arterial flows for
specific corridors, or even on classes of facilities in a metropolitan
area.
It is difficult to conceive of extensive applications for
non-computerized traffic flow relationships in transportation planning.
At the sketch planning stage, measures are too broadly defined to permit
an analysis of specific facilities, and in the design stage, most
analysis needs are too complex for calculator or worksheet methods.
There are occasional circumstances in which traffic flow relationships
would suffice—for example, a unique facility, such as a toll bridge with
metering, could be evaluated with simple queuing models--but most
realistic analyses would be burdensome without the aid of a computer.
-------
3-9
SEGMENT OF
ARTERIAL
ARTERIAL TRAVEL TIMES (SECONDS)
FIELD
MEASUREMENTS
111
132
145
82
100
24
28
DAVIDSON
EQUATION RESULTS
110
129
155
76
25
98
24
33
TABLE 3.1
Comparison of Field Measurements with
Davidson Equation Computations
(Location: San Pablo Avenue in Richmond, El Cerrito, Albany, and
Berkeley, California)
Source: (39)
-------
3-10
3.1.2 Graphical Techniques
One of the few non-computer methods which have had widespread,
legitimate application is a graphical technique for determining the best
signal progression along an arterial. The analyst begins with a diagram
of the arterial and its intersections represented in time and space, as
in Figure 3.1. Cycle lengths and splits are determined previously, using
Webster or a similar method (cycle length should be the same for all
intersections, to permit a consistent progression), and are plotted
horizontally, starting at the location of each intersection on the
distance axis. Figure 3.1 shows the red phases beginning at the same
time for all three intersections. A line representing the average speed
on the arterial (speed=distance/time) reveals that vehicles would stop at
every intersection under this scheme. For travel in one direction, the
progression is improved by changing the cycle start times to maximize the
bandwidth—the time window drivers would have for getting through all
three intersections without stopping—for a speed of 25 miles per hour,
as in Figure 3.2.
If the arterial has approximately equal flows in both directions, a
two-way progression is necessary. In Figure 3.2, the start times appear
to be correct for two-way flow, but generally they would require further
adjustment to provide an acceptable bandwidth in each direction.
1For an in-depth discussion of the use of graphical techniques, see
(38).
-------
3-11
TIME
FIGURE 3.1
Basic Intersection Diagram
( | green
bandwidths in each
direction
TIME
FIGURE 3.2
Traffic Signal Progression
red
green
-------
3-12
This graphical technique is a simple design tool for a specific
transportation-air quality measure. It does not compute travel times,
impacts, or flows as an adjunct to signal optimization, but it does
provide a reliable non-computer-based method for designing flow
improvements. A plot of the existing signal settings on an arterial also
can indicate the potential for improvement through optimization.
-------
3-13
3.1.3 Manual Areawide Traffic Engineering Analysis Method (71*)
The majority of methods for facility analysis were designed for
site-specific or project level studies. Much less work has been done on
tools for estimating the cumulative, areawide impacts of ubiquitous
implementation of a site-specific measure, or for examining the
interactions of travel demand with facility capacity and operations.
Wagner has developed a simple approach for addressing both of these
issues. The analysis approach is as follows:
• Given; An urban area highway system which serves an existing
quantity of travel demand (VMT) at an existing level of
travel quality (average travel time per vehicle mile), and
• A plan for implementing one or more major traffic engineering
actions on different types of highways in the total system
for the principal purpose of improving the quality of highway
travel,
• Find; The resultant impact of the action(s) on total areawide
highway travel demand (VMT), average travel time, and various
impact measures.
The analysis is carried out in a sequence of steps, performed
separately for work and non-work travel (corresponding approximately to
peak period travel and off-peak period travel). The steps are:
• segmenting the highway network into functional classes having
significantly different travel time characteristics;
• determining the fraction of base year VMT on each functional
class;
• specifying an approximate travel time/volume relation for
each functional class, expressed as the elasticity of travel
time to changes in VMT;
• identifying the measures which affect each class, and
estimating the potential improvements by drawing on
transferred experience and model results;
-------
3-lt
• collapsing the class-by-class data into single impact
estimates for each measure.
The method, which exploits the wealth of descriptive regional
highway network data which is available to most MPO's, can be used as a
first-cut tool for assessing the areawide impacts of traffic management
and control strategies and combined packages of measures. The method is
illustrated in Case Study VI of Volume II.
-------
3-15
3.1.4 Transfer of Experience
Evidence from previously implemented highway measures and other
analytical efforts can be invaluable in sketch-planning analyses. Most
analytical tools, as discussed above, apply to site-specific, project
level impacts and entail considerable effort and expense. In many
circumstances, detailed analyses will be infeasible because of resource
constraints, or because the measures are too vaguely defined. In these
cases, experience on other similar projects may apply, or the "typical"
impacts of measures may be inferred from a sample of existing projects.
There are many sources of data about the effects of highway
measures. Published compilations document a wide variety of before and
after studies, but focus principally on demand effects and fail to
discuss the underlying assumptions in sufficient detail. Primary
documents from actual projects around the country are the major source of
information: they outline the travel time improvements and reductions in
congestion which have resulted from specific highway operational
improvements. Researchers are beginning to consolidate these data into
concise, accessible formats. For example, Wagner gathered data on
signalization improvements and freeway control, among other things.
Typical data from that effort are shown in Tables 3.2 and 3.3 (74).
In transferring these results to another project, one would not
simply apply the sample average impact blindly. There is wide variation
in the data, indicating different underlying circumstances for each
project. In Table 3.2, two of the cities, Toronto and San Jose, already
had computerized control systems and relatively thorough programs
-------
3-16
TABLE 3.2
Traffic Signal Timing Optimization Impacts
Location
Toronto Central Area
Toronto Suburban Area
San Jose - CBD
Los Angeles - Inner City
(Broadway - Figueroa)
Los Angeles - Inner City
(Pico Boulevard)
Los Angeles - Inner City
(Wilshire Boulevard)
Macon, Georgia CBD
Ingleuood, California
Citywide
Montgomery, Alabama CBD
Charlotte, NC CBD Fringe
Washington, DC CBD
Number of
Intersections
68
51
46
26
6
AS
54
60
50
10
40
Study Method
Field Measurement
Field Measurement
TRANS Simulation
TRANS Simulation
TRAKS Simulation
TRANS Simulation
SICO? Simulation
SIGOP Simulation
TRAKSYT
Simulation
TRANS Simulation
Tine of Day
7-9
10-12
1-3
4-6
7-9
10-12
1-3
4-6
4-6
3-4, 5:33-6
4-5:30
2:30-3:30
4:30-5:30
AM Fetk
7:45-8:45
4:45-5:45
7-10
3-6
AM Peak
Off Peak
PM Peak
5-6
UTCS-1 CNETSIM) 1 Off Peak
Averaze Speed, Mph
Before
15.8
17.1
15.5
13.7
21.3
28.2
27.5
21.7
15.4
17.4
15.4
21.1
20.2
13.1
12.7
11.7
22.9
22.0
16.31
19.09
17.94
7.68
11.97
After
16.5
17.5
16.3
13.7
21.0
29.1
28.1
20.9
15.7
20.6
18.9
24.9
21.5
14.4
14.4
13.7
30.9
30.0
20.24
20.26
19.87
8.66
13.22
Percent
Change
+ 4.4
+ 2.3
+ 5.2
0
- 1.4
+ 3.2
+ 2.2
- 3.7
+ 1.9
+21.1
+22.7
+18.0
+ 6.4
+ 9.9
+13.4
+17.1
+35.0
+36.0
+24.1
+ 6.1
+10.8
+25.8
+10.4
Average, All Locations:
Average, Toronto & San Jose
Average, All Others:
+11.8
+ 1.6
+18.4
Sote: Toronto and San Jose were aggressively managed, computerized signal systems in the BEFORE case
Source: (74)
-------
3-17
TABLE -3.3
Freeway Ramp Control System Impacts on Average Speed
Location
Minneapolis I-35W
Northbound
(Inbound)
Southbound
(Outbound)
Chicago, Eisenhouwer
Expressway
East bound
(Inbound)
Los Angeles, Santa
Monica Freeway
Eastbound
(Inbound)
Houston, Gulf Freeway
Northbound
(Inbound)
Los Angeles Harbor
Freeway
Southbound
(Outbound)
Detroit, Lodge Freeway
Northbound
(Outbound)
Length,
Miles
16.6
16.6
12.7
12.7
9.4
9.4
13.5
6
4
4
6
Time of Day
7:15-8:15
6:30-9:00
4 : 30-5 : 30
3:3—6:30
2 hour
AM Peak
4 Hour
AM Peak
6:30-9:30
7:00-8:00
3:45-6:15
3:45-6:15
2:30-6:30
Averages All Data
Averages for Data
Including Ramp Delays
Before
Ramp
Control
33.8
43.9
33.7
38.5
30.3
37.7
36.2
20.4
25.9
25.9
27.3
32.8
34.2
After
Ramp
Control
45.5
50.1
40.1
45.7
33.0
39.7
50.6
32.6
40.3
40.3
36.4
41.4
Percent
Difference
vs.
Before
34
14
19
19
9
5
40
60
55
55
33
26
After,
Including
Ramp
Delays
43.0
48.5
38.6
44.4
41.4
37.4
37.4
32.6
40.8
Percent
Difference
vs.
Before
27
10
15
15
14
44
44
19
19
Source: (74).
-------
3-18
of ongoing signal timing work. Their impacts thus represent a refinement
of high quality timing plans, and the resulting percent improvements are
markedly smaller than for the other cities. For judging the potential
improvement from computerized signal optimization, the status of the
existing system would be an important factor in determining which results
were most applicable to the new situation.
Ideally, experience within a metropolitan area could be used to
judge the potential of other, similar measures in that area. For
example, the impacts of a preferential lane in one freeway corridor may
indicate the likely effects for similar corridors in the region; signal
optimization in one city may mirror the potential for improvement on
equivalent facilities in nearby cities. Where there is no experience
with similar measures, modelling results from planning studies in the
area may suffice. Thus, if ramp metering had been studied for one or
more corridors, the predicted impacts might be transferred to other
corridors, taking into consideration the similarities and differences
which might affect the analysis.
Model results also can be used to develop approximate relations
between travel time and traffic levels for entire facilities. In a
freeway corridor, for example, successive runs of FREQ (40) for different
flow assumptions would measure the elasticity of travel time with respect
to volume. These elasticities then could provide a "first cut"
approximation of the effect on level of service caused by a change in
travel demand in the corridor. By extension of this approach to other
major freeway corridors in the region, and to principal arterials through
-------
3-19
applications of the TRANSYT (57) package, a more comprehensive set of
elasticities could be developed. Alternatively, the elasticity relations
from a handful of model applications might be applied region-wide, if
initial results for different facilities were in close agreement.
Previous experience and model results remain difficult to apply at
the most general level of policy articulation—regionwide strategies.
Yet such broad initiatives are characteristic in the preliminary phases
of planning. In assigning priorities for further study to various
measures, it helps to know their likely relative impacts. For example,
how much would a comprehensive signal optimization plan reduce the
average trip time for all travellers? Data shown in Tables 3«2 and 3«3
cannot address this question directly, because the fraction of travel
occurring on heavily signalized facilities is not known. In fact, this
problem applies generally to analysis at the regional scale—it is
difficult to translate highway-related impacts from specific projects to
aggregate regional measures of effectiveness.
-------
3-20
3.2 Calculator Methods
To date there are very few calculator methods available for facility
operations analysis, since a level of complexity requiring computer
approaches is reached quickly in many applications. The set of programs
developed by Murphy (52) shows the potential of programmable calculators
for these analyses. These programs emphasize traffic engineering methods
for localized application, such as signal intersection delay computation,
signal cycle length optimization, signal synchronization for maximum
bandwidth, and traffic counting distribution selection.
Transit operations calculator methods are just beginning to be
developed. A program for analyzing bus system operations is described in
the following section; an example application is contained in Case Study
III in Volume II.
-------
3-21
3.2.1 Program BUS (48)
Program BUS simulates the operation of one or several buses over a
specific route. It is used to predict running times, actual headways and
passengers on board as consequences of traffic congestion and bunching,
given scheduled headways and operating speeds. Various strategies such
as exclusive bus lanes, intersection controls, bus schedules and speeds
may be evaluated using the program.
Calculation Procedure—The logic of the program involves moving
buses one-by-one along a route, simulating each vehicle's performance.
The route is broken up into three types of segments: line segments,
street intersections, and stations. Vehicles are moved through a
succession of route segments specified by the user.
The time spent in each section depends on traffic volumes, passenger
arrival rates, and the operating characteristics of the vehicles.
Vehicle acceleration, operating speed, and prevailing speed limits are
inputs to the model. Travel time is calculated employing basic kinematic
relationships. The Webster relationship is used to calculate
intersection delay as a function of vehicle volumes and signal operation
(76), and a simple queuing model allows congestion to be modelled as well.
Passenger service delays at stations depend on the number of
passengers boarding and alighting. The number of alighting passengers is
assumed to be a fixed fraction of the volume on board the buses as they
arrive at each station. Boarding passenger volumes depend on the
passenger arrival rate at each station, the time elapsed since the
previous bus departed from the station, and the number of passengers left
-------
3-22
unbearded by the previous bus. Passenger arrivals are random, with a
Poisson distribution; the rates at which passengers arrive at stations
must be determined by an independent estimate of bus passenger demand on
the route and their distribution along the route.
As the bus is run through the route, segment by segment, the program
keeps track of the number of passengers on board and the passage of
time. A simulation record is provided to allow the analyst to simulate
route operations by running a number of buses along the route in sequence.
Suggested Applications—The major use for program BUS in air quality
planning is the translation of measures to improve bus service into
changes in transportation level-of-service for use as input to travel
demand models. Specifically, in-vehicle travel time and waiting time for
passengers may be predicted. The program may also be used to determine
the operating characteristics of proposed new bus routes given existing
traffic volumes and intersection operations.
Equilibration or balancing of passenger demand and bus route
supply—an important aspect of developing accurate impact estimates—may
be accomplished by iteratively applying Program BUS and a travel demand
forecasting procedure to determine actual operating characteristics and
demand levels. Case Study III in Volume II demonstrates the use of
program BUS in this way. Program BUS would have limited application to
policy level or areawide analysis; it is better suited to the evaluation
of specific routes and service options.
-------
3-23
3.3 Computer Methods
Computerized highway models commonly are categorized by several
attributes. "Microscopic" models represent the trajectories of
individual vehicles moving through the highway system, while
"macroscopic" models deal with the behavior of vehicle streams along
different segments of the system, much as fluid flow is studied in
mechanical engineering. "Stochastic" models assume that vehicle arrivals
and trajectories will obey probabilistic principles, while
"deterministic" models treat arrivals and trajectories as constants
within each time period. These distinctions have relevance in planning,
because they influence accuracy and expense; microscopic and stochastic
models tend to be more complex and thus more expensive, while macroscopic
and deterministic models are potentially less accurate.
In practice, no highway model is purely representative of any one
category. Highly microscopic computer simulation models of traffic flow
incorporating stochastic properties have been developed and used for
analyzing the interactions between traffic performance, roadway
geometries, and traffic control features; while less microscopic,
deterministic models of freeways and surface arterial networks have been
applied to the full spectrum of transportation-air quality measures.
Some of the most important of these models are discussed briefly below.
More detailed descriptions of selected models are also provided in the
sections which follow.
Micro-assignment is a deterministic tool for detailed analysis of
urban road networks (20). It uses an incremental traffic assignment
technique and a block-by-block origin-destination (0-D) table to simulate
-------
3-24
traffic operations in a study area of up to 1000 city blocks. It is most
appropriate for evaluation of local road modifications, such as one-way
streets, auto restricted zones, neighborhood traffic restraints, and the
like. Micro-assignment employs a conventional approach with three main
steps: network construction, path tracing, and traffic assignment. Each
distinct traffic movement is represented by a separte link, and, unlike
other network programs, any node can be an origin or destination for
traffic. All types of intersection control are represented explicitly,
including stop signs, yield signs, signals, and freeway ramps. Different
origin-destination tables are provided by time period, and the model
performs a full network assignment for each table. The program outputs
include link volumes and traffic times, from which other impacts can be
computed. Variations of the model have been used in several U.S.
metropolitan areas. These applications have resulted in modifications
which improve the representation of factors important to traffic flow,
such as transit vehicles, pedestrians, and weaving. Nevertheless, the
effort and expense of implementing the program have been deterrents to
more widespread application, and severely limit the potential of
micro-assignment for planning. Areas which already use such a model or
can justify it on grounds other than air quality, however, may find it to
be a useful tool in transportation-air quality analysis, as well.
NETSIM, like its predecessor UTCS-1, is a microscopic, network-based
simulation model for developing and evaluating complex traffic management
strategies (*45). The model can treat all major forms of urban traffic
control, including stop or yield signs, simple fixed-time traffic signals
-------
3-25
operating either independently or as part of a coordinated system,
vehicle actuated signals, and signal systems operating under dynamic
real-time control. NETSIM views each vehicle in the network as a
separate entity. Vehicle motion is determined from performance
characteristics which are input by the user, in conjunction with a set of
algorithms covering car-following, queuing-discharge, and
lane-switching. The exact position of each vehicle is updated for
extremely short time intervals. Output of the model includes a
comprehensive set of traffic performance mesures for each individual link
and for the network as a whole, as well as data on emissions and fuel
consumption. Although NETSIM is relatively new, UTCS-1 has been
extensively appied and gives reasonable results.
TRANSYT is a technique for optimizing signal offsets and splits in a
local street and arterial system (56,57). Like its predecessors, the
most recent version of the model, TRANSYT6, provides a macroscopic
deterministic simulation of traffic flow. Vehicles are represented by
platoons which change as they proceed through signals and disperse along
the roadway links. The street network is represented as a series of
nodes connected by uni-directional links, and traffic volumes are given
in a node-to-node origin-destination table. Signal settings are chosen
to minimize a linear function of travel time and number of vehicle stops
in the system. TRANSYT6 can model preferential treatment schemes for
arterials, such as exclusive lanes and priority signalization. In
practice, the program has been useful both as a design and as an
-------
3-26
evaluation tool. Because it requires a relatively small level of coding
effort, and relies on explicit and reasonable assumptions, TRANSITS has
been and continues to be implemented extensively.
TRANSYT6C is a modification of the existing TRANSYT6 model to
include environmental consequences and certain traveller responses (39).
In this effort, procedures were developed to account for fuel consumption
and vehicle emissions and to model path and mode shifts, and the
performance index was expanded to include a broader range of variables in
the signal optimization. The new model has been tested extensively for
applications in California.
SIGOP II is a signal optimization program developed for FHWA and is
a descendant of TRANSYT and SIGOP I, also a signal optimization program
(66). Like TRANSYT, it utilizes a macroscopic model to select signal
timings which minimize a function of travel times and stops. Unlike
TRANSYT, it employs a dynamic programming technique in which traffic
streams are represented as "states" with specific properties. These
states are transformed at intersections according to the decision
variables of the program (i.e., the signalization parameters). SIGOP II
has several attractive features for transportation-air quality planning.
It can calculate the capacity of each intersection approach, which
reduces the level of effort in data preparation, and it incorporates
error messages and diagnostic tests for the user operating the program.
However, SIGOP has not been applied as widely as TRANSYT, and the latest,
most appropriate version is still undergoing field tests by FHWA.
-------
3-27
FREQ6PE was developed for analyzing freeway design improvements and
entry control strategies (MO). It consists of two major parts. The
first, SIMFRE, is a deterministic macroscopic simulation model which
predicts traffic performance as a function of freeway design and
ramp-to-ramp freeway volumes. SIMFRE generates traffic performance data
in terms of miles of travel, travel time, fuel consumption, vehicle
exhaust emissions, and vehicle noise emissions. The second, PREFO, is a
linear programming optimization that is used to determine an entry
control plan for a given freeway origin-destination pattern. The linear
program allows a mixture of normal and priority entry control as well as
special ramp treatments such as buses only.
These two routines are extremely useful in assessing the short-term
impacts of strategies such as ramp metering. In a broader sense, the
simulation model alone could be used to assess virtually any freeay
configuration under analysis. In either case, the data requirements and
run expenses are relatively modest.
To represent the longer-term effects of ramp metering, FREQ6PE
incorporates an ad hoc route and mode shift procedure. Elasticities are
used to determine the effects of changes in travel time on mode choice,
and minimum path assumptions are invoked to determine the extent of
diversion to downstream ramps and parallel arterials. The data
requirements and run costs in this case are somewhat higher, and the
program results are more difficult to validate.
!fiun costs range from $5 to $50 depending on how many of the
program features are used.
-------
3-28
FREQ6PL, an alternative version, incorporating the basic simulation
program, was developed for analyzing freeway priority lanes (16). In an
application for the Santa Monica Freeway, it came remarkably close to
replicating the actual short- and medium-term impacts of the ill-fated
"diamond" lane. (The project was terminated before its long-term impacts
could be determined.)
From these few brief descriptions, it should be apparent that none
of the computer-based techniques is truly a "sketch planning" tool.
Nevertheless, some are more appropriate to the planning and design of
transportation-air quality measures than others. Highly microscopic
approaches like NETSIM yield powerful general purpose models with a wide
range of applications to such pervasive measures as advanced traffic
signal control strategies, bus lanes and bus priority signals,
right-turn-on-red, etc. In addition, they are capable of detailed energy
and emissions estimates based on individual vehicle trajectories, which
permits estimation of a variable emissions profile along a line source.
However, data preparation and computational requirements are unreasonably
high for most planning applications, even for relatively small networks
or linear arterials.
Less microscopic, deterministic computer models show a good deal
more promise for small-scale planning applications. Two of these--the
TRANSYT and FREQ series of programs—are particularly attractive in terms
of the strategies which can be modelled, the quality of support
documentation, and the relative ease of application. While they are not
sketch planning tools in the strictest sense, TRANSYT and FREQ do have
-------
3-29
significant potential because of their reduced running costs and wide
distribution to user groups. Once the investment in set-up and initial
coding of selected freeways or portions of the arterial network has been
made, many alternatives can be tested quickly and economically. Since
both models are design (FREQ for designing freeway control strategies and
TRANSYT for optimizing signal timing plans) as well as evaluation tools,
the initial investment can double as a planning and as an implementation
measure. It would not make sense to apply these tools comprehensively
throughout the network, but they could be used for critical proglem
corridors and subareas, or for a relatively small set of representative
locations—which, in turn, could be extrapolated to assess areawide
implications of traffic management and control actions.
Other deterministic models—notably micro-assignment—have been
suggested for small-scale applications, especially for evaluating
neighborhood traffic management actions. In some cases, the models, such
as SIGOP, differ little from TRANSYT and FREQ and have been used much
less extensively. In other cases, such as micro-assignment, the data and
computational requirements are out of line with small-scale planning, and
efforts to create more tractable versions are several years from
fruition. In the remainder of this section, more detailed descriptions
of TRANSYT and FREQ are presented as the best available computer methods
of highway operations analysis for transportation-air quality planning.
However, the reader should be aware that other existing models may be
appropriate for specific purposes, and that future developments may
broaden considerably the range of computer applcations for short-range
analysis.
-------
3-30
3.3-1 TRANSYT
TRANSYT is a model which simulates traffic behavior in a network of
urban arterials and optimizes the settings (offsets, cycle lengths and
green time) of traffic signals. Six generations of the model have
emerged from the transport and Road Research Laboratory (U.K.) since
1967, with successively more detailed representations of network and
vehicle characteristics and an increasing range of traffic control
strategies from which to choose (39, 56, 57). Recent versions have been
adapted to deal explicitly with traffic management measures, to represent
demand responses, and to allow flexibility in optimization of criteria
(39).
TRANSYT requires a detailed specification of network geometry, link
flows, and signalization criteria, including:
• a node number for each intersection;
• a separate link for each direction of flow feeding into an
intersection. More than one link may be required for special
lanes, such as for transit priority or for turning movements;
• saturation flows (in vehicles per hour) for each link, based
on roadway widths and on simple principles of intersection
capacity found in most traffic engineering texts;
• link travel times for uncongested flow, for both buses and
cars;
• average flows in vehicles per hour for all links;
• amount of time required for pedestrians to safely cross at
each intersection (considering width of intersection and
number of pedestrians);
• existing signalization plan, with cycle lengths, green and
amber times, and progression (if any);
• miscellaneous other parameters which specify run options,
optimization criteria, fuel consumption characteristics, and
output formats.
-------
3-31
The TRANSYT program has two key sub-models: simulation and
optimization. Their relation to overall program logic is shown in
Figure 3.3. The simulation model focuses on the behavior of platoons
travelling through the street system. Flow into the system from outside
is assumed to arrive at constant rates, but platoons of vehicles form
during the red phases at signalized intersections. In travelling
downstream from an intersection, these platoons tend to disperse, but
inevitably, some intersections receive incoming platoons rather than
fully spread-out vehicle streams. If these platoons are "caught" by a
red light, a great deal more delay (in passenger hours) will result than
if they pass unimpeded through an intersection during the green phase.
The simulation portion of TRANSYT simply estimates the extent to which
platoons are caught at each intersection and the resulting total delay in
the system, given the fixed characteristics of flow, signal timings, and
network geometry. Used by itself, it can depict the consequences of
user-specified changes to each of these system features, but its primary
application is in conduction with the optimzation sub-model.
The TRANSYT optimizer essentially seeks to minimize delay in the
system by selecting the best signal settings for platoons travelling in
each direction. It does so by systematically varying the signal timing
parameters, including progression and red/green splits, in search of the
"global" minimum for a performance index. The British version uses a
simple performance measure:
-------
3-32
PHYSICAL
CHARACTERISTICS
STOP
FLOW PATTERN
SIMULATION
SUB-MODEL
YES
OPTIMIZATION
SUB-MODEL
END
SIGNAL TIMINGS
SIMULATION
OUTPUT
OPTIMIZATION
OUTPUT
FIGURE 3.3
TRANSYT Program Logic
-------
3-33
n
PI = 2 (d. + ks.)
1=1
where:
PI = performance index
i = link index
n = number of links
d. = delay on link i (i.e., travel time)
s. = stops on link i
k = weighting factor
At several intermediate points and after finishing the optimization,
TRANSIT re-simulates system performance for the revised signal settings.
The basic model has been extended in several important ways (39):
• expansion of the performance index to distinguish the delay
and number of stops for priority versus non-priority
vehicles, and to represent fuel consumption and emissions
explicitly;
• addition of fuel consumption and air pollution sub-models to
the basic simulation model;
• inclusion of a "demand response" component to reflect the
likely route and mode shifts resulting from certain
strategies. The revised program logic with demand response
is shown in Figure 3.4.
These modifications have been directed at making TRANSIT a more flexible,
comprehensive, and accurate tool for both planning and design.
TRANSYT outputs are designed to assist the analyst in identifying
the most serious design deficiencies and in formulating remedial
transportation measures. In addition to optimal signal settings at each
node, a full range of impacts and travel characteristics is provided for
-------
3-34
PHYSICAL
CHARACTERISTICS
c
STOP
C
STOP
C
STOP
FLOW PATTERN
SIMULATION
SUB-MODEL
OPTIMIZATION
SUB-MODEL
DEMAND RESPONSE
SUB-MODEL
OPTIMIZATION
SUB-MODEL
C
END
SIGNAL TIMINGS
SIMULATION OUTPUT)
OPTIMIZATION
OUTPUT
DEMAND RESPONS
OUTPUT
OPTIMIZATION
OUTPUT
FIGURE 3.4
Revised TRANSYT Program Logic
-------
3-35
each link in the system, including actual travel time, fuel consumed, HC,
CO, and NOx emissions, number of stops, and maximum queue length. A
pictorial version of the platooning on each link also is provided, as
shown in Figure 3.5. This can be used to examine how well the signal
settings have managed to prevent platoon arrival during the red phase.
For example, Figure 3.5 shows that the minimum arrival rates occur during
the red phase, with the platoon, or what remains of it, reaching the
intersection during the green time.
TRANSYT outputs and program documentation are geared toward certain
transit/shared ride incentives for local streets, and bus contra-flow
lanes. However, the program applies equally to a much broader range of
street and intersection-related measures. The basic simulation/
optimization model can represent any traffic engineering improvements
which affects intersection capacity, including turning lanes, additional
signal phases, stop signs, and the like. In addition, it can reflect
changes in exogeneous factors, such as general increases in flow and new
traffic generators. Table 3.4 illustrates some of the output capability
of TRANSYT for a range of measures.
A number of assumptions and limitations are important for some
applications of TRANSYT:
• only signalized intersections and side street stop sign
controlled intersections are modelled;
• all intersections must have the same cycle length and only
one cycle length can be evaluated in each computer run;
• traffic enters the network at a constant uniform rate;
-------
3-36
5000
3750
§
O
W
2500
1250
Green
Phase
.Red Phase.
20
;';!•
'.••fl
I
Vehicle departure
pattern
Green
Phase
Vehicle arrival
pattern
30 40 50
TIME(SECONDS)
60
70
80
Legend:
Vehicles arriving during "red" phase
Vehicles arriving during "green" phase
Vehicles which arrived during "red" phase and waited
in a queue
FIGURE 3.5
Display of Platoon Formation at a Traffic Signal Adapted from TRANSYT Output
-------
3-37
IMPACTS1
ALTERNATIVE
Signal
Optimization
Vehicle
Delay
Signal
Optimization
Passenger
Delay
Reversible
Lane
Bus
Lanes
Time Spent
(pass-hr . )
ST
-9
-9
-7
+58
LT
+1
+1
+1
+2
Fuel
Consump.
(gal.)
ST
-6
-5
-3
+32
LT
+3
+2
+4
+3
Vehicle
Emission
(kg.)
ST
-9
-9
-7
+50
LT
0
+1
+2
+2
Mode
Shift
(pass-mi.)
ST
—
— .
—
—
LT
0
0
0
+1
Product.
(pass-mi. )
ST
—
—
—
—
LT
+15
+15
+16
-2
Bus
Time
Spent
(pass-hr.)
ST
_c
-6
-1
-13
LT
0
-1
0
-13
Bus
Fuel
Consump .
(gal.)
ST
-6
-9
-1
-22
LT
0
-A
-1
-23,
1
Definitions of Impacts:
All percentage changes are with respect to existing conditions:
Time Spent
Puel Consump.
Vehicle Emission
Mode Shift
Product.
Bus Time Spent
Bus Fuel Consump.
ST
LT
% change in passenger-hours for all vehicles
TL change in gallons of fuel consumed for all vehicles
% change in total vehicle emissions
2 change in passenger—miles for buses
% change in productivity (passenger-miles) on Study arterial
Z change in bus time spent in passenger hours
Z change in gallons of diesel fuel consumed by buses
short term results
longer term results
TABLE 3.4
Impact Changes Due to Traffic Management Strategies
For Base Conditions
-------
3-38
• percentage of turning traffic at all nodes remains constant
throughout the simulation period;
• total traffic entering a node equals total traffic leaving a
node;
• vehicles queued at the intersection are assumed to be stored
vertically; i.e., queues from one intersection are assumed
not to interfere with the upstream intersection;
• the demand response has not been validated thoroughly against
field data;
• the emissions and fuel consumption models have not yet been
updated to 1979 values.
The basic TRANSYT simulation/optimization model has proven accurate and
efficient in applications all over the world. These limitations do not
alter that basic fact, but they should serve as a reminder that each use
of an anlysis method must be accompanied by an appraisal of the
assumptions which may be violated in a specific problem context.
-------
3-39
3.3.2 FREQ
FREQ is a computer model which simulates traffic behavior in a
freeway corridor (MO, 16). Six generations and many sub-variations of
the model have emerged over the past eight years, with successively more
comprehensive representations of the freeway corridor, including parallel
arterials, demand responses, and fuel and emissions sub-models. Recent
versions have been adapted to deal explicitly with measures such as
preferential lanes and ramp metering.
As with TRANSYT, FREQ requires a detailed specification of network
geometry, origin-destination flows, and freeway operating policies.
Input supplied by the user include:
• "time slices" defined to capture the distinct phases of
freeway flow—typically 15 minutes each, so that a peak
period might consist of 10 to 12 time slices.
• "subsections" defined as the distinct segments of the freeway
under study, as in Figure 3.6. Only one direction of freeway
flow is studied at a time, and no more than UO subsections
may be included.
• for each subsection:
- number of lanes
- capacity
- length
- a truck factor (percentage of vehicles which are trucks)
- a volume delay curve
- subsection gradient
- subsection surface quality index
- ramp capacities
• origin-destination volumes for autos and buses.
• miscellaneous run parameters, depending on the exact FREQ
version and analysis circumstances.
• if a parallel arterial is being analyzed:
- subsection data for each of its subsections
- base case flow on each of its subsections
-------
3-10
Origins and destinations are defined as in Figure 3.6. For this
simple example, four flow values would be supplied (1-1, 1-2, 2-1, 2-2).
Because it is not easy to obtain these data in practice, methods are
available elsewhere for synthesizing the 0-D tables from traffic counts
at each of the on- and off-ramps and the mainline entry and exit points.
The FREQ simulation focuses on vehicle behavior in each time slice.
It first computes the seotion-by-seetion flows which are implied by the
origin-destination tables. Then it analyzes subsection capcities in
light of limiting factors such as weaving and merging. Then, given the
flows and modified capacities, it computes travel times by subsection, as
well as speeds and vehicle densities at points along the freeway.
Finally, it produces fuel consumption and emissions calculations by
subsection, and prints the results for the current time slice. After
processing all time slices (sequentially) in tbis manner, it prints a run
summary. This sequence is outlined in Figure 3.7.
Simulation alone is sufficient for most FREQ applications, but there
are some measures which require a more design-oriented use of the
program. One such measure is ramp metering, for which FREQ can predict
the necessary metering rates at the various ramps, and the effect on
parallel routes. This entails a considerable increase in the program's
complexity, as shown in Figure 3.8. The basic model now resides in a
loop which is executed four times after the initial simulation of base
conditions. On the first pass, optimal metering rates are computed based
on flows from the initial simulation, and the flows are re-simulated
using, in effect, the new ramp capacities. On subsequent passes, three
-------
DIRECTION OF FLOW
ORIGIN 1 I SUBSECTION 1
SUBSECTION 2
SUBSECTION 3 I DESTINATION 2
FIGURE 3.6
Freeway Geometry Definitions for FREQ
-------
3-42
INPUT AND INITIAL CALCULATIONS
'"I
FREEWAY GEOMETRY
FLOWS
I
READ ONE SET OF
TIME-SLICE DATA
SIMULATION
"1
RAMP CAPACITY
ANALYSIS
CALCULATE SUB-
SECTION DEMANDS
MERGING ANALYSIS
WEAVING ANALYSIS
MAINLINE CAPACITY
LIMITING ANALYSIS
TRAVEL TIME
CALCULATIONS
FUEL CONSUMPTION AND
EMISSIONS CALCULATIONS
FOR:
Vehicles on Freeway
Delayed Vehicles
Diverted Vehicles
NO
PRINTOUT RESULTS \
FOR CURRENT TIME
SLICE
FIGURE 3.7
Schematic View of FREQ Simulation Procedure
-------
3-43
FLOWS
DEMAND
RESPONSE:
1. no response
2. shift to
arterial
3. shift to down-
stream ramp
4. mode shift
SIMULATION
YES
OPTIMIZATION
YES
FREEWAY
GEOMETRY
FINAL
OUTPUTS
FJDR JACK JIME_SLICE
FIGURE 3.8
FREQ Priority Entry Optimization Procedure
-------
different demand responses are assumed to occur, one at a time, with a
ramp metering optimization and simulation for each case. After the final
simulation (for revised mode shares), the run terminates. Another
version of FREQ for freeway priority lanes is similarly complex, although
it employes a different demand-responsive scheme.
FREQ outputs are intended to help the analyst in identifying the
most serious freeway bottlenecks and in formulating remedial
improvements. The program produces "speed-contour" maps which depict the
average speed on each subsection in each time slice (see Figure 3.9).
Speeds of 25 and under on such a map imply demand in excess of capacity
(and guaranteed queue formation), and the downtream boundary of the
queued region invariably indicates the bottleneck's location. Other FREQ
outputs include travel time, fuel consumption, emissions, and number of
queued vehicles for each subsection and for the freeway as a whole. In
the optimization loops, these outputs are available after each simulation
step, or for the initial and final steps only.
In one of its many forms, FREQ can model virtually any
freeway-oriented strategy, but it is suited particularly for three
applications: ramp entry control, priority treatments, and sensitivity
analysis of freeway travel times to flows and design features. In past
appliations, FREQ has required extensive data collection and calibration
(which consists of running the model and "fine-tuning" the capacities to
achieve correspondence with observed patterns of congestion), but these
efforts have paid off in an extremely accurate and inexpensive (for
simulation) analysis tool. In fact, a before-and-after analysis of the
-------
3-45
w
,J
V)
H
H
FLOW
10
9
8
7
6
5
4
3
2
1
50
49
48
47
45
44
42
40
48
50
50 50
50 50
48 47
46 40
44 38
40 36
38 34
38 35
48 48
50 50
50
50
46
38
36
32
33
36
48
50
50
50
46
34
34
f
23
/
36
38
49
50
50 50
50 50
46 47
30 28
/ / /
19 18
/ ///
21 23
/ / / /
35 33
34 35
50 49
50 50
50
50
49
26
/ /
17
'A
19
,/ /
V /
[21
^
36
48
50
50 50
50 50
47 47
\ Q
JL o
j
f /
18
' /
20
A
22
j—
33
31
«,
32
34
33
47 49
50 50
50 50
50 49
47 47
38 46
36 46
37 44
38 40
39 46
48 48
50 50
50
48
47
46
46
44
40
46
48
50
LOCATIONS AND
-TIMES OF
CONGESTION
1 2 3 4 5 6 7 8 9 10 11 12 13
SUBSECTION
FIGURE 3.9
An Idealized Speed Contour Map From FREQ
(Each number in the map represents the predicted
speed (in miles per hour) at a specific time
and location.)
-------
3-46
Santa Monica Diamond Lane in Los Angeles was remarkably close to what
actually occurred (16). Thus, FREQ has considerable "up front" costs,
but is a perfectly valid sketch planning and design tool afterwards.
-------
CHAPTER 4 IMPACT ANALYSIS TECHNIQUES
A number of sketch planning techniques are available for estimating
the emissions and energy impacts of measures, once their transportation
impacts have been determined using the techniques described in earlier
chapters. In fact, a number of the methods described in the previous
chapters have been integrated with procedures for estimating emissions
and energy impacts. For example, the manual pivot-point worksheets (see
Appendix A), the 3MODE(VAN)-AGG calculator method (see Appendix D), and
SRGP (6) all have capabilities for producing these impact estimates. For
example, a version of SRGP recently has been integrated with the MOBILE1
model (30), which is described in more detail in this chapter. Other
procedures, including manual approaches, have been documented by USDOT
and EPA (70).
Fewer truly sketch planning approaches exist for estimating other
social, economic, and environmental effects. Often the transfer of
results from other areas will be the only practical approach unless the
resources are available to address these impacts in more detail. A
number of studies have dealt with a broad range of impacts of various
measures; and again, EPA has provided references to many of these
analyses (70).
Four methods for estimating emissions or energy impacts are discussed
in this handbook:
-------
4-2
• a manual automotive emission estimation method;
• a manaual automotive fuel consumption/operating cost technique;
• calculator programs for bus emission and energy consumption;
• the MOBILE1 emission factor computer program.
The first two methods illustrate how impact worksheet methods can be
integrated with manual methods of estimating changes in travel demand.
They are summarized in this chapter and presented in complete detail in
Appendices D and E. The other two methods are summarized briefly.
-------
1-3
*t.l Automotive Emissions Estimation Procedure
A method is presented in a set of worksheets for calculating the
hydrocarbon, carbon monoxide, and nitrogen oxide emissions associated
with changes in automobile travel. The emissions estimation technique
can be used in conjunction with manual or other travel demand methods to
forecast the impacts on emission levels of a variety of transportation
control strategies. It is based on procedures recommended by FHWA and
EPA (70). The technique produces estimates based both on the number of
trips and the number of vehicle miles of travel in the area under study.
Since transportation control policies are not expected to result in
significant changes in the number of automobiles owned or operated within
a given study area, emissions related to auto ownership levels are not
considered.
Incorporated into the technique are calculations of emissions
associated with three aspects of vehicle operations: (1) trip-related
cold-start emissions; (2) trip-related hot-soak or evaporative emissions;
(3) emissions related to vehicle miles of travel.
The method is designed to accommodate data which has been
disaggregated by market segment or population subgroup. In addition,
stratification into work and non-work trips is recommended for each
component of the estimation technique. VMT-related emissions should be
calculated separately in the case of subgroups for which average trip
distances and/or speeds vary.
Appendix D presents complete information, including worksheets,
tables and calculation procedures, which can be used to exercise this
emissions estimation procedure.
-------
4.2 Automotive Fuel Consumption and Operating Cost Estimation Procedure
Detailed procedures for estimating automotive fuel consumption and
operating costs in an urban analysis area context, as a function of the
distribution of vehicle weights, average trip length and speed, and
ambient temperature have been developed previously and programmed as part
of computerized travel prediction methods (8). These procedures have
been simplified to be suitable for hand calculation for inclusion in this
handbook. In Appendix E, the method is presented in a step-by-step
fashion which can be followed to obtain estimates of automobile fuel
consumption and operating costs for base cases and policy alternatives
which have been analyzed using the pivot-point demand estimation
worksheets,1 or other procedures which provide average trip distances
and times. The method can be used for various breakdowns of total
travel, such as for separate population subgroup and/or trip purpose
segments, or for total travel. The basic procedure calculates fuel
consumption and operating costs per trip; the results can be multiplied
by number of trips to obtain totals per analysis area.
The inputs required to apply the method are listed in Table 4.1,
along with default values based on 1976 conditions. Various sources from
which these input values may be obtained for particular localities and
analysis years are also indicated. The method is based on urban area
travel data in the 10 to 40 mile per hour speed range, and does not
represent the decreased fuel economy which occurs at speeds greater than
40 miles per hour.
iThese worksheets are included in Appendix A and are described in
Section 3.1.1.
-------
4-5
TABLE 4.1
Inputs to the Automobile Fuel Consumption and Operating
Cost Procedure
Variable
Name
I . Genera
WT
RFC
TEMP
II. Genet
GCOST
RMTCOST
III. Popt
TIME
DIST
TPHH
HHS
Description
il, Related to Fuel Consumption
Average passenger vehicle weight (1000 Ibs]
Impacts of all technological factors but
Defaults
(1976
Conditions)
3.96
1.00
vehicle weight on fuel consumption, relative
to 1975 technology (ratio of analysis year
fuel consumption to 1975 fuel consumption
when vehicle weight reamins constant)
Average ambient temperature (degrees celsius) 10
al, Related to Operating Costs
Gasoline cost (dollars/gallon in analysis
year dollars)
Repair, maintenance, and tire costs
(dollars /mile in analysis year dollars)
lation Subgroup and Trip Purpose Segment-Spe
Average one-way trip time (minutes)
Average one-way trip distance (miles)
Daily trips per household made by the group
(daily one-way trips/household)
Total households in the group (households)
0.60
0.03 x97
cific
-
-
_
_
Source
local vehicle
fleet data
( 6,3)
EPA data on
fuel economy
vs. vehicle
weights
local weather
data
local gasoline
sales data
Bureau of
Labor Statis-
tics (68)
Worksheet VI
Worksheet I
Worksheet I
or VI-A
local data,
Worksheet I
-------
4.3 BUSPOL and ENERGY (48)—Environmental and Energy Impacts of Bus
Operations
Two calculator programs have been developed for use in predicting the
pollution and energy consumption impacts of existing or proposed bus
operations. Program ENERGY computes:
• the number of vehicles required on a route given its travel time
and the desired schedule
• layover time, given the schedule and number of buses
• total round trip time for each vehicle as well as per mile fuel
consumption rates
Vehicle-miles travelled per hour and total fuel consumption per hour of
operation are then calculated. The per-mile fuel consumption is
determined by route, grade, and operating speed. The effective operating
speed is computed based on the specified values of the distance and
travel time on the route. The program does not calculate the energy
impacts of operating policies on auto or other modes.
Program BUSPOL calculates bus air pollutant emissions as a function
of bus operating speed and total vehicle miles operated, on a system or
route basis. The emissions rates for diesel buses used in the model are
based on EPA data (31). The current version of the model assumes that
all buses are operating in the "warm-up" condition, and that all buses in
the fleet were manufactured prior to the 1973 model year. Both of these
programs are illustrated in Case Study III, Volume II.
-------
t.t MOBILE1 (30) - Motor Vehicle Emissions
MOBILE1 is a computer program that calculates composite emission
factors for hydrocarbons (EC), carbon monoxide (CO), and oxides of
nitrogen (NOx) from motor vehicles. The program includes the most recent
emissions factors and methodology developed by EPA and yet provides the
user with the flexibility to vary basic assumptions about deterioration
rates for emission control devices and the timetable by which new vehicle
emission standards are implemented. The program calculates composite
emission factors for three regions:
• low altitude,
• California,
• high altitude (higher than 4000 feet),
and considers six vehicle types. Composite emission factors can be
generated for each year from 1970-1999.
Required input data include:
• year of desired emission factor
• percentage of vehicles operating cold (with and without
catalysts)
• percentage of vehicles operating in the hot transition phase
• speed
• temperature
Results are quite sensitive to the assumed percentage of vehicles
-------
4-8
operating cold, moderately sensitive to speed and temperature
assumptions, and relatively insensitive to the assumed percentage of
vehicles operating in the hot transition phase. Reasonable information
on cold operation percentages is available (28) and the percentage of
daily VMT operating cold can range from 15 percent in larger cities to as
much as MO percent in smaller urban areas.
If good local data are available, MOBILE1 also allows the user to
input such data as vehicle mix and vehicle age distribution. Default
values are based on national data and represent reasonable assumptions
where local data are not available.
-------
CHAPTER 5 DESIGNING AN ANALYSIS APPROACH
The preceding chapters in this handbook describe a variety of sketch
planning techniques which are available for transportation-air quality
analysis. The usefulness of each technique depends on many factors,
including data availability, staff skills, study schedule and deadlines,
and the particular measures being considered. Thus, one should not
expect that one or two methods will be able to handle al of the analysis
tasks facing the typical MPO. Rather, an overall program for planning
and analysis must be developed, utilizing an array of sketch planning
approaches to meet the specific conditions at hand. To do a credible job
of analyzing a range of transportation-air quality measures requires that
careful priorities be set for both the specific measures, or combinations
of measures, that are considered and the specific techniques, or
combinations of techniques, that are used at various stages of the
planning process.
This chapter provides guidance on how to develop a work program for
the analysis of transportation-air quality measures and, in particular,
how to judge the appropriateness of the various methods described in this
handbook for a specific analysis task. A number of topics are addressed,
including:
• developing an analysis strategy and selecting techniques
• representing transportation system changes in analysis techniques
• market segmentation
• data sources
• sensitivity and accuracy
-------
5-2
While it is impossible to develop a list of techniques which are always
appropriate for analyzing a given transportation measure, each of the
factors listed above influences the selection process and provides the
analyst with a set of criteria for judging the appropriateness of one or
more methods for a specific analysis task.
-------
5-3
5.1 Developing an Analysis Strategy and Selecting Techniques
A host of factors must be addressed in developing an overall
strategy. These factors include:
• overall analysis budget, including staff resources
• staff skills
• methodologies currently in use
• basic analysis approach (e.g., successive cycles of analysis and
screening of potential measures)
• study schedule
• data availability
• range of measures under consideration and relative priority of
measures
• stage of planning for each measure (i.e., initial consideration
versus detailed design)
• type of impacts expected and impacts of most concern
• level of detail required including degree of market segmentation,
facility versus corridor or areawide impacts, etc.
Generally, all of these factors will place constraints on the techniques
which can be used in a particular context and will dictate the subset of
techniques from which the most appropriate method, or methods, for
analyzing specific measures will be chosen.
The first five factors represent key constraints on the development
of an overall analysis strategy. Budget and staff resources may limit
the number of measures examined, the impacts considered, and the level of
analysis detail, as well as the specific methodologies employed.
Staff skills and the analysis methods currently being used also may
influence the anaysis strategy adopted. If there is limited familiarity
with computer techniques, the simpler calculator and manual methods will
-------
5-4
be more appropriate. Similarly, if staff has some experience with the
analysis or implementation of particular measures, the transfer of
experience may suffice for these measures while more sophisticated
analysis methods are used for less familiar measures. Methodologies
already being used, whether true "sketch planning" approaches or not,
will probably be adequate for addressing some measures and again may
dictate the priorities for acquiring additional analysis methods.
The basic analysis approach may vary from a detailed consideration
of a few measures to a broader initial screening of a range of measures.
In many cases, several cycles of analysis, each successively more
detailed, will be an efficient means of considering a broad range of
measures while reserving sufficient resources to analyze the most
promising measures in more detail. Again, the appropriate analysis
methods will vary depending on the approach adopted. The range of
methods documented in this handbook is sufficient to tailor the
techniques used to the cycle of analysis being performed. Transfer of
experience and other manual and simple calculator approches can be used
for an initial screening while other calculator and computer assisted
approaches may be used for more detailed "valuation. Obviously, schedule
constraints may limit the number of cycles of analysis that are
appropriate and therefore the methods that can be used. However, even
where time constraints are tight, an initial "broad brush" analysis may
be critical to setting priorities for more detailed evaluation.
Data availability is also a critical determinant of the types of
analyses that can be performed. The methods presented in earlier
-------
5-5
chapters vary widely in their data requirements, and some of the
techniques are oriented toward developing certain base case data items
were none exist. In addition, many data sources not typically used in
longer range transportation planning analyses are available for the more
flexible analysis requirements of a sketch planning approach.
Section 5.5 discusses some of these sources in more detail.
The range of measures being considered and the stage of planning
(from initial consideration to detailed design) for each measure has a
large impact on the set of analysis methods that will be required.
Figure 5.1 illustrates eight broad classes of transportation-air quality
measures and some of the specific policies or options that are included
in each category. These specific options range in scale from localized
projects to areawide programs, reflect a number of different objectives
and cover a variety of potential impacts. The broader the range of
measures being considered, the fewer the analysis resources that will be
available for any one measure. On the other hand, the closer a measure
is to being implemented, the more detailed the analysis ought to be.
Again, the benefits of performing a number of cycles of analysis,
tailored to the range of measures being considered and the implementation
status of each measure, are indicated. In general, a broad range of
measures will have to be given at least some consideration and the
implementation status of each measure will vary. Thus, a variety of
analysis methods will have to be employed. Often, a measure showing
sufficient potential during an initial screening will have to be analyzed
with increasingly detailed methods in subsequent analysis cycles.
-------
5-6
FIGURE 5.1
Representative Actions Within Eight Classes of Transportation-Air Quality
Measures
1. Automobile-Restricted Zones
• auto-free zones/pedestrian malls
• peak-period auto restrictions in congested areas
• auto restrictions with transit and pedestrian improvements
2. Priority Treatment for High Occupancy Vehicles (HOVs)
• HOV lanes on freeways
• parking privileges for HOVs
• preferential treatment for HOVs at toll booths
• signal pre-emption for transit vehicles
3. Traffic Flow Improvements
• signal timing improvements
• ramp metering
4. Transit System Improvements
• fare reductions
• bus lanes or other priority treatment
• increased frequency of service
• improvements in operations
5. Parking Programs
• peak period parking restrictions
• freeze on parking supply
• parking price structure to favor short term occupants over all-day
occupants
6. Pricing Policies to Discourage Low-Occupancy Auto Travel
• differential tolls or parking prices favoring HOVs
• congestion passes
• increased gasoline tax
7. Programs to Encourage Carpooling and Vanpooling
• areawide ridesharing promotion and matching programs
• employer-based ridesharing programs
• incentives for carpools and vanpools, including preferential parking,
preferential treatment, no-toll policies, etc.
8. Alternate Work Hours
• four day, 40-hour work week
• flextime
• staggered work hours
-------
5-7
In reviewing the range of measures that must be analyzed in a
particular study, key criteria for selecting a set of analysis methods
are the type of impacts anticipated from the measures and the degree of
detail desired in the results of the analysis. Throughout this handbook,
techniques have been classified as being primarily demand (travel
behavior), supply (facility operations^, or non-transportation impact
(e.g. emissions, energy, etc.) oriented. As described in Chapter 1, this
classification was adopted because many transportation-air quality
measures can be analyzed by either primarily focusing on demand effects
with less emphasis on facility operations, or primarily on facility
operations with less attraction to traveller behavior. For example,
traffic flow improvements and certain HOV priority treatment measures may
affect primarily facility operations and vehicular flow, while areawide
parking programs, pricing policies, and carpool/vanpool promotion
programs may affect primarily traveller behavior. If either facility
operating or traveller behavior effects are dominant, particularly for
initial screening of a range of measures, one or more of the specific
demand or facility operations techniques described in this handbook can
be selected.
Analyses of traffic operations and travel demand often will proceed
independently. As a result, demand estimates may depend on
level-of-service variables derived from network assignment approaches
which employ only order-of-magnitude values of speed stratified by a few
broad functional highway classes, or a set of volume versus (average)
speed curves representing different highway classes. Often, these
-------
5-8
estimates are based on personal judgment and scanty empirical data (often
from outdated traffic studies) or are obtained from previously coded
transportation networks; seldom are rigorous validations of the speed
estimates undertaken. On the other hand, highway analyses usually rely
on entry to exit flows within the study segment, without regard to actual
points of origin and destination. It is difficult to obtain the flow
data without detailed field measurements and virtually impossible to
predict the effects of operational changes on travel demand. In many
cases, these simplifications are justified in a sketch planning
analysis. However, for measures for which both traveller behavior and
facility operations impacts are expected to be significant, an approach
to analyzing supply/demand equilibrium must be developed. For example,
designation of an HOV lane may increase both transit and shared-ride mode
shares while creating significant traffic flow impacts in the lanes left
open to all traffic.
Two approaches for developing an equilibrium analysis framework are:
• use a supply-oriented method and a demand-oriented method
iteratively until demand and facility operating impact
estimates are consistent
• use one of several methods with a more integrated treatment
of supply/demand effects
To represent a measures such as the HOV lane discussed above, for
example, initial travel time estimates for high occupancy vehicles and
other vehicles using the facility would be used to predict changes in the
number of vehicles in each category that would use the facility. From
these revised volumes, new travel times could be estimated, and changes
in volumes predicted again. This procedure would continue until the
-------
5-9
difference in travel time estimates for two successive iterations was
reduced to within an acceptably small value. Other measures requiring an
iterative procedure are those involving resource constraints, such as
parking supply, gasoline supply, etc. With these measures, a "shadow
time or dollar price" is usually introduced to discourage use of the
constrained resource. The magnitude of this shadow price is increased
until the demand for the particular resource being constrained matches
the amount supplied.
Several of the methods described in this handbook provide a more
integrated treatment of facility operations and demand analysis. For
example:
• The FREQ series of models uses specified origin-destination
flows for the base case, but predicts demand impacts through
a set of elasticities derived from logit mode choice models
enumerated on a sample of corridor travellers (16, Uo).
• A recent application uses demand and supply elasticities at
the regional level to predict the joint consequences (and the
new equilibrium) resulting from improvements in highway
supply (71). With the proper determination of elasticities,
it also could be applied to corridors or to more
broadly-defined sub-areas of the region.
• A recently developed corridor analysis approach uses a
FREQ-type facility model (with multimodal representation),
approximate service functions for local streets, and an
equilibration scheme (6l).
These recent developments reveal a wide range of possibilities for
integrating demand and facility operations analysis. Depending upon time
and resource availability, and on the type of measure under
consideration, a wide range of approaches might be appropriate.
All of the factors discussed above must be considered in selecting
techniques to analyze specific measures. In general, it will be
-------
5-10
desirable to utilize a range of techniques in several cycles of
analysis. Obviously, in selecting methods that are consistent with the
overall analysis strategy adopted, it is necessary that the techniques be
able to represent the measures being considered. In short, the variables
included in a method must be capable of representing the significant
changes caused by a measure (e.g., change in in-vehicle travel time,
out-of-vehicle travel time, out-of-pocket costs, etc.). Table 5.1
provides a summary of the measures that each method, described in detail
in this handbook, can address. The best method in any particular context
will depend on all the issues discussed earlier as well as the specific
design of the measure being considered. Thus Table 5.1 is intended only
as a rough guide.
The manual demand methods are generally not appropriate for
auto-restricted zones, parking measures (especially supply constraints)
and staggered work hours. Some manual demand methods do not explicitly
deal with carpooling and/or shared-ride auto usage as a unique mode.
This capability is a necessity for the study of high occupancy vehicle
priorities, pricing policies to discourage low occupancy vehicles, and
carpool/vanpool incentives. The work mode choice worksheet methods are
applicable and appropriate for analyzing the largest number of
strategies, largely due to their explicit treatment of shared-ride auto
travel. Both pivot point (to predict changes from a base case) and
synthetic (to predict absolute levels) methods are useful, depending on
data availability.
-------
5-11
TABLE 5.1
Applicability of Demand and Facility Operations Analysis Methods
AL'TO
RESTRICTED
METHOD ZONES
Manua 1 Demand
PIVOT POINT MODE CHOICE (£
SYNTHETIC MODE CHOICE (£
QUICK RESPONSE ESTIMATION (£
SYSTEMATIC DATA ANALYSIS (J
Calculator Demand
HHCEN 0)
2MODE-AGC (£
3MODE(VAN)-AGG (£
Computer Demand
CAPM 3
SRGP O
TRANSIT SKETCH PLANNING (£
Manual Facility Operations
TRAFFIC FLOW FORMULAE 3
GRAPHICAL TECHNIQUES (B
AREAWIDE TRAFFIC ENGINEERING (J
TRANSFER OF EXPERIENCE (J
Calculator Facility Operations
BUS •
Computer Facility Operations
TRANSYT (J
FREQ Q
Manual Impacts
EMISSIONS WORKSHEETS ()
FUEL CONSUMPTION/OPERATING COSTS Q
Calculator Impacts
BUSPOL/F.NERGY 3
Computer Impacts
MOBILE1 O
HUV
PRIORITIES
O
3
•
3
3
•
3
3
G
•
3
d
O
3
O
O
G
0
0
G
O
TRAFFIC
FLOW
IMPROVE-
MENTS
3
3
«
3
G
G
G
3
G
G
3
3
G
«
•
G
G
O
O
3
O
TRANSIT
SYSTEM
IMPROVE-
MENTS
3
G
»
3
G
G
G
•
G
G
«
«
0
3
O
G
G
G
G
O
G
PARKING
PROCRAM3
3
•
«
3
3
3
3
•
3
3
3
3
«
« "i
•
G
G
3
3
«
G
PRICING
POLICIES
G
G
•
3
G
3
G
•
G
•
3
3
•
0
•
G
G
G
G
3
G
CARPOOL/
VANPOOL
INCENTIVES
G
3
•
3
G
•
G
•
G
•
«
^9
«
3
•
G
G
O
O
3
0
STAGGERED
WORK HOURS
«
9
•
3
3
3
3
•
G
3
«
1 3
«
3
•
O
G
G
G
O
3
KEY
GENERALLY APPLICABLE
GENERALLY NOT APPLICABLE
-------
5-12
-------
5-13
The calculator demand methods are useful for analyzing HOV priority
treatments, traffic flow improvements, transit system improvements, and
pricing policies. The effects of these measures generally can be
translated readily into changes in modal levels of service. The mode
choice methods are well-equipped to predict the effects of modal level of
service changes.
For several of the measures, calculator methods can provide only
partial results. Parking programs are a good example. On one hand, most
parking programs can be translated easily into changes in modal level of
service; on the other hand, it is difficult to account for the effects of
parking supply on modal level of service changes. Carpool and vanpool
incentives generally are handled by calculator methods, but there are
some difficulties in specifying the impacts of carpool incentives which
do not directly change travel times and costs.
Computerized demand methods are judged to be generally appropriate
in terms of the variables and relationships represented, for the widest
range of measures. Measures oriented toward high-occupancy vehicles have
the fewest applicable methods. The method which deals specifically with
the most general breadth of applicability is SRGP. However, SRGP is
limited, generally, to large-area measures rather than localized changes.
Facility operations analysis methods are critical for assessing
measures and often are important in evaluating ARZ's, transit system
improvements, and staggered work hours. Both the manual and calculator
supply methods are only sometimes appropriate for most measures,
primarily because of their orientation to isolated locations rather than
-------
5-14
to interrelated systems covering the larger geographic areas of concern
in a transportation-air quality analysis. Methods of areawide
integration of spot or project-oriented p-ocedures are valuable in
aggregating their results up to the level required for use in larger
geographic scale analyses, but may be limited in appropriateness in some
cases if calculator and manual methods are relied upon, rather than
computerized techniques, for developing necessary input elasticities.
The computerized supply methods are all generally appropriate for
analyzing all measures, but it is recognized that the costs of data
collection for these models still limits their value for true sketch
planning purposes.
-------
5-15
5.2 Representing Transportation System Changes in Sketch Planning Methods
An important criterion for the selection of specific sketch planning
methods is the ability of a method to adequately represent the
transportation system changes caused by the measure or measures being
studied. In order to adequately represent a measure, the model variables
in a method must encompass all the significant transportation changes
anticipated. For example, the analysis of the travel demand response to
a transit fare or schedule change must be done using a method which
includes these transit 1eve1-of-service variables. The fare increase
would be represented as an increase in transit out-of-pocket costs while
an increase in frequency of service would be represented as a decrease in
travel time (i.e., a decrease in out-of-vehicle wait time).
If a method does not include those variables likely to be affected
by a particular measure, then in general it cannot "represent" such a
measure and is an inappropriate technique. In some cases, however, it is
possible to represent a measure adequately even if the precise variables
affected are not included. In such instances, the variables included in
a method must be used as proxies for other changes that occur. For
example, parking management measures may affect the supply of parking as
well as out-of-pocket costs and travel time. Such supply constraints
often can be represented as "shadow cost or dollar prices" on locations
where supply is limited. Thus an artificially high parking cost or
travel time increment is assumed for the affected locations until the
number of travellers selecting those parking locations is consistent with
the supply constraint assumed. Travel time increments will be preferred
-------
5-16
where travel costs have different impacts on different income groups; for
example, in models which obtain a cost variable as travel cost divided by
household income. In these cases, shadow time costs can be based on
available research on the value of time for the average traveller,
providing a time/cost tradeoff in which travel time penalities are used
to represent "shadow price" increases. In the same way, different
components of travel time can be adequately represented by understanding
the tradeoff between in-vehicle and out-of-vehicle time.
For those measures that can be expressed in terms of changes in
travel time and cost, the representation is relatively straightforward
and consists of the following three steps:
1) Identify those travellers affected by the measure (by income,
mode, corridor facility, etc.)
2) Identify how these travellers are affected in terms of changes
in specific level of service variables.
3) Estimate the magnitude of the level of service changes and the
resulting changes in travel demand or facility operating char-
acteristics.
For most measures, those travellers directly affected are easily
identified, although with some techniques complications may arise if the
measure is to be implemented on a very small scale (e.g. a single
facility, one or two blocks in the CBD, etc.). Some measures, though,
may have significant indirect effects as well. For example, CBD parking
restrictions would have a direct impact on those individuals currently
parking within the CBD. While it would be reasonable to expect that many
of these individuals would switch to modes other than auto, it would also
\
be reasonable to expect that some would use less convenient parking in
-------
5-17
fringe areas, increasing the demand for spaces in those areas. Those
currently parking in these fringe areas, then, would experience an
indirect impact of CBD parking restrictions as fringe area parking became
increasingly difficult to find.
When identifying how individuals are affected by a particular
measure, it is important to consider the alternative choices available to
these individuals. For example, if parking rates are increased for
off-street facilities in an area with little free parking available, the
appropriate model representation would be the increased rate. If,
however, a significant amount of free parking were available in less
convenient locations, many individuals may choose to walk farther rather
than pay the rate increase. For these individuals, the appropriate
representation would be an increase in walk time.
Consideration of alternative responses is also important when
estimating the magnitude of level of service changes. For example, if an
existing lane on a heavily congested, major expressway is to be reserved
for high occupancy vehicles during peak periods, there are a number of
ways that those directly affected could respond. One response, of
course, would be that intended by such a measure: people would form
carpools or use transit to take advantage of decreased travel times
afforded by the reserved lane. If this were the only response allowed,
the resulting changes in travel times might be quite large. It is
likely, though, that many individuals would prefer to use an alternative
route rather than change modes. Still others may have the flexibility in
choosing their work hours to be able to travel outside the time period of
-------
5-18
the HOV restriction. To the extent that these alternative responses
occur, the impact of the HOV lane in terms of travel time changes would
be diminished.
There are a number of measures whose effects cannot be completely
expressed in terms of travel time and cost. For example, while auto
restricted zones typically result in travel time changes, the enhanced
attractiveness of such areas is difficult to quantify. Similarly,
variable work hours measures are difficult to represent only using
available sketch planning techniques. To analyze such measures, then,
empirical evidence such as operational experience also must be relied
upon.
-------
5-19
5.3 Market Segmentation Guidelines
Various factors apart from cost and level of service can cause
different subgroups within the population to respond differently to
transportation alternatives. Market segmentation is a procedure for
dividing, or classifying, the population into a number of discrete groups
which exhibit at least some degree of homogeneity with respect to the
characteristic(s) of interest. Such classification facilitates analysis
of the differential impacts of various measures across different
subgroups within the population. The impact of a transportation measure
on the travel behavior of individual subgroups or market segments is
estimated as a function of changes in cost or level of service. The
total aggregate forecast is the weighted sum of the forecasts for each
group, where the weights are the fractions of the total analysis
population represented by each group. The calculation of separate
predictions for various population subgroups not only improves
forecasting accuracy but also enables the user to identify measures that
would have an inequitable distribution of impacts (e.g. on low income
classes).
Among the manual demand techniques which have been reviewed, market
segmentation as discussed here is feasible only in the case of the
pivot-point and synthetic worksheet methods. The nomographic/formula
techniques are not sufficiently flexible to allow for further market
segmentation beyond that which is already incorporated into the
-------
5-20
procedure, and more informal estimation techniques generally are
applicable only at a broader, more aggregate level than that suggested in
this discussion.
The development of a market segmentation approach for applying the
worksheet methods involves a basic tradeoff. On the one hand, the more
dimensions along which an urban area population can be classified and the
finer the degree of classification, the greater the accuracy and
reliability of the resulting predictions. On the other hand, very
complex classification schemes require a corresponding increase in the
number of computations, increasing the likelihood of an error being made
and possibly becoming too complicated and time consuming. Further, base
modal shares or even the size of individual classes may not be known with
any degree of certainty for a very detailed classification.
Among the classification schemes most appropriate for the
pivot-point mode choice method are:
• Modal availability (e.g. transit and vanpool availability)
• Socioeconomic characteristics (e.g. income, auto ownership)
• Trip orientation (e.g. CBD-destined, suburban)
• Pattern of level of service impacts
Modal Availability--Under this classification scheme, the analysis
population is categorized according to the travel modes available to the
household. For example,
• Drive-alone and shared-ride
• Transit, drive alone, and shared-ride
• Transit and shared-ride
• Drive-alone, shared-ride and vanpool
-------
5-21
Classification by modal availability is particularly important. If a
classification of this nature is not used, the implicit assumption is
that all modes are available to the entire population. In instances
where this assumption is invalid, the failure to classify market segments
by modal availability can severely bias the analysis results.
Socioeconomic Characteristics--Classification by relevant
socioeconomic characteristics reduces aggregation bias and allows
explicit evaluation of differential impacts by factors such as income
level or auto ownership. To be of maximum effectiveness, the
classifications developed should be based on characteristics which
produce the greatest variation in travel patterns. For example, if work
trip mode choice varies significantly across groups with different auto
ownership levels but remains relatively unchanged across different income
groups, then auto ownership levels should be used to classify groups.
Regardless of the classification system selected for socioeconomic
characteristics, data on existing shares and on changes in transportation
level of service for each group are required. Practical problems of data
availability may then influence formulation of a classification system.
Trip Orientation—If there is considerable geographic variation in
travel patterns, classification of the analysis population by trip
orientation can prove effective. While the trip orientation
classifications most appropriate for a particular urban area will depend
on the area's unique distribution of travel patterns, the greatest
differences for many urban areas are found between suburban-to-CBD
-------
5-22
oriented trips and suburban-to-suburban trips. The former generally have
a higher transit share due to better transit level of service relative to
auto. Similarly, some differences in travel patterns may exist between
individual travel corridors.
Pattern of Level of Service Impacts—The level of service impacts of
a particular measure may differentially affect various transportation
facilities or services and thus differentially impact their users.
Market segmentation by users of specific facilities or services may be
appropriate in such cases, particularly if facility operating impacts are
expected to be significant or of primary concern. For example, when a
high-occupancy vehicle (HOV) lane alternative is analyzed in a travel
corridor, separate market segments for freeway users and non-freeway auto
users should be considered, since those not using the freeway will only
be affected if they change their travel route as a result of the new HOV
lane.
-------
5-23
5.U Factors Influencing the Accuracy of Sketch Planning Techniques
Any planning decision based on a forecast of future conditions must
consider both the predicted impacts of alternative actions and the degree
of confidence which can be assigned to the prediction. The analyst must
keep in mind the limitations of the techniques being used and the factors
which may determine the reliability of the results of the analysis.
Four factors determine the accuracy of impact predictions:
• Accuracy of the input data items
• Level of detail of the analysis
• Appropriateness and reliability of the technique being used
• Transferability of techniques
Depending on the urban area, the measures being analyzed, and the
technique in use, the relative importance of the above factors may vary.
However, in any analysis, each potentially can play a key role in
determining the usefulness of the results.
-------
5-"'
5.^.1 Input Data Accuracy
Although the specific data elements required by the various sketch
planning techniques may differ, each requires data from the following
categories:
• Travel market segment or population subgroup proportions
• Base case data
• Transportation system changes
Travel market segmentation is a key to the development of an
accurate analysis approach because different segments of the population
may be expected to be affected in different ways by transportation-air
quality measures. The accuracy with which the population can be divided
into subgroups or market segments will have a direct influence on the
accuracy of the resulting impact forecasts. The accuracy of market
segment proportions depends on the availability of up-to-date
socioeconomic data (particularly auto ownership and income breakdowns)
and a careful assessment of the groups which will potentially be affected
by a given measure. For future-year analyses, accurate forecasts of the
key variables defining market segments are required. If forecasts must
be used, it is likely that the market segmentation approach will be
limited by data availability.
Once market segments have been identified, an accurate
representation of the base case (existing conditions, or future
conditions before any of the measures under analysis take effect) is
required. Depending on the availability of data or previous forecasts,
some elements of the base case data set may have to be estimated as part
of the sketch planning procedure. Some of the techniques listed in this
-------
5-25
handbook include procedures for estimating base-case travel behavior from
socioeconomic land use and transportation level-of-service data.
Inaccuracies in the data used in these procedures and simplifications
inherent in the models themselves are potential sources of error in the
base data and the subsequent impact analysis.
The use of pivot-point or incremental techniques avoids the problem
of predicting base case travel behavior within the sketch planning
procedure and removes the requirement for base transportation
level-of-service estimates--both of which are potential sources of
significant error. However, their use depends on the availability of
data on travel mode choice, and other trip characteristics which are
truly representative of the base conditions for the impact analysis. An
assessment of the likely accuracy of the results of any analysis must be
based on the confidence which may be placed in the base case travel
behavior and socioeconomic data, regardless of whether its source is a
sketch planning analysis or previous planning work.
The final class of input data required by the sketch planning
techniques is the transportation level-of-service changes which are
associated with the measures being analyzed. Potential sources of error
in the level-of-service impact estimates include:
• under- or over-estimation of the magnitude of the level-of-
service impacts of measures,
• failure to consider certain types of impact, and
• inaccuracies in the allocation of impacts among the different
groups (market segments) under analysis.
These errors can be minimized by a careful review of the implications of
each measure for transportation system performance.
-------
5-26
5.1.2 Level of Analysis Detail
In general, if the data are available to support it, a more detailed
analysis will be more accurate (and also more expensive) than a simple
one. For sketch planning techniques, the level of analysis detail is
determined largely by the degree to which the travel market is
segmented. Given a constant level of accuracy in the input data, the
more markets or segments identified, the more accurate the resulting
forecast.
Three considerations jointly determine the degree to which market
segmentation is used in an analysis:
• Required degree of accuracy
• Budget available for analysis
• Data available for market segments
In many cases the constraint on market segmentation may be the ultimate
determinant of analysis detail. Although identifying additional market
segments may be beneficial, if too many assumptions must be made to
develop the base data for these market segments, little or nothing will
be gained in terms of accuracy over a simpler market segmentation
scheme. This is not to say that additional market segments should not be
identified if assumptions are required to develop base case data for each
market segment. If most of the required data items are available and
simple assumptions can be used to calculate the missing items from more
aggregate base data, it will no doubt be beneficial to proceed with the
more extensive segmentation scheme. A certain degree of judgment must go
into the design of the market segmentation scheme. The validity of the
analyst's judgment may be tested in later stages of the analysis through
the use of sensitivity tests, as described in a later section.
-------
5-27
Along with the number of market segments identified, the type of
classification scheme employed can have a significant impact on the
accuracy of predictions developed with sketch planning techniques. As an
illustration, consider a proposal to increase bus frequency by 50 percent
on all routes in a large city. Although classifying the travel market by
auto ownership level would be useful, in this case a more appropriate
segmentation scheme would identify and separate those groups which would
be affected by the service improvement from those who would not.
(Households living in sections of the city not served by transit would
not be affected by the proposed change at all.) In developing a market
segmentation scheme, it is important to identify those variables which
determine how the population will be affected by the measures being
evaluated and to classify the market using those variables.
-------
5-28
5.4.3 The Reliability of Travel Models
Many analysis methods are based on travel models estimated using
empirical data. But no model can represent the real world perfectly.
Even given very accurate and detailed input data, some error is inherent
in the use of any model. Assessing the magnitude of this error is
difficult because of the interrelation between input data accuracy and
model reliability in determining the accuracy of resulting predictions.
Each of the techniques described in this manual has been tested and found
to be reasonably accurate relative to other techniques available for
sketch planning analysis. In terms of their reliability, many compare
quite favorably with any existing technique, as they are based on the
same theory found in extensive computer-based model systems. In most
instances, the accuracy and level of detail of the data available for use
in an analysis will have by far the greatest impact on the accuracy of
sketch planning impact estimates.
Although the error magnitudes of models are difficult, if not
impossible to assess, the error associated with the models described here
can be controlled to some degree by ensuring that the most appropriate
model is applied to the measure being analyzed. Ensuring that the model
is sensitive to both the 1evel-of-service and travel behavior changes
implicit in a given transportation-air quality measure is the first step
in finding the most appropriate technique. Section 5.2 provides a more
in-depth discussion of the applicability of the sketch planning
techniques to specific measures.
-------
5-29
S.M.U Sensitivity Testing
While error and uncertainty can be controlled to some degree through
careful application of the sketch planning models, they can never be
eliminated. In order to assess the probable magnitude of the error of a
given impact prediction, sensitivity analyses may be employed.
Sensitivity analysis involves redoing all or part of a completed analysis
using different values for specific data items or model coefficients
while holding all other values constant. In this way, the sensitivity of
the impact assessment to an assumption which was necessary to perform the
*
analysis or the value of an input data item which is uncertain can be
assessed. A sensitivity analysis can be conducted for a single item or
several simultaneously. (Best case and worst case analyses are examples
of simultaneous sensitivity tests.)
Although sensitivity tests can be conducted for all of the data
items or all the model coefficients used in a particular analysis, it is
likely that time and budget constraints will preclude such an approach.
In order to assure that the sensitivity analyses conducted are useful,
several guidelines should be followed:
• Test those data items which were estimated with the highest
degree of uncertainty or for which the most significant
assumptions were required.
• Test those variables which will have the greatest impact on
the predicted results, as reflected by the size of the model
coefficients. Other important items include market segment
sizes and base mode shares. Highest priority for sensitivity
testing should be assigned to items fitting into both of the
above classifications.
• Confine the sensitivity tests to values which may reasonably
be expected to occur for the input data items.
-------
5-30
• Test only the model coefficients relating directly to the
level-of-service change anticipated for the measure under
analysis.
• Where two variables are expected to be of similar importance,
test the variable coefficient which is less statistically
significant first.
A final guideline for sensitivity testing, and for the evaluation of
analysis accuracy in general, is to match the level of effort to the
requirements of the task being done. For example, during the initial
stages of an evaluation of a number of transportation-air quality
measures, less accuracy may be required than when detailed feasibility
f
studies are being conducted for one or only a few measures.
-------
5-31
5.1.5 Transferability
The reliability of transportation analyses depends to a great extent
on the quality of travel demand predictions. Many urban areas would find
it desirable to improve their travel demand modelling capabilities in
this regard. However, in many cases, the development of new models would
require a substantial proportion of the total short-range analysis
budget—in some cases, it could exceed it. Such resource considerations
have motivated the development of a methodology for using a travel demand
model developed for one urban area in other settings. This process is
called "transferring" a model.
To be transferable, it is not enough that a model merely "fit"
existing data; the model must be able to explain how travel behavior will
change as conditions change. Explaining how behavior will change is much
more likely if we understand why it changes; in other words, we wish to
have an understanding (or at least a theory) of the underlying
causality. Thus, rather than just correlating existing travel behavior
with socioeconomic characteristics (e.g., retail employment in a zone)
and transportation level of service, the model "specification", that is,
the set of variables included in the model and the relationships among
the variables, must represent the causal relationships between these
variables. Otherwise, there would be no reason to expect a model to
perform well under different conditions, for even a major change in a
spuriously correlated variable would not change travel behavior, while a
change in an omitted causal factor could be significant (but would not be
captured). Thus, causal specification of a model would seem to be a
precondition to its consideration for transfer.
-------
5-32
From a practical point of view, however, no model is perfectly
specified. Often, variables which should be included in the model must
be excluded (e.g., when not enough data are available). When data for
model development are taken from a single urban area, there may be some
cultural differences between urban areas which are not explicitly
represented in the model. The peculiarities of the data will be
implicitly hidden in the model coefficients which are estimated with the
data. Thus, no model developed on such data can be expected to be
completely transferable.
Luckily, "perfect" transferability is unnecessary; the model needs
only to be sufficiently accurate for practical transportation planning
needs. Hence, it is possible that models which are statistically
"non-transferable" may be quite adequate for some practical
applications. In this sense, transferability is not an absolute to be
accepted or rejected, rather, it is a relative attribute dependent not
only on differences between coefficients but also on the particular
application for which the model is intended.
As an example of a test of the transferability of disaggregate
demand models, the coefficients of a mode choice model developed on 1968
Washington, DC data were re-estimated on data sets representative of New
Bedford, Massachusetts in 1963, and Los Angeles, California in 1967 (2).
The model coefficients resulting from this re-estimation were compared to
those originally estimated for the model. A multinomial logit mode
choice model was used in this test (5). This model predicts the
probability of a commuter driving alone, sharing a ride (i.e., two or
-------
5-33
more persons in a car), or using transit for the home-to-work trip. The
model specification is given in Figure 5.2. The coefficients were
estimated using the maximum likelihood method with HIM observations from
a 1968 home-interview survey conducted in Washington, DC. The model
coefficients and statistical properties are shown in Table 5.2.
In "transferring" this model to both the New Bedford and Los Angeles
data sets, the specification of the independent variables is identical to
that of the original model, with the exception that both CBD variables
and the government worker variable are excluded. In the case of the CBD
variables, the congestion and inconvenience associated with driving into
the CBD of a large, dense city such as Washington, are very real factors
in choosing between auto modes and transit. In a small city such as New
Bedford or a very diffuse city such as Los Angeles, however, the
distinction between CBD and non-CBD trips would probably have little
added effect on this choice. Therefore, the CBD variables are assumed to
have a value of zero. Similarly, the effects of large organizations
offering carpool incentives did not exist in either New Bedford or
Los Angeles at the time the surveys were administered and therefore the
government worker variable takes the value of zero for both of these
cities.
The estimation results, the coefficients and statistics of the
models estimated on the New Bedford and Los Angeles data sets, are shown
in Table 5.2 along with the original Washington coefficients. The
coefficients of the New Bedford model all have the correct signs. The
»t' statistics, however, are not nearly as large as those of the original
-------
5-34
FIGURE 5.2
Work Mode Choice Model;
Definition of Variables
Variable Code
1. D
c
2. D
s
3. OPTC/INC
4. 1VTT
5. OVTT/DIST
6. AALD
c
7. AALD
s
8. BW
c
9. GW
s
10. DCITY
c
11. DCITY
Definition
s
12. DINC
c,s
13. NWORK
14. DTECA
s
1, for drive alone
0, otherwise
1, for shared ride
0, otherwise
round trip out-of-pocket travel cost (in cents)/
household annual income (in dollars)
round trip in-vehicle travel time (in minutes)
round trip out-of-vehicle travel time (in minutes)/
one way distance (in miles)
# of autos/licensed drivers, for drive alone
0, otherwise
# of autos/licensed drivers, for shared ride
0, otherwise
1, if worker is the breadwinner, for drive alone
0, otherwise
1, if worker is a civilian employee of the federal
government, for shared ride
0, otherwise
1, if work place is in the CBD, for drive alone
0, otherwise
1, if work place is in the CBD, for shared ride
0, otherwise
household annual income—80 x # of persons in the
household (in $), for drive alone and shared ride
0, otherwise
# of workers in the household, for shared ride
0, otherwise
employment density at the work zone (employees per
commercial acre) x one way distance (in miles), for
shared ride
0, otherwise
Alternative: c = drive alone, s = shared ride (car pool), t = transit
-------
Transferability
Variable
New Bedford
D -2.198
C (-2.648)
D -1.535
8 (-1.535)
OPTC/INC -87.33
(-1.576)
IVTT - .0199
(- .4849)
OVTT/DIST - .1013
(-2.903)
AALD 2.541
C (3.674)
AALD . 4499
S (.8478)
BW 1.026
C (3.769)
GW
s
DCITY
c
DCITY
s
DING .000072
°'S (1.279)
NWORK .1874
S (1.249)
DTECA .00060
S (.7665)
// of observations 453
# of alternatives 1208
Log Likelihood
at zero -436.4
Log Likelihood at
convergence -256.5
5-35
TABLE 5.2
of Work Mode Choice Model to Different
Coef f icient/ (t-statistic)
Cities
Washington Los Angeles
-3.24
(-6.86)
-2.24
(-5.60)
-28.8
(-2.26)
- .0154
(-2.67)
- .160
(-4.08)
3.99
(10.08)
1.62
(5.31)
.890
(4.79)
.287
(1.78)
- .854
(-2.75)
- .404
(-1.36)
.00007
(3.46)
.0983
(1.03)
.00063
(1.34)
1114
2924
-1054.0
-727.4
-2.746
(-4.85)
-1.830
(-3.95)
-24.37
(-2.07)
- .01465
(-2.25)
- .1860
(-4.02)
3.741
(7.19)
.6093
(1.58)
.8101
(3.28)
-
-
—
.000083
(2.31)
.0810
(.46)
.00027
(2.23)
879
2549
-930.0
-391.2
-------
5-36
model, although for only three coefficients (in-vehicle travel time,
shared ride auto availability, and employment density times distance) are
they seriously low. This relatively poor statistical performance may be
partially linked to smaller' sample size (453 versus 1114 observations)
and the much lower variability observed for several of the level of
service variables due to data limitations. (Data were available only for
those trips with both origin and destination within the city of
New Bedford itself.) For example, average one-way distance (as a proxy
for travel time and cost) in the Washington sample was about eight
miles. In the New Bedford sample, however, average one-way distance was
only two miles. In this case, overall statistical performance is much
better than the New Bedford model, due in part to the larger sample size
(879 observations).
A comparison of the three sets of coefficients indicates that they
are remarkably similar. The signficance of the differences in
coefficient values that do exist between models can be evaluated by
transforming the coefficient values into elasticities. The elasticities
which correspond to the Washington, New Bedford and Los Angeles
coefficients are presented in Table 5.3.
Except for the travel cost elasticity for the New Bedford data, all
other elasticities are sufficiently similar to warrant the conclusion
that even if the model as a whole (and in particular the model constants
and socioeconomic variables' coefficients) is not transferable, the
level-of-service elasticities evaluated using the Washington coefficients
are transferable.
-------
TABLE 5.3
Comparison of Elasticities for Three Mode Choice Models
mmmmmmmmmmmmmmmmmmmmmmmmmm
Elasticities Calculated for the
New Bedford Sample Means
Variable
Out-of-Pocket Cost
Drive Alone
Shared Ride
Transit
In-Vehicle Travel Time
Drive Alone
Shared Ride
Transit
New Bedford
Coefficients
-.092
-.057
-.852
-.138
-.315
-.373
Washington
Coefficients
-.030
-.019
-.281
-.107
-.243
-.288
•••••••••••••••••
Elasticities Calculated for the
Los Angeles Sample Means
Los Angeles
Coefficients
-.047
-.092
-.378
-.026
-.664
-.653
Washington
Coefficients
-.055
-.109
-.447
-.027
-.695
-.684
Out-of -Vehicle Time
Drive Alone
Shared Ride
Transit
-.138
-.189
-1.55
-.217
-.299
-2.46
-.027
-.122
-1.024
-.023
-.105
-.881
Ul
I
OJ
-------
5-38
In summary, the fact that the original specification resulted in a
reasonable model in other areas, together with the fact that both the
sets of coefficients taken together and key elasticities are not
significantly different, is encouraging. The differences between several
of the coefficients indicate areas where more research on improved
specification may prove fruitful, and demonstrate that the comparison of
coefficients estimated for two different data sets is a very powerful
method of detecting specification errors. But, whatever improvements are
implemented, no model will be perfectly specified and therefore perfectly
transferable, and hence the motivation for the application of "updating
procedures"--correction factors--for the model coefficients.
By far the simplest approach for transferring a model from one urban
area to another, and certainly the one requiring a minimum level of
effort, would be to use the existing model with its original
coefficients. In doing this, the assumption being made is that all
factors relevant to the choice process are embodied in the model, an
assumption which will never by fully justified. For example, the
specifications of most models available today contain constant terms to
account for all other factors not explicitly explained by the model
(comfort, convenience, safety, etc.). The presence of these constants
indicates that in fact the model has not captured all aspects of the
choice process. Because these "other factors" can vary between areas,
the value of such a constant estimated in one area may or may not be
appropriate for another area. Therefore, although there is theoretical
justification for transferring the relationships estimated between time,
-------
5-39
cost, income, auto availability, etc.—the variables which theory says
influence travel behavior—no theoretical basis exists for transferring
these constant tenns.
Fortunately, in most applications data are available on existing
conditions; the model will be used to predict changes in travel behavior
as a result of changes in the independent variables. In these
applications, only those coefficients associated with the specific
variables to be changed need be used. For incremental predictions such
as the manual and calculator-based "pivot-point" techniques, there is no
need to transfer constant terms.
In some situations, however, data on the existing conditions are not
uniformly available at the required level of detail and the constants
must be modified. For example, observed transit mode share may be 7
percent, but the transferred model predicts 2 percent. If extensive
information were available, it could be used to determine which variables
were causing the problem, and therefore should be adjusted. Without such
data, a suitable approach might be to use the existing model with
adjustments of the alternative-specific constants, variables 1 and 2 in
Figure 5.2. In this approach, the coefficients are accepted without
change, and either aggregate data on travel patterns or a disaggregate
sample with observed choices from the new area would be used to adjust
the constants, better reflecting the existing situation. The adjustment
is performed by applying the model to the new area in the same way in
which it will be applied for forecasting. With aggregate data, model
predictions are aggregated to the level for which observed data are
-------
5-40
available. The constants are then adjusted until the model replicates
existing aggregate data. With disaggregate data, choice probabilities
for each observation in the sample are summed and compared with the
summed observed choices. Constants then are adjusted until predicted
shares match those observed. This is the procedure that has been used in
l
adapting SRGP to a number of urban areas (7). A number of other, more
complex updating procedures are available, but their application in a
sketch-planning context is somewhat limited (2).
following equation can be used to obtain an approximation of
the required change in a constant term:
„ DBS PREDK
Ar ~ On E • E.
ACm * ^n PREDm OBSb
ACm a the required change in the constant term
associated with alternative m.
b = observed shares of alternative m, and of
alternative b, which is that alternative with
no constant term.
b = predicted shares of alternatives m and b.
-------
5.5 Data Sources for Air Quality Analysis
Many of the techniques described in this handbook are flexible in
terms of their analytical detail and input data requirements. In order
to make the best use of these techniques, the analyst must be prepared to
look beyond the traditional sources of transportation planning data. The
key to using these techniques is to make the maximum possible use of all
available data sources, by tailoring the analysis to the available data
as well as the transportation-air quality measure being analyzed.
This section briefly reviews the alternative sources of data which
may be available to urban area planners for use in air quality analyses
and suggest the ways in which they may be employed. However, because
every urban area is unique in terms of the data available for air quality
analyses as well as the transportation measures most appropriate for
evaluation, it is not possible to specify a step-by-step approach to data
gathering and impact analysis.
The guidelines presented in this section are supplemented by the
evaluation case studies presented in Volume II, which serve to illustrate
the use of a wide variety of data sources and associated analytical
techniques. Together, these general guidelines and case studies are
intended to define a style of approach to air quality impact evaluation,
the details of which are developed by the analyst to fit the requirements
of the particular problem to be addressed.
Several basic types of data generally are required for
transportation-air quality analyses:
• Existing or base case travel behavior—including travel mode
shares, trip lengths, and trip-making frequencies for different
trip types.
-------
5-42
• Vehicle miles of travel (VMT)—areawide total VMT may be of
interest, or it may be necessary to have VMT broken down by trip
type, time of day, highway facility type, location, or some
combination of the above.
• Socioeconomic data—household auto ownership, income and other
socioeconomic data may be of interest, particularly in defining
population subgroups for analysis.
• Employment data—many of the transportation measures with
potential to address air quality problems are employer-based. For
example, ridesharing programs are typically applicable at large
employment sites: in this case, employer size data is necessary
for air quality impact evaluation.
• Transportation level of service—depending on the analysis
technique being used, either base case and revised transportation
1eve1-of-service values or level-of-service changes are required.
In many ways this can be the most formidable of the data
requirements if base case data must be developed. However, this
difficulty can be overcome by using the incremental or
"pivot-point" techniques described in this handbook which rely
strictly on changes in level qf service. These changes are
relatively easy to estimate based on the specific measures being
analyzed.
• Transportation facility characteristics—the analysis of traffic
engineering actions as air quality improvement measures requires
data on the characteristics of highways and traffic control
devices. Also, planned future changes in the transportation
system must be accounted for in the analysis of any
transportation-air quality measure.
Table 5.M indicates potential sources for each of the above
evaluation data elements. Within particular urban areas, still other
sources of data may be available for air quality planning use and should
be pursued as appropriate.
The most comprehensive source of planning data in any urban area
will be previous planning and transportation modelling work. The quality
and level of detail of data available from previous efforts will in many
cases be the primary determinant of the analysis techniques to be used in
the evaluation of transportation air quality measures. For example, if
-------
TABLE 5.4
Transportation-Air Quality Data Requirements and Sources
^^\^ Data Sources
Data ^~-^
Elements ^\_
TRAVEL MODE SHARES
TRIP LENGTHS
TRIPMAKING FREQUENCIES
VMT
SOCIOECONOMIC DATA
EMPLOYMENT DATA
TRANSPORTATION
LEVEL OF SERVICE
TRANSPORTATION FACILITY
CHARACTERISTICS
AUTO OWNERSHIP
CENSUS
•
•
•
•
•
HOME
INTERVIEW
SURVEYS
•
•
•
•
•
•
•
LAND USE
PLANNING
•
•
EMPLOYERS
•
STATE
EMPLOY-
MENT •
AGENCIES
•
TRAFFIC
COUNTS
•
•
•
•
•
TRANSIT
OPERATING
DATA
•
•
•
ESTIMATION
USING
MODELS
•
•
•
•
•
•
NOMO-
GRAPHIC
FORMULAS
•
•
•
•
PREVIOUS
PLANNING
AND
MODELLING
•
•
•
•
•
•
•
•
•
L_n
U)
-------
extensive and up-to-date data is available in UTPS format and access to a
computer is readily available, SRGP (see section 2.3.2) could be
effectively employed to analyze a large number of transportation
measures. However, if relatively little or no UTPS work has been done,
setting up SRGP would be costly and inefficient, suggesting that one of
the manual analysis techniques should be employed.
While previous analytical work provides an important input to the
air quality planning process, the other data sources shown in Table 5.4
may be used to support reliable analyses, and in many cases will be
required to supplement the data from previous work. Each source is
discussed briefly below.
Census Data—Information on work travel mode choices and work trip
destinations as well as basic socioeconomic data including auto ownership
are available from census reports. Case Study III in Volume II provides
an example of the use of census data in the evaluation of transit
preferential treatment. Additionally, the County Business Patterns
report provides employer size and industry employment figures at the
county level. At present, the primary weakness of census data is its
age. In areas experiencing rapid growth, ten year old census data from
1970 may not be available in time for use in analyzing measures to be
included in the 1982 SIP revision. However, in many cases, for general
assessment purposes, census data, supplemented by more up-to-date
estimates wherever possible, will be adequate.
Home Interview Surveys—A recently concluded home interview survey
(HIS) can provide all of the travel data needed for transportation-air
-------
5-U5
quality analysis. Typically, however, it will have been several years
since a comprehensive'HIS has been conducted. (In many areas the last
HIS was conducted in the early- to mid-sixties.) Techniques are
available to update survey data to conform with more current data from
other sources and even a 10- to 15-year-old HIS can provide useful
information on which to base an evaluation.
Lane Use, Economic, and Population Projections—These studies will
provide the primary sources of future population and employment
projections which must be used to develop base data for other than
current or past year analyses. These studies are typically conducted by
metropolitan planning organizations as part of their ongoing planning
activities. The information available from future-year projections can
be used to update any data items not explicitly forecast for future years
by using simple relationships between the forecast and non-forecast data
items.
Employers—Employers can provide the most comprehensive and
up-to-date information on employment in an urban area. Of course,
contacting each employer in an area can be quite time consuming, and
there is no guarantee that the employers contacted will provide the
requested information. An alternative source of employment data,
particularly for the CBD of large cities, is a downtown business
association. Such an organization may be able to provide an estimated
breakdown of employment by category as well as the total employment in
the downtown area. This information can be used to supplement data
available from the Census' County Business Patterns.
-------
5-46
State Employment Services—These agencies may be able to provide
more recent county-level employment data than that available from the
census, although in somewhat less detail with respect to industries and
employer size categories.
Traffic Counts—Traffic counts are typically conducted by city
traffic departments as a part of their traffic engineering program. They
also are conducted by state highway personnel in many states on major
facilities. The coverage and frequency of these counts is variable among
urban areas and among cities within urban areas. Traffic counts may also
have been conducted in support of transportation modelling efforts in the
past.
Transit Operating Data—Transit operators receiving federal funding
support are required to maintain operating, revenue and cost data under
the Uniform System of Accounts and Records. Among the data which must be
maintained are passenger-miles and unlinked passenger trips. Many
operators maintain more extensive operating and passenger data for their
own use in evaluating service. These more detailed data may be of use in
corridor and subarea analyses of transportation-air qualiy measures.
Estimation Using Models—In cases in which previous planning and
modelling efforts do not provide the required information, both computer
and manual models are available for estimating base conditions,
particlarly trip-making characteristics. Some of these models are
described in this handbook; examples of their use in developing base case
travel behavior estimates are outlined in Volume II. In general, because
-------
5-17
of the expense and effort involved, these methods would only be used to
develop base data if no other source of travel behavior estimates were
available.
Nomographs and Formulas—An alternative to the use of models in
developing base travel behavior estimates is the transfer of data from
similar urban areas through the use of nomograph or formula
relationships, This approach may be more desirable in the initial stages
of the evaluation of transportation-air quality measures since the use of
models can be expensive and time-consuming. An excellent source of data
and basic relationships for transfer to an urban area is NCHRP Report 18?
(18). The techniques described in this report correspond to the
traditional four-step travel behavior estimation process (trip
generation, trip distribution, mode choice and assignment) and include
time-of-day, auto occupancy, and other supplementary estimation
procedures.
It is likely that different data sources will be most appropriate
during the various stages of the air quality planning process. At the
outset, a large number of candidate measures will be under analysis, and
general analytical techniques with limited data requirements will be used
to screen out those measures with little or no potential. As the
evaluation of candidate actions proceeds, more detailed analysis
techniques and supporting data will be required to identify the most
desirable transportation measures and to develop effective combinations,
or packages, of measures. As the options are narrowed, the impacts of
the proposed measures on specific segments of the population may also be
of concern, requiring still more extensive data.
-------
5-U8
In developing an evaluation approach, these varying data
requirements must be balanced against the availability of specific data
items. In many cases, it will be necessary to estimate some of the more
detailed data elements based on the information which is available. If
accurate and up-to-date information is not available for certain key data
items, sensitivity analyses may be used to bracket the range of impacts
which may be used to bracket the range of impacts which may be expected.
Again, flexibility in analysis approach and data utilization will be a
primary requirement for the development of an effective
transportation-air quality program.
-------
APPENDIX A
Worksheets for the Manual Pivot-Point Mode Choice Method
The pages which follow consist of copies of the worksheets
originally included in Ref. (14), for analyzing the impacts of changes in
transportation 1eve1-of-service on work mode choice. The use of these
worksheets is discussed in Section 2.1.1.a and illustrated in Case Study
I, Volume II.
-------
POLICY:
Population
Subgroup
~
OS1
?*
2t§
ia
f
Average Household Data
Annual
Income $
Number
of Workers
Number of
Non-Work
Auto Trips
Base Work Trip Modal
Shares
!§
§«
3) V>
H
s1
9
1
Average
Carpool Size
Average
Trip Length
®£
jl
Non-Work
(one way)
Average Daily VMT
I
*•
1
!
00
m
s
-------
A-2
I-A. BASE VMT
1. BASE HOUSEHOLD WORK TRIP VMT
Base Base
Shared Ride Base Average Drive Alone
Modal Share Carpool Size Modal Share
-r +
Base
Autos Used No. of Workers Work Trip
per Workers per Household Length
X X
POLICY:
SUBGROUP:
Base
Autos Used
C Worker
Base
Household
Work VMT
X 2X) =
2. BASE HOUSEHOLD NON-WORK TRIP VMT
No. of Non-Work Non -Work
Auto Trips Trip Length
3. BASE HOUSEHOLD TOTAL VMT
Base
Household Non-
Work VMT
Base
Household
Total VMT
-------
A-3
II. CHANGES IN TRANSPORTATION LEVEL OF SERVICE (ROUND TRIP)
+*
CO
1
0>
•o
oc
1
TO
5
0>
o
<
.1
0
Out-of- Pocket
Travel Cost
AOPTCt *
Out-of-Vehicle
Travel Time
AOVTTt min.
In-Vehide
Travel Time
AIVTTt min.
Carpool Prom-
otion & Match-
ing Incentives
0,1
Out-of- Pocket
Travel Cost
AOPTCs, 4
Out-of-Vehicle
Travel Time
AOPTCsr rnin.
In-Vehide
Travel Time
A IVTT^ min.
Out-of-Pocket
Travel Cost
AOPTCda *
Out-of-Vehicle
Travel Time
AOVTTda min.
In-Vehicle
Travel Time
AIVTT^ min.
Population
Subgroup
2
-------
A-4
11-A. CHANGES IN TRANSPORTATION LEVEL OF SERVICE BY CARPOOL SIZE
"S
£
oc
o>
o
Changes in Transportation Level of Servi
1
Q.
(3
c
Q.
+
«
2 Person Carpool
Promotion
Matching
Incentives
and
0,1
Out-of-Pocket
Travel Cost
AOPTC3*
Out-of-Vehicle
Travel Time
AOVTT3t
In-Vehicle
Travel Time
AIVTT3+
Promotion and
Matching
Incentives 0,1
Out-of-Pocket
Travel Cost
AOPTC2 <^
Out-of-Vehicle
Travel Time
AOVTT2 min.
In-Vehide
Travel Time
AIVTT2 min.
Average Size of
3+ Person Carpools
is
si
" s
TO .2
GO (A
Population
Subgroup
3+ Person
§
CNJ
O
Q.
-------
III. ESTIMATION OF REVISED WORK-TRIP
MODAL SHARES
POLICY:
A-5
SUBGROUP:
1 CHANGE IN UTILITY FOR EACH MODE
Drive Alone
AUTILITY =
Shared Ride
Trip Length AOVTTda
TOTAL CHANGE
Trip Length AOVTTsr
-.16 -r-
Average Car
ool Size
Income AOPTCsr
+ -29.0 H-
-H .29 |XI
TOTAL CHANGE
I -.015 X|
TOTAL CHANGE
2. REVISED MODAL SHARE
Base Modal Share AUtility
Revised Share
Drive Alone —
Shared Ride ~~
Transit
Total =[
-------
III-A. ESTIMATION OF CHANGES IN CARPOOL
SIZE
1. CHANGE IN UTILITY FOR EACH CARPOOL SIZE
A-6
POLICY:
SUBGROUP:
AlVTTg
AUTILITY = -.015
2 Person Carpool
Trip Length AOVTT2
2 Person
Carpool Size
-I- -29.0 *
TOTAL CHANGE
3+ Person Carpool
AUTILITY =
Trip Length AOVTT3
Income AOPTC3 Carpool Size
TOTAL CHANGE
2. REVISED CARPOOL-SIZE SHARES
Total Share Revised Share
3. REVISED CHANGES IN SHARED
RIDE LEVEL-OF-SERVICE (LOS)
Revised 2 2 Person Revised 3+ 3+ Person Shared Ride
Person Share ALOS
Person Share ALOS
4 REVISED AVERAGE Revised 2 2 Person Revised 3+ Average 3+ Revised Average
Person Share Carpool Size Person Share Carpool Size
Carpool Size
CARPOOL SIZE
-------
A-7
IV. ESTIMATION OF CHANGES IN VMT
1. REVISED HOUSEHOLD WORK TRIP VMT
POLICY.
SUBGROUP
Revised Revised Revised Revised
Shared Ride Average Drive Alone Autos Used
Modal Share Carpool Size Modal Share Per Worker
Total
Household
Work VMT
Auto VMT
2. CHANGE IN HOUSEHOLD WORK TRIP VMT
Revised Household Base Work Change in
Work-Trip VMT Trip VMT Work Trip VMT
Base Work
Trip VMT
X
100
i
% Change in
Work Trip VMT
3. REVISED HOUSEHOLD NON-WORK TRIP VMT
Change m ....
Autos Used No. of Workers Autos Remaining
per Worker \ per Household per Household
Change in
Autos Remaining
per Household
Revised Non-
Work VMT
.08
4. CHANGE IN TOTAL HOUSEHOLD VMT FROM BASE
Revised Work Revised Non- Revised
Trip VMT Work VMT Total VMT
Change in
Total VMT
% Change in
Total VMT
-------
A-f
V. SUMMARY OF CHANGES IN VMT
Percent Change
in VMT
75
1
i
NORMALIZED HOUSEHOLD VMT
h-
75
5
.0.
«
I
<
I
II
Subgroup
Fraction
of Total Pop.
X
II
W
II
Ul
I
Q. >
AVERAGE HOUSEHOLD VMT
h-
>
75
1
H
Q.
•s
1
<
O
&
<
8
£
Population
Subgroup
,
II
I
-------
A-9
V-A. SUMMARY OF CHANGES IN MODAL SHARES
2
NORMALIZED MODAL SHARE
I
1
1
6
w
£
Shared Ride
o>
I
<
cc
m
cc
m
DC
03
cc
m
cc
CD
Fraction
of Ibtal Pop.
u
W
UJ
cc
X
to
1
5
UJ
u
<
?
o
1
h-
6
1
,55
-§
£
i
w
1
.1
o
oc
03
oc
03
oc
03
OC
03
'«
*03
Population
Subgroup
UJ
1
03 OC
*
-------
APPENDIX B
Worksheets for the Manual Synthetic Mode Choice Method
The pages which follow consist of copies of the worksheets orig-
inally included in Ref. (18), for analyzing the impacts of transportation
measures on work mode choice. The use of these worksheets is also
discussed in Section 2.1.1.b.
The travel demand model incorporated in these worksheets was
developed as part of a project for the U.S. Department of Transportation
using household-level socioeconomic and travel information from the 1968
Washington, D.C. Council of Governments' home interview survey (5). It
is a multinomial logit model which predicts mode shares for three modes:
drive alone, shared ride, and transit.
The definitions of variables used in this model and its coefficients
are presented in Figure 5.2 and Table 5.2 in this report. Note that the
cost coefficients have been estimated in 1968 dollars (or cents). These
coefficients should be adjusted for use with data for later years if the
relative inflation rates for income and transportation costs are known.
-------
B-2
B.I SYNTHETIC MODE CHOICE MODEL -
DRIVE ALONE UTILITY
OVTT Term
OVTT,,
Socioeoonomic terms
Base Alternative
Revised Alternative
Policy:
DIST
INCOME
= total DA utility =
-------
B-3
B.2 SYNTHETIC MODE CHOICE MODEL -
SHARED-RIDE UTILITY
OVTT Term
OVTTC
Socioeconomic terms
DIST
INCOME
DIST
Base Alternative
Revised Alternative
Policy:
Shared Ride
Occupancy
USR = total SR utility =
-------
B-U
B.3 SYNTHETIC MODE CHOICE MODEL -
TRANSIT UTILITY
Base Alternative
Revised Alternative
Policy:
OVTT Term
-.1599
*
OVTTT
DIST
IVTT term
IVTT,,
INCOME
= total T utility =
-------
B-5
B.4 SYNTHETIC MODE CHOICE MODEL -
MODE SHARES
Base Alternative
Revised Alternative
Policy:
Exp
Exp
Exp
UDA (B-
SR '
(B.3)
Exp UL
Drive Alone Mode Share
Exp (UM)
DA
Shared-Ride Mode Share
Exp (USR)
SR
Transit Mode Share
Exp (UT)
Exp (Uk)
characters enclosed in parentheses identify source worksheets.
-------
C-l
APPENDIX C
INCREMENTAL WORK TRIP MODE CHOICE PROGRAM DOCUMENTATION
This appendix consists of two sections which together contain a
complete description of the program 3MODE(VAN)-AGG. The first section,
User Documentation, is directed to the analyst who will establish the
analysis procedures and to the technician who will actually complete
the analyses. The second section is the detailed documentation which
lists the program and other technical information. This will be valu-
able for anyone who wishes to write a new program or modify this one.
The user documentation consists of the following nine sections:
Purpose
Calculator
Keywords
Symbols and Codes
General Description
Data Requirements
Procedure
Comments
Example
The person who sets up the analysis process should pay particular
attention to the General Description (which includes limitations and
capabilities), Data Requirements, and the first two worksheets (base
data and changes in level of service) under Procedure. If a user-supplied
model is desired in place of the default model, the last two sections
under Comments (User-Supplied Coefficients and User-Supplied Model
Specification and Coefficients) must be studied. The person who actually
performs an analysis with this program will need the worksheets in the
Procedure section and the first part of Comments - Use of the Default
-------
C-2
Coefficients. Figure C.I illustrates the steps of an analysis using
this calculator program.
The detailed documentation consists of eight sections:
Detailed Logic
Registers
Symbols and Codes
Flow Chart
Labels
Flags
Card Formats
Program
These sections list the actual program and describe information which
is of interest to a calculator programmer, but is not needed for the
use of the program. To revise this program, it will be necessary to
refer to the list of program steps and, of course, to understand the
programming instructions for whichever calculator is being used.
Program 3MODE(VAN)-AGG is an example of a typical program which
can be written for the calculator. It is a prototype of what can easily
be done to adapt lengthy manual analysis techniques to calculator
techniques.
-------
C-3
FIGURE C.1
Analysis Steps
Determine Policies and Policy Packages for Analysis
For each policy or policy package:
• Determine market segmentation
• For each market segment:
• Determine number of workers
• Determine average round trip distance
• Determine base mode shares
• Determine changes in level of service
• Determine occupancies - carpool and vanpool
• Determine circuity factor - vanpool
• Run program
• Record aggregated results (over all market segments)
-------
0-4
I. USERS' DOCUMENTATION; Incremental Work Trip Mode Choice Model
(3MODE-VAN-AGG)
A. PURPOSE
This program is designed to revise work trip modal shares for three
modes (drive alone, shared ride or carpool, and transit) with an option
for including vanpool as a new or revised mode. In the default version
of the model, revised mode shares reflect changes in any or all of the
following level-of-service variables:
• in-vehicle travel time
• out-of-vehicle travel time
• out-of-pocket travel cost
• carpool/vanpool incentives
In addition to revised modal shares, the program produces data on revised
modal volumes and vehicle miles of travel. Emissions impacts of policies
under analysis can then be determined by using these outputs as inputs
to an emissions calculation procedure such as that described in AppendixD.
The program can be used to analyze an unlimited number of market
segments for any policy and of aggregating the modal volumes and (non-
transit) vehicle miles of travel over all the segments.
The program has been written to provide as a default the specifica-
tion of a multinomial logit work mode choice model estimated from 1968
Washington, D.C. area data. This specification requires only four
coefficients in the incremental form, but capacity for five coefficients
per mode (a total of fifteen) has been provided. This was done for ease
in revision of the default model or substitution of a user-specified
model, if desired.
See below for the specification of this model, which is documented fully
in Cambridge Systematics, Inc., "A Behavioral Model of Automobile Owner-
ship and Mode of Travel", U.S. Department of Transportation, Office of
the Secretary, Federal Highway Administration, 1974.
-------
C-5
B. CALCULATOR
TI59; with PC100A printer optional
C. KEYWORDS
Mode choice, incremental logit model, aggregation, vehicle
occupancy, vanpools
-------
C-6
D. GENERAL DESCRIPTION
1. Main Program
This program is capable of analyzing an unlimited number of market
segments and of aggregating the modal volumes and non-transit VMT results
over all the segments. Each market segment can have up to five modes
which are defined as the user wishes with two restrictions.*
• The fourth mode (vanpool, if default coefficients are used) uses
the same coefficients as the second mode (carpool).
• There is no analysis of the fifth mode ("other"). Level of
service for "other" modes is expected to remain constant and the
share is included so that the sum of modal shares always equals one.
Each of the first three modes can have up to five coefficients
(unique to the mode, if desired). Extra coefficient storage registers
may be left equal to zero (to remain unused) or used as constants to
modify the effects of the level-of-service coefficients.**
The program can handle three different types of vanpool mode share
changes:
• base share equal to zero and new share equal to zero;
• base share equal to zero and new share greater than zero;
• base share greater than zero and new share greater than zero.
* Use of the default coefficients suggests use of the default modes:
drive alone, carpool, transit, vanpool, and other (walk, cycle, taxi).
** This modification of constants is typically done in transferring entire
models, and is not often used in pivot point applications. See Ben-Akiva,
Moshe and Terry Atherton, Transferability and Updating of Disaggregate
Travel Demand Models, Transportation Research Board, 1976.
-------
C-7
In the first and third cases, the actual base vanpool share is entered
by the user and the new shares are determined from the changes in modal
levels of service. In the second case, the vanpool subroutine described
below is used to create a base share on which to pivot.
The calculation of estimated changes in transportation modal shares
incorporated into the program is based on the incremental form of the
multinomial logit model.* Based on probabilistic choice theory, this
form is used to pivot about an existing situation. The approach pre-
dicts revised travel behavior based on existing travel behavior and
changes in level of service rather than employing a full transportation
demand model system to recalculate modal shares based on detailed house-
hold, zonal, and level-of-service data. By employing a "pivot-point"
approach, data requirements are greatly reduced: no knowledge of detailed
socioeconomic and level-of-service data for each household or zone is
required. Only existing estimates of modal shares (probabilities) and
proposed changes in level of service are necessary.
The logit model predicts the probability that a behavioral unit will
make choice "i" from the set of alternatives "A" and is expressed in the
following exponential form:
* Background on the theory and derivation of this model is presented in:
Richard, Martin G., and Moshe E. Ben-Akiva, A Disaggregate Travel
Demand Model (Westmead, Farnborough, Hants, England: Saxon House/
Lexington Books) 1975.
-------
C-8
-C (C-l)
Z e
m£A
where :
m,i = travel mode alternatives
A = the set of possible choices
P(i:A) = the probability of choosing alternative i out of the
set of the available alternatives A
U , U. = the utility of alternatives m and i
This particular model form exhibits many favorable properties which
make it desirable to use. First, the exponential form is in agreement
with consumer behavior theory (i.e., if an alternative begins with a choice
probability of 0.5 or greater, the initial response to an increment of
utility improvement is greater than the response to a later increment
of utility improvment of the same magnitude). Second, the probabilities
necessarily sum to one as they should. Third, the curve of P(i:A) versus
U. has the general shape of the curve of "diminishing returns" at both
ends. This reflects known travel behavior nicely; i.e., no matter how
good (or bad) a service is, all of the ridership is never captured (or
lost), and ridership is most susceptible to diversion to some other alter-
native in the highly competitive middle range. Fourth, the logit model
is capable of extension to any number of travel alternatives. Finally,
the logit form is mathematically tractable, leading to simplicity in
calibration, transformation, and application.
-------
09
Any change in transportation will result in a change in the level
of service X , and therefore a change in the utility function as expressed
ij
by the weighted sum of transportation and socioeconomic variables
(or U. = ^i-iX..). The revised probability resulting from changes in
j
modal utilities is given by:
U± + Au±
Z e
meA
where:
AU. = the change in utility for alternative i
= ze..AX..
J 1J 1J
6.. = a calibrated model coefficient
AX.JJ = a change in an independent variable
P'(i:A) = the predicted probability of choosing alternative i
when U^ changes by ATJ^
Equation (C-2) can be represented by:
e • e
Um
Dividing both the numerator and denominator of (C-3) by Z e yields
meA
-------
C-10
U, Um AU.
[e V Z e m] • e X
m£A
. fn ,\
u u AU (c~4)
r m. v m, m
E [e / L e ] e
meA meA
which can be written as:
AU.
AUm
Z P(m:A)e
meA
(C-5)
Note that whenever p(i:A) equals zero, P'(i:A) will equal zero regardless
of the change in utility or probabilities of other modes.
To be completed, the probabilities obtained from this equation must
be expanded into an aggregate forecast of travel behiavior. A straight-
forward and simple approach is to use average values of existing choice
probabilities and of changes in level of service for a given population
subgroup. The incremental model form expressed in the program then
simply replaces P(i:A), the probability of an individual choosing alter-
native i with the base modal share for that alternative, and X.., the
change in transportation level of service for an individual, with the
average change in that variable for all trips. Thus, aggregate or popu-
lation subgroup modal shares become the basic units for which transpor-
tation-related impacts are predicted.
-------
c-ii
The travel demand modal choice model incorporated as a default
into the program is of the logit type described above. It was developed
as part of a project for the U.S. Department of Transportation.* The
calibrated model predicts the probability of a person driving alone,
sharing a ride, or using transit for the home-to-work trip. Using this
definition of alternative modes, all travellers who are part of a carpool
or vanpool are treated as shared riders, regardless of who actually drives
the vehicle; i.e., drivers with passengers, as well as passengers, are
classified as shared riders.
The model was calibrated by the method of maximum likelihood esti-
2
iiiation*- using household-level socioeconomic and travel information from
the 1968 Washington, DC, Council of Governments home interview survey.
The model specification, coefficients, and statistical properties are
shown in Table C.I. All important coefficients are significant with low
standard errors (and correspondingly high "t" statistics). The variables
used in specifying the model can be classifeid into three types: socio-
economic (e.g., household income, automobile ownership, number of
workers); land use and locational (e.g., distance, workplace, employment
Cambridge Systematics, Inc., A Behavioral Model of Automobile Owner-
ship and Mode of Travel, U.S. Department of Transportation, Office of
the Secretary, and the Federal Highway Administration, Washington DC 1974.
2
"Maximum likelihood estimation" is a technique for curvefitting or
calibration, similar to least-squares or regression techniques but
more sophisticated and compatible with the non-linear logit form.
-------
C-12
TABLE C.I
Washington Work Mode Choice Model
xi' st
Variable
Dd
Ds
OCTC/INC
IVTT
OVTT/DIST
AALDd
AALDS
BWd
GWS
DCITYd
DCITYS
MNCd,s
NWORKg
DTECA,,
o
Definition
Drive alone constant
Shared ride constant
Out-of-pocket travel
cost divided by
income (round trip)
In-vehicle travel
time (round trip)
Out-of-vehicle travel
time divided by
distance (round trip)
Auto availability
(drive alone only)
Auto availability
(shared ride only)
Breadwinner (drive
alone only)
Government worker
(shared ride only)
CBD workplace (drive
alone only)
CBD workplace
(shared ride only)
Disposable income
(drive alone &
shared ride only)
No. of workers
(shared ride only)
Employment density
(shared ride only)
Drive
Alone
-3.24
-28.8
-.0154
-.320
3.99
.890
-.854
.000071
Shared
Ride
-2.24
-28.8
-.0154
-.320
1.62
.287
-.404
.000071
.0983
.00065
Transit
-28.8
-.0154
-.320
Standard
Error
.473
.401
12.7
.0057
.078
.395
.305
.186
.161
.311
.298
.00002
.095
.00049
11 *_ M
Statistic
-6.86
-5.60
-2.26
-2.67
-4.08
10.08
5.31
4.79
1.78
-2.75
-1.36
3.46
1.03
1.34
-------
C-13
density); and transportation level-of-service (e.g., out-of-pocket travel
cost, in-vehicle travel time, out-of-vehicle travel time).
For use in the program, the calibrated model has been expanded in
two ways:
• The shared ride mode has been further subdivided into two modes,
carpool and vanpool, depending on the type of vehicle used. Both
of these modes have utility functions with the same structure,
equal to that calibrated for the shared-ride mode.
• A fifth mode, "other", is included to represent all modes of
travel not included in the four other modes. The utility of this
mode is never allowed to change. It is included so that all base
mode shares sum to one.
When revised modal shares have been determined by the program, modal
volumes and VMT are determined. The equation for modal volume is:
VOL = S * Pop (C-6)
mm
where:
Vol = Volume of mode m
m
S = Share of mode m
m
Pop = Number of workers in the market segment
The equation for VMT for one market segment is:
VOL VOLvp * CIRC
VMTTOT = VOLDA + OCC + "»» > * DIST
where:
VOL_. = Drive alone volume
VOL p = Carpool volume
= Vanpool volume
OCC = Carpool occupancy
\j"
OCC.jp = Vanpool occupancy
CIRC - Vanpool circuity factor
DIST = Round-trip distance
-------
C-14
2. Vanpool Subroutine
The purpose of this subroutine is to "create" a non-zero base on
which to pivot the vanpool mode share. * Adjustments are made to the base
shares in other modes so that the adjusted base mode shares (including
vanpool) sum to one. If there is a non-zero base vanpool share, it
can be used directly in the pivot point program without employing this
subroutine. The program equations are:
SDA - SDA - -5 * SVP(1 -
sa
CP = Scp - .5 * S(! - S) (C-9)
ST(1 -
where:
SDA • Base drive alone share
a
SDA = Adjusted drive alone share
SCP = Base shared ride share
a
SCP = Adjusted shared ride share
Sj = Base transit share
o
ST = Adjusted transit share
Syp = Specified base vanpool share
These incorporate the assumptions that:
• the proportion of the new vanpool mode share made up of commuters
previously using transit is equal to the existing transit modal
share, and
• an equal number of drive-alone and carpool commuters are attracted
to vanpools.
A zero base share will always give a zero revised share. An assumption
must be made about an adjusted vanpool share if the revised share is
expected to be non-zero.
-------
C-15
For example, if the existing modal share of transit were 15 percent,
15 percent of the new vanpool modal share would be from transit, and the
transit modal share would be reduced accordingly. Note that because the
carpool modal share is generally less than the drive-alone share, the
contribution of carpool as a percentage of its modal share will be
greater than the corresponding percentage for drive-alone. Because
"other" modes, such as walk, taxi, bicycle, etc., are assumed to be
unavailable to the population subgroup with vanpooling available, these
modes remain unaffected by the introduction of vanpooling. This is con-
sistent with the fact that vanpooling is usually only offered to and
attractive for employees with long commuting distances and poor access
to transit.
3. Carpool Subroutine
This subroutine is used in the analysis of any policy which treats
carpools differentially based on the number of passengers in the vehicle.
Only two classes of carpools can be handled; normally, if the distinction
is not between two and three-or-more (3+) person carpools, it will be
between two- and three- (2/3) person carpools and four-or-more (4+) person
carpools.
When changes in level of service attributable to the policy under
analysis have been determined for the two classes of carpools, they are
entered in the subroutine along with the average vehicle occupancy and
share of carpool trips for each class. The subroutine is an incremental
logit model which treats carpool as the only available mode and computes
-------
C-16
revised carpool shares, average changes in level of service for all
carpools, and a revised average vehicle occupancy for all carpools.
The series of equations for this are:
where:
AUCP1
CP1 ' e
SCPI • -
s
CP2
SCP2
*-« VlA 1
SCPi ' e
ALOSCP - SCP1 • ALOSCP1 + SCP2 ' ALOSCP2
OCCcp - l/[(Sjpl/OCCcpl)
S _. = base carpool share for class i
or 1
S' = revised carpool share for class i
AU = change in utility for carpool class i
U-t i
LOS p . = level of service for carpool class i
LOS p = level of service averaged for all carpools
L
-------
C-17
E. SYMBOLS AND CODES
Modes:
Variables:
DA
SR
m
m
T
0
TOT
IVTT
H
OVTT
r
OPTC
m
INCENT
DIST
Y
occ
m
CIRC
POP
U
m
VOL , VOL1
m m
VMT , VMT'
m m
s , sa, s'
m m m
A/W, A/W
Ix
Other:
6n
i
An
m
m
Drive Alone
Shared Ride
CP = Carpool
CP1 = Carpool class 1
CP2 = Carpool class 2
VP = Vanpool
Transit
Other
All modes
In-vehicle travel time for mode m
Out-of-vehicle travel time for mode m
Out-of-pocket travel cost for mode m
Incentives for the carpool and vanpool modes only
Round-trip distance
Average household income for a market segment
Average vehicle occupancy for mode m
Vanpool circuity factor
Market segment population
Utility of mode m
Base and new volume of mode m
Base and new vehicle-miles of travel for mode m
Base, adjusted, and new share of mode m
Base and new autos per worker
Variable x, summed over all market segments
Coefficient of variable n for mode m
Change in variable n for mode m
-------
C-18
F. DATA REQUIREMENTS
1. Main Program
For each market segment:
• round trip distance in miles
• household average annual income in miles
• average carpool occupancy
• number of workers in the market segment
• average vanpool occupancy
• vanpool circuity factor - the ratio of vanpool travel distance
to drive-alone travel distance. This ratio is normally greater
than one reflecting increases for the purpose of picking up
passengers.
• base mode shares:
- drive alone
- carpool
- transit
- vanpool
- other
• round trip changes in level of service for each mode:
- in-vehicle travel time
- out-of-vehicle travel time
- out-of-pocket travel cost
- incentives (carpool and vanpool only)
2. Additional Da-ta for Vanpool Subroutine
• specified base mode share for vanpool for each market segment
-------
C-19
3. Additional Data for Carpool Subroutines
For each market segment:
• average occupancy for carpool classes 1 and 2
• share of total carpool volume in classes 1 and 2
• changes in level of service for each segment:
- in-vehicle travel time
- out-of-vehicle travel time
- out-of-pocket travel cost
- incentives
These data items can be organized and documented on the three
worksheets which follow this section:
C-l Base Data
C-2 Changes in Transportation Level of Service
C-3 Changes in Transportation Level of Service by Carpool Size
-------
C-20
WORKSHEET C-1
BASE DATA
POLICY:
Population
Subgroup
Average round
Trip Length (mi)
Annual
Household
Income ($)
Average
Carpool Size
Population
Average
Vanpool Size
2Q.$
s£3
3E-o
«< o_
Base Work Trip Modal Shares
fs
§$
o
Q>
I
31
^
1
o_
9
1
-------
WORKSHEET C-2
POLICY:
CHANGES IN TRANSPORTATION LEVEL OF SERVICE
(all data represent round trips)
'opulation
Subgroup
Drivi
-I?
^ o
su
^M
n
t> * 0
0^5
ss|
° ^ Q
*3" «
*-» S
goo
!fi
a * s.
3. £ -o
^!l
i^ * *
£21
5if
-H=!i
-^ 3 *
3 «
5
Transit
OHO
OS T
< 2. o
_^H|
OHO
0 g £
5io
J> 0 4,
a?
A S
^ H ?
<^ H 0
ir
Vanpool
Out-of-Vehicl»
Travel Time
AOVTTvp(min.)
I??
3SJ
^8 £
3" »
ftf
S «• *
1 - i1
^ 51
o
I
-------
WORKSHEET C'3
CHANGES IN TRANSPORTATION LEVEL OF SERVICE BY CARPOOL SIZE
POLICY-'
Class 1=
Class 2=
person carpools
person carpools
Population
Subgroup
Base Shares
by Carpool
Size Segment
O
_»o>
u>
VI
O
to$
to
Average Size
of Carpool
by Segment
O
^ST
"S
O
M5T
rocn
«
Changes in Transportation Level of Service (Round Trip)
Class 1
>!?
HSL|
- i*!
- 3 (P
^s*?
91 •
^aa
"H^ &
"O 3 3l
g*f
QHHS
'i-s 5' *
3 £
£
^a1?
il •
n-9.
o-^H^
I1!
D>*g
0 < r
3ia
"°£-o
-a O rt
-ftS
a 1
|||
jSo'
1 I
n
i
NJ
-------
C-23
G. PROCEDURE
Travellers who are affected by the policy under consideration
should be grouped in market segments which are internally homogeneous.
(See Section 5.3.) The classification criteria generally appropriate are:
modal availability, socioeconomic characteristics, and trip orientation.
The classification employed will depend on the nature of the air quality
policy and the groups which are affected.
When market segments have been classified and all necessary data
compiled, the program should be run using the procedure outlined on the
worksheets which follow. These worksheets include:
C-4 Program Steps
C-5 Carpool Subroutine
C-6 Results by Market Segment
The next section (H) contains supplementary discussion for selected steps,
-------
C-24
WORKSHEET C-4 PROGRAM STEPS
639.39
(3MODE(VAN)-AGG-2(B)/7900110/ESE)
USER INSTRUCTIONS
STEP
1
2
PROCEDURE
Read card(s) -Partitioning is 639.39
Initialize Storage Registers, and store' -
default Coefficients.. ..
If_. default coefficients_jare used* skip
to. STEP 4 ... . . _ ._
ENTER
-
1
A
=RESS
DISPLAY
.29
r
3a
3b
3c
3d
3e
3f
?
3h
31
31
3k
31
3m
3n
3o
4
4a
4b
4c
4d
OPTIONAL - Enter user-supplied coefficients
coefficient of IVTT— DA mode
coefficient of OVTT/DIST-DA mode
coefficient of OPTC/Y-DA mode
coefficient 4-DA mode
coefficient 5-DA mode
coefficient of IVTT-SR mode
coefficient of OVTT/DIST-SR mode
coefficient of OPTC/Y-SR mode
coefficient of incentives-SR mode
coefficient 5-SR mode
coefficient of IVTT-T mode
coefficient of OVTT/DIST-T mode
coefficient of OPTC/Y-T mode
coefficient 4-T mode
coefficient 5— T mod@
IF A MISTAKE IS MADE^ BEGIN AGAIN AT STEP 3a
Enter market segment data (from worksheet 13
round trip distance in miles ^ 0
household average annual income in $ ? 0
enter "1" if income is not needed for this
segment.
Average carpool occupancy 4 0
Default =2.5
Market segment population
™™
9 -,DA =
9 2DA =
8 3DA =
9 /,DA =
8sDA =
e^R-
flnt!R =
e,SR =
0ASR =
'e^sR="
Q!T =
92T =
63T =
6 ,T =
4
95T =
'
DIST =
Y =
occcp=
POP =
i
STO
STO
STO
STO
STO
STO
STO
STO
STO
STO
S1Q
STO
STO
R/S
R/S
R/S
R/S
*1
00
m
02
03
04
05
07
08
09
10
11
1?
13
14
«
-
---
TO™
' •" **
—
•
.....
" ':_::;""
- -— -.
. , _
-------
C-25
WORKSHEET C-4
(continued)
(3MODE(VAN)-AGG-2(B)/790110/ESE)
USER INSTRUCTIONS
STEP
4e
4f
5
5a
5b
PROCEDURE
Average vanpool occupancy 4 0
default — 10
Vanpool circuity factor ^ 0
default — 1.5 • . .
If a mistake is made, press [cjand begin
with STEP 4a
If carpool subroutine is not used,
go to STEP 6
OPTIONAL - Carpool Subroutine (data from woi
Enter average occupancies of two classes of
carpools :
1. Average occupancy carpool class 1
2. Average occupancy carpool class 2
Enter level of service and mode share data
1. AlVTT - carpool class 1
2. AOVTT - carpool class 1
3. AOPTC - carpool class 1
4. AINCENT - carpool class 1 (0/1)
5. A^ - carpool class 1**
6.-- carpool class 1 share as -a -fraction
of all carpools *
If a mistake is made, press |C' | and begin
with STEP 5a
7. AIVTT- carpool class '2
8. AOVTT - carpool class 2
9. AOPTC- carpool class 2
10. A INCENT - carpool class 2 (0/1)
11. Ac - carpool class 2**
12. carpool class 2 share as a fraction
of all carpools*
If a mistake is made, press |c' | and begin
with STEP 5 a
* These two values should sum to 1.0
** If default coefficients are used, these
A's must equal zero.
ENTER
OCCyp =
CIRC =
ksheet II-3)
occcpl=
occcp2= —
A =
A =
A =
A =
A =
sr.Pi =
A =
A =
A =
A =
A = •
c =
r>T>o
CP2
R/S
R/S
-'
R/S
R/S
R/S
R/S
R/S
R/S
R/S
R/S
R/S
R/S
R/S
R/S
R/S
R/S
=RESS
'
2nd
\
E'
DISPLAY
— -
38
-------
C-26
WORKSHEET C-4
(continued)
(3MODE(VAN)-AGG-2(B)/790110/ESE)
USER INSTRUCTIONS
STEP
5c
-
6
6a
6b
6c
6d
6e
6f
6g
7
PROCEDURE
carpool subroutine results:
1. New average occupancy for all carpools
2. New carpool change in level of
service (for all carpools)
Enter A LOS for carpool class 1
Enter A LOS for carpool class 2
To repeat for another change in level of
service, press |D'| and repeat STEP 5c2.
Step 5 results can also be recorded on Work
Xo^can^inue^ith^STEP^^^^^^^^^^^
ENTER
ArPi =
ACP2=
sheet 11-5 .
^^K <^K ^^ ^^_ .^_ — 1 1.
Store modal shares for this market segment
(from Worksheet II-l)
Base drive-alone share as a fraction **
(i.e., .645, not 64.5)
Base carpool.share as a fraction **
Base transit share as a fraction **
Base vanpool share as a fraction **
Base "other" share as a fraction **
Base autos per worker
If a mistake is made, press |GO TO 1/X
and begin with STEP 6a •
Base VMT
If vanpool subroutine is not used, go to -
STU =
S_
CP ~
ST =
SVP=
S0 =
- - - —
STEP 8
OPTIONAL Vanpool Subroutine
If VP = 0 in 6d and the revised. vanpool
share ^ 0, enter a "base" vanpool share on
which to pivot. Defaults —
.14 for firms with < 2000 employees
.06 for firms with >_ 2000 employees
If a mistake is made, press |B'| and repeat
STEP 7
* printed but not displayed
** these five fractions must sum to 1
:SVF =
PRESS
. ,.
R/S
R/S
R/S
2nd
\M&
D1
L-,
R/S
R/S
R/S
R/S
R/S
s/s
R/S
R/S
.
DISPLAY
*OCC,p =
\ =
Li
A/W -=
VMT =
TOT
-
-------
C-27
WORKSHEET C-4
(continued)
(3MODE(VAN)-AGG-2(B)/790110/ESE)
USER INSTRUCTIONS
STEP
8
8a
8b
8c
8d
8s
8f
8g
8h
8L
8j
8k
81
8m
8n
80
9
9a
9b
9c
9d
9e
PROCEDURE
Enter changes in level of service from Works
If there are NO changes in level of service,
Press [S]and go to STEP 10; otherwise,
Drive Alone changes:
AIVTTM
AOVTTM
AOPTCM
AAf-A **
A'5tu **
If there are no additional changes in LOS,
press [f] and go to STEP 10
Carpool changes from Step 5 or Worksheet 11
if optional STEP 5 was used:
AIVTTCP
AOVTTrp
\*C
AOPTCcp
AINCENTCP *
A5CP **
If there are no additional changes in LOS,
press [B] and go to STEP 10
Transit changes :
~h iVTT_
A OVTTT
A OPTCT
A4T **
A , _
A5T **
If a mistake is made, press [GO Tp| |l/x)
and begin at STEP 6a '
If the vanpool share in 6d - 0 and STEP 7 -
was not used, GO TO STEP 10
OPTIONAL: Enter vanpool change in L0s""ftrpn
A IVTT
AOVTTyp
A OPTCyp
A INCENTyp ***
Ac **
a 5VP
* This must equal 0 or 1 unless the carpoo]
must be between 0 and 1 inclusive.
** If default coefficients are used, these £
*** 9d must equal zero if STEP 7 was used.
ENTER
heet II-2)
A =
A =
A =
A =
A _
-5
Ast
___
A =
A *
A =
A = _
A -
A .
A =
A =
A =
"
"VorTTsKeetT
A ™ _
A =
A = '
A =
A =
subroutine j
's must equa
PRESS
.D
B
R/S
R/S
R/S
R/S
R/S
B
R/S
R/S
R/S
R/S
R/S
B
r\ / ft
R/S
R/S
R/S
R/S
R/S
-2)
R/S
R/S
R/S
R/S
R/S
s us
0.
Jd> j
.
n wh
DISPLAY
_ . . _. —
..
-
.ch case it
-------
C-28
WORKSHEET C-4
(continued)
(3MODE(VAN)-AGG-2(B)/790110/ESE)
USER INSTRUCTIONS
STEP
10
lOa
lOb
lOc
lOd
lOe
lOf
10g
lOh
10 i
10j
10k
103
10m
lOn
ll
PROCEDURE
Market sejgment results
New DA share *
New DA volume
New DA VMT
New CP share *
New CP volume
New CP VMT
New T share *
New T volume
New VP share *
New-VP volume
New VP VMT
New Other share * "
New VMT for this market segment
New Autos per worker *
These results can also be recorded on Works
To analyze another market segment and have
the results aggregated with previously
analyzed market segments, press [C3 and GO
TO STEP 4
To print** aggregated results of all pre-_
vious market segments, pressfE*)
- •
.._.
These results can also be recorded on Works
To analyze a new policy (no aggregation wit
previous market segments), press ^TJand GO
TO STEP 2. (Since this sets memories to zero
data from previously tested policies must b
copied before A is pressed, and user-
supplied coefficients, if any, must be re-
entered in STEP 3.
* Printed but not displayed
**See Comments (Section 4) for retrieving
data without a printer.
ENTER
eet II-6
leet II-6
P
B
R/S
R/S
R/S
R/S
R/S
R/S
R/S
R/S
C
E
A
RESS
-
DISPLAY
c '
b DA =
VOL'm=
VMT'DA=
S'PP =
VOL'cp=
VMT'CP=
s'T
VOL'T =
c '
VP =
VOL'vp-
VMT 'TO-
s'o -
VMT^nT=
A/W =
£VOLV=
ZVMT'n&=
EVOL'cp=
IVMT'cp-
£VOL'T =
£VOL'vp=
ZVMT'yps:
£VMT-j>oT=
IVMT^,OT=
A£VMTTOT= '
%ALVMTTOT=
—
-------
C-29
WORKSHEET
C-5
Policy:
CARPOOL SUBROUTINE
Output from Step 5 - Input to Step 8
(all data items represent round trips)
Population Subgroup
New Shared
Ride
Occupancy
Average
Shared Ride
IVTT
Average
Shared Ride
OVTT
Average
Shared Ride
OPTC
Average
Shared Ride
"Incentive"
-------
. WORKSHEET
C-6
Policy:
RESULTS BY MARKET SEGMENT
o
i
Population
Subgroup
[
TOTALS
Base
Autos
Per
Worker
X
Base
VMT
New Drive Alone
Share
X
Vol
VMT
New Carpool
Share
X
Vol
VMT
New Transit
Share
X
Vol
New Vanpool
Share
X
Vol
VMT
New
Other
Share
X
New
VMT
New
Autos
Per
Worker
X
Change
in VMT
Percent
Change
in VMT
-------
C-31
H. COMMENTS
1. Use of Default Coefficients
If a printer is not used, the values which are printed/displayed
must be copied from the calculator display. See note for Step 11 to get
cumulative VMT values.
Step 4 Copy data from Worksheet C-l
If income is not needed for this market segment (i.e., there is no
change in cost for any mode), then a "1" should be entered for income.
Market segment population should be entered as the total number of
work trips. If a zero value is entered for any market segment data, the
program output will be meaningless.
«*
Step 5 Copy data from Worksheet C-3
If the policy being analyzed affects carpools differently based on
occupancy, the carpool subroutine should be used.* This subroutine gives
a revised carpool occupancy and average changes in level of service for
all carpools, regardless of occupancy.
The subroutine handles only two classes of carpools. In many cases,
policies which discriminate by carpool occupancy do so on a two-person
versus three-or-more person basis. If a policy affects four-or-more
person carpools differently than two- and three-person carpools, the
subroutine should be run with these two carpool classes.
* For example, a policy of free bridge tolls for three-or-more person
carpools or preferential highway lanes for four-or-more person
carpools.
-------
C-32
The occupancy of two-person carpools is 2.0 and this should be
eneterd as OCC, unless two- and three-person carpools are considered
as a group. In the latter case, an occupancy value between two and
three will be entered. The second occupancy is of three-or-tnore or
four-or-more person carpools, as appropriate.
Changes in level of service represent round trips. They should
be entered in the order shown, first for carpool class 1 and then for
carpool class 2. The shares for the two classes of carpools (Steps 5b6
and 5bl2) must sum to one.
Step 5cl gives the new average occupancy of all carpools, which is
automatically stored for use in the main program. Step 5c2 (repeated
as many times as necessary) gives new average changes in level of
service for all carpools. These values must be manually input in
Steps 8f to 8i.
Step 6 Copy data from Worksheet C-1
The sum of the five mode shares must be one. Although "autos per
worker" is not required as a variable in this model, provision is made
for the input of a base value, so that changes in autos available
for non-work travel, useful in some non-work models, can be computed
by the program.
-------
033
Step 7
If the base vanpool share equals zero but the tested policy will
result in a non-zero vanpool share, the vanpool subroutine must be
used. A non-zero base share roust be used on which to pivot the vanpool
mode share. When the "base" vanpool share is entered in the subroutine
(Step 7), the other base shares are automatically revised so that they
still sum to one. The assumptions and logic of these changes are found
above in Section D.2.
Step 8 Copy data from Worksheet C—2 and Worksheet C-5
The changes in level of service are entered in minutes or cents, as
appropriate, from the values on worksheet C-2 and/or worksheet C-5 (if
the carpool subroutine is used). Note that when the default coefficients
are used, Steps 8d, 8e, 8j, 8n, and 80 equal zero and must be entered as
such.
If there are no changes in level of service for the market segment,
[B] should be pressed just after [5] and the user can continue with Step 10,
If there is no change in level of service for transit and vanpool, the
user may press [IT] after Step 8j is completed and then continue with
Step 10. Similarly, if there is no change in level of service for car-
pool , transit, and vanpool, [B~| may be pressed after Step 8e,
Market segments with no changes in level of service will often be used
to ensure that the total base VMT can be determined, even though these
market segments will have no changes in mode shares.
-------
C-34
Incentives for carpool (Step 8i) is a variable which usually
equals either zero or one. If the carpool subroutine is used, the car-
pool incentive may take on any value from zero to one inclusive.
Step 9 Copy data from Worksheet C-2
If the base vanpool share is zero and Step 7 was not used, the
program automatically skips Step 9. Vanpool incentives (Step 9d) should
equal zero if Step 7 was used to create a base vanpool share. Step 9e
must equal zero when default coefficients are used.
Step 10
If a printer is not used, each value should be copied to Worksheet C-4
as it is displayed. The revised modal shares by market segment can also be
displayed by pressing:
RCL 24 -
-------
C-35
2. User-Supplied Coefficients
If user-supplied coefficients are desired for the analysis of any
policy, they can be easily entered in Step 3. All coefficients or as
few as one may be changed; any coefficient not specifically changed
in Step 3 will have the default value assigned to it in Step 2. If
the model specification is not changed, all instructions for Step 4 and
so on, are valid. That is, if the first value (STO 00) is the coeffi-
cient of in-vehicle travel time (IVTT) for the drive-alone mode, the
second value (STO 01) is the coefficient of out-of-vehicle travel
time divided by distance (OVTT/DIST) for the drive-alone mode, etc;
then the user need not be concerned with the explanations below and can
follow all instructions for use of the default coefficients.
3.User-supplied Model Specification and Coefficients
The program was written for ease of execution using the default
coefficients, while allowing for the possible substitution of alternative
model specifications and coefficients. Basically, the program is capable
of handling up to five coefficients for each of the three modes, with the
coefficients of the fourth mode constrained to equal those of the second
mode. Some of the coefficients stored in registers 00 through 14
are modified in the program by socioeconomic values. (See Table C.I
User-supplied coefficients may be stored in any order in Step 3 as
long as the appropriate STO XX is entered after the coefficient value.
-------
C-36
above, variables 3, 4, and 5.) Any change in program specification
must take this into account.
The easiest method of avoiding the effect of socioeconomic data in
the utility equations is the storage of user-supplied coefficients in
the first, fourth, and fifth places for each mode. (This corresponds
to storage registers 0, 3, 4, 5, 8, 9, 10, 13, and 14, which are unaffec-
ted by income, or trip distance values in the utility equation.)
The program user must take care to enter zero for the change in level of
service for the second and third coefficient places, if this method is
used.
An alternate method (which must be used when there are more than
three relevant coefficients for any one mode) is to compensate for the
division of a variable by a socioeconomic value by multiplying the
input value by the same value. For instance, the default model takes
AOVTT and automatically divides it by round-trip distance. If this
modification is not desired, the true value of AOVTT should be multiplied
by the round-trip distance so that the division done in the program
will result in:
AOVTT * round-trip distance
= AOVTT
round-trip distance
The user who wishes to change the default specification is advised
to study section D.I of this Appendix and to have a thorough understanding
of the capacity and limits of the program as written.
-------
C-37
I. EXAMPLE
The following completed worksheets illustrate the analysis of
preferential treatment for buses which decreased the average peak period
transit travel time to downtown by ten minutes for in-vehicle time and
two minutes for out-of-vehicle time. Three market segments represent
all workers in the urban area:
1. Downtown workplace; all modes available
2. Downtown workplace, drive-alone not available
3. Workplace not downtown — unaffected
The analysis results in Figure C.2 show that for market segments one
and two, the transit mode shares increase, while all other non-zero mode
shares decrease in proportion to their original share. The total VMT
for these market segments decreases since transit ridership does not
contribute to vehicle-miles travelled.
There are no changes in level of service for market segment three
which consists of all unaffected workers. This market segment is
included so that its VMT is part of the urban-wide base VMT used to
calculate a final percent change in VMT attributable to the policy. For
this analysis, the absolute change in VMT is -2200 and the percent
change is -3.66.
-------
C-38
FIGURE C.2
Printed Analysis Results
Market Segment 1*
0.53
37100.
4246662528
2972.66377
29726.6377
1887405568
2045. 560384
0.
0.
0.
.0943702784
35011.37329
.5001624756
Market Segment 3*
0. 79
22120.
0. 65
1300.
18200.
0. 35
700.
3920.
0.
0.
0.
0.
0.
0.
22120.
0. 79
Market Segment 2*
Aggregated Results**
0. 16
960.
0.
0.
0.
.3534579374
353.4579374
848.2990498
.5139953361
513.9953361
0.
0.
0.
. 1325467265
848.2990498
0. 141383175
4272. 66377
47926.6377
2374. 64183=;
10053.03464
2559. 55572
0.
0.
60180.
57979.67234
-2200. 327661
-3.656244036
30
31
34
36
38
Printer Output
* These data values correspond to Worksheet C-4 Step 10
**These data values correspond to Worksheet C-4 Step 11
-------
C-39
II. DETAILED DOCUMENTATION; INCREMENTAL WORK TRIP MODE CHOICE MODEL
(3MODE-VAN-AGG)
A. DETAILED LOGIC
Step 1
Step 2
Step 3
(optional)
Step 4
Step 5
(optional)
- Read program from cards
- Load register 0-14 — default coefficients
Set registeres 15 - 39 equal to zero
- Load registers 0-14 — user-supplied coefficients
- Load registers 15 - 20 — market segment data
- CARPOOL SUBROUTINE
Load registers 21 - 22 — carpool class occupancies
Calculate average carpool occupancy and levels of service with
shares and changes in level of service by carpool class:
Registers
Contents
AU
CP1
AU,
CP2
Calculations
AH/IT *
- CPI
+ AOVTT/DIST * 60VTT/DIST,,--
Crf-L .L OA J.
+ AOPTC/YCPI *
-1- AINCENTcpl * 6INCENTcpl
AIVTTcp2 * 6IVTTCp2
AOVTT/DISTCp2 * 60VTT/DISTCp2
AOPTC/YCp2 * 90PTC/Ycp2
AINCENTcp2 * 0INCENTCp2
-------
C-40
Step 5 (continued)
Registers
-
38
39
17
(display)
Contents
DENOM
-------
C-41
Step 8
- Calculate revised modal utilities due to changes in
level of service.
Register
Contents
Calculation
AU
DA
AIV1T * 6IVTT
JJA JJA
AOVTT/DISTDA * 00VTT/DISTDA
+ AOPTC/Y_ . *
DA
^.
DA
MDA * 64DA + A5DA * 95DA
AU
CP
AlVTTcp * 6IVTTcp
+ AOVTT/DIST * 60VTT/DIST
AOPTC/YCP *
* 6INCENT
C_
-------
042
Step 9 - Calculate revised vanpool utility using changes in vanpool
(optional) level of service.
Register
Content
Calculation
AU,
VP
+ AOVTT/DIST * 90VTT/DIST
AOPTC/YW * 60PTC/Yvp
* 9lNCENT
vp
A5VP * 95VP
Step 10 - Calculate new modal shares using base modal shares and
changes in utilities (AU )
m
Register
-
24
25
26
27
28
Content
DENOM
e »
b DA
S'CP
s'T
S'VP
s'o
Calculation
AU.
Z (base share ) * e
i
AUn.
o DA
(ST. .[or S^J * e ) /DENOM
DA DA
AUCP
(S [or S^ ] * e ) /DENOM
AU
(ST[or Sj] * e *) /DENOM
AUVP
(S^ * e ) /DENOM
SQ/DENOM
Display or print market segment revised modal shares,
volumes, VMT, total VMT, and revised autos per worker
Step 11
- Display or print aggregated results
-------
C-43
B. REGISTERS ;
Tnput
oo 9rvxTDA
01 00VTT/DISTDA
02 eOPTC/YDA
03 94DA
04 65DA
05 eiVTTCP
06 eOVTT/DISTCp
07 eOPTC/YCP
08 84CP
09 65CP
10 9IVTTT
11 90VTT/DISTT
12 60PTC/YT
13 04T
14 95T
24
25
26
27
28
SDA
SCP
ST>
SVP
so
29
30
31
32
33
34
ZVOL'M
ZVMT'DA
ZVOL'CP
ZVMT'Cp
ZVOL'T
ZVOL'yp
Internal
22
23
24
25
£>TOT~ interim
SQP - interim
Sj)A - interim
Spp - interim
35
36
37
38
39
zvMT'yp
ZVMT_nT
ZVMT ' Tnn
i\J J
AZVMTTnn
xU J
%AZVMTrr,
26 ST - interim
27
38
- interim
CP1
39 SCP2~ interim
15
16
17
18
19
20
21
DIST
Y
occcp
POP
OCCyp
CIRC
OCCPTn
Output
21 W
24 S'
25 ' S'
26 S1
27 S'
28 S'
rrT(
DA
CP
T
VP
r\
TOT
22 OCC
CP-2
-------
C-44
C. SYMBOLS AND CODES
Modes:
Variables:
DA
SR
T
0
TOT
IVTT
m
OVTT
m
OPTC
m
INCENT
DIST
Y
OCC
m
CIRC
POP
U
m
VOL , VOL'
m m
VMT , VMT'
m m
s , sa, s'
m m m
A/W, A/W1
Ex
Other:
m
An
m
Drive Alone
Shared Ride
CP = Carpool
CP1 = Carpool class 1
CP2 = Carpool class 2
VP = Vanpool
Transit
Other
All modes
In-vehicle travel time for mode m.
Out-of-vehicle travel time for mode m
Out-of-pocket travel cost for mode m
Incentives for the carpool and vanpool modes only
Round-trip distance
Average household income for a market segment
Average vehicle occupancy for mode m
Vanpool circuity factor
Market segment population
Utility of mode m
Base and new volume of mode m
Base and new vehicle-miles of travel for mode m
Base, adjusted, and new share of mode m
Base and new autos per worker
Variable x, summed over all market segments
Coefficient of variable n for mode m
Change in variable n for mode m
-------
C-45
p. FLOW CHART; (3MODE(VAN)-AGG-2(B)/790110/ESE)
8/9
READ PROGRAM CARDS
INITIALIZE REGISTERS
& STORE DEFAULT 6's
USER-
SUPPLIED
6's?
yes
ENTER USER-
SUPPLIED 6's
ENTER MARKET SEGMENT
DATA
CARPOOLS
TREATED
IFFERENTL
yes
USE CARPOOL
SUBROUTINE
\
\
ENTER
MODE
/
no
BASE
SHARES
ENTER
OCC,
BASE
GET NEW
OCC AND
CARPOOL
ALOS,
SHARES
CARPOOL
ALOS
VANPOOL
SUBROUTINE
NEEDED?
yes
ENTER 'NEW BASE
VANPOOL SHARE
ENTER ALOS
-------
C-46
10
lla
lib
lie
GET MARKET
SEGMENT DATA
PRINT
AGGREGATED
RESULTS TO THIS
POINT?
ANALYZE
ANOTHER
MARKET SEGMENT AND
AGGREGATE WITH
ABOVE SEGMENTS
ANALYZE
NEW POLICY
SET ALL MEMORIES
TO ZERO AND 9's
TO DEFAULT?
yes
PRESS E
yes
PRESS C
GO TO STEP 4
yes
PRESS A
GO TO STEP 2
END
-------
E. LABELS;
C-47
Program Step Number
Clear memories
Store default coefficients
Store socioeconomic values
Store base modal shares
Begin vanpool subroutine
Determine new utilities
Determine new modal shares,
volumes, VMT
Print cumulative volumes
and VMT
Begin carpool subroutine
Determine new carpool utilities
'Determine revised average
carpool level of service
F. FLAGS:
None used
G. CARD FORMATS:
None used
-------
C-48
H. PROGRAM LISTING
000 76 LBL 053 18 IF: me, --:b xt>
001 11 H 054 91 R/S in? 99 PPT
002 47 CMS 055 42 STD HIS 76 LBL
003 76 LBL 056 19 19 in9 17 £;•
004 16 H' 057 91 R/S lift 91 p/c;
005 93 . 058 42 STD 111 42 STD
006 00 0 059 20 20 112 '~'7 •'-?
007 01 1 060 76 LBL 113 43 RCL
008 05 5 061 35 1/X 114 26 26
009 94 +/- 062 91 R/S 115 75 -
010 42 STD 063 42 STD 116 53 <
Oil 00 00 064 24 24 117 43 RCL
012 42 STD 065 91 R/S 118 26 26
013 05 05 066 42 STD 119 65 v
014 42 STD 067 25 25 120 43 RCL
015 10 in 068 91 R/S 121 27 27
016 93 . 069 42 STD 122 54 ':>
017 03 3 070 26 26 123 42 STD
018 02 2 071 91 R/S 124 22 22
019 94 +/- 072 42 STD 125 95 =
020 42 STD 073 27 27 126 4'"1 STD
01 01 074 91 R/S 127 26 ~26
42 STD 075 42 STD 128 43 RCL
023 06 06 076 28 28 129 24 24
024 42 STD 077 43 RCL 130
025 11 11 078 25 25 131
026 02 2 079 55 -f- 132
027 09 9 080 43 RCL 133 43 pCL
081 17 17 134 27 27
. i .-i
029 42 STD 082 85 + 135 ?5
r!0n r-~' 02 083 43 RCL 136 43 RCL
STD 084 24 24 137 22 22
07 085 95 = 138 54 ':>
42 STD 086 99 PRT 139 55 -f-
12 12 087 91 R/S 140 02 2
93 „ 088 85 + 141 54 :<
02 2 089 43 RCL 142 95 =
09 9 090 27 27 143 42 STD
42 STD 091 55 + 144 24 24
08 08 092 43 RCL 145 43 RCL
040 76 LBL 093 19 19 146 25 25
041 13 c 094 65 x 147 7=;
042 91 R/S 095 43 RCL 148 53 <
043 42 STD 096 20 20 149 53 •'
044 15 15 097 95 = 15H 43 RCL
045 91 R/S 098 65 x 151 27 27
046 42 STD 099 43 RCL 152 75 -
047 16 16 100 18 18 153 43 RCL
048 91 R/S 101 65 x 154 22 22
049 42 STD 102 43 RCL 155 54 t
050 17 17 103 15 15 156 55 +
051 91 R,--S 104 95 = 157 H2 2
052 42 STD 105 44 SUM 15ft 54 ':<
-------
Ct:.
Li:
.....J •'*
e.'.i ••—-i ii
ii.'
_J '.D IZ1 •-£) I--- _J P-- O CO .....J LP
o C'-j ii i— C'-J o ••••'• e.'i C'-j L.U LU '-•-. x e..i o -i-
a;: co x ii. ii: lit.'
I •(••
...I ' C'
Ct
..J CO
O C'
Ct.'
|j-.i i LP CO CO U'"i •••-I LP CO ''tf If.i C'-J CO LP CO '£i LP C'J '•£' O C'-J CO P-- P- OJ T-I U'!' C'"1 Ll"'1 U".1 •• ••« LP CO ••£' U'' i'0 Ll'".1 IP •••••' LP >.'"' I' - LP '"O '•£' LP CO 0"'' LP ••—i LP CO '..0 L
CO CT'' '.£i rj1 T~I co CT'' '•£' '':J" ••—< CT'> C'J C'J '•£• ^t' C'-J 0''> '"d" C'-J O CO '-.'I" C'-J '£> ••••••' Cr'' '-.D ''"!' O CO '7'1 '£' ''"1" O LP ^1" ••••••i OJ '7'1 '£' ''"I" O LP ^t' i—• U'"1 '"3" ->~i 0"' C1'' '•£' nJ" O C
LP '-£i P-™ CO 0'-' O -'-"i C'-J CO "tfr LP '-.0 P-~ CO CT'' O i—i 0-.I CO ''d" LP '-.D P-- CO >7~> O T--I C'-J CO '*•¥ LP '•£' P-- CC> CT'1 O T-I C'J i'0 '^l" LP '.£i I''-- CCi 0"-' O T-i C'-J CO 'f-f LP '-£i P-- C
'•£i '-£i '-£i '•£> '-£i P-» P-» P-- P-- P-- t'-- P-- P-- P-~ P-- CO CO CO CO CO CO CO CO CO CO 0~'" 'If'' O~'' CT'' CT'' O''1 CT'1 CT'' 0" • CT'1 O O O O O O O O O O i—ii—ii—ii—ii—IT-H-I—1-1—ii
C'-J C'-J C'J C'-J C'J C'J C'J C'J C'J C'-J C'J C'J C'J C'-J C'-J C'-J C'-J C'-J C'-J C'-J C'-J C'-J C'-J C'-J C'-J C'-J C'-J C'-J C'-J C'.J C'.J C'.J C'.J C'.J C'.J CO CO C1"' CO CO CO CO CO CO CO CO CO CO CO CO CO CO CO '.'
LP
_J
•I- O
ii:
o -i- o
_j P-- co
•I- O T-H 4- •'•-.
ii; ii:
_J CO
C~i O
—I IT) E
O C'.J II I— C'-J
ii:
l__l
a:
O
ii:
-t-
Lt.
_l C'-J
O T-l -I-
ct:
_J '
ii:
U"' CO U~i U1" 1-1 U1" CO P-- LP CO '•£' Llli CO P-» U~.i i-< U'".i CO CO LP -"--i U":i CO '•£' '•*• O Lil' -"-I" i— ' LO ^1" •*-* CO 'J'> '•£.> ^1" O CO O'1- '-£i '•*• O 0"'- C'-J C'-J ••£' ^ C'J 0''- T]" C'-J 0''- '-£i ''-'I" i— i CC' '7'' '•£' ^J"
" U"" CO LP LP I~H U~i C'"' C'J U"'.i CO '.£i
1 Lil' rr 1-1 CO O"'- '-£' •<]" i--1 LP •*!" ••~1
o
C'J CO -^1- Uli '•£> P-~ CO i7-' O 1-1 C'J CO '•*• 111" '•£> P-- CO '7-. O ••-< C'J C'li ^J- L)~i '•£" r-- CO "7'- O ••-•> C'J CO ''"I" Llli '•£> P-- CO CT'' O ••••••< C'-J CO '•:!• Li"" ••£' P-- CO "7'> O i-« C'-J CO •*•
-T-H -,-1 -T-H 1-1 -r~i -v-< -r~i -^-H OJ C'J C'-J C'J C'.J C'J C'J C'-J C'-J C'-J CO CO CO CO CO CO CO CO CO CO -*-f ••*• •* •* -r-t ••*• •*!• •* ••#• ^t' Llli U~i Li"' Llli Llli Li".' Lll" U1" LP LP '•£" '-£i '-£i '•£' '•£"
OJ C'-J C'J C'J C'J C'J C'J C'J OJ OJ C'-J C'-J C'J C'-J C'J C'-J C'J C'-J C'J C'J C'.J C'.J C'.J C'J C'J C'J C'J C'J C'J C'J C'J C'.J C'J C'J C'J C'J C'J C'J C'-J C'J C'-J C'J C'J C'.J C'J C'J C'J C'J C'-J C'-J C'J C'-J C'J
n tp_i
II I— C'-J Oil
co _i
LP C'J Llli '£i
i7'' •*• C'.J P'--
Ltl
I O
o o +
ii:
co
•._
lii:
o
Ct:
._J ID CO
O •<—i + X
a; ii:
_J C'J
O O •!•
ii:
>_i
LE.
_J CO CO _J -1*' ~-- X ,_J -^t' O -^J" CO _J U~i CO _J '-£i
o o + x x. e;> o n 12:121 x o o-j u i— o-J '••-. x o o + x x o o
ii: lit: ii: 1—1 _J ii: CO lit: lit: Ct: ii:
U'l' CO O 111" T-I LP CO T-I U~.i CO LP LP •'-' U~i CO C'J ID CO '•£> LP ••--' U") CO CO LP T-I LP CO '^J" IT" C'J CO LP CO ^J" Uli C'-J n:l" T-H LP CO LP LP T-I LP CO '-£i
'•£' ••*• O CO '7'' '•£" '^1" O LD -^t" T-I co '7'' '-£i "^1" O Ltl" 1d" 1-1 CO "7'' '•£' -"tf" O CO '7'' '-£i ^1" O '7'> C'-J C'-J '-£i "d" C'J '7'' TJ" C'J "7'' '•£' ^ O CO >7'> '-£' •& O
'7'- O T-I C'-J CO ''-T Lll' '-£' P-- CO i7'- O ••-> C'J CO ''-'J" Ii"' '•£' P--- CO f.T' O *-» C'-J CO Tf \j~) '•£> P-~- CO CT'- O T-H OJ ro -':j- |.O -.£i p-- CO i7'> O T-I C'-J CO ^J" Uli '-£> P-- 01" O"1- i_J T-I
LP '-£i '•£! '•£• '-£i '•£' '•£' '•£' '•£' '•£' '•£' P-- P-- P-- P-- P-- P-~ r-- P-- r-- P-- CO Oj CO CO CO CO CO CO CO CO Cr-' i7'> 0'> '7'> i7'' "7'' "7'> i7'> CT'' 0"' O L_I i_" "__i O i_i "_' O O '_' T-I -r-i
^ 1 ^.H ^-| ^_| ., 1 Y-l ^ 1 ., 1 -r~l ^ 1 T-l I—I T-H 1—I 1 1 T-H T-l T-l -* 1 T-l 'I-—I 1—I T-H 1 1 -I—I 1—I 1—I 1—I T-l T-l T-l T~l 1 1 T-l -I-H T-l TH 1—I -I—H T-l T~H OJ I."'J if'J |"'J 0-J C'-J C'J C'-J C'-J C'J C'-J C'J
-------
Ida
i5a
s/a
iad
13
i d a
iad
O '— •
D1S
=
«g — ,
i5a
T-
fi"7
'_' '-•
i.da
ii. o
wns
13
Ida
r '-•
wns
S£
wns
s/a
i a a
=
03
Ida
SI
Ida
X
61
Ida
-±
*£
wns
s/a
iaa
—
81
ida
iad
i3
a is
=
T1 ""?
i d a
-i-
?C
£*
16
66
13
£*
66
83
3*
S6
r"1 ^
£*
SS
,-, — *
O O
£ •fr-
iii t-
**
13
£*
13
**
S£
**
16
66
S6
03
£*
S9
SI
£*
S9
61
£*
SS
*£
**
16
66
S6
81
£1?
S9
66
i3
3*
S6
-n —y
O C-
£1?
SS
*2*
£i*
Sit7
lit
Oi*
69*
89*
•* "7 K
"7 . K
[3 • F
*9*
£9*
39*
19*
0 9 *
6S*
8S*
ZS*
9S*
SS*
*s*
£S*
3S*
IS*
OS*
6**
8**
Z**
9**
S**
***
£**
3**
I T? T?
i 1 1 p
6£*
8£*
i£*
9£*
S£*
*£*
££*
3£*
I£*
0£*
63*
83*
Z3*
93*
S3*
££
wns
s/a
iad
-
SI
ida
X
iad
93
D1S
=
33
ida
•r
93
ida
13
wns
3£
wns
s/a
iad
=
SI
Ida
X
l_\
Ida
-r
I£
wns
s/a
iaa
—
81
Ida
X
iad
S3
D1S
=
33
ida
-:-
S3
Ida
13
D1S
0£
wns
O C-
16
66
S6
81
£*
S9
66
93
3*
S6
33
£*
SS
93
£*
13
**
3£
**
16
66
S6
SI
£*
S9
^1
£*
SS
I£
**
16
66
S6
81
£*
S9
66
S3
3*
S6
3-3
£*
SS
S3
£*
13
3*
6e
**
£3*
13*
03*
61*
81*
2l*
91 *
SI*
*I*
£1*
31*
II*
01*
60*
80*
iO*
90*
SO*
*0*
£0*
30*
10*
00*
66 £
86 £
i6£
96£
S6£
*6£
£6£
36£
I6£
06£
68£
88£
Z8£
98E
SSE
*8£
£8£
7.ft£
I8£
08E
6i£
8i£
ii£
9i£
Si£
*i£
£i£
3 L £
J^ i**^ i J
~"1 •"*! **jj
i •— * C3
•:.:-
63
wns
s/a
iad
=
81
i d a
X
iad
*3
D1S
n
33
Ida
-^
*3
ida
33
D1S
=
83
ida
+
93
i d a
+
*3
Ida
+
S3
Ida
+
i3
a Is
=
LZ
Ida
a
lai
X
XHl
AH I
=
60
Ida
X
s/a
66
S6
SI
£*
S9
63
**
16
66
S6
81
£*
S9
66
*3
3*
S6
33
£*
SS
*3
£*
33
3*
S6
83
£*
SS
93
£*
SS
*3
£*
SS
S3
£*
SS
i3
3*
S6
2.3
£*
31
9i
S9
£3
-^ -^
•i •-•
S6
60
£*
S9
16
69£
89 £
i9£
99 £
S9£
*9£
£9£
39£
T9£
09E
6S£
8S£
tl S £
9S£
SS£
*S£
£S£
3SE
TS£
OS£
6*£
8*£
i * £
9*£
S*£
**£
£*£
3*£
I*£
0*£
6££
8££
i££
9££
S££
* £ £
'I- O t-
3££
I££
0££
63£
C- T ~-
iil O C-
93£
S3£
*3£
£3£
33£
I3£
03£
6 1 £
16
-------
en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en 4*. 4>- 4*- ••£•• 4*. 4*. 4*. 4* 4* 4* 4* 4* 4*. 4x 4^ 4*. 4>. •*•• 4*. 4*. 4* **• 4*
ro r-o ro r-o r-o ro ro r-o ro r-o *-» >-* *-* *-«•>-* •-•• i-^ i-^ (-* i~» o o o o o o o o o o ••£' \o ••£' ••£' ••£' •£> ••£' ••£' ••£' ••£' co cc> cc> co co co co ro ro co ~\i -\i ~--J
••£> co ---j 17-- en 4*- co ro H-* o ••£< co ---j cr. en 4^- co ro i--1- o \D co ~--j 17-' en 4*. co r-o •-* o >.£> co ~-.j 17-' en -u co r-o i-* o --£1 co --.j 17-- en 4* co r-o >-* o vo co ---j
17-, i.£i i-* -.j ro 4± '-£i r-0 41- '-£i TO 4*- O O PO 4^- O O i-* --J '•£' '•£' ro O O i-* "--J '-£i '-£i '•£' '-£i TO 4* CO i-* 4*- en rO 4^- CO 4*- '.£i O O O '7'' CO 4^ en CO 4^ '•£' CO
en i--- o 17-- ro r-o •-* *-* ro i-* cr. r-o co co en ro >-*• ro co cr. •-* o ro ••£' ro en 17-' H* co ••£' en 4^- co en -\i co en en co ••& ro en o o *-* en 17-- co en co r-o en 17-'
77 m r~ co TO co 70 co co
x '••-. • co ro — H \ ro -H \ ro — i co co ro — i •
co r~ ro n co •— • a co cr. a en a
o r~ TO r~ i~-i r~ TO ID ~o TO TO TO co TO co
ro - co ••-.. co -£ ...Q ro m co x e TO u ro o + H* o -i- ro o co -H n o o ^ x co o -i- co -H u co
r~ co -H -c r~ co -c — i 4^ r~ ----j r~ en r~ '•£• n cr- r~ co n cr.
en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en en
co co co --J --j -\i -j -M --j --.j --.j --.j --J a-- cr-. cr. a-- a-- a-, a-- a-- a-, a-, en en en en en en en en en en 4*- 4* *• *. 4^- 4* 4* 4± 4-- 4*. co co co co co co co co co co
ro *-• o ••£• co -j a-, en *. co ro »-. o ••£" co --J a-, en 4- co ro •— o •-£• co -j o". en 4- co ro *-* o ••£• co --J CO O 4*. i7'' ••£' CO TO ---J en i-* 4*. en O 4*- "7'. •-£" CO i— *• en O 4i i7-- '.£• CO O *•
en 4^- ^ a-, ro en ^ en co ro en ••£' co en •— en co co en H- en en co en cr. co en -j co en •— en en co en a-, co en •— en en co ,
CO 70 70 70 CO CO CO TO l~
ro -H ii co o + co o x ro cr ro cr •-• ro — i it x x 2:
co a ••£• r~ co r~ co a-. 3 en 3 a-. * co x
70 TO 70 TO 70 TO 70 TO
u o o x x + o o x x + ro o -i- »-* o -i- o o x x
••£• r~ co co r~ co en * c* r~ --J r~ co
TO 70 70 70
»-• o -i- o o x x + o o
en r~ a-. r~ co en r~
••. |T. 17-, 17-, 17-, iT, 17-, -*• i—» o o o o o o o o o o '•£• vo \o '•£' '-£! '•£' ••£' ••£' '-o ••£! co co co co co co co
j ••£! co -M a-, en 41. co ro >-j- o '-£i co -\i a-, en 4^. co r-':' *-* o '•£' co ---j 0-. en 4*- co ro "-•• o -.£1 co -\i cr-. en 4* co ro "-1 o ••£! co --.j a-, en *. co
I1 CO '7'. '•£> '•£' '-£i '•£' '7'. CO 4*- CO '•£' i7'> CO 4^- "—' ~-J i—*• 4^ '•£< CO '.£i PO 4* '-/I CO 4*- CO TO 4*- en CO 4^- CO 4* '-£i TO 4± en CO 4*. CO 4^ CO '-£i CO 4^- en
ji c/i i-* (-* ..£i 17-. ~-j ro '•£• en en ro co en ••£' co en >•-••• co en ro co ••£' ro en co co en ••£' co ro ro en en ro co en
i-* O 70 "0 70 70 70
:> '••-. —i '••-. TO u '•••. x co o + •••-.
":•:! d iTi —I I'd '.£i I I'M
70 TO 70 .-^i co ;ri •& co >-^ ;ri
ro o -i- co o + ro o -i- co o co —i u ro o -i- co o co —i x u co o -i-
ro r~ '-.o r~ •— r~ co r~ •-£' a co r~ ••£' r~ co a x co r~
-------
APPENDIX D
An Automotive Emissions Estimation Procedure
This appendix is structured into four parts:
• Calculation Procedure; an overview of the procedure with references
to each Table and Worksheet used;
• Tables; data on percent cold-starts and factors for cold-starts,
hot-soaks, and VMT-related emissions, to be applied as indicated in
the Worksheets;
• Worksheets; forms which guide the user through the step-by-step
calculation of automotive emissions;
• Interpolation Procedures; instructions for two- and
three-dimensional interpolation, to be used if desired in conjunction
with emission factor tables.
-------
D-2
D.I Calculation Procedure
The method requires, for both a base case and any number of policy
alternatives, levels of VMT and the number of trips in the study area.
These data items can be obtained through the use of manual, calculator, or
computerized-demand estimation techniques. The worksheets have been
designed specifically, however, to be used in conjunction with the
pivot-point demand estimation Worksheets I through V. Since the output
from the demand estimation worksheets is in terms of VMT only, Worksheet
VT-A of the emissions estimation technique is used to calculate the number
of trips for each appropriate market segment, based upon VMT data and
average trip distance values. The average trip distance values appear as
input data in the demand estimation worksheets.
Two alternative methods of estimating the percentage of trips beginning
with a cold start are presented in Worksheet VI-B. Part 1 can be used only
if parking duration data of the type described below are obtainable; Part 2
should be used in the more likely event that adequate parking data are not
readily available.
The parking duration procedure included in Part 1 of Worksheet VI-B is
based on the assumptions that vehicles equipped with a catalytic converter
will start cold after the engine has been turned off one hour, and that a
non-catalyst vehicle will start cold after four hours of disuse. The
worksheet refers to Table D.I, which provides estimates of catalyst-equipped
^These worksheets are included in Appendix A and described in
Section 3.1.1.
-------
D-3
autos for a number of forecast years. If information is available
indicating the percentage of trips involving automobiles which have been
parked longer than one hour and longer than four hours, this part can be
used. Depending on the level of aggregation of the data, this estimation
procedure can be used in repeated applications to predict emissions produced
by different market segments or by work and non-work trips. If the only
available parking duration data are for total travel, then this worksheet
can be used just once per policy.
Part 2, which should be used if parking duration data are not
available, incorporates an explicit disaggregation of work and non-work
trips. The fractions of trips with cold starts to be used in the emissions
calculation in this worksheet are given in Table D.2. The frequency of cold
starts differs among work and non-work trips, since the duration of parking
time preceding the start of a non-work trip is normally shorter than that
for work trips.
Worksheet VI-C is used to calculate total auto starting and evaporative
emissions for both work and non-work travel by pollutant (HC, CO, and NOx).
For each pollutant and purpose, start-up emissions are determined by
multiplying the total trips by the HC, CO, and NOx factors in Tables D.3,
D.I, and D.5. The emission factors in Tables D.3 and D.1* vary by percent
cold starts, temperature, and forecast year. To use these tables, it is
therefore first necessary to find the percent cold starts, applying the
procedure presented in Worksheet VI-B. Users will in most cases find it
necessary to interpolate in three dimensions to find the correct factor in
-------
D-4
Tables D.3 and D.4 for specific percentages of cold starts, average
temperatures, and forecast years. The NOx emission factors in Table D.5
do not vary by temperature, although they do vary by year. Interpolation
will generally be necessary to account for the percent of cold starts and
forecast year.
Because they are related to the number of trips, evaporative HC
emissions are also calculated using Worksheet VI-C. The quantity of
hot-soak emissions is found by multiplying the number of auto trips by the
appropriate hot-soak emission factors in Table D.6. The worksheet also
provides for the summation of emissions by pollutant and by trip purpose,
and of total by pollutant.
The next stage of the emissions estimation technique is calculation of
VMT-related emissions, using Worksheet VI-D. This simply requires the
multiplication of vehicle miles by travel by population subgroup and purpose
by the factors given in Table D.7. Interpolation by speed and forecast year
will generally be required.
As a final step, the total change in emission levels is calculated by
summing the trip-related and VMT-related emissions using Worksheet VI-E.
Inputs are obtained from Worksheets VI-C and VI-D for both base and revised
alternatives. The worksheet also allows percentage changes by population
subgroup and pollutant, and for all subgroups, to be computed.
1-Exact methods of linear interpolation in two and three dimensions
appear in Section D.U.
-------
D-5
D.2 Tables
All emission factors appearing in Tables D.I through D.7 come from the
same U.S. DOT/EPA source (3D. These tables include no correction factors
for differences in the following variables, the values of which are held
constant in all applications of the method:1
• percentage of travel by air-conditioned vehicles
• vehicle load
• percentage of travel by trailer-towing vehicles
• humidity
The tables also make no allowance for the effects of government-mandated
vehicle inspection and maintenance programs, which will reduce emission
levels relative to values given in the tables.
•'•More detailed estimation procedures recommended by EPA do include
correction factors (29) for these variables.
-------
D-6
TABLE D.I
Percent of Auto Trips by Catalyst-Equipped Vehicles
Forecast Year
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
Percent of Auto
Catalyst-Equipped
38%
47
57
66
76
85
88
90
93
95
98
Trips by
Vehicles
Source: (70)
-------
D-7
TABLE 0.2
Percentage of Trips with Cold Starts
Year
1979
1980
1981
1982
1983
1984
1985
1986
1987
Work Trips
86%
88
89
90
91
91
92
92
92
Non-work Trips
49%
51
54
57
58
58
59
59
59
Sources:
percent cold starts by trip purpose data (69)
2) percent catalytic and non-catalytic-equipped
vehicles per year (70)
3) trip frequency by trip purpose data (15)
-------
D-8
TABLE D.3
Start-up Hydrocarbon Emission Factors (Grams/Trip)
Y
E
A
R
1
9
7
7
1
9
8
2
I
9
6
7
• •
PERCENT
OF TRIPS
STARTING
COLO
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
70
80
90
100
tfmmmmm»~mm
40 1
3,9
6,1
8,2
10,4
12.5
14.7
16.8
19.0
21.1
23,3
25,4
3.0
«.3
5.6
6.8
8,1
9,4
10,7
12,0
13,3
14,6
15,8
3.2
4,0
4,9
5. ft
6,6
7.5
8,3
9,2
10,0
10,9
11.8
^•••••••«
T E »•
50 1
3.9
5,7
7,4
9.1
10.8
12.6
14.3
J6.0
17.8
19,5
21.2
3.0
4.0
5.0
6,1
7.1
8.1
9.2
10.2
11.2
12.3
13.3
3.2
3.9
4.6
5.2
5.9
6.6
7.3
8.0
8.7
9.3
10.0
y mmmmmmm
1 P E R 4
60 1
3.9
5.3
6.7
8.0
9.4
10.8
12.2
13,6
14.9
16.3
17.7
3.0
3,8
4.6
5,4
6.3
7.1
7.9
8.7
9.6
10.4
11.2
3.2
3.7
4.3
4.8
5.4
5,9
6.5
7.0
7.5
8.1
8,6
k •••••••
k T U R E
70 1
3.9
5.0
6.1
7.2
8.2
9.3
10,4
11.5
12.5
13.6
14.7
3.0
3,6
4.3
4,9
5.6
6.2
•6.9
7,5
8,2
8,8
9.5
3.2
3.6
4.1
4.5
4.9
5.4
5.8
6.2
6.7
7.1
7.5
^•••••*B
( D E
80 1
4.0
4.9
5.7
6.6
7.4
8.3
9.1
10,0
10,8
11.7
12.5
3.1
3.6
4.1
4.7
5,2
5.7
6.2
6,8
7,3
7,8
8,4
3.3
3.6
4.0
«,3
4,7
5,1
5,4
5,8
6,1
6.5
6.8
*•>»*»»•«
c. F )
90 1
4.1
4,7
5,3
5,9
6.6
7,2
7,8
8,5
9,1
9,7
10,4
3,1
3,5
3,9
4.3
4.7
5,2
5,6
6,0
6,4
6.8
7,2
3,3
3,6
3,9
4,2
4,4
4,7
5,0
5,3
5,6
5.8
6,1
f w»«r»»»«
100 I
4.1
4.5
4.9
5.4
5.8
6,3
6,7
7,2
7.6
8,0
8.5
3.1
3,4
3,7
4,0
4.4
4.7
5.0
5,3
5,6
6,0
6.3
3,3
3.5
3.7
4.0
4.2
4,4
4,6
4.9
5,1
5.3
5.5
*....*.«4
Source: (70)
-------
D-9
TABLE D.4
Start-up Carbon Monoxide Emission Factors (Grams/Trip)
y
E
A
R
MM 1
• V
1
9
7
7
t
9
e
2
1
9
8
7
• «•<
PERCENT
OF TRIPS
STARTING
COLD
0
10
20
30
40
50
60
70
60
90
100
0
10
20
30
40
50
60
70
60
90
100
0
10
20
30
40
50
60
70
60
90
100
k •••••>•»•«<
10 1
15.5
91.6
166.
244.
320.
396.
472.
546.
624.
700.
776.
9.8
61.7
114.
165.
217.
269.
321.
373.
425.
477.
526.
9.7
35.5
61,4
87,2
113.
139.
165.
191.
217.
242.
266.
^mmmmmmm^
T E *
20
15.5
76.6
138.
199,
261.
322.
383.
445.
506.
567.
629.
9.8
50.8
91.9
133.
174.
215.
256.
297.
338.
379.
420.
9.7
31.2
52.7
74.3
95.8
117.
139.
160.
182.
203.
225.
• ••»••••<
1 P E R /
30
15.5
64.9
114*
164.
213,
262^
312^
361.
411.
460.
509.
9.8
42.5
75.1
108.
140.
173.
206,
238.
271.
304.
336.
9.7
27.8
45.8
63.9
82.0
100.
118.
136.
154.
172.
191,
.»••»••••.
^ T U R E
40 j
15,5
55.2
95,0
135.
174.
214,
254.
29«.
333.
373.
413,
9.8
36,0
62.1
88,2
114.
141 ;
167^
193^
219.
245,
271.
9.7
25.0
40.4
55.7
71.1
86,4
102^
117.
133.
148.
163.
• «•••»•••<
: ( D E
50
15.5
47.4
79.3
111.
143.
175.
207.
239,
271,
303.
334.
9.8
30,9
51.9
73.0
94.1
115^
136.
157.
17*.
199.
220.
9.7
22.8
36,0
49.2
62.3
75.5
88.7
102.
115.
128.
1«1.
lmm»mm*mi
G. F )
60
15.5
41.0
66,5
92,0
117.
143.
168.
194,
219.
245.
270.
9.8
26,9
44,0
61,1
78.1
95.2
112,
129.
146.
164.
181,
9.7
21.1
32,5
43.9
55,3
66,7
76,2
89.6
101,
112.
124.
• 9mm»mmm4
70 1
15.5
35.7
56.0
76,2
96.5
117.
137.
157.
177,
198,
218,
9.8
23.8
37,7
51.7
65,6
79,6
93,5
107.
121,
135.
149,
9.7
19.7
29,7
39,7
49,7
59,7
69,7
79,7
89,7
99,7
110,
imwmmmmmi
Source: (79)
-------
D-10
TABLED. 5
Start-up NOx Emission Factors (Grams/Trip)
% of Trips
Starting
Cold
0
10
20
30
40
50
60
70
80
90
100
Year
1977
4.6
4.6
4.6
4.5
4.5
4.4
4.4
4.4
4.3
4.3
4.3
1982
2.1
2.3
2.5
2.6
2.8
2.9
3.1
3.2
3.4
3.5
3.7
1987
1.2
1.4
1.6
1.8
2.0
2.1
2.3
2.5
2.7
2.9
3.1
Source: (70)
-------
D-ll
TABLE D.6
Hot Soak Emission Factors
Forecast Year
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
Hot Soak Factor
11.8
10.6
9.4
8.3
7.1
6.0
5.2
4.4
3.6
2.8
2.2
(HCgrams/trip)
Source: (70)
-------
D-12
TABLE D.7
Auto Travel Emission Factors (Grams/Mile)
SPD
5,0
7.5
10,0
12.5
15.0
17.5
20.0
22.5
25.0
27.5
30.0
32.5
35,0
37,5
40,0
42.5
45,0
47.5
50,0
52.5
55,0
57.5
60.0
1977
13^0
*l*
6,8
5*5
-------
D-13
D.3 Worksheets
Worksheets VI-A through VI-E, desribed in Section D.I, appear on the
pages which follow.
-------
VI-A. INPUT TRAVEL DATA SUMMARY FOR EMISSIONS
ESTIMATION
Base Alternative
| [Revised Alternative
Forecast Year:
Policy:
Population
Subgroup
£ =
Work VMT
(I or IV)1
Average
Work Trip
Distance
(miles)(I)
TT-> __
Number of
One-Way
Work Trips
Non-Work
VMT
(I or IV)
«
Average
Non-Work
Trip
Distance
(miles) (I)
=
£ =
Non-Work
Trips
u
TOTAL WORK VMT
TOTAL WORK TRIPS
TOTAL NON-WORK VMT
TOTAL NON-WORK TRIPS
Source Worksheets are indicated in parentheses where applicable.
VMT and trips on worksheets I and IV in Appendix A must be multiplied
by the number of households per population subgroup.
TOTAL TRIPS
-------
VI-B. COLD START FRACTIONS
Base Alternative
Revised Alternative
Policy:
1. If Daily Parking Duration Data are Available:
Subgroup:
Forecast Year:
% of Auto Trips by % of Auto Trips with % of Auto Trips by % of Auto Trips with
Catalyst-Equipped Parking Duration Non-Catalyst Equip- Parking Duration % of Auto Trips
Vehicles (Table D.I) 1 hour ped Vehicles 4 hours with Cold Starts
2. If Daily Parking Duration Data are not Available
of Work Trip Cold Starts (Table D.2)
Non-Work Trip Cold Starts (Table D.2)
-------
VI-C. AUTO START-UP AND EVAPORATIVE EMISSIONS
Base Alternative
Revised Alternative
Policy:
Forecast Year:
Temperature:
Work Trips
I
Non-Work Trips
(1)
Population
Subgroup
TOTALS
(2)
Z Cold
Starts.
(VI-B)1
(3)
Trips
(IV-A)
Pollut-
ant2
HC(c)
C0(c)
N0x(c
HC(h)
«C(c)
C0(c)
N0x(c
HC(h)
HC(c)
C0(c)
>)0x(c
*C(h)
CO
ll
D.3
D.4
D.5
D.6
D.3
D.4
D.5
D.6
D.3
D.4
D.5
D.6
TOTALS
(4)
Start-Up
Factors
HC
X, co
Subgroups NOx
(5)
Emissions "
Col. 3 X Col. 4
(grams)
(6)
% Cold
Starts
(VI-B)
(7)
Trips
(VI-A)
1
(8)
Start-Up
Factors
HC
^ CO
f *
Subgroups NOx
(9)
Emissions -
Col. 3 X Col. 4
(grams)
Source Worksheets are indicated in parentheses
where applicable
2
(c) Indicates cold start factor
(h) indicates hot soak factor
both work and non-work start-up factors
obtained from the indicated tables
Work Trip Start-
Up Emissions
(grams)
Non-Work Trip
Start-Up
Emissions
(grams)
Total Start-Up
Emissions
(grams)
-------
VI-D. AUTO TRAVEL EMISSIONS
Base Alternative
Revised Alternative
Policy:
Forecast Year:
Work Trips
I
Non-Work Trips
I
\ /
(1)
Population
Subgroup
(2)
Average
Speed
(3)
m
(VI-A)1
TOTALS
(4)
Auto Travel
Factors
(Table D.7)
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
v^ Hc
2_j co
NOx
SubcrouDB
(5)
Emissions "
Col. 3 X Col. 4
(grams)
(6)
Average
Speed
J
1
(7)
VMT
(VI-A)
-
1
Subff
(8)
Auto Travel
Factors
(Table D.7)
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
T-^ HC
2^t co
NOx
roups
(9)
Emissions "
Col. 7 X Col. 8
(grams)
Source Worksheets are Indicated in
parentheses where applicable
Work Trip Travel
Emissions
(grams)
Non-Work Trip
Travel Emissions
(grams)
Total VMT Travel
Emissions
(grams)
-------
VI-E. SUMMARY OF CHANGES IN EMISSIONS
Revised Alternative
Policy:
Forecast Year:
(1)
Population
Subgroup
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
HC
CO
NOX
Base Emissions
(2)
Trip-Related
(VI-C)1
TOTALS
Source Worksheets are indicated in
parentheses where applicable
O)
Travel
(VI-D)
HC
ZCO
NOx
Sub-
groups
(4)
Total
(Col. 2 + Col. 3)
Total Base
Emissions
(grains)
Revised Emissior
(5)
Trip-Related
(VI-C)
(6)
Travel
(VI-D)
Z
s
(7)
Total
[CoLS+Coi. 6)
HC
CO
NOx
(8)
Change in
Total
Emissions
;Col. 4-CoL 7)
Sub- Total Change
groups in Emissions
(grans)
(9)
Percent
Change in
Emissions
(Col. 8/CtoU)
x 100
Percent
Change,
Total
Emissions
o
fc~•
oo
-------
D-19
D.^ Interpolation Methods for Multi-Dimensional Tables
D.il.l Two-Dimensional Interpolation
This procedure can be used to obtain interpolated values of emissions
factors from Table D.5 and D.7. In each table, let y equal the year, z
equal the vertical dimension (percent of trips starting cold in Table D.5,
and speed in Table D.7), and e equal the emission factors. Thus, yf and
zf represent the specified input levels; e_, the desired output factor.
a. Determine the bracketing values for both y and z; the largest values
less than yf and zf (y, and z,) and the smallest values greater
than y and zf (y^ and z2).
b. Determine the emission factors for each combination of bracketing
values. Arrange these in a table of the following form:
combination emissions factors
yl'Zl _ *!!_
yi'Z2 _ *12_
y2'Zl e21
e22
c. Calculate the following ratios:
f = f =
- yl Z2 - Zl
-------
D-20
d. Compute the interpolated value:
e»- n _ r wi _ f ^ • P
f - ^ zy;u Iz; ell
+ (1 - V * fz ' e!2
* V1 - V ' e21
+ fy ' fz ' e22
The following example shows the use of the procedure to obtain an
interpolated value from Table D.5 corresponding to y_ = 1979,
zf = 32.
a. Determine bracketing values:
x-j^ = 1977, x2 = 1982
ZL = 30, z2 = 40
b. Determine emissions factors:
combination
1977,
1977,
1982,
1982,
30
40
30
40
emissions factor
p _ h c
1 1 ~ *
e12 = 4.5
e21 = 2.6
e22 = 2.8
c. Calculate ratios:
f =
y 1982 - 1977
32 - 30
f s
40-30
-------
D-21
d. Compute the interpolated value:
ef = 0.6 • 0.8 • 4.5 +
0.6 • 0.2 • H.5 +
0.4 • 0.8 • 2.6 +
O.JJ • 0.2 • 2.8 = 3.76
D.U.2 Three-Dimensional Interpolation
The two-dimensional procedure can be extended to three dimensions for
use with Tables D.3 and D.4. In each table, let t = temperature, p =
percent of trips starting cold, y = year, and e = emission factor. As
before, the subscript f deontes the desired values, and the subscripts 1 and
2 the bracketing values. The same steps are included in this procedure as
those in the two-dimensional procedure. In each case, however, changes
exist to reflect the number of dimensions.
An example from Table D.3 is shown integrated in each step. In this
example, tf = 52, pf = 36, and yf = 1979.
a. Determine the bracketing values t,, t-, p., p?, y.., y .
Example:
tlf t2 = 50, 60
Px, p2 = 30, UO
yr y2 = 1977, 1982
-------
D-22
b. Determine emission factors for each combination of bracketing values;
Example
c.
combination emission factor combination
tj_ P! y1 e111 50, 30, 1977
tj_ Px Y2 e112 50, 30, 1982
*! P2 yI e121 50, 40, 1977
tj_ P2 Y2 e122 50, 40, 1982
t? PJ_ yx egll 60, 30, 1977
t2 Pj_ y? e212 60, 30, 1982
t2 P2 y1 e221 60, 40, 1977
t2 P2 Y2 e222 60, 40, 1982
Calculate ratios: Example
t - t
f - - f -
t ~ t '
t - t
t2 - tx
Pf - P!
f - f -
p p
P2 - PI
f _ .. ,L_ ^ f -
emission factor
9.1
6.1
10.8
7.1
8.0
5.4
9.4
6.3
50-50
60 - 50
36 - 30 06
40 - 30
1979 - 1977 n ,.
y2 -
1982 - 1977
-------
D-23
d. Compute the interpolated value:
Example;
ef = (1 - ft)(l - f )1 - f ) • ein + ef = 0.8 • 0.4 • 0.6 • 9.1
(1 - ffc)(l - f ) • f • en2 + 0.8 • 0.4 • 0.4 • 6.1 +
(1 - f. ) • f (1 - f ) • e.-, + 0.8 • 0.4 • 0.6 • 10.8
t p y J.
-------
APPENDIX E
An Automobile Fuel Consumption and Operating Cost Estimation Procedure
The inputs required to apply the manual method of estimating
automotive fuel consumption and operating costs are listed in Table E.I,
along with default values based on 1976 conditions. Various sources from
which these input values may be obtained for particular localities and
analysis years are also indicated. The method is based on urban-area
travel data in the 10 to UO mile per hour speed range and does not
represent the decreased fuel economy which occurs at speeds greater than
40 miles per hour.
After determination of the input values for each population
subgroup/trip purpose segment, the following steps are carried out:
1. Compute Fuel Consumption for Warmed-up Operation
A = 0.009133 + 0.008U26 . WT
B = -O.OOUnoM + 0.003637 . WT
FWARM = (A . DIST + B . TIME) . RFC
2. Compute Temperature/Distance Correction Factor
TDFAC = 1.0 + (TEMP - 10) . 0.00503 . DIST"0'3738
-------
E-2
TABLE E.I
Inputs to the Automobile Fuel Consumption and Operating
Cost Procedure
Variable
Name
I . Genera
WT
RFC
TEMP
II . Gener
GCOST
RMTCOST
III. Popu
TIME
DIST
TPHH
HHS
Description
1, Related to Fuel Consumption
Average passenger vehicle weight (1000 Ibs]
Impacts of all technological factors but
vehicle weight on fuel consumption, relativ
to 1975 technology (ratio of analysis year
fuel consumption to 1975 fuel consumption
when vehicle weight reamins constant)
Average ambient temperature (degrees celsiu
al, Related to Operating Costs
Gasoline cost (dollars/gallon in analysis
year dollars)
Repair, maintenance, and tire costs
(dollars/mile in analysis year dollars)
lation Subgroup and Trip Purpose Segment-Spe
Average one-way trip time (minutes)
Average one-way trip distance (miles)
Daily trips per household made by the group
(daily one-way trips/household)
Total households in the group (households)
Defaults
(1976
Conditions)
3.96
1.00
e
s) 10
0.60
0.03 x97
cific
-
-
—
Source
local vehicle
fleet data
( 6,3)
EPA data on
fuel economy
vs. vehicle
weights
local weather
data
local gasoline
sales data
Bureau of
Labor Statis-
tics (68)
Worksheet VI
Worksheet I
Worksheet I
or VI-A
local data,
Worksheet I
-------
E-3
3. Compute Cold Start Correction Factor
a. If DIST is less than or equal to 3 miles,
RCW = 0.5689 + 0.11447 . In (DIST)
b. If DIST is greater than 3 miles, and less than or equal to 25
miles,
RCW = 0.6187 + 0.1022 . In (DIST)
c. If DIST is greater than 25 miles,
RCW = 0.95
1. Compute Fuel Consumption for the Segment (FUEL is in gallons per day)
FUEL = FWARM . TPHH . HHS/(RCW . TDFAC)
5. Compute Operating Costs for the Segment (OCOST is in analysis year
dollars per day)
OCOST = FUEL . GCOST + DIST . RMTCST . TPHH . HHS
Values obtained for separate population subgroups and trip purpose
segments can be summed to obtain estimates of analysis area totals.
As an example of the procedure's application, consider the case in
which all general input variables take on their default values, and the
following values have been estimated for all work trips in an urban area
following implementation of a policy alternative oriented to improving
air quality:
TIME = 25 minutes
DIST = 14 miles
TPHH = 2.1 "daily work trips/household
HHS = 15,700 households
-------
E-4
The steps described above follow:
1. A = 0.009133 + 0.008426 . 3.96
= 0.0425
B = -0.004404 + 0.003637 . 3.96
= 0.01
FWARM = (0.0425 . 14 + 0.01 . 25) . 1.0
= 0.845
2. TDFAC = 1.0 + (10 - 10) . 0.00503 . 14~°'3738
= 1.0
3. DIST = 14, so 3b. is used:
RCW = 0.6187 + 0.1022 In (14)
* 0.8884
4. FUEL = 0.845 . 2.1 . 15,700/(0.8884 . 1.0)
= 31,360 gallons/day
5. OCOST = 31,360 . 0.60 + 14 . 0.0397 . 2.1 . 15,700
= 37,140 dollars/day (1976 $)
-------
APPENDIX F
Bibliography
1. Adler, T., and M. Ben-Akiva, "A Joint Frequency, Destination, and
Mode Choice Model for Shopping Trips," Transportation Research
Record 569, Transportation Research Board, 1975.
2. Atherton, Te^ry J and Moshe E. Ben-Akiva, "Transferability and
Updating of Disaggregate Travel Demand Models," Transportation
Research Record 610, Transportation Research Board, Washington,
D.C., 1977.
3. Austin, T.C. e_t a_l., "Passenger Car Fuel Economy Trends Through
1976," Automobile Fuel Economy, Progress in Technology Series, Vol.
15, Society o* Automotive Engineers, 1976.
M. Barton-Aschman Associates, Bicycling in Pennsylvania; Recommended
State Policies for Providing Bicycle Facilities and Programs,
prepared for the Pennsylvania Department of Transportation, 1976.
5. Cambridge Systematics, Inc., A Behavioral Model of Automobile
Ownership and Mode of Travel, U.S. Department of Transportation,
Office of the Secretary, and the Federal Highway Administration,
Wshington, D.C., 1974.
6. Cambridge Systematics, Inc., Urban Transportation Energy
Conservation: Final Report, Volume II, "Analytical Procedures for
Estimating Changes in Travel Demand and Fuel Consumption," prepared
for the U.S. Department of Energy, Washington, D.C., 1978.
7. Cambridge Systematics, Inc., Urban Transportation Energy
Conservation, Final Report, Volume III, "Case City Application of
Analysis Methodologies," prepared for the U.S. Department of Energy,
Washington, D.C., 1978.
8. Cambridge Systematics, Inc., Urban Transportation Energy
Conservation, Final Report, Volume V, "SRGP Operating Instructions
and Program Documentation," prepared for the Department of Energy,
Washington,T).C., 1978.
-------
F-2
9. Cambridge Systematics, Inc., "Parking Management in the Denver
Region - Analysis of Transportation Impacts and Implementation
Measures," prepared for the Denver Regional Council of Governments,
Denver, Colorado, 1978.
10. Cambridge Systematics, Inc., "Travel Model Development Project,
Volume I, Summary Report," prepared for the Metropolitan
Transportation Commission, Berkeley, California, 1977.
11. Cambridge Svstematics, Inc., "Dual Mode Planning Case Study, Final
Report, Volume 3," prepared for the U.S. Department of
Transportation, 1977.
12. Cambridge Systematics, Inc. and Multisystems, Inc., Method for
Estimating Patronage of Demand Responsive Transportation Systems,
Final Report prepared for U.S. DOT, 1977.
13. Cambridge Systematics, Inc., A Behavioral Analysis of Automobile
Ownership and Modes of Travel, Volumes II, III, IV, prepared for the
U.S. Department of Transportation, Office of the Secretary, and the
Federal Highway Administration, Washington, D.C., 1976.
14. Cambridge Systematics, Inc., "Guidelines for Travel Demand Analysis
of Program Measures to Promote Carpools, Vanpools and Public
Transportation," prepared for the Federal Energy Administration,
Washington, D.C., 1976.
15. Cambridge Systematics, Inc., "Analysis of Motor Vehicle Ownership
and Household Travel Patterns," prepared for the National Science
Foundation, Washington, D.C., 1979.
16. Cilliers, Matthys P., Reed Cooper and Adolf D. May, FREQ6PL - A
Freeway Priority Lane Simulation Model, Institute of Transportation
Studies, University of California, Berkeley, prepared for the
California Department of Transportation, 1978.
17. Chan, Yupo, "Review and Compilation of Demand Forecasting
Experiences: An Aggregation of Estimation Procedures," prepared for
the U.S. Department of Transportation, Washington, D.C., 1977.
18. COMSIS Corporation, Quick Response Urban Travel Estimation
Techniques and Transferable Parameters, NCHRP Report 187,
Transportation Research Board, Washington, D.C. 1978.
19. Courage, Kenneth, "Description of Program Library for Portable
Calculator - Signalized Intersection Operations Analysis Workshop,"
Traffic Engineering Enhancement Program, University of Florida,
Gainesville.
-------
F-3
20. Creighton-Hamburg, Inc., Micro-Assignment - Final Report, Bethesda,
Maryland, 1969.
21. Creighton-Hamburg, Inc., Freeway Surface Arterial VMT Splitter, and
Schneider-Scott Direct Assignment VMT Splitter Program, prepared for
the U.S. Department of Transportation, FHWA, Contract FH-11-7585
(Phase A), July 1971.
22. Davidson, K.B., "A Flow-Travel Time Relationship for Use in
Transportation Planning," Proceedings, Third Conference of the
Australian Road Research Board, Volume 3, Part 1.
23. De Leuw, Gather & Company, A.M. Voorhees and Associates, and R.H.
Pratt, Inc., Transit Corridor Analysis - A Manual Sketch Planning
Technique, a UTPS manual available from FHWA or UMTA, Washington,
D.C.
2U. De Leuw, Gather & Company, Characteristics of Urban Transportation
Systems; A Handbook for Transit Planners, prepared for the U.S.
Department of Transportation, Washington, D.C., 1974.
25. Domencich and McFadden, Urban Travel Demand; A Behavioral Analysis,
North-Holland Press, Amsterdam, 1975.
26. Duguay, B., Woo Jung, and D. McFadden, "SYNSAM: A Method for
Synthesizing Household Transportation Survey Data," working paper
No. 7618, Urban Travel Demand Forecasting Project, Institute for
Transportation Studies, University of California, Berkeley, 1976.
27. Easa, Said, Alan Willis, and A.D. May, Traffic Management of Dense
Networks, working paper, Institute of Transportation Studies,
University of California, Berkeley, December, 1978.
28. Ellis, e_t al., "Estimating Hot/Cold Transient Phase in Motor Vehicle
Operation," prepared for the Federal Highway Administration, 1978.
29. Environmental Protection Agency, "Mobile Source Emission Factors,
Final Document," 1978.
30. Environmental Protection Agency, "User's Guide to MOBILE1: Mobile
Source Emissions Model," 1978.
31. Environmental Protection Agency, "Compilation of Air Pollutant
Emission Factors," EPA Report No. AP-42, Second Edition, 1976.
32. Goldner, William et al., "Theory and Application: Projetive Land
Use Model," Institute of Transportation Studies, University of
California, Berkeley, prepared for the Federal Highway
Administration, March 1972.
-------
F-4
33. John Hamburg and Associates, "North Central Texas Council of
Governments Thoroughfare Analysis Process User's Manual," 1977.
34. John Hamburg and Associates, and R.H. Pratt and Associates,
"Functional Classification for the North Central Texas Thoroughfare
Planning System," 1976.
35. Haney, Dan G., SRI Network Analysis Program (SNAP), Stanford
Research Institute Project ^0-456531-02^0?, 1971.
36. Highway Research Board, Highway Capacity Manual, Special Report 87,
Washington, D.C., 1965.
37. Ingram, G.K., G.R. Fauth, and F.A. Kroch, TASSIM; A Transportation
and Air Shed Simulation Model - Volume I; Case Study of the Boston
Region; Volume II; Program User's Guide; prepared for the U.S.
Department of Transportation, 1974.
38. Institute of Traffic Engineers, Transportation and Traffic
Engineering Handbook, 1976.
39. Jovanis, Paul P. and A.D. May, Further Analysis and Evaluation of
Selected Impacts of Traffic Management Strategies on Surface
Streets, Institute of Transportation Studies, University of
California, Berkeley, October 1977.
40. Jovanis, Paul P., A.D. May, and Waiki Yip, FREQ6PE - A Freeway
Priority Entry Control Simulation Model, Research Report No. 78-9,
Institute of Transportation Studies, University of California,
Berkeley, 1978.
41. Kaplan, J.A. and L.D. Powers, "Results of SIGOP-TRANSYT Comparison
Studies," Traffic Engineering, September 1973.
42. Kassoff, H., and D.S. Gendell, "An Approach to Multi-Regional Urban
Transportation Policy Planning," Highway Research Record 348, pp.
76-93, 1971.
43. Koppelman, Frank S., Travel Prediction with Models of Individual
Choice Behavior, Ph.D. Dissertation, M.I.T. Department of Civil
Engineering, Cambridge, Massachusetts, 1975.
44. Lieberman, E.B., et al., "Logical Design and Demonstration of the
UTCS-1 Simulation Model," Highway Research Record 409, 1972.
45. Lieberman, E.B., R.D. Worall, D. Wicks, and J. Woo, NETSIM Model;
Volume 4 - User's Guide, U.S. Department of Transportation, FHWA,
Washington, D.C., 1977.
-------
F-5
46. Mackinder, I.H., "COMPACT - A Simple Transportation Planning
Package," Journal of Institution of Municipal Engineers (U.K.),
Volume 100, No. 8, August 1973.
47. Manheim, Marvin L., Fundamentals of Transportation Systems Analysis,
Volume 1: Basic Concepts, M.I.T. Press, Cambridge, Massachusetts,
1979.
48. Manheim, Marvin L., Furth, and Solomon, "Responsive Transportation
Analyses: Pocket Calculator Methods," Volumes 1-3, Department of
Civil Engineering, M.I.T., Cambridge, Massachusetts, 1978.
49. May, A.D., "Some Fixed-Time Signal Control Computer Programs,"
Traffic Control Symposium, Monte Carlo, September 1974.
50. Metropolitan Washington Council of Governments, "TRIMS Model User's
Manual," Technical Report No. 9, 1976.
51. Mulder, T.E., "A Temporal Micro-Assignment Model," Chicago Area
Transportation Study, Report 344-02, 1973.
52. Murphy, Jeremiah F., "The Fully Programmable Hand-Held Calculator:
A Tool for Transportation Engineers," Transportation Engineering,
1977.
53. N.D. Lea and Associates, Ltd., "Transportation Policy and Transit
Improvement Study for the City of Regina," Final Report, Volume 2,
prepared for the City of Regina, 1977.
54. Pecknold, Wayne M. and John H. Suhrbier, "Tests of Transferability
and Validation of Disaggregate Behavior Demand Models for Evaluating
the Energy Conservation Potential of Alternative Transportation
Policies in Nine U.S. Cities," Final Report, prepared for the
Federal Energy Administration by Cambridge Systematics, Inc., 1977.
55. R.H. Pratt Associates, Inc., "Traveller Response to Transportation
System Changes," prepared for the U.S. Department of Transportation,
Washington, D.C., 1977.
56. Robertson, D.I. , "TRANSYT: A Traffic Network Study Tool," U.K.
Ministry of Transport, TRRL Report LR 253, Crowthorne, U.K., 1969.
57. Robertson, D.I. and P. Gower, "User Guide to TRANSYT Version 6,"
U.K. Ministry of Transport, TRRL Supplementary Report 255. 1977.
58. Schleifer, H. and S.L. Zimmerman, "Distribution of VMT to Large
Zones for Sketch Planning," Transportation Research, February 1977.
59. Schleifer, H., S.L. Zimmerman, and D.S. Gendell, "The Community
Aggregate Planning Model, An Urban Transportation Sketch Planning
Procedure," Transportation Research Record 582, 1976.
-------
F-6
60. Schneider, M., Direct Estimation of Traffic at a Point, Tri-State
Transportation Committee Interim Technical Report, July 1965.
61. Talvitie, Antti, et_ al., Policy Evaluation in a Transportation
Corridor, Final Report Series, Vol. XI, Urban Travel Demand
Forecasting Project, University of California, Berkeley, 1978.
62. Tanner, J.C., L.C. Cyenes, D.V. Lynam, S.V. Magee, and A.H. Tulpule,
"Development and Calibration of the CRISTAL Transport Planning
Model," RRL Report, Transport and Road Research Laboratory,
Department of the Environment, (U.K.), 1972.
63. Traffic Research Corporation, "Final Report: Development and
Calibration of the EMPIRIC Land Use Forecasting Model," prepared for
the Metropolitan Area Planning Council, Boston, Massachusetts, 1967.
6U. Traffic Research Corporation, "Traffic Signal Optimization Program,"
prepared for the Bureau of Public Roads, U.S. Department of
Commerce, Washington, D.C., 1966.
65. Transport and Road Research Laboratory (U.K.), Mathematical Advisory
Unit, Department of the Environment, "An Integrated Multimode Trip
Distribution and Assignment Program," 1973.
66. U.S. Bureau of Public Roads, SIGOP; Traffic Signal Optimization
Program User's Manual, NTIS No. PB 182835, 1968.
67. U.S. Department of Commerce, "Calibrating and Testing a Gravity
Model for Any Size Urban Area," Washington, D.C., 1965.
68. U.S. Department of Labor, Bureau of Labor Statistics, Monthly Labor
Review, published monthly.
69. U.S. Department of Transportation, "Determination of Vehicular Cold
and Hot Operating Fractions for Estimating Highway Emissions," 1976.
70. U.S. Department of Transportation, "How to Prepare the
Transportation Portion of Your State Air Quality Implementation
Plan," Washington, D.C. , November 1978.
71. U.S. Department of Transportation, Urban Mass Transportation
Administration, "Introduction to UTPS," Washington, D.C.
72. U.S. Department of Transportation, Urban Mass Transportation
Administration, "UTPS Reference Manual," Washington, D.C.
73. University of Illinois at Chicago Circle, The Service Area
Identification Methodology, Volume U, Final Report, prepared for the
U.S. DOT, 1977.
-------
F-7
7*4. Wagner, Frederick A., Urban Transportation Energy Conservation,
Volume IV; Analysis of Traffic Engineering Actions, prepared under
subcontract to Cambridge Systematics, Inc., for the U.S. Department
of Energy, 1978.
75. Watanatada, T., Application of Disaggregate Choice Models to Urban
Transportation Sketch Planning, Ph.D. Thesis, Department of Civil
Engineering, Massachusetts Institute of Technology, Cambridge,
Massachusetts, 1977.
76. Webster, F.V. , Traffic Signal Settings, Road Research Technical
Paper No. 39, Great Britain Road Research Laboratory, 1958.
77. Webster, F.V. and B.M. Cobbe, "Traffic Signals," Road Research Paper
No. 56, London, England, 1966.
' "U.S. GOVERNMENT PRINTING OFFICE ! 1980 0-311-726/3705
------- |