United States
Environmental Protection
Agency ^^^
Office of Mobile Sources
Washington, DC 20460
EPA-42Q-R-94-Q02
July 1994
Air
EPA Methodologies for Estimating Emission
and Travel Activity Effects of TCMs
-------
(This page intentionally blank for 2-sided printing)
-------
SYSAPP94-92/096
Final Report
METHODOLOGIES FOR ESTIMATING EMISSION
AND TRAVEL ACTIVITY EFFECTS OF TCMS
Systems Applications International
July 1994
-------
(This page intentionally blank for 2-sided printing)
-------
Final Report
METHODOLOGIES FOR ESTIMATING
EMISSION AND TRAVEL ACTIVITY
EFFECTS OF TCMS
July 1994
SYSAPP94-92/096
Prepared for
Ms. Valerie Broadwell
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
and
Mr. Mark Simons
U.S. Environmental Protection Agency
Office of Mobile Sources
Prepared by
B. S. Austin
J. G. Heiken
S. B. Shepard
L. L. Duvall
Systems Applications International
101 Lucas Valley Road
San Rafael, CA 94903
415/507-7100
1521092093
-------
(This page intentionally blank for 2-sided printing)
-------
NOTICE
The contents of this document reflect the view of the
authors, who are responsible for the facts and accuracy of
the information and data presented herein. The United States
Government does not endorse products or manufacturers
mentioned herein. Trademarks or names of specific products
or manufacturers appear only because they are considered
essential to the objectives of this document.
The contents do not necessarily reflect the official policy
of the U.S. Environmental Protection Agency, This document
does not constitute a standard, specification, or regulation.
It does not replace or supersede official guidance of the
United States Government, nor does use of its contents
relieve any party of its obligations or responsibilities to
meet any governmental requirement.
-------
(This page intentionally blank for 2-sided printing)
-------
Contents
List of Tables m
Lisf of Figures 'v
1 INTRODUCTION 1-1
Puipose of this Document . 1-1
This Document in Relation to Other TCM Evaluation
Methodologies 1-2
Methodologies Provided in This Document 1-2
2 ESTIMATING TRAVEL ACTIVITY EFFECTS FROM INDIVIDUAL
TCMS . 2-1
Overview of Individual Measure Methodologies 2-1
Data Requirements for Applying Methodologies 2-5
Step 1: Identify the Potential Direct Trip Effect and
the Trip Type Affected 2-13
Step 2: Calculate the Direct Trip Reductions 2-17
Step 3: Calculate the Indirect Trip Increases 2-22
Step 4: Determine Direct Peak/Off-Peak Period Trip Shifts 2-26
Step 5: Calculate the Total Trip Changes 2-33
Step 6; Calculate the VMT Changes Due to Trip Changes 2-34
Step 7; Calculate the VMT Changes Due to Trip Length Changes 2-35
Step 8: Determine the Total VMT Changes . . 2-36
Step 9: Calculate Speed Changes 2-36
3 METHODOLOGY FOR CALCULATING EMISSION CHANGES FROM TCM
ACTIVITY EFFECTS 3-1
Overview 3-1
Step 1: Emission Analysis of Trip Changes 3-6
Step 2: Emission Analysis of VMT Changes 3-18
Step 3: Emission Analysis of Fleet Speed Changes 3-25
Step 4: Total Emission Changes Due to TCM Implementation 3-28
4 TCM INTERACTIONS AND MODE CHOICE DEPENDENCE ON MULTIPLE
ATTRIBUTES 4-1
Introduction 4-1
Overview of the Packaging Methodology 4-3
Mode Choice Considerations 4-4
Step-By-Step Approach to Conducting a TCM Package Analysis ...... 4-6
Example Application of Packaging Methodology 4-15
92093.01 i
-------
References R-i
Appendix A: Summary of Recent TCM Methodologies Developed
in California
Appendk B; Methodology to Evaluate Peak Period Trip Shifts of Flextime
and Compressed Work Participants
92093.01 11
-------
Tables
1-1 TCMs listed in the Clean Air Act §108(f) 1-4
2-1 Data used in TCM methodologies 2-6
2-2 Selected sources of general travel data 2-10
2-3 Methods for computing elasticities 2-12
2-4 Equations for identifying potential trips per day affected 2-15
2-5 Fraction of direct tip effects assumed to be work related (a) by TCM , , , 2-18
2-6 TCM adjustment factor, a, defined in Equation (2-3) for each TCM . . . 2-19
2-7 Indirect trip effects 2-24
2-8 Fraction of trips removed from peak period by peak period length and
increase in travel period of flextime participants 2-29
2-9 Fraction of trips removed from peak period by original peak period
length and increase in peak period of compressed work week
participants 2-32
2-10 Trip length changes 2-37
92093.cn
-------
(This page intentionally blank for 2-sided printing)
-------
Figures
2-1 TCM analysis overview 2-3
4-1 Attributes of travel choices 4-7
4-2 Sample hypothetical attribute profile for two measures, rideshare
and transit 4-11
4-3 Value function for each attribute for rideshare and transit 4-12
4-4 San Francisco base case input file 4-17
4-5 Base case simulation illustration of calibration technique and selection
of reasonable weight profiles for the city of San Francisco 4-18
4-6 Control scenario input file for the city of San Francisco 4-21
4-7 Control simulation 1.0% increase in transit fares in San Francisco 4-22
4-8 Demand elasticities with respect to travel cost 4-23
4-9 Base case input file for the Maricopa County base case simulation 4-25
4-10 Base case simulation for Maricopa County metropolitan area 4-26
4-11 Control scenario input file for the Maricopa County metropolitan area . . 4-29
4-12 Control scenario 1,0% transit fare increase in the Maricopa County
metropolitan area 4-30
92093,01
-------
(This page intentionally blank for 2-sided printing)
-------
1 INTRODUCTION
The 1990 Clean Air Act Amendments (CAAA) created a range of new, more stringent
transportation control requirements. Major federal agencies such as EPA and the
Department of Transportation must work together to ensure that transportation projects
further attainment of air quality goals (conformity); the private sector must market a new
slate of alternative, less polluting fuels; states must take action in the more serious
nonattainment areas to offset any emissions growth related to increased vehicle miles
travelled (VMT); and many state and local government agencies must implement
transportation control measures (TCMs) that modify driving behavior and limit emissions
resulting from traffic congestion.
To help understand and meet the new Clean Air Act's requirements, Congress instructed
the EPA to publish a number of guidance documents related to transportation control.
This document is one of the many EPA-sponsored publications which state and local
governments may find useful as they work to achieve their transportation planning goals.
The document provides a step-by-step approach for quantitatively estimating the travel
and emissions changes that are possible from implementing a number of TCMs suggested
in the CAAA.
PURPOSE OF THIS DOCUMENT
This workbook was developed as a tool for applying the Clean Air Act Amendments'
(CAAAs) TCM provisions. Title I of the Amendments (provisions for attainment and
maintenance of national ambient air quality standards) states that within one year from the
enactment of the 1990 amendments, EPA must publish information regarding the
formulation of and emission reduction potential of the TCMs listed in §108(f) of the
CAA. This document fulfills part of this requirement. Table 1-1 lists the 16 broad TCM
categories included in the Clean Air Act Amendments.
Air quality and transportation agencies will have a more focused interest in the §108(f)
measures depending upon an area's nonattainment status, the extent to which TCMs may
will be relied upon for emission reductions, and the degree of existing implementation.
92W3.10 \,\
-------
THIS DOCUMENT IN RELATION TO OTHER
TCM EVALUATION METHODOLOGIES
A number of methodologies for calculating the effects of TCMs on travel activity and
emissions are available. Ranging from traditional transportation modeling approaches to
sketch planning approaches developed in the late 1970s as well as several more recent
methodologies developed in California, many currently available techniques have been
criticized as being too complex, too optimistic, or not sufficiently linked to appropriate
emission categories to be satisfactory for use in air quality planning applications. The
methodologies presented here address a number of these criticisms and attempt to provide
more useful approaches for estimating the effects of TCMs on travel activity and
emissions. The methodologies build upon past efforts but include a number of innovative
techniques designed to produce more reliable estimates of TCM effectiveness. It must be
stressed that these methodologies are sketch planning techniques and will calculate
approximate effects. They generally utilize region-wide estimates of existing travel
characteristics and calculate region-scale effects. If corridor, facility, or traffic analysis
zone level data is available, it can be used to obtain more precise estimates, particularly
with respect to speed changes.
Transportation modeling approaches may provide more detailed and accurate estimates.
The methodologies presented here may be most applicable in two general circumstances:
(1) regions which do not have transportation modeling tools calibrated for their area, and
(2) regions which desire approximate TCM results for the purpose of deciding whether
transportation modeling is indicated. In many cases, the effects of TCMs are expected to
be much smaller than the uncertainties in the transportation models themselves. In such
cases the use of transportation models may not be an appropriate use of scarce resources.
More recently, a number of TCM analysis methodologies have been developed in
California. A summary of several of these is provided in Appendix A. All require the
use of software and a number do not address factors which may offset TCM benefits.
Others provide enhanced analytical techniques but require a transportation modeling
environment and are not adequately documented and/or publically available for
widespread use. The methodologies presented in this document can be applied using a
hand calculator if desired. They also provide specific equations for calculating effects
that may partially offset TCM benefits.
METHODOLOGIES PROVIDED IN THIS DOCUMENT
This document includes three main chapters. Chapter 2 presents quantitative
methodologies for individual TCMs. Equations for calculating trip reductions, vehicle
miles travelled (VMT) reductions, and speed increases are provided for seven example
92093.10 1-2
-------
TCMs. Equations are provided for calculating both direct and indirect effects1. VMT
reductions resulting from trip reductions are calculated separately from trip length
reductions. A number of such procedures are provided in order to ensure that emission
calculations are accurate2. All TCM effects are quantified in a way that facilitates
emission calculations. The methodologies are also structured and presented in a manner
that is intended to encourage analysts to adapt them to TCMs and situations other than
those specifically presented here. It is important to note that the user will need to input
the number of TCM participants before beginning analysis of some of the measures.
Chapter 3 provides a discussion of the emission categories affected by TCMs and
provides a quantitative methodology for estimating the emission effects of TCMs. The
emission analysis methodology calculates total mass emission changes resulting from the
travel activity effects calculated in Chapter 2. If desired, emission reductions from TCM
travel effects calculated using procedures other than those presented here can also be
quantified using the techniques provided in chapter 3. The procedures focus on the use
of the EPA MOBILE emission model and are most directly appropriate for regional scale
analysis of hydrocarbons (HC), oxides of nitrogen (NOX), and carbon monoxide (CO)
unless detailed corridor or facility specific data are available for inputs. Some discussion
of calculations for paniculate matter and microscale carbon monoxide concentrations is
also provided.
Chapter 4 presents a methodology for analyzing TCM packages. If a telecommuting and
a compressed work week program are instituted at the same workplace and employees
can choose either one, how many employees will choose telecommuting? If a parking
price increase is implemented together with a ridematching service, how much more
ridesharing will result than if just a ridematching program were implemented? The
approach presented here provides a method for roughly approximating how TCMs will
interact with one another in such situations. The approach is based upon the principal
attributes of travel modes and the comparative values of the different modes for each
attribute. The only data required are current mode choices, travel costs by mode, and
travel times by mode. Chapter 4 presents two example applications of the methodology
for two urban areas with very different mode splits. The approach appears to perform
well in both cases. It should be stressed that the methodology is new and has not been
extensively tested. It is likely that further empirical data will lead to improvements in the
future. However, it may prove to be a powerful tool for evaluating other areas and TCM
Direct effects refer to the primary effect of a TCM. For example, telecommuting seeks to
reduce employee work trips. Indirect effects refer to secondary effects resulting from TCM
implementation. For example, when a telecommuter works from home, the telecommuter or a member
of their household may wish to use the vehicle; thus a potential secondary effect of telecommuting is to
increase trips by household members.
For example, if VMT changes due to trip length (i.e., ridesharers drive to a park and ride lot)
are summed together with VMT changes due to trips that are eliminated (i.e., ridesharers who are picked
up at home), it is difficult to calculate trip start emission changes separately from exhaust emissions.
92093.10 1-3
-------
combinations. At a minimum, it provides a framework within which to identify and
analyze the many possible TCM interactions and combinations.
Appendix A briefly reviews a number of other methodologies developed recently in
California. Appendix B provides detailed mathematical documentation on the
methodological techniques used to calculate how many trips are shifted from peak to off-
peak periods by TCMs such as flextime and compressed work weeks.
TABLE 1-1. TCMs listed in the Clean Air Act §108(f).
1. programs for improved public transit;
2. HOV and bus lanes (construction of and conversion of existing lanes to);
3. employer based transportation management plans, including incentives;
4. trip-reduction ordinances;
5. traffic flow improvement programs that reduce emissions;
6. parking facilities for multiple occupancy vehicle programs or transit service;
7. vehicle use restrictions in downtown or other high emission areas, especially during peak
use periods;
8. programs providing for all forms of high-occupancy and shared ride services;
9. programs limiting portions of roads or sections of metropolitan areas to non-motorized
vehicular use or pedestrian use (both temporal and spatial restrictions);
10. bicycle use incentives in both private and public areas;
11. idling restrictions;
12. cold-start emission restrictions (in accordance with Title II);
13. employer-sponsored programs to permit flexible work schedules;
14. programs and restrictions to promote non-single occupant automobile travel as part of the
transportation planning and development efforts of a locality (new shopping centers,
special events and other centers of vehicle activity included);
15. programs for new construction of and major reconstructions of paths, tracks, or areas
solely for the use by pedestrian or non-motorized means of transportation when
economically feasible and in the public interest; and
16. programs to encourage the voluntary removal from use and the marketplace of pre-1980
model year light duty vehicles and pre-1980 model light duty trucks.
92093.10 1-4
-------
2 ESTIMATING TRAVEL ACTIVITY EFFECTS FROM INDIVIDUAL TCMS
This chapter presents screening methodologies for calculating travel activity changes from
TCMs. These activity changes include trip, vehicle miles travelled (VMT), and speed
changes. The approach draws partly from methodologies developed for the California
Air Resources Board (CARB) (Austin, et al., 1991) and the California Department of
Transportation (CalTrans) (Sierra Research, 1991). The methodologies offer three
particularly important features: (1) they provide an approach for calculating effects that
may partly offset TCM benefits (i.e., increased driving as a result of a carpooler or
telecommuter leaving a vehicle at home), (2) they explicitly link TCM effects to motor
vehicle emissions categories included in the EPA MOBILE emission factor model, and
(3) recognizing that the vast number of TCMs and potential implementation strategies
makes it impossible to develop and present methodologies to cover every possible
situation, the methodologies are structured in a manner that will allow the analyst to
quickly adapt them to TCMs and situations other than those specifically presented here.
OVERVIEW OF INDIVIDUAL MEASURE METHODOLOGIES
Many nonattainment areas rely or plan to rely upon TCMs for achieving some portion of
emission reductions needed for attainment and maintenance of air quality standards. The
procedures used to estimate the effectiveness of TCMs in achieving such reductions must
yield realistic results that do not exaggerate the potential benefits of TCMs. Procedures
developed in the past typically do not provide techniques for considering offsetting
effects, and do not properly link travel changes to emissions. *
METHODOLOGY HIGHLJGHTS
VMtth hand cotculotofs - not a block box.
Offsetting effects generally not considered
include (1) TCM participants who do not
reduce trips (100 new ridesharers does not
result in 100 fewer trips as each carpool
has a driver, and since some ridesharers
drive to park and ride lots) (2) increased
driving by household members of
carpoolers, telecommuters, or other TCM
participants when the vehicle is left at
home, (3) increased driving by
telecommuters or compressed work week
employees on the days they do not
commute to work, and (4) increased travel due to reduced congestion. This chapter
presents methodologies which address such factors.
^ effects such as heiecsed use at
oatos loft at home by itoesheiws, tetecomfnutere
-------
Improper linkages of travel activity changes to emissions occur often. A common
approach has been to simply assume emission reductions are proportional to activity
changes. For example if TCMs are calculated to reduce VMT by 2 percent, then
emissions are assumed to be reduced by 2 percent. However, motor vehicle emissions
result from a number of different vehicle types (i.e., autos, vans, heavy trucks), and from
different vehicle operating modes (i.e., start emissions, exhaust emissions, and
evaporative emissions). The vehicle types and emission categories affected by a TCM
need to be considered. For example, a ridesharing program may reduce auto use,
increase van use slightly, but is not likely to have any effect on heavy duty truck travel.
A reduction in VMT will reduce exhaust emissions, but not necessarily start emissions.
Further, TCM implementation can significantly affect the timing and location of
emissions. For example, diurnal evaporative emissions while vehicles are not in use may
occur more often in outlying areas near residences or park and ride lots given widespread
implementation of ridesharing, telecommuting, and compressed work week programs.
Such changes may not affect the amount of emissions, but may be important in
nonattainment areas where emission locations affect pollutant concentrations. The
methodologies presented in this chapter produce results that are easily and explicitly
linked to emission categories and vehicle classes. Chapter three explains the emission
linkages in detail, and provides procedures for estimating emission effects after travel
activity effects have been quantified.
The methodologies presented here provide specific, quantitative screening techniques for
calculating net trip, VMT, and speed changes for the peak and off-peak periods of an
average day. The methodologies for calculating travel activity changes consist of ten
steps listed in Figure 2-1. The steps begin by identifying the maximum number of trips
that may be reduced by a TCM, refining this and related estimates, and then calculating
the resulting emission changes.
Each of the are explained in more detail in the following sections. The steps are
presented in a series of tables which accompany explanatory text. Both the tables and
text present general equations covering each step. The purpose of presenting the
methodologies in this manner is to highlight the patterns and various similarities between
methodologies for very different TCMs. A key goal in developing the tables and the
equations contained in them was to reduce the methodologies to a level where such
patterns are visible in order to facilitate the creation of additional methodologies for
TCMs not specifically covered here. It is not possible, or even desirable to write down
the equations for every possible TCM. However, once the analyst is familiar with the
techniques contained in this document, he or she will have many of the tools necessary
for developing additional methodologies for other TCMs.
For each general equation we include a TCM-specific term that will differ for each
individual TCM. For example, TCM-specific terms applicable to ridesharing may
include the average carpool size and the percent of carpoolers who use park and ride lots.
Specific terms for HOV lanes might include the time period of operation as an HOV
lane, and the number of trips along the route where the HOV lane is added. While such
terms may differ for individual TCMs, the way in which they are handled after defining
them is very similar. The tables provide equations for calculating tMs TCM-specific term
for each individual TCM included in this document. Also included in the tables are
92093.05 2-2
-------
TCM Analysis Overview
STEPS
Calculate emission
changes
Calculate speed changes
Determine total VMT changes
;Calculate VMT changes due to trip length
changes
Calculate VMT changes due to trip changes
Calculate total trip changes
Determine peak/off-peak period trip shifts
Calculate indirect trip increases
Calculate direct trip reductions
Identify potential direct trip effect and affected trip type
EXAMPI£S
Based on changes in trips (start
emissions), VMT (exhaust
emissions! and speeds
Speed increases dm to
decreased volumes
I Net VMT changes
Multiply difference betwteen
average work trip length and
satellite work stations by number
of telecommuters working at
satellite, stations
Multiply average work trip length
by number of work trips, reduced '•
Net effect of measures and
decreases
Portion of trip changes that
occur in ike peak versus the.
off-peak periods
Number of additional trips due
to availability of auto to
member of telecommuters
household
Number of trips eliminated
through telecommuting
Number of employees allowed
to teteeommute
HGU1E 2-1
2-3
cdrVtsaOOl
-------
example sources for information on the parameters required and example values for the
parameters. An example accompanies each step; most examples use real world data from
the San Francisco Bay Area. This data is used because it was recent and readily
accessible; its use does not signify that it is representative of other urban areas.
Individual TCMs Addressed in This Document
As discussed above, while it is impossible to present detailed methodologies for every
possible TCM, the methodologies in this document have been designed and structured in
a manner that facilitates the development of additional methodologies by the user. The
particular TCMs addressed in this workbook include:
1. Telecommuting
2. Flextime
3. Compressed Work Weeks
4. Ridesharing
5. Parking Management
6. Transit Improvements (one methodology for decreased fares and another for
increased service)
The following briefly summarizes each of these TCMs and their primary effects on travel
activity.
• Telecommuting: Telecommuting is an employer-sponsored change in work
location to either the home or a satellite center. The direct effects of
telecommuting are to reduce work trips for those who opt to work at home or
to reduce work-related VMT for those who opt to work at a satellite center.
• Flextime: Flextime is an employer-sponsored flexible work scheduling
program to reduce peak period travel. Schedules are designed to avoid travel
during the most congested times of day. The direct effects of this TCM are to
shift work-related trips and VMT from the peak period to the off-peak period
in order to reduce traffic congestion.
• Compressed work weeks: Compressed work weeks is another employer-
sponsored work scheduling program. Its design is to have employees work a
10 hour day in order to eliminate one work day a week or a 9 hour day in
order to eliminate one work day every two weeks. The direct effects are to
reduce work trips on the eliminated work days and to also shift the time of
work travel to the off-peak period on the working days (due to the increased
daily hours of work).
92093.05 2-4
-------
• Ridesharine: Ridesharing or carpooling to the work place has a direct effect
on the work-related trips. In this analysis, the distinction is made between
ridesharers who join an existing caipool, ridesharers who form new carpools,
and ridesharers who utilize park and ride lots. In each of these cases the trip
and VMT reduction analysis is unique.
• Parking management: Parking management refers to employer based parking
programs to reduce the use of single occupant vehicles. The direct effects
include the elimination of work trips due to an increase in transit and
ridesharing modes. There is also a percentage of the work force who can opt
to use fringe parking facilities.
• Transit Improvements: Transit improvements can encourage individuals to
ride buses instead of driving their own vehicles. Two types of transit
improvements are covered here: (1) a decrease in fares and (2) an increase in
service. Transit improvements can reduce both work and non-work trips as
well as VMT. Typically, transit use will result in fewer trip reductions than
programs such as ridesharing, as individuals will often need to drive a vehicle
to a transit station. The number of trips saved depends on the proximity of the
transit stations and stops to individual's homes.
These particular TCMs were chosen from among the many that are possible based in part
upon their frequency of application. In addition, a goal was to include specific
methodologies for a range of TCMs involving price, behavioral, and system changes.
This range is hoped to provide enough specific examples of the ways in which the
methodologies can be applied to provide the user with sufficient tools for applying or
extending the methodologies to additional TCMs.
DATA REQUIREMENTS FOR APPLYING METHODOLOGIES
The most time-consuming step in TCM
analysis is the collection of the data
needed to conduct the analysis. The
quality of the data used in the analysis
affects the results more than any other
factor. The availability of local data is
crucial for calculating reliable results. If
no transportation demand model has been
developed for the geographic area under
study, the data collection process may be
challenging, as the needed data may need
to be collected from multiple (and
possibly conflicting) sources. In this document, example data values are used throughout
the text. It must be stressed that these are presented for illustrative purposes only. These
example values should not be used in other geographic areas.
Results ore only as good as 1he hput data;
use of locally representative data whenever
possible b crucial
ff Time spent comparing Input values reported
c by various soirees can be very Instructive
92093.05
2-5
-------
Applying the methodologies in this document will require data from a variety of sources.
Such data include travel data (i.e., VMT and trips for peak and off-peak periods), data on
existing TCM implementation (i.e., average number of people per caipool, average
distances to park and ride lots), census data (i.e., number of employed persons, average
number of people per household, or vehicle ownership). A list of these data
requirements and example data sources is presented in Table 2-1. Not all data is needed
for all TCMs; for example, the number of lane miles of new HOV capacity is not used in
the telecommuting methodology. In addition, some data not listed here may be necessary
for TCMs other than those specifically covered in this document. Finally, the user needs
to be cautious in using such data for an entire region. The detail provided in
transportation models, which can consider small scale changes in trip and socioeconomic
characteristics, is more appropriate (if the data used in calibrating the model is relatively
recent). When possible, such detailed estimates should be used to calculate baseline
(before additional TCM implementation) conditions from which to calculate TCM effects.
As the methodologies presented here will often be used in areas that do not have such
models, this data will need to be collected from other sources. Such approaches are
characteristic of sketch planning methods which have been in use for some time (for
example, approaches developed in 1979 by Cambridge Systematics [CSI, 1979]).
Data Used in Examples and Need For Region-Specific
Data When Applying the Methodologies
Example values and example applications of the methodologies are provided in many of
the tables and in all of the worked examples. A worked example is provided after each
analytical step is presented. These examples utilize travel activity data primarily from the
San Francisco Bay Area, since this information was relatively recent and readily
accessible to the authors. The use of these values is not meant to imply that they
represent a "standard" set of data that could be applied to any location. Individual
agencies applying these methodologies need to be sensitive to the wide range of observed
values that have been documented for various areas. For instance, the percent of home-
based work trips that are made using single occupant vehicles (SOVs) is reported to be
58.8 percent in the San Francisco Bay Area and 71.8 percent in Phoenix, Arizona (DOT,
1988). In addition, different sources of travel activity data may report different values
for the same variable; while the Department of Transportation reports that the percent of
work trips made using an SOV is 45.1 percent in New York City, the Bureau of the
Census reports this value to be 26.3 percent. To obtain the most accurate estimate of
TCM effectiveness for a specific region, travel data specific to that region must be used.
Values such as regional VMT, trips per person, and regional trips by mode are values
that will vary based upon characteristics of the region itself. These characteristics include
the availability of differing transit modes, previous levels of TCM implementation, land
use patterns and geographic characteristics of the area, and socio-economic
characteristics. TCM-related travel, VMT and mobile source emissions changes will vary
according to these characteristics.
92093.05 2-6
-------
TABLE 2-1. Dita used in TCM methodologies.
Data
Type
Example Sources
Travel
data
Single occupant vehicle work and non-work trips
per day
Shared vehicle work and non-work trips per day
(transit separately from carpools)
Percent of work and non-work trips occurring in
peak period of day
VMT by trip type (either for study region as a
whole or for subareas) in peak and off-peak periods
Average work trip distances (if data available for
subareas, particular demographic groups, or other
subgroups such as ridesharers this should be used)
Average non-work trip distances
Average speeds for peak and off-peak periods. It is
preferable to have these data for major roadway
types or subareas
Relative costs of different modes as well as cost
ranges (i.e. highest and lowest costs possible for a
mode)
Elasticity of mode choice with respect to cost
Elasticity of speed with respect to volume
Length of peak period (number of hours)
Average Vehicle Occupancy (should collect two
numbers: one total and one without transit)
(1) Generally the best source of data is from
local transportation planning agency data
and/or projections. In larger urban
areas, a designated Metropolitan
Planning Organization (MPO) is
responsible for collecting and updating
transportation-related data. In smaller
areas, planning districts or commissions,
government associations and the like may
collect data.
(2) Local ridesharing agencies and local
employers who have conducted surveys
on employee driving patterns. Early in
the data collection process, the need for
additional surveys should be evaluated.
(3) Regional FHWA office. The nine
FHWA regional offices are particularly
useful for obtaining relevant traffic count
data from the Highway Performance
Monitoring System, and for guidance on
proper interpretation of these data.
(4) National Personal Transportation Survey,
Conducted by the Department of
Transportation, the most recent survey
was conducted in 1990 and interviewed
almost 22,000 households. Tne survey
estimates VMT, trips, temporal travel
characteristics and many other
parameters. Because of the small sample
sizes, these data need to be used with
caution in specific geographic areas.
(5) Publications from the Institute of
Transportation Engineers such as the
Transportation and Traffic Engineering
Handbook, manual of trip generation
rates, and others.
(6) Jouraey-to-work data from the U.S.
census.
92093.06
2-7
-------
TABLE 2-1. Concluded.
Data
Type
Example Sources
TCMdata
Average number of people per carpool
Fraction of carpoolers who do not drive to park and
ride lots
Fraction of carpoolers who join existing carpools
Fraction of carpoolers who form new carpools
Average distance to park and ride lots
Frequency of ridesharing, telecommuting
Fraction of telecommuters who work from satellite
centers
Average distance to satellite centers
Local ridesharing organization statistics (see, for
example, RIDES, 1990, or Maltzmann, 1987), or
MPO. National data (i.e. census or NPTS) can
be used if nothing local is available
Same as above
Same as above
Same as above
Same as above or from literature
Either user specified (if specified as a
programmatic element) or from literature
Participating employers
Participating employers
Census
data
Number of individuals over 16
Number of employed persons
Total population in study region
Number of people per household
Percent of population of driving age that does not
own a vehicle
Census or local statistics (i.e., State finance
department or labor department)
Same as above
Same as above
Same as above
Same as above
92093.06
2-8
-------
The values used in the examples are frequently from a travel demand model (in this case,
the MTCFCAST model developed for the Metropolitan Transportation Commission in
California). This highlights the manner in which these screening methodologies arc used:
baseline travel activity data or output from base case transportation modeling is used as
the starting point from which to calculate TCM effects. The travel demand model and
methodology used to generate these values will obviously affect the values used to
calculate the effects of the various TCMs and the subsequent results.
Not all agencies will have region-specific information for all of the required variables. In
these cases, several sources of "standard" values are available that may be used as
substitutes if no better data are available. A number of these sources are listed in
Table 2-2.
Data Relating Changes in Price or Time to Changes in Travel Behavior
Some of the methodologies in this workbook employ elasticity measures in order to
predict transportation demand responses to system changes, such as fare increases, tolls,
and travel time increases or decreases. The economics concept of "price elasticity" is the
informal ancestor of transportation elasticities. This concept, put simply, is "...the
percentage change in quantity of commodity or service demand in response to a 1 percent
change in price (DOT51981)." This means that a price elasticity of -0.3 indicates that for
a 1 percent increase in price of a good or service there is an 0.3 percent decrease in the
demand for that good or service. The negative sign indicates that there is an inverse
relationship between demand and price (as the price increases, demand decreases). For
example, a 1 percent increase in parking prices might result in a 0.3 percent decrease in
parking demand. Transportation elasticities are computed in three ways: point
elasticity, arc elasticity, and shrinkage factor methods. These three methods are
summarized in Table 2-3.
Users of elasticities should keep in mind that in order for elasticities to be applicable, the
change in the transportation system must be a relative one. That is, it must involve a
quantifiable percentage change in the system parameter involved. Put another way,
elasticities can be used to compute the change in transit system use as a result of a change
in the overall price of service, but they cannot be applied to predict response to a new
transit line. Elasticities are not meant to be used as precise predictive measures. They
are intended to serve as an indicator of the likely order of magnitude of response to a
change in the transportation system and are very useful in providing first-order, aggregate
response estimates (DOT, 1981), This is one reason that elasticities are used in these
methodologies, since they are intended to provide information on the relative effects of
TCMS. These estimates should then be confirmed through more extensive analysis
methods.
Elasticity values are available from a number of sources, including the Transportation and
Traffic Engineering Handbook (TTE, 1982) and Traveler Response to Transportation
System Changes. (DOT, 1981), or they may be calculated using the formulas given in
Table 2-3 using region-specific data.
92093.05 2-9
-------
TABLE 2-2. Selected sources of general travel data.
SOURCE
AUTHOR(S)
INFORMATION
OFFERED
National Personal Transportation Survey
(NPTS1
U.S. Department of Transportation, 1993 Office
of Highway Information Management (for
summary reports) Electronic files are available
from the Volpe National Transportation Systems
Center in Massachusetts,
Presents data from a 1990 survey of almost
22,000 households on total travel, determinants
of travel, person trips and miles of travel,
¥ehicle trips and vehicle miles of travel,
journey to work and work-related trips, ride
sharing and vehicle occupancy, and others.
Characteristics of Urban Transportation
Demand
U.S. Department of Transportation, 1988.
Presents data on a wide variety of statistics
relaated to urban travel, comprised almosi
exclusively of post-1970 data.
Citv and County Data Book
U.S. Department of Commerce, Bureau of the
Census, 1988.
Data from the 1980 U.S. Census regarding
mode of travel to work and number of workers.
I
>—1
o
Commuting in America - A National Report on
Commuting Patterns and Trends
A. Pisarski, Eno Foundation for Transportation,
Inc. 1987.
Describes patterns in commuting over Ihe past
thirty years, including changes that have
occurred that affect current transportation
policy. Many statistics related to commuting,
including number of workers, relationships
between urban development and commuting
behavior, and mode of travel to work.
Highway Capacity Manual. Special Report 209
Transportation Research Board, 1985.
Provides techniques for estimating highway
capacity and level of service. Includes
information on traffic characteristics and
performance and new procedures for capacity
analysis of freeways and rural roads. Discusses
pedestrian traffic flow and facilities and the
effect of bicycles in the traffic stream.
Travel Characteristics at Large-Scale Suburban
Activity Centers
JHK & Associates for the Transportation
Research Board, National Research Council,
1989,
Travel activity data (trip generation, travel
time, etc.) are summarized for 6 geographically
representative suburban sites.
92091,06
-------
TABLE 2-2, Concluded,
SOURCE
AUTHOR(S)
INFORMATION
OFFERED
Transportation and Traffic Engineering
Handbook. Second Edition
W,S. Homburger, Editor
Institute of Transportation Engineers (ITE),
1987.
Various general values related to transportation
(including elasticities, mode split, general
impacts), along with explanations of many
widely used concepts in traffic engineering.
1990 Census of Population - summary
publications and data sets
U.S. Bureau of the Census.
Release of summary information from the 1990
Census, including worker statistics and journey-
to-work, are being made available continuously.
The Census Bureau should be contacted via
telephone or modem to determine the
availability of updated information.
1980 Census of Population. Vol. 2. Journey to
Work: Characteristics of Workers in
Metropolitan Areas (PC80-2-6D)
U.S. Bureau of the Census, My 1984.
Similar to the City and County Data Book, this
volume focuses more specifically on work
travel, type of work, number of workers,
commute time, etc.
92093.06
-------
TALBE 2-3. Methods for computing elasticities.
METHOD
FORMULA
SUMMARY
Point Elasticity
Derived directly from the economist's definition of elasticity.
Lack of information on the functional relationship between P
and Q (the shape of the demand curve) precludes ihe
computation of point elasticites from empirical data (DOT,
1981).
£p = elasticity
P = price
Q = quantity demanded at price P
Arc Elasticity
to
Alogg_
AlogP logP2-logP1
« = elasticity
Qj, Q2 = demand before and after
Pi, ?2 = price or service before and after
This logarithmic formulation most nearly approximates point
elasticity and is frequently employed (DOT, 1981).
Shrinkage Factor
(Shrinkage Ratio)
JL
0i _
This form of elasticity is historically used in reporting
response to transportation system changes. There are certain
conceptual difficulties with this method. For instance,
consider a specific experimental irasportation price reduction
and accompanying travel volume increase. Assume that the
demand returns to its original level if the price is raised back
to its original state as a second experiment. Intuitively, the
elasticity in this hypothetical example should be the same for
both experiments (it is if arc elasticity is computed).
However, if ihe changes in price are moderately large, Ihe
corresponding shrinkage factors will be different. While this
method and the arc elasticity equation will yield very similar
results when changes are small, discrepancies arise and
values differ when changes are large (DOT, 1981),
92093.06
-------
It must be stressed that eksticities are very approximate. In a given region, they will
vary widely depending on the base costs, travel times, and mode shares.
The remainder of this chapter presents the methodologies. Readers should note that they
are presented in a step-by-step fashion and that each step covers all TCMs. If a
particular TCM is being evaluated, simply use the equations applicable to that particular
TCM.
STEP 1: IDENTIFY THE POTENTIAL DIRECT TRIP EFFECT
AND THE TRIP TYPE AFFECTED
Step 1 determines the potential number of vehicle trips1 affected (PT) by the TCM and
the distribution of effects by trip type (work and non-work).
This step identifies the total number of vehicle trips that might be reduced. Subsequent
steps are used to adjust this estimate to represent the actual reduction which may be
achieved. In this first step one also defines the fraction of trip changes which are work
related. While many TCMs such as ridesharing and telecommuting aim to reduce work
travel, others, such as HOV lanes, may affect both work and non-work travel. Still
others, such as school-based trip reduction programs (not specifically covered in this
document) may not affect work trips at all. Distinguishing between trip types is
important in the context of these methodologies because the trip type affected is used to
determine the allocation of trip changes between the peak and off-peak periods.
Potential Direct Trip Effects (PT)
Potential direct trip effects are the maximum number of vehicle trips per day affected.
Two possibilities for calculating potential direct trip effects (PT) arc presented: (1) the
user supplies the number of participants and the frequency of participation (Le.f the
number of ridesharers and the number of days per week the average carpooler shares a
ride), or (2) the user can use elasticities to determine PT. Eksticities express the
percentage change in a variable (i.e., number of transit users) given a change in cost and
can be used when a TCM directly influences travel costs (i.e., parking management or
transit fare decreases). Elasticities can be thought of as rough approximations for
calculations by mode choice models incorporated into traditional transportation demand
models. A third alternative sketch planning technique is to use 'utilities', which can
relate changes in the desirability of a travel mode (due to changes in cost or travel time)
to mode shifts. This method is discussed in detail in (CSI, 1979). It is not included here
because the method requires the use of mathematical coefficients derived from detailed
surveys and regression analysis of travel behavior. A variant of this approach is used in
the methodology for evaluating effects of TCM packages. The techniques described in
this document are for use when such detailed data is not available; however, their use
Unless otherwise noted, "trips" in this document refer to vehicle trips rather than person trips.
92093.05 2-13
-------
may be enhanced by such data when they exist, as they provide a way of addressing
factors not considered in earlier documents.
To directly specify the participation rate in units of people per day, the user needs to
identify the target number of participants. This approach can be used for TCMs such as
ridesharing or telecommuting, which target particular people. For work-related TCMs
the number of potential trips can be calculated as:
PT = AT * FID * 2 f2"1)
where N is the number of participants (people), F is the frequency of participation (days
per week), D is the average number of commute days in a week, and the factor of "2"
adjusts for trips to and from work. For example, if 500 new ridesharers were expected
to result from a new ridesharing program and carpooled an average of 3 days per week,
then the average number of new ridesharing trips per day would be 600.
To calculate the potential number of trip effects using elasticities, one can use the
following equation
PT = e * AV * P0 (2-2)
where e is the elasticity of the change in participation level (FT) with respect to a
changing variable (AV) such as cost or time (both e and AV should be expressed as a
percent change), and Po is the number of individuals experiencing the change in cost.
Elasticity approaches such as equation (2-2) may be used if the TCM primarily involves a
cost or travel time change. For example, suppose a transit agency is willing to
implement a SO percent reduction in transit fares on routes affecting 10,000 people. The
agency estimates the elasticity of ridership with respect to fare to be -0.2. The potential
number of new transit users may be estimated as 1000.
Table 2-4 provides an equation and a summary description of key parameters for
determining the potential trip effect for each TCM specifically covered in this document.
Fraction of work related travel («)
The fraction of work related travel of a TCM ,« , represents the fraction of direct trips
associated with TCMs which influence work trips. Thus to is a number between 0 and 1
where 1 indicates that only work travel is directly affected by a given TCM and 0
indicates no work travel is affected by a given TCM. Since only two types of travel are
addressed in these methodologies, the fraction of non-work travel changes equals one
minus u. For a TCM which influences work and non-work travel about equally (e.g.
transit increases), w is assumed to equal a study region's base work travel fraction.
Recommended values for u are listed in Table 2-5.
92093.05 2-14
-------
TABLE 2-4. Equations for identifying potential trips per day affected.
TCM
Method to determine potential trip effect
(FT)
Parameter Description
Telecommuting
Flextirm
Gompreised work week
Rideshare
Transit (decreased fares)
Transit (increased
service)
FT = N*F/D*2
FT = N*F/D*2
PT = N*F/D*2
FT = N*F/D*2
FT = E * MASS. * P0
User supplied
Parking management
PT = [NSPACE-ALTSPei * e * APRC
F = the number of telecommute
days per week,
F = the number of days per
week flextime is in operation.
F = the number of work days
eliminated per week.
F = the number of days per
week that are earpaaled-
« = percent change in ridership
given a percent change in fare or
travel time
AFAKE = percent change in
transit fare
MPOs or transit organizations
typically can provide estimates of
increased ridership
NSPACE = # of pMklng places
subject to prince increase.
ALTSPC = # of "spillover"
parking places available.
APRC = percent change in
parking price
AVO = Average Vehicle
Occupancy
92093-06
2-15
-------
TABLE 2-5. Fraction of direct trip effects assumed to be work related («) by TCM.
TCM
Telecommuting
Flextime
Compressed wort; week
Rideshare
Transit (decreased fares)1
Transit (increased service)
Parking Management
Recommended value of u
1
i
1
i
WORK
WORK
i
Comments
Only work trips influenced
Only work trips influenced
Only wortc trips influenced
Only work trips influenced
WORK = work trip fraction (work
trips/total trips)
work trip fraction
Only work trips influenced
WORKpj. ia the peak period work
trip fraction (peak period work trips/
total peak period work trips)2
1 This is for general transit fare decreases available to all transit users; for the case of
employer subsidized transit passes the value of « = 1 should be used.
92093.06
2-16
-------
EXAMPLE 2-1. PT (Potential Trip Effects)
RTOESHARING
FT = N* F/D*2
FT = 10,000 * 3/5 * 2
PT'-~ 12,000 (vehicle trips per day)
Discussion
10,000 new ridesharers (N) carpool three times per week (F). Therefore the
maximum trip reduction is equal to 12,000 vehicle tr^s per day. Subsequent steps in
this methodology show how this value issreduced toy faet^^ driving
to park and ride lots, increased use«of ithe veMcle by ri^
and considerations such ^^^ carpoolers
still drive). ' •' :-: • - •• ' • :••:••-•• •. • •'
STEP 2 - CALCULATE THE DIRECT TRIP REDUCTIONS
This step is used to calculate the direct trip reductions resulting from the potential trip
reductions calculated in step 1. (such as work trips reduced by telecommuters). Indirect
trip changes are secondary trip effects such as increased nonwork trips by telecommuters
on their days off and are estimated in Step 3. The following equations calculate the daily
average direct trip reduction for work and non-work trips:
AXRIPSD = a * PT (2-3)
AixiPSDiW = w * &napsD (2-4)
AIMP5DjJSW « (1 - w) * ATRIPSD (2-5)
where:
= Total trip reduction for work and non-work trips
a = TCM specific factor equal to the fraction of participants who make
a direct trip change (trip changes per participant)
PT = is the potential trips affected per day
ATRIPSDjW = Direct work trip reduction
ATRBPSDjNW = Direct non-work trip reduction
In Equation 2-3, a accounts for the fraction of potential trip reductions (PT) which may
actually be eliminated. This variable is used to offset the potential trip reductions by
considering issues that may reduce the effect of a TCM. For example, some ridesharers
drive to park and ride lots, and some may previously be transit users, a is specific to
each TCM and the procedure for calculating it is provided in Table 2-6 for each TCM
covered in this document.
52093.05 2-17
-------
It is important to account for reduced work trips separately from reduced non-work trips
(equations (2-4) and (2-5)). This becomes critical in later steps when speed effects are
calculated as a function of reduced VMT in the peak and off-peak periods (speeds are
characteristically slower in peak periods than in off-peak periods). Most work trip
reductions affect peak period travel while fewer non-work trips do so.
The equations listed in Table 2-6 address a number of issues not covered in other TCM
methodologies. The logic behind each is discussed for each TCM below.
Telecommuting
While most telecommuting programs allow employees to work from their homes, an
increasing number employ either satellite work centers, or aim to reduce work-to-work
trips by using teleconferencing for meetings. The equation in Table 2-6 adjusts the
potential trip reductions by the percent of telecommuters who will work from home rather
than from satellite work centers. In addition, it adjusts PT to account for the fact that
some telecommuters may be transit users or carpoolers. The equation assumes that
employees will choose telecommuting days that are not in conflict with their ridesharing
days (i.e., that telecommuting will not affect AVO by breaking up carpools). This is
reasonable as most individuals share rides only two or three times a week and generally
telecommute once a week. Needs for a vehicle such as running errands or making work-
to-work trips can be accommodated in a manner similar to before the telecommuting
program was instituted.
This equation does not directly address teleconferencing. It can be used to estimate the
impact of teleconferencing programs by estimating the number of work-to-work trips
saved by such a program (i.e., by calculating the average attendance and frequency of
meetings targeted) and substituting a work-to-work trip AVO (employees from the same
office frequently share rides to off-site meetings so that AVO for work-to-work trips is
higher than average) into the equation listed in table 2-6. The variable "SAT" would be
removed from the equation in this case.
Flextime
Flextime does not reduce trip making, but simply influences the time of day trips are
made. It has been argued that flextime has the potential to break up carpools (indicating
that flextime could potentially increase trips) but evidence on this issue has been mixed.
If the employees targeted for a flextime program also rideshare, or if a ridematching or
carpool incentive program is instituted at the same time, and there is concern that the two
programs may be antagonistic, the TCM packaging approach presented in chapter four
may be applied to roughly quantify such effects.
92093.05 2-18
-------
TABLE 2-6. TCM adjustment factor, a, defined in equation (2-3) for each TCM.
TCM
Telecommuting
Reuime
Camptenal work
week
lidesharing
Tttruh
Piflcini MiQigemcnl
gt
-(1-SAT)/AVO
0
-I/AVO
-[NOLD -UNEW • (NCAJt-l)\NCARlttVO]
-IMVO
-[THAN -f (NOLD'HD) +
[NEW"MS*((*eA8-l}«CAR] -
FRNG)
Variables
SAT = fraction of telccamrnutcrs who work in
satellite office.
AVO " Average Vehicle Occupancy (with Baiuit)
Not applicable for trips
AVO - Average Vehicle occupancy (with transit)
NOLD " friction of ridesteret* whs Join esMng
eifpools and don't drfvc to park ind ride lots
NCAR - Average number n( people per carpod
NEW ** Fraction of ridesharers who form new
nrpoots arid don't drive !o park ml fide
lots
AVO »* Average vehicle occupancy
THAN - Fraclion of PT who will use transit
RD - Fraction of PT who will rttatare
NOLD = Defined above
NCAR - Defined above
FRNO ** Friction of PT wta uril! me fringe
parking facUiu'ei
Same n above fa parking management
Description
Takes into account drivers who did not go to
satellite offices or who did not use SOV mode
ptter la lelecommuttag.
a U zero to show lhal no dkecl irtp reductions
occur
Takes into acctma that drivers who switch from a
non-SOV mode do not ratoee trip,
MOLD accounts for UK fact that each ridesharer
who joins an existing carpod saves a trip
NEW accounts fat (he fact (at for every new
carpool, one trip (by driver) stiil Ukcs place
tales into account mode ghlfl between different
non-SOV modes , (Some ri&sharers mny swteh to
feinsh
"PT" is equivalent to the number of people subject
10 fhe parking manigernefli program who wfll uje
itared nodes in reipsmse. Of these, some will
switch to transit and some to Timesharing. We
assume (hat the proportional switching to each will
be in the same proportion is the enisling mode
split
Although variables are Ihe same as for parking
management, ihe values of the variables could be
different (espec billy since fringe parking b not an
option for HOV tines).
Example Values
SAT = 2$ (WSEO. W89)
71.4% (MTC, 1990)
NOLD is approximately 355
(Pjdes, 1990), (UMTA, 1985}
(Mabmin, 1987)
NEW is approximately 62% (same
references as for NOLD)
1.126 (MTC, 1990)
TRAN - 37 A1 of stared rides
in San Francisco Bay Ares {MTC,
1990)
RD = 62.6 X of shared rides In
San Francisco (MTC. 1990)
FRNO <= 0.0* for (his eiample
K)
1As noted in the text, the example values shown should not be used in other study regions.
-------
Compressed Work Weeks
The potential number of trip reductions due to compressed work weeks is divided by
AVO to adjust for participants who are ridesharers.
Ridesharing
The ridesharing equation is somewhat more complex than the others. It accounts for the
driver in a new caipool and for the fact that some caipoolers drive to a central location
such as a park and ride lot (thus not eliminating a trip). Most methodologies have
assumed that each ridesharer reduces a trip. In effect, they assume that the variable
"NOLD" (the fraction of ridesharers who join existing carpools and who do not drive to a
park and ride lot or other central location) equals one. In reality, many ridesharing
programs encourage the formation of new carpools while also increasing participation in
existing pools.
The first part of the equation adjusts for the tact that each caipool has a driver; if a new
caipool contains three passengers, then only two trips will be saved. To account for this,
the equation multiplies the fraction of caipoolers forming new carpools by the percent of
passengers that will reduce trips (NCAR-1)/NCAR, or 75 percent in a caipool containing
four passengers). Next, the whole equation should be divided by the AVO to account for
existing carpoolers or transit users. If a ridesharing program is targeting only SOV
users, one should not divide by AVO.
Parking Management
The number of individuals who would shift from SOV to a shared travel mode was
roughly estimated using elasticities in the equation given for FT for a parking price
increase. If an alternate methodology is used (i.e., a more sophisticated mode choice
algorithm) the value should be reported in units of individuals shifting from SOV to
shared ride modes (or, ideally, by the number of individuals who will shift to transit and
the number that will shift to carpools).
The equation builds upon the same concepts given in the ridesharing equation. Of the
individuals who rideshare, some will join existing and some will form new carpools. All
transit users are assumed to reduce trips (alternatively, the percent of transit users who
drive to bus stops may be used to adjust the variable "THAN"). It is difficult to calculate
the variables "THAN" and "KD" without using sophisticated mode choice techniques that
may require considerable time and resources. However, for an approximation it is
probably reasonable to use the base distribution of mode splits between ridesharing and
transit (i.e., if 65 percent of shared rides are carpools and 35 percent are transit, then
"RD" is assumed to be 0.65, and "THAN" is assumed to be 0.35. The fraction of
individuals who can drive to fringe parking lots near work sites and share rides to work
should be subtracted from the result (if any data are available for this). The equation is
not divided by AVO in this case, because the parking management program assumed is
92093.05 2-20
-------
one that only institutes parking charges for SOV users. If a parking program institutes
charges for all vehicles, the equation needs to be divided by AVO. In addition, such a
program may make transit more attractive with respect to carpooling. The analyst may
wish to approximate the change by using the elasticity of transit use with respect to price.
The change in price would be the average parking charge for a carpooler, and the
increased fraction of transit use would be used to calculate a new "IRAN" and "RD" for
the individuals affected by the parking program.
Example Application of Equation in Table 2-6
An example application of the equations in Table 2-6 is provided below for ridesharing.
First, determine a:
EXAMPLE 2-2. Determining a (TCM Specific Factor to Adjust "FT")
Some ridesliaimfQ^e^ while
some;fbntt new*^
(XMitM a drawer. ;TJie:$averag$;snu^ :
^ instfe-paitieutefe ofstfae //
;:
'
92093.05 2-21
-------
then direct trip effects:
EXAMPLE 2-3. TRIPp (Direct Trip Reduction)
RIDESHARING
= -0;61 * 72,000
- -7,320
Discussion
In this example, 10,000 ridesharers reduce 7,320 trips per day (direct trips; indirect
trip effects calculated in later ::stq>S:::wiQ::::tend::to'ofEiset:thiS::aroount''SomewhatrmoiB). It
should be noted that the impact biGonsidering past mode share, whether ridesharers will
drive to park and ride lots, and whether they will join existing or fonn new carpools
reduces the estimated benefit by about 40 percent.
and finally the direct woik trip effects:
EXAMPLE 2-4. TRIPPW (Direct Work Trip Reduction)
• .-:•-.;•>.: • ..v.'^/•••^\ : .-1ODESHARING --
AJ1 direct strip:.; effects ;df iicideshaiing? ;are work :ielated.
STEP 3 - CALCULATE THE INDIRECT TRIP INCREASES
It is important to also consider effects that may offset TCM benefits. For example, when
TCM participants leave their vehicles at home, members of their households may use
them for either work or non-work trips. In addition, TCM participants may experience
an increase in their 'travel budget' by participating in a TCM program. For instance, a
telecommuter saves time and driving costs on telecommuting days. The telecommuter
may desire to 'spend' some of these decreased costs on extra non-work related travel.
92093.05 2-22
-------
Table 2-7 presents an initial list of equations that may be used to evaluate such offsetting
effects.
The equations in Table 2-7 provide methods for approximating trip increases due to
vehicles being left at home. The third equation roughly estimates trip increases due to
decreased roadway congestion.
The first equation can be interpreted as follows: The rate at which work trips may
increase because a vehicle is left at home depends on: (1) the fraction of the population
that does not own a vehicle (and therefore may wish to use one for work travel); (2) the
fraction of the population that shares a ride to work, (individuals who do not have access
to vehicles and who share rides to work may prefer to drive their own vehicle to work);
(3) the household size minus 1 (1 is subtracted to account for the fact that one of the
household members is the new ridesharer affected by the TCM); (4) the employment rate
(unemployed household members will not need to commute to work); and (5) the trip
generation rate for SOV drivers. The second equation can be interpreted in a similar
fashion except that it considers a different segment of the population (unemployed
household members of driving age). As noted in the example, the work trip increase
equation is conservative in that it assumes that all household members without a vehicle
who share a ride to work (implying that they work too far away to walk or bicycle) will
use the vehicle left at home.
If both equations are used, a small overestimation of trip increases may result since
double counting of vehicle use may occur. The likelihood that an unemployed individual
over 16 who does not own a vehicle lives in the same household as an employed transit
or shared ride user without a vehicle could be used to reduce the potential for double
counting. The total trip increase can be adjusted by multiplying the sum of INC^^j and
by one minus this probability.
An example application for determining the value of INC^rg and the total trip effect is
provided in the box below.
92093.05 2-23
-------
TABLE 2-7. Indirect Trip Effects,
I
to
Equation
INCW H = NV * SHR * (SIZE - 1) * BMP
Explanations
This equation estimates the rate of
trip increases due to household
members of TCM participnnls who
leave their vehicles at home. The
total number of trip increases
resulting is calculated by
multiplying a factor similar to a for
direct trips (INCV u) by the direct
trip reduction calculated in Step 2,
Variable Definition^
= Rate of increased SOV work trip
making by household members of TCM
participants who leave their vehicle! at home
NV = Fraction of population that does not own a
vehicle (census dote)
SHR = Fraction of trips made via shared mode
(28.6 in Bay Area)
SIZE = Average household size (Approximately
2.56 in the San Fraaciaco Bay Area)
EM P = fraction of population that is employed
(and over 16) (About 53%)
Work trip generation rate for SOV users
(trips per day) (about 1.71 in the San Francisco
Bay Area)
Trip Effect
INCWH *TPJPD_w/2
This equation results in
the number of trip
increases. These are
divided by 2 because the
number of vehicles left
at home is equal to the
number of trips saved as
rideshareif! leave their
vehicles at home/2
(assuming that each
ridesharer makes two
work Iripi per day - one
from home to work and
one from work to home
(continued)
As noted in the text, example values should not be used in other study regions.
92093,06
-------
TABLE 2-7. (Concluded). Indirect Trip Effects,
Equation
Explanations
Variable Definition
Trip Effect
INCN»H =
* SHE * - 1) *
UNEMP * T%
This equation estimates tbe rate at
which non-work trips will increase
due to increased availability of
vehicles for non-work trips
previously made via transit or
shared ride.
jj || = Rate of increased non-work trip
making by unemployed household members of
TCM participants who leave their vehicles at
home.
UNEMP = fraction of population over 16 that is
unemployed.
INCNH *TRIPDpW/2
- C,)/C0 * SHE * TRPs
I
K3
Ui
Vehicle trips could increase if the
TCM or TCMta implemented in an
area reduce congestion sufficiently
enough to encourage individuals to
shift from non-SOV to SOV
model, Thii equation veiy roughly
approximates 'latent demand'
possibilities such as shared ride
commuter* who would prefer to
use SOVs
tim = Elasticity of mode choice wilh respect to
COlt.
Cfl = Pre-TCM cost of work travel (for this
equation the cost is equal to out-of-pocket costs
+ regional avenge hourly wage rale applied to
total tiwel time,)'
Cl - Post-TCM travel cost (this cost cannot be
calculated until the speed changes am calculated
in step f and then translated into a cost change
due to reduced travel time. In most TCM
calculations (Me effect wilt not be considered but
the equation i« included for completeness.
TIPs = Total trips per day affected by the speed
increase.
1 Studies have shown that in-vehicle travel time is not weighted as heavily as access time (e.g. waiting for the bus). If desired, the travel
time could be weighted to adjust for such factors
92093.06
-------
EXAMPLE 2-5 INC^ (Indirect Work Trip Effect)
RIDESHARING
INCWH = NV* SHR •* (SIZE - 1) * EMP *
~~~ ' = 0.13 x .107 * (2.56 -1) * 0.528 * 1,705
INCWH = .02
Table 2-7 includes a column labeled "trip effect" to show how the rate of increase is
used to calculate a number of trips :
ATZIPSIfW =
= -02 * (-<7,
= 73
Discussion
Approximately 2S (in this example) of the vehicles left at home niay be used for work
trips by household members who previously were ridesharing or using transit. Note
that ttm is ;a veiy iconservative e
vehicles who need to commute to work wiU use the ^
STEP 4 - DETERMINE DIRECT PEAK/OFF-PEAK PERIOD TRIP SHIFTS
Many TCMs, including peak period delivery restrictions, flextime, compressed work
weeks, and HOV lanes shift travel between peak and off-peak periods. The amount of
travel shifted depends on a number of factors, including the length of the peak period and
the number of hours individuals are willing (or allowed) to shift the time of their travel.
Such shifts are important because of the potential for congestion relief when travel is
spread more evenly throughout the day. In general, the higher speeds typical of lower
congestion can result in lower emission rates. Step four explains how to calculate the net
shifts in work and non-work trips to and from peak and off-peak periods. It should be
noted that this step is used only for TCMs which directly shift travel times (i.e.,
flextime). The allocation of trip reductions calculated in steps 1 - 3 for TCMs such as
ridesharing or telecommuting to the peak and off-peak period is accomplished in step 5.
The following two variables are determined in this step (4):
• ATRIPSfp = the change in peak period trips (the subscript S signifies the
category of trip shifts), and
• ATRIPSjOp = the change in off-peak period trips.
92093.05 2-26
-------
As an example, a compressed work week plan may eliminate work trips by allowing
employees to work four 10 hour days instead of five 8 hour days. Additionally, this
TCM also redistributes existing work travel between peak and off-peak periods due to the
change in travel times of the participants. The amount of travel that is redistributed
depends on the length of the peak period. If the peak period is very long there is a
smaller probability that a one or two hour change in the time one departs for work will
shift travel to the off-peak periods.
The procedures for flextime and compressed work weeks are presented below.
Flextime
Flextime allows for a broader period of travel to and from work resulting in a shift of
work trips from the peak period to the off-peak period. Of the total potential trips
affected by a flextime program, only some will actually shift from the peak to the off-
peak period. Conceptually, if employees are supposed to be at work by 8:30 a.m. and a
full flextime program is instituted wherein employees can arrive and leave from work at
any time, as long as they put in a full eight hour day, only some will actually shift their
travel out of the peak period. If the peak period runs from 7:00 - 10:00 a,m,, these
employees would have to shift their travel time by close to two or more hours in order to
travel outside the peak period. While some employees may be willing to do so, many
may not be. If the peak period is shorter, then obviously more employees will shift their
travel to outside the peak period. In the many urban areas which experience long peak
periods, the impact of TCMs such as flextime is not likely to be significant.
The effect of peak period length and the fraction of individuals who will actually change
their pre-flextime travel patterns so that they shift out of the peak period can be evaluated
for both the a.m. and the p.m. period using the following equations:
PT PT
— - &fLEX,PM* -y P"7*
where ATRIPg^p is the change in peak period trips (the two terms of this equation are to
distinguish between AM and PM peak periods), FT is the potential trips identified in Step
1 of this chapter (the factor of 2 is used to divide PT equally between the AM and PM
peak periods), and S^LEX is the fraction of the flextime potential trips which will shift
from the peak period to the off-peak period, A table of values for 5 is provided below,
along with a discussion of how these values are derived. The peak period subscript on £
is necessary since its value can differ between AM and PM peak periods. The negative
signs in Equation 2-7 indicate a decrease in peak period trips due to flextime
implementation.
92093.05 2-27
-------
After ATRIPS j» is determined, the change in off-peak trips ATRJPSjOP can then be
obtained from Equation 2-8:
ATKIPSSOP = -^TRIPSSP (2-8)
In other words the decrease in peak period trips equals the increase in off-peak period
trips.
Intuitively, the fraction of potential trips removed from the peak period, %LEX> w^ vajy
according to the length of the peak period. For regions with longer peak periods,
flextime work scheduling would be expected to have less of an effect when compared to a
region of relatively short observed peak period. Moreover, the wider range of flextime
travel period (i.e., the period of hours allowed for flextime travel) the higher the
probability that more trips would be removed from the peak period.
*s approximated here by assuming (1) participants are equally as likely to travel
earlier or later (than before flextime implementation); (2) assuming a normal distribution
(Gaussian distribution) of work trips; and (3) establishing the average increase in travel
period for flextime participants. The average time increase of the flextime participant,
can be established from employee surveys. If this is not possible, other estimations can
be used. Table 2-8 lists a range of possible values for $FLEX for a range of peak period
lengths and average time period increase of flextime participants. Hie peak period length
is based on data. The average time period increase is either 'guessed* at, or derived
from employee survey data. For example, one might assume that employees will travel
either a half hour earlier or a half hour later (equalling a total increase of one hour as in
row one of Table 2-8). Alternatively, one may reason that employees would be willing
to travel up to an hour earlier or later (the fraction of travel shifted out of the peak is
listed in row two of table 2-8). A detailed explanation of how such assumptions are
translated into values is provided in Appendix B. Appendix B can also be used as a
guideline for developing values of JFLEX otner tnan ^e examples provided in Table 2-8.
Example 2-6 provides an application of the methodology to estimate trip shifts due to
flextime implementation.
92093.05 2-28
-------
TABLE 2-8. Fraction of trips removed
in travel period of flextime participants.
from peak period by peak period length and increase
Average travel
period increase for
flextime
participants
1 hour
2 hours
3 hours
Peak period length.
2 hours
.139
.475
.812
2.5 hours
.094
.323
.627
3 hours
.060
.233
.475
3.5 hours
.046
.139
.287
4 hours
.094
.233
.365%
92093.06
2-29
-------
(This page intentionally blank for 2-sided printing)
-------
EXAMPLE 2-6. Trip Shifts of Ftartime Participants
Evaluate a flextime program of 10,000 participants.
(1) From Step 1, it is evaluated that number of potential trips, FT = 20,000. This
value of FT assumes F/D = 1 where F is the number of days per week the flextime
program is in effect and D is the number of commute days per week. F/D = 1
indicates that the flextime program is in operation every commute day,
(2) Using Equation 1-1 and Table 2-^6 to evaluate ^EX' assume a AM peak period of
2 hours, a PM of 2.5 hours, and an average increase in flextime travel period of 1 hour
for both the AM and PM peak periods,
>
ATWP = -ae9 * WmG12 - &1Q94 -* 20&00/2
= -2,330 (nips per day)
Discussion
application shows a net sh^ The
resulting shift of iithe off-peak period (from E^ 2-8) would then be 4-2,330
Compressed Work Weeks
Compressed work week scheduling as generally implemented adds one or two working
hours to every four days in order to eliminate one or two days every two weeks from the
work schedule. This daily extension of working hours from a compressed work week
schedule results in a percentage of participants experiencing trips shifts outside the peak
period as commuters travel earlier to work and return home later.
As for flextime, only a fraction of the total participants will shift out of the peak period.
For compressed work week participants, this occurs only on the days which travel has
been extended. The number of trip shifts of compressed work week participants can be
determined from the following equation:
« „ F SHIFT n.Q\
- 8cww,PM*N* D l/ yj
where ATRIPSyp is the change in peak period trips, N is the number of participants
(identified in Equation 2-2 of Step 1), Fggjpj, is the number of days per week the
participant experiences extended hours (this value is discussed below), D = the number
of days per week of commuting (identified in Equation 2-2), and 5CWW is the fraction of
compressed work week participants removed from the peak period (this value is discussed
below) for each peak period.
92093.05 2-30
-------
The negative sign in Equation 2-9 indicates a decrease in peak period trips due to
compressed work week implementation. In general, Pg^^ equals four (i.e.,
compressed work week scheduling adds one or two working hours to every four days).
The total decrease in peak period trips is equal to the increase in off-peak period trips.
Thus after ATRIPSjp is determined from Equation 2-9, the increase in off-peak trips
ATRIPSjQp can be determined from Equation 2-8.
Equations 2-8 and 2-9 can be used to determine the trip shifts due to compressed work
week implementation once dfjww, the fraction of participants removed from the peak
period, has been identified. Intuitively, 6CWW varies by the length of the peak period
and by the increase in the travel period.
An approximation of 5CWW can be made by (1) assuming the increase in travel period is
equally distributed between the AM peak period and the PM peak period; (2) assuming a
normal distribution (Gaussian distribution) of work trips; and (3) establishing the average
time shift of the program. The time shift of the program depends on whether the
compressed work week schedule extends working hours 1 or 2 hours every 4 days. If the
program extends working hours one hour every four days the anticipated shift would be
1/2 hour earlier in the AM peak period and 1/2 hour later in the PM peak period. If the
program extends working hours 2 hours every four days, then the corresponding shifts
would be 1 hour in the AM peak period and 1 hour in the PM peak period. Table 2-9
incorporates these assumptions into tabulated values 6Cww for a range °f peak period
lengths and average time shifts of compressed work week participants. A detailed
explanation of the methodology used to determine the tabulated values of 5CWW is given
in Appendix B.
Example 2-7 is an application of the estimation of trip shifts due to compressed work
week implementation.
92093.05 2-31
-------
TABLE 2-9. Fraction of trips removed (6cwiv) from peak period by original peak period length and
increase in peak period of compressed work week participants, A nine-hour expanded work day
corresponds to a 1/2 hour increase in travel period in the AM and the PM peak periods; a 10-hour
expanded work day corresponds to a 1 hour increase in each peak period.
Increase travel
period length (per
peak period) for
compressed work
week participants
1/2 hour
1 hour
Peak period length
2 hours
0.139
0,475
2.5 hours
0.094
0.323
3 hours
0.060
0.233
3.5 hours
0.056
0.175
4 hours
0.046
0.139
92093.06
2-32
-------
EXAMPLE 2-7. Trip Shifts of Compressed Work Week Participants
Evaluate a compressed work week program of 5,000 participants with 2-hour extended
woik days ..every four days to eliminate one work day per week.
(1) N is the number of participants (5,000), FsfflFT is the days per week of extended
hours (4), D is the number of days per week commuting (4, recall the 5th day is
eliminated), and 50$rw identified in Table 2-7 (assume an AM peak period length of 2
hours and a EM of 2.5 hours):
SHIFT®)
= -0. 475*5, QQO*(4/4) - 0.323*5 ,QQQ*(4/4)
= -3,990 trips per compressed work week day*)
' "":-: "•'••' '-:'": ' Discussion
This applications shows a net shift of 3,990 trips per day from the peak period to the
offspeak period.
* To calculate average weekday changes in trips, multiply 3 ;!990 by 4/5.
STEP 5 - CALCULATE THE TOTAL TRIP CHANGES
Total net trip changes are determined from the change in trips determined from Steps 2-
4. Four totals are distinguished:
• ANETRPWjF = total work peak trip changes
• ANETRPW[OP = total work off-peak changes
p = total non-work peak changes
p = total non-work off-peak changes
The total trip changes can be estimated as follows:
* RIP ++&miPS ) (2-10)
(2-11)
(2-12)
, op
92093.05 2-33
-------
where w is the fraction of total trips that are work trips determined in Step 1, PKW and
PK,^ are the fraction of TCM affected work (subscript W) and non-work trips (subscript
NW) trips which occur in the peak period, ATRIPSjP and ATRIPS>P are the trip shifts
determined in Step 4, ATRIPj w and ATWPifNW are the indirect work and non-work trip
effects determined in Step 3, and ATRIP,, w and ATRIPjj ^ are the direct work and non-
work trip effects determined in Step 2,
In equations (2-10) through (2-13), the values for PKW and PKpjw arc the observed fraction
of work and non-work trips during the peak period. The values of PKW and PKj^ are the
same as the fraction of work trips and non-work trips of the total trips for the modeling
region except for HOV lanes, flextime, and compressed work weeks (since these three TCMs
can change this fraction). Values of PKW and PKNW are region-specific and should be
obtained by the TCM modeler. Example values are 0.608 for PKW and 0.288 for PK^y
(in San Diego in 1986). For HOV lanes flextime, and compressed work weeks, PKW and
P&NW should be set to 1.0 as the direct and indirect trip effects of these TCMs occur only at
peak periods.
The following example illustrates these concepts:
EXAMPLE 2-8. Allocating Trip Changes Between Peak and Off-peak Periods
:.;-:. ,Y. --V ^ ^WMMlB&fyjf %f||
£ '•.•; ^Mj^?^^^&E^&-^'WS§^^^WK^^-':;'' -Wj^^S
:toM:iwp;5irip;;€eiui|tep^
:iiO^iiitrlps|jlw^5icll^
pas calcuktedgdu^feiise oJ^
^MisJM^.Jsjetjua^
;du^g|pel^|pencfeiiien;i^ ^offflie;
sve4^:sifia::T€!Mi|MDgiani;-:t^^
:ShiS5:C€pJHtJOn?JSJMlU^^^
STEP 6 - CALCULATE THE VMT CHANGES DUE TO TRIP CHANGES
As discussed above, VMT changes occur as a result of trip reductions and changes in trip
length. As discussed in detail in Chapter 3, it is important to distinguish between the two
kinds of VMT changes. Trip reductions affect vehicle start emissions and exhaust emissions,
while VMT trip length changes only affect exhaust and related emissions. These two types
of VMT changes are calculated in Steps 6 (trip reduction) and Step 7 (trip length changes).
92093.05 2-34
-------
The net VMT reduction resulting from trip reductions determined in Step 5, (for peak and
off-peak periods) can be calculated as follows:
= (ANKTRPWP * DISTW) - (ANKTRPNP * DISTNW) (2-14)
&VMTOP = (ANETRPWOP * DISTW) - (ANETRPNOP * DISTNW) (2-15)
where DISTW and DISTN are the average VMT per trip for work and non-work trips (units
of miles per trip).
EXAMPLE 2-9. AVMTp (Change in Peak VMT due to Trip Reductions)
&VMT = (NETRPWPxDISTw) -
AVMT = ~{4>406x 27) - (0)
AVMTn = -1Z%
No example is given for the non-woik trip increase (see table 2-8 for equation) so this part
of the equation is not used in this example; The net wori^
the average woik?trip;distance to;calculate the VMT Auction. StJdstic^ have shown
work distances tend to be; longer for ridesharers on av^ra^, so if possible, data on trip
distances for the population participating in a TGM should be used. Here we have used the
average distance to woifc for ridesharers (27 miles) in me San Francisco Bay Area (T^
1990): • •••:: ; •'. ''-•-•• -*>\ '•' - -...'"" -;'-:Si { ; ' •''' ••:/ "•- ••'••/_ • . .- • •-.-;:-•
STEP 7 - CALCULATE THE VMT CHANGES DUE TO TRIP LENGTH CHANGES
An additional category of VMT changes includes trip length changes. If a telecommuter
works from a satellite work center, no trip has been eliminated, but the length of the work
trip may be substantially reduced. Some of the TCMs discussed in this document cause trip
length changes. Non-woik trip lengths are assumed not to change for the TCMs discussed
here. VMT changes due to trip length changes can be estimated as follows:
AVMTL w = jS * FT * -(DIST^-DIST^ (2-16)
where PT is the number of potential trips reduced (calculated in Step 1), 0 represents the
fraction of those participants who change their trip length (rather than eliminate a trip),
DISTW equals the average work trip length, and DISTnew equals the new worktrip length.
The new work trip length corresponds to variables such as the average distance to park and
92093.05 2-35
-------
ride lots, or to satellite work centers. Suggestions for calculating the variable & are
presented in Table 2-10.
For example, all ridesharers who drive to park and ride lots would be changing their trip
lengths rather than eliminating trips and the factor a represents the fraction of ridesharers
who do not drive to park and ride lots (and therefore eliminate trips rather than change then-
trip length). Similarly, for telecommuters SAT represents the number of telecommuters who
work from satellite work centers. The number of telecommuters who work from satellite
centers reduce their trip length. DISTW is the unadjusted average trip per (miles) for work
trips, and DISTnew is the new work trip distance (e.g., the distance to the park and ride lot
or to the satellite work center).
STEP 8 - DETERMINE THE TOTAL VMT CHANGES
Total VMT changes can be determined from the sum of the VMT changes determined in
Steps 6 and 7. This- is illustrated by the following equations:
ANETVMTp = AVMTTP+PKW*AVMTLW (2-17)
ANETVMTOP = &VMTTOP+(l-PKw) * AVMTLjW (2-18)
where
AVMTLjW = the net change in VMT due to trip length changes (Step 7),
AVMTT[p = the net change in peak period VMT due to trip changes (Step 6),
AVMTTjOp = the net change in off-peak period VMT due to trip changes (Step 6),
and
PKW = is the fraction of work VMT that occurs in the peak period.
STEP 9 - CALCULATE SPEED CHANGES
The change in speeds associated with the VMT decreases can be calculated in several ways:
volume to capacity relationships, network models, or elasticities of speed with respect to
volume. The latter method, shown here, is approximate and it cannot be stressed enough
that the elasticities used here are examples only. If elasticities are used, every effort should
be made to ensure that they are representative of the study region and circumstances. In
particular, elasticities vary widely depending on base conditions (i.e., speeds, mode shares,
travel costs). Other considerations of the relationships of speed to traffic volume due to flow
and traffic density are described in an Appendix to methodologies developed for the
California Air Resources Board (Austin, et al., 1991) .
92093.05 2-36
-------
TABLE 2-10. Trip length changes.
TCM
Telecommuting
Flextirne
Compressed Woric
Weak
Rideaharc
Transit
Paiicing
Management
Value of ft
SAT
0
0
1-NOLD-NEW
DMVTRAN
TRANS*DWVTRAN
+ FRNG
+ RD*(1-NQLD-NEW)
Calculation/Explanation
SAT — fraction of people who drive to satellite work
stations.
Flextime does not change trip lengths.
Compressed work weeks do not change trip lengths.
Accounts for people who drive to park and ride lota from the
previously used variables of:
MOLD = fraction of rideaharers who join existing carpools
and don't drive to park and ride lots, and
NEW = fraction of ridesharere who form new carpools
and don't drive to paik and ride lots.
DRIVTRAN = fraction of people who drive to the public
transit station.
Accounts for people who use transit and drive to transit stop,
people who use fringe parking facilities, and people who use
ridesharing who drive to park and ride lots:
TRAN = fraction of participants who will use transit .
FRNG = fraction of participants who will use fringe
parking facilities.
RD = fraction of participants who will use ride sharing.
MOLD, NEW, D1TVTRAN: lame as defined above.
92093.06
2-37
-------
Change in peak speeds can be determined from the following:
where
€p = elasticity of peak speed with respect to volume,
TOTVMTj, = total VMT in peak period, and
ANETVMTp = the net change in peak VMT determined in Step 8 of this analysis.
The change in off-peak speeds is calculated by the same method:
ANETVMTOP _ _m
A5PDrtP = - — * €OF (2-20)
OF TOTVMTOP op
where
eop = elasticity of off-peak speed with respect to volume,
TOTVMTOP = total VMT in off-peak period, and
ANETVMT0P = the net change in off-peak VMT determined in Step 8.
EXAMPLE 2-10. ASFDp (Change in Peak Speeds)
=:
^
92093.05 2-38
-------
METHODOLOGY FOR CALCULATING EMISSION CHANGES
FROM TCM ACTIVITY EFFECTS
This chapter (1) reviews TCM effects on emission categories addressed in emission factor
models and (2) provides a methodology for estimating emission effects of TCMs. The
methodologies developed in this document quantify how TCMs affect travel behavior and
vehicle emissions. From an air quality perspective, TCM evaluations must quantify
mobile source emission reductions associated with TCM-induced changes in activity level
variables such as trips, VMT, and speeds. Quantifying these changes in travel behavior
is the most difficult challenge facing the TCM analyst and is addressed in Chapter 2 of
this document. Once changes in travel variables are appropriately quantified, a motor
vehicle emission factor model, such as MOBILE, can be used to quantify emission
changes.
OVERVIEW
This chapter presents a methodology for calculating emission benefits resulting from the
travel activity changes calculated in Chapter 2. The methodology may also be used in
conjunction with TCM travel effects generated in another manner. The emission analysis
methodology translates the travel activity level changes into total mass emissions through
the use of emission factors. Emission factors are expressed in the units of mass per
activity level, such as grams per trip or grams per mile, and are calculated by the EPA's
motor vehicle emission factor model, MOBILE1. Emission changes can be calculated
from the activity level changes by multiplying the activity level changes by the
appropriate emission factors. The activity level changes are calculated in a manner which
links them explicitly to emission categories by addressing the particular vehicle class and
activity types considered by MOBILE. In this document, examples of emission factors
are provided to illustrate the calculations necessary for the emission analysis of TCMs,
Users are reminded that these emission factors are for illustrative purposes only, and any
analysis of TCMs will require the use of region-specific emissions factors derived from
the most recent MOBILE model. Note that the examples used in this document have not
been updated to reflect new MOBILE releases.
1 The MOBILE model, referred to singularly in this document, is actually a series of
models continually being updated and reviled. Use of the model should be restricted to
the latest released version. For illustrative examples, this document used the third release
of MOBILE version 4.1 dated November 1991,
92093.07 3-1
-------
This chapter focuses on the use of MOBILE - an emission factor developed for states
other than California. The EMFAC model is the California equivalent to MOBILE
developed by the California Air Resources Board. In general, the methodology presented
in this chapter can also be used with EMFAC emission factors, although it is
recommended that the user take the time to ensure the compatibility of the units used in
reporting emission factors units (EMFAC and MOBILE report some emission categories
in different units),
The emissions methodology focuses on HC, CO, and NOx, the three pollutants reported
by MOBILE. A separate model for PM-10 emissions developed by SAI for the EPA to
replace the 1985 EPA paniculate emission factor model is under review. PM-10
emission factors from this model can be used in conjunction with the methodology of this
document without much additional effort. PM-10 emission changes are simpler to
calculate than HC, CO, and NOx because PM-10 emission factors do not vary by speed2
or trip-type and are therefore, proportional to VMT changes.
The remainder of this overview discusses:
* The MOBILE emission factor model,
* A summary of key emission effects,
• Considerations for micro-scale modeling, and
* An overview of the emission analysis methodology,
and is followed by the detailed, step-by-step emissions analysis methodology.
The MOBILE Emission Factor Model
MOBILE produces motor vehicle emission factors for hydrocarbons (HC), carbon
monoxide (CO) and oxides of nitrogen (NOX) at conditions specified by the user. Inputs
include vehicle fleet information, vehicle fuel information, vehicle operating conditions,
temperature data, and vehicle inspection data. In some instances, the model has available
national average data for use as default values if no regional or local data are available.
The use of regional or local data is strongly recommended as all of the MOBILE input
parameters have a significant effect on the predicted emission factors, EPA has provided
guidance documentation outlining the recommended usage of MOBILE and the input data
it requires, Draft User's Guide to MOBILE 5a (EPA, 1993) and Procedures For Emission
Inventory Preparation, Volume IV: Mobile Sources (EPA, 1989). Note: these documents
are updated frequently. The latter document describes the EPA recommended input data
for MOBILE and is currently under revision (draft versions of the update of this report
can be obtained from regional EPA offices). Both of these documents should be
consulted prior to using the model.
r\ ^^^
* The current PM-10 model, developed in 1985, did not incorporate speed adjustment
factors into the emissions analysis; however, future updates of the PM-10 emission factor
model could incorporate vehicle speeds.
92093.07 3-2
-------
Emission factors are reported by vehicle class and emission category. Currently,
MOBILE defines nine vehicle classes:
* Light-duty gasoline vehicle (LDGV),
» Light-duty gasoline tracks less than 6000 Ibs GVW (LDGT1),
• Light-duty gasoline tracks more than 6000 Ibs GVW (LDGT2),
* Light-duty gasoline tracks, the total composite of LDGT1 and LDGT2,
(LDGT),
* Heavy-duty gasoline vehicles (HDGV),
« Heavy-duty diesel vehicles (HDDV),
• Light-duty diesel vehicles (LDDV),
« Light-duty diesel tracks (LDDT), and
» Motorcycles (MC),
It also reports a fleet average emission factor which is the composite of all vehicle
classes. Most TCMs affect trips and VMT of LDGVs and LDGTls.3 All vehicle
classes are affected by changes in speed.
In addition to vehicle classes, emission factors are reported for the following emission
categories:
Exhaust - Vehicle tailpipe HC, NO^, and CO emissions which occur during the
operation. Exhaust emissions are further categorized (according to the operating
condition of the vehicle) into start-up emissions (cold and hot) and warmed-up
stabilized emissions. These are commonly referred to as cold-start, hot-start and
hot-stabilized emissions, respectively.
Hot soak - HC emissions which consist of the evaporation of emissions from the
engine and fuel lines immediately following the end of a trip.
Diumals - Evaporative HC emissions resulting from temperature fluctuations
occurring when the vehicle is not in use. These are categorized into partial-day,
full-day and multiple-day diumals according to the period of vehicle non-
operation.
Crankcase - HC emissions from the vehicle crankcase during operation, significant
only for older model-year vehicles.
Running Loyses - HC evaporative emissions which occur during the operation of
the vehicle.
3 There are TCMs which specifically target heavy-duty vehicles, such as the peak
period restriction of heavy-duty vehicles in central business districts; however, these
TCMs are not addressed in this report.
92093.07 3-3
-------
Resting Losses - HC emissions resulting from permeation of Eon-metaJlc
evaporative emission control equipment occurring at all times (when a vehicle is
in-use and when it is not in-use).
Refueling - HC emissions resulting from vapor displacement from the vehicle
gasoline tank and from gasoline spillage during vehicle refueling.
Emission categories will be treated individually in the following emission analysis with
the exception of resting losses. Resting loss emissions occur 24 hours a day and would
not be affected by TCM implementation unless a TCM produced fewer vehicles in the
vehicle fleet. Although wide-spread and extensive TCM implementation can affect
vehicle ownership patterns, this is not addressed in the methodologies presented in this
document. For this reason resting losses will not be included in the emissions analysis of
this chapter.
Summary of Key Emission Effects
The following summarizes how each component of motor vehicle emissions may be
affected by TCMs,
Cold and Hot Start Emissions
Changes in cold and hot start emissions resulting from TCMs are proportional to changes
in trips. The number of TCM participants and the number of days per week they
participate are good indicators of changes in start emissions.
The average speeds driven in hot and cold start modes may also change as will the
relative proportions of trips taken in various operating modes. Conceptually this can
change emissions. For a given trip, the number of miles driven in cold start mode as
opposed to hot stabilized mode may change. The MOBILE model calculates emission
factors for a user-specified distribution of cold-start, hot-start, and hot-stabilized
emissions allowing explicit consideration of such changes.
Exhaust and Running Loss Emissions
Exhaust and running loss emissions would change due to a TCM's effects on VMT and
trip speeds. A frequent "back of the envelope" approach to estimating TCM emission
changes is to linearly link emissions with VMT. However, there are some serious flaws
in such an approach. Assume, for example, that a telecommuter reduced Ms or her total
work trip VMT from a 50-mile round-trip commute to a 5-mile round-trip commute to a
nearby satellite work center (i.e., a 90 percent VMT reduction). A rough emissions
reduction estimate that assumed emissions changes were proportional to VMT reductions
would fail to account for the fact that trip end emissions from cold starts and hot soaks
would continue to occur. Trip end emissions are a substantial fraction of the total
92093.07 3-4
-------
emissions associated with shorter trips; merely linking estimated emissions reductions to
VMT reductions would be a poor approximation of the resulting change. In addition, it
is also important to consider any shifts in the timing of VMT since exhaust and running
loss emissions can be temperature-sensitive.
Speed increases along affected roadways may also result in significant emissions benefits.
Exhaust HC and CO emissions drop sharply between speeds of zero to and about 50
mph. NOx emissions decrease until about 20 mph, after which they increase,
particularly after 50 mph. On the other hand, running loss evaporative emissions
decrease consistently with increasing speed.
Hot Soak. Diurnal.
Hot soak emissions will change in accordance with a change in vehicle trips, and
refueling emissions drop proportionately to decreased VMT. Diurnal emissions are more
difficult to analyze than hot soak or refueling emissions. There will be some shift in the
number of partial-day, full-day, and multiple-day diurnals related to when vehicles are
operated which is influenced by TCMs. Since nearly all vehicles experience some type
of diurnal cycle, the net emission change of shift within the diurnal categories can be less
significant than the other emission categories.
Considerations of TCM Effects on Microscale Modeling
The current EPA guidance for conducting intersection hotspot carbon monoxide (CO)
modeling is contained in the EPA document, Guideline for Modeling Carbon Monoxide
for Roadway Intersections (Schewe et al, 1990). This guidance contains information
regarding evaluation of air quality impacts at one or more roadway intersections where
vehicular traffic will cause or contribute to increased emissions of CO, It recommends
CAL3QHC as the intersection model of choice. CAL3QHC is a microcomputer-based
modeling methodology developed to predict the level of carbon monoxide (CO) or other
inert pollutant concentrations from motor vehicles traveling near roadway intersection.
Based on the assumption that vehicles at an intersection are either in motion or in an
idling state, the program is designed to predict air pollution levels by combining the
emissions from both moving and idling vehicles. CAL3QHC is a consolidation of the
CALINE-3 line source dispersion model and an algorithm that internally estimates the
length of the queues formed by idling vehicles at signalized intersections. Other models
available are CAIINE4, developed by the California Department of Transportation for
use in California, and the TEXIN2/MOBJULE4 model, often used in the state of Texas.
The TCM analysis methodology presented in this document is generally of a regional
scale and may not be suitable for microscale analysis unless traffic zone or corridor-
specific inputs are used instead of regional inputs. Regional values for several
parameters and vehicle characteristics used in this analysis may have significant local
variation.
92093.07 3-5
-------
Overview of Emissions Analysis Methodology
The emissions analysis methodology is categorized according into the changes in travel
activity levels (trips, VMT and speed). Emission categories influenced by trip changes
are: hot-start and cold-start exhaust, hot soak, and diurnal emissions; emission categories
affected by VMT changes are: hot-stabilized exhaust, running loss, crankcase, and
refueling emissions; and speed changes affect the categories of hot-stabilized exhaust and
running loss emissions. The detailed treatment of each emission category affected by
changes in vehicle activity is presented in the remainder of this chapter and is divided
into the following four steps:
(1) Emission analysis of trip changes resulting from TCM implementation,
(2) Emission analysis of VMT changes resulting from TCM implementation.
(3) Emission analysis of changes due to an overall fleet speed changes.
(4) The total emission change (sum of steps 1 through 3).
To use this methodology, one must calculate emission factors by running the MOBILE
program for various scenarios corresponding to conditions identified later in this chapter.
It is important to use MOBILE input values representative of the study region. If future-
year emission controls are not implemented correctly (i.e. no future-year emission
controls), emissions benefits calculated by the methodology presented here will be too
optimistic. Guidance on proper estimation of future-year emission factors can be obtained
from regional EPA offices.
STEP 1: Emission Analysis of Trip Changes
In this step the emission changes due to the change in trips am evaluated. The emission
categories influenced by this evaluation are hot-start and cold-start exhaust, hot soak, and
diurnal emissions. Of these categories, the hot-start and cold-start exhaust and hot soak
emissions are directly related to the number of trips while diurnal emissions are indirectly
influenced by the number of trips. Diurnals result from temperature fluctuations
occurring when the vehicle is not in use and can be affected according to the portion of
the day the vehicle is not used. The number of trips will also affect the diurnal
emissions, as the number of full or multi-day diumals may increase as trips are forgone.
Each emission category mentioned above is evaluated and discussed separately below,
Determine the Distribution of Trip Changes
Prior to evaluating the emissions, it is necessary to determine the distribution of trips
among the affected vehicle classes. Most TCMs analyzed in this document affect trips by
LDGVs and LDGTls - these include ridesharing, telecommuting, alternative work
3-6 Emissions Analysis of Trips Changes
-------
schedules and compressed work weeks. Other TCMs such as transit improvements
implemented as route additions may affect trips made by heavy-duty vehicles as well.
The equations provided account for changes in LDGVs and LDGTls, If necessary, the
adaptation of the equations to other vehicle classes should be obvious. The fraction of
trips associated with the LDGVs can be determined for any region from the total trips
which are due to the LDGVs (TRIPLpQv) and the total number of trips which are due to
the LDGTls
LDGT1
where TnopJDGV "SP16861118 tne fraction of trips which are LDGV, and since there are
only two vehicle classes being analyzed TTRIPLDGTI can ^ Determined from:
Equations 3-1 and 3-2 assume that a given TCM will influence both LDGV and LDGT
according to their trip representation in the vehicle fleet. It is recommended that region-
specific trip totals be used in these equations.
If trip data for Equations 3-1 and 3-2 are not available, an approximate value can be
obtained from the default vehicle VMT fraction data from the MOBILE model output.
The VMT vehicle fractions for LDGV and 1DGT1 can be substituted into Equation 3-1 in
place of the TRIP data. This approximation assumes that the vehicle trip distribution is
equivalent to the vehicle VMT distribution. Note that the MOBILE vehicle VMT
fractions are a function of calendar year. For example, in 1990 MOBUJE4. 1 reports a
LDGV VMT fraction of 0.626 (i.e., 62.6% of the total fleet VMT is from LDGVs) and
a LDGT1 VMT fraction of 0.171 (national average default values). The approximate
value of TTRTP Gm be obtained from substituting these values into Equations 3-1 and 3-2
to yield YTXBP^IJJGV = 0.785 and T'IISIPXDGTI = ®'215. These values would be
interpreted as follows: if a telecommuting program reduced 100 trips per day, 78 of these
would be LDGV and 22 would be LDGT1.
Hot-Start Trip Changes
First calculate the total number of trip changes due to TCM implementation as the sum of
the four trip-type totals determined in Step 5 of Chapter 2. The four trip types are:
» ANETRPW}P = total work peak trip changes
• ANETRPW op = total work off-peak changes
= total non-work peak changes
OP = total non-work off-peak changes
and the equation for total trip changes equals the sum of the four trip types listed above:
S1ZJ* 1 3-7 Emissions Analysis of Trips Changes
-------
ATRIPTOTAL = Amim^^p^ANETTiPy^op^AmTRP^^ANETRPf^op (3-3)
Second, it is necessary to calculate the total trips which began with the vehicle engine
cold (cold-start trip) and trips which began with the vehicle engine wann (hot-start trip).
The following equations determine the number of hot-start and cold-start trips changes
from the total trip changes:
ATRIPCST =
AIWPHST = (1 -TCSJ.W) * (ANEmpWtF+ANzniPWi0p} + (3.5)
where the subscripts CST and HST refer to "cold-start" and "hot-start" respectively and
7csr *s tne fraction of trips begun in the cold-start operating mode. This fraction
depends on the trip type.
In general, work trips involve mostly cold-start trips (i.e, y^ST ** ^)- ^or non*work
trips, the value of tcsrjw ^ Burned to correspond to the fraction of cold starts to total
starts for the study region. It is ideal to use local values for the fraction of starts which
are cold; however, these data are generally unavailable. In the absence of local data, the
MOBILE default fraction of cold starts can be used. This fraction is based on the
Federal Testing Procedure (FTP) driving cycle. The default fraction of cold starts is
0.43. It is suggested that a fraction of 1.0 cold starts be assumed for work trips and the
default fraction of 0.43 for non-work trips.
Determine Hot-Start and Cold-Start Emission Factors
Hot-start and cold-start emissions are the exhaust emissions which occur at the initiation
of a vehicle trip. Hot-start and cold-start emission factors need to be determined for each
of the three pollutants by running MOBILE for 3 scenarios (100% cold start, 100% hot
start and 100% hot stabilized). The results of which are substituted into Equations 3-6
and 3-7 (identified below) to calculate separate hot and cold start emission factors. This
produces six trip-start emission factors:
Exhaust hydrocarbon, hot-start mode
Exhaust hydrocarbon, cold-start mode
Exhaust carbon monoxide, hot-start mode
Exhaust carbon monoxide, cold-start mode
Exhaust oxides of nitrogen, hot-start mode
Exhaust oxides of nitrogen, cold-start mode
STEP 1 3_g Emissions Analysis of fHps Changes
-------
It is necessary to calculate the gram per trip emission factors for each of the categories
listed above. The MOBILE model does not explicitly calculate start-up emission in
grams per trip, but rather a gram per mile exhaust emission rate combining the start-up
and hot-stabilized portions of the emissions. Hot-start and cold-start emission factors in
grams per trip can be determined from the following equations using the MOBILE model:
CST = (EXHiQQ%cST,26MPH ~ EXHIQO%SJE,26MPH) * 3. 59 '3~6-'
HST = (EXHiQQ%fiST,26MPH ~ E%HlQQ%sm,26MPH) * ^-59 (3-7)
where CST and HST are the cold and hot-start emission factors in grams per trip (which
need to be determined for all three pollutants and both vehicle classes), EXH is the
MOBILE emission factor in grams per mile, and 3.59 is the FTP driving cycle trip-start
miles per trip, and 26 mph is the speed at which the start portion of the FTP cycle is
driven. The subscripts of 100% CST, 26MPH, 100% HST, 26MPH, and 100% STB,
26MPH of EXH indicate the operating conditions and the speed at which EXH is
evaluated by MOBILE. 100% CST, 26MPH indicates 100% cold-start operating mode
at 26 mph vehicle speed; 100% HST, 26MPH indicates 100% hot-start operating mode
at 26 mph vehicle speed; and 100% STB, 26MPH indicates 100% hot-stabilized
operating mode at 26 mph vehicle speed,
Equations 3-6 and 3-7 assume the trip-start driving conditions are uniform and
comparable to the trip-start driving conditions of the FTP driving cycle. As noted above,
the 26 mph and 3.59 miles per trip start represent the average speed and the length
respectively of the trip-start portion of the FIT'. These values should always be used in
Equations 3-6 and 3-7. Example 3-1, illustrating the calculation of trip-start emission
factors, is presented at the end of Step 1 of this chapter.
Petjennine the Hot-Start ancLCold-Start Emission Changes
Once the start emission factors are calculated emission changes due to trip reductions are
determined by multiplying the trip changes by the start emission factors for each of the
exhaust pollutants (HC, CO, NO^) and vehicle classes:
CST
(3_g)
(3,9)
STEP1 3.9 Emissions Analysts of Trips Changes
-------
ACOCST = (ATMPScsr * ymPXDGV * CSTLOGV co) +
(ATRIPScsT*ympJLDGT1
&COHST -
(&TRIPSHST* JTRIP^LDGTI * HSTLDGTLCO)
ANOXCST = (A.TMPSCST* JjWPJDGV* CSTLDGVJ*Ox) + (3-12)
ANOXHST = (&TRIPSHsT*'YmP,LDGV*HSTLDGVtNQx} + (3-13)
In Equations 3-8 through 3-13, the variables KST and CST are the hot-start and cold-
start emission factors (grams per trip) for the subscripted vehicle class and poEutant,
ATRIPCST and ATRIP,^ are defined in Equations 3-4 and 3-5, and THOPJLDGV anc^
TTRIPJLDGTI were defined in Equations 3-1 and 3-2. The emission factors tor all three
pollutants are determined from Equations 3-6 and 3-7 using region-specific MOBILE
emission factors. Example 3-2 (provided at the end of Step 1) demonstrates how to
derive trip-start emissions using Equations 3-8 through 3-13.
Determine Hot Soak Emission Changes
Hot soak emissions are the HC evaporative emissions associated with a vehicle trip end.
Equation 3-14 can be used to calculate the change in hot soak emissions (AHCIISK) by
multiplying the change in total trips by the emission factor predicted by MOBILE:
* " * (3-14)
* *
where HSK is the hot soak emission factor (grams per trip) for the subscripted vehicle
class reported by MOBILE, ATRIPTOTAL was defined in Equation 3-3 and TTMPJJKJV
and TTOIPJUJGTI were defined in Equations 3-1 and 3-2. The hot soak emission factor
in grams per trip can be directly taken from the MOBILE model using version 4. 1 or
later. Earlier versions of the model do not report individual hot soak emission rates. An
example application of Equation 3-14 is included in Example 3-3,
Determine Diurnal Emission Changes
Diurnal HC emissions occur from the daily temperature changes while a car is not in use.
MOBILE distinguishes three different types of diurnal emissions depending on the period
STEP 1 3-10 Emissions Analysis of Trips Changes
-------
of day the vehicle is unused:
• Multiple-day diurnal (MDB - vehicle is unused for two or more consecutive
days.
* Full-day diurnal (FDI) - vehicle is unused from Sam to 5pm or is unused all
day, but was driven during the previous day.
* Partial-dav diurnal (FDD - vehicle remains unused for only a portion of a day,
MOBILE output combines the partial-day and the full-day diumals into a combined
weighted diurnal (WDI).
Diurnal emissions occur whether a car is or is not driven during a given day, and
MOBILE*. 1 assumes that 94.3% of the LDGVs and LDGTs experience one type of
diurnal. Only vehicles driven during enough intervals through out the day as to not
experience a significant temperature rise do not undergo a diurnal. Vehicles which are
driven during a given day may undergo a partial-day or full-day diurnal. Vehicles which
are not driven during a given day will undergo either a full-day or multiple-day diurnal.
TCMs which affect trip making may also affect the distributions of diurnal types.
To evaluate diurnal emission changes one needs to determine when the vehicle is unused.
An approximation can be made by making a few assumptions4 distinguishing between
the diumals for cars driven during a given day and those not driven during a given day.
It is assumed that vehicles not driven would experience an increase in multiple-day
diumals relative to the number of full-day and partial-day diumals. MOBILE currently
assumes an average of 23,8% of the LDGVs and LDGTs are not driven during a given
day, and the model also assumes 16.1% of the LDGVs and LDGTs experience a
multiple-day diurnal. Since the multiple-day diurnal vehicles are a subset of the vehicles
not driven, it can be stated that 67.6% (16,1 % divided by 23.8%) of the cars not driven
experience a multiple-day diurnal based on data within the MOBILE algorithms.
For a given TCM, the number of vehicles unused in a day can be approximated from the
net trip changes divided by the number of trips per day. Assuming that 67.6% of the
unused vehicles experience a multi-day diurnal, the following four equations (separated
by vehicle and trip type) approximate the change in diurnal emissions:
4 As is indicated, diurnal emissions occur whether or not a vehicle is driven so that
the change in diurnal emissions due to trip activity changes is calculated from the
difference of two types of diumals. Therefore diurnal emissions should have a less
significant impact than other emission categories because only a portion of the diurnal
emissions is affected, and any assumptions made with respect to diurnal emissions are
expected to have minor influences on the results of this analysis. This assumption is
verified in the example applications presented at the end of Step 1.
5ZXP1 3-11 Emissions Analysis of Trips Changes
-------
'w
7TRIPJJ)GV * (WDILDGV~MDILDGV)
NW
,OP
(3-16)
ANE1RP
= 0-676*
Vv
(3-1 g)
* (WDILDGTl~MDIWGTl)
where AHCjji^ is the change in diurnal emissions for the subscripted vehicle class and
trip type (W = work trip, NW = non-work trip), ANETRPWpP is the net trip changes
for the indicated trip type and period (P = peak period, OP = off-peak period)
determined in Step 5 of Chapter 2, MDI is the multi-day diurnal emission factor for the
subscripted vehicle class determined by MOBILE, WDI is the weighted diurnal emission
factor for the subscripted vehicle class determined by MOBILE, -YTKEP is defined in
Equations 3-1 and 3-2 for the indicated vehicle class, TPDW is the number of work trips
per vehicle commute day (i.e. a commuter makes two trips to and from work on the days
commuting by personal vehicle, TPDW = 2), and THIj^ is the number of non-work
trips per day per vehicle (TPD^ values are region dependent; example TFDj^r values
are illustrated in Table 2-7).
The MDI and WDI emiision factors (grams per vehicle) for Equations 3-15 through 3-18
are determined using MOBILE. Equations 3-15 through 3-18 evaluate the change in
diurnal emissions due to a change in vehicle trips as the difference between the multiple-
day diurnal and the weighted diurnal. If trips decrease, these equations determine the
emission increase due to an increase in multi-day diumals and a decrease in weighted
diurnals which would be observed if fewer vehicles were in-use. Alternatively, if vehicle
trips increase multiple-day diurnals would decrease and weighted diurnals would increase.
The net diurnal emission change is then the sum of the changes calculated from Equations
3-13 through 3-15:
3-12 Emissions Analysis of Trips Changes
-------
(3-19)
An example application of the calculation for diurnal emission changes is given in
Example 3-3.
Total Emission Changes Due to Trip Changes
The following equations can be used to determine the total HC, CO and NOx changes
due to trip changes resulting from TCM implementation:
(3-20)
(3-21)
(3-22)
The values of AHGfmp, ACOjmp, and ANOx-pmp determined in Equations 3-20
through 3-22 are required later in Step 4 to calculate the total emission change. Example
3-4 demonstrates the use of Equations 3-20 through 3-22.
STEP 1 3-13 Emissions Analysis of Trips Changes
-------
EXAMPLE 3-1: Hot-Start and Cold-Start Emission Factors
Example MOBILE4. 1 Emission Factor Data
National Default Fleet, 75°F, 9.0 psi, 26 mph, 1990 Calendar Year,
No I/M Program
Vehicle Operating Mode
100% Cold Start
100% Hot Start
100% Hot Stabilized
LDGV Emission Factor
(grams/mile)
HC
2i55
135
0.95
CO
30-33
14,12
1L03
NOX
1.88
1.72
1:09
LDGT1 Emission Factor
(grams/mile)
HC
3:59
1;99
134
CO
42.25
18.41
13.68
NOX
2.42
2.14
L39
(1) Using Equation 3-6 for LDGV exhaust hydrocarbons:
CSTLDGVrHc = (235 -0.95) *3.S9 -= 5,74 (grams per trip)
(2) : Using Equation :;3^7 for LDGViexhaiistshydrocarbons: :
H^lDGVrco = I7-35 --OSS5) * 3,:59 = 1.44 (gramxper trip)
^Similarly fbr:the other pollutants aad vehicles:
i«S» (g/trip)
=2can;be cale^
reported m^ranwp^;wife.vnie;f^
start mites per trip. Emission factors presented^
STEP1
3-14
Emissions Analysis of Trips Changes
-------
EXAMPLE 3-2: Hot-Start and Cold-Start Emissions Changes
(1) Using MOBILE VMT distributions data in Equation 3-1 and 3-2 to determine the trip
distribution:
"fllUPfLDGV = 0.626 t (0*626 + 0.171) = 0.785
= $ -0.785) = 0.215
(2) Using Equation 3-3 to determine total • trip changes with the trip change data from
Example 2^-8 (this example illustrated the implementation of a rideshare program) ;
mp •=» - 4*406 '(trips)
o
-•- o
(3) Using ^Equations 3-4 and .';$&;*&; determine the number of ihot-start and coldrstart trips
^
C4) lJsing|3Equations 3^ through!3£I3 to d^m
i| exainple illustrates the redurtion in hot-start and cold-start emissioas due to trip
In this example is for illixstrativepu^
STEP 1 3-15 Emissions Analysis of Dips Changes
-------
EXAMPLE 3-3: Hot Soak and Diurnal Emissions Changes
Example MQBILE4. 1 Emission Factor Data
National Default Fleet, 60°F to 84PF Temperature Range, 9.0 psi, 26 mph,
1990 Calendar Year, No I/M Program
Emission Category (units)
Hot Soak (g/trip)
Weighted Diurnal (pains)
Multi-day Diurnal (grams)
LDGV Emission
Factor
3;06
3,30
6.04
LDGT1 Emission
Factor
3.60
5.11
15.33
(1) Using Equation 3-14 to determine hot soak emission changes with example total trip
changes and trip distributions calculated in Example 3-2 and with emission factors taken
from the data shown; above:
= 0,785
TTRIP,LDGTI = 0-215
- -7,247
example trip change data calculated m Example 2-2/wim emission te
data above, and with: 2 ; work;tirips?pec:; work day that ;a vehicle :is medsto commute
'
-•-4j406 (trips)
= 0:
4NE3nB2Rw,OP = -2*84:1 (trips)
-6.04)
'
-(grams)-
(3)?Using Equations S^Wsta sum the diurnal componentsn
:+ 9?t5$c
d3:(grams)-
seen in
; Discussion '••;•.. •' •'::':"i :;.....•; • ''; ''•;
:^ As can;be
, a decrease in trips causes an increase; in vehicles left at home tesulting hi a
emission ;change(^
:
STEP1
3-16
Emissions Analysis of Dips Changes
-------
EXAMPLE 3-4: Total Emissions Changes Due to Trip Changes
Using Equations 3-20 through 3-22 to sum; all components of the trip emission changes with
the data of the components taken from Examples 3-2 and 3-3:
= -45.3 x 103 (grams) AHCHsr = 0 (grams)
= --554 x 103 (grams)
•= 0: (grams)
= -21 J9 x 103 (grams)
jj^j. = 0 (grams)
AHCDNL = + 14.4x 103 (grams)
= -23,0 x Id3 (grams)
^^m
Discussion
Note that the diurnal emissions: merease with decreased trips and the other emission
categories decrease with decreased trips. As is discussed m the text of this chapter^ the
observed diurnal emission change is smaller than the other emission categories. Moreover,
when consideruig the omerHC emission categories such as hotTStabilized exhaust, which is
done later in this chapter, the diurnal contribution to line overaUHGestunation becomes
less sigmficant; ; ':-f:..:-:'.':-:.'.' .. .-;:' '"•:.. •-;.-,•• .--T. . ; -.
STEP 1 3-17 Emissions Analysis of Trips Changes
-------
STEP 2; Emission Analysis of VMT Changes
This step evaluates emission changes due to VMT changes. The emission categories
influenced by VMT include hot-stabilized exhaust, ranning loss, crank case, and refueling
emissions. For this analysis it is not necessary to distinguish between the last three
categories. In this report they are summed into one category termed "VMT-related
evaporative" emissions.
Identify Distribution of VMT Changes
Analogous to the need to define TTHP, & *s necessary to determine the vehicle
distribution of VMT affected by a given TCM (TVMT)- T^*8 parameter is similar to
tTMp except it k based on total VMT and not total trips. Equations 3-23 and 3-24
define the distribution of VMT for the affected vehicle classes (assuming LDGVs and
LDGTs):
7VMT&DVT1
(3-24)
where TvMTJLDGV represents the fraction of VMT which are from LDGVs, "YVMT,LDGTI
is the fraction of VMT from LDGT1, VMT is the total VMT for the subscripted vehicle
class. Regional VMT estimates should be used in Equations 3-23 and 3-24.
If VMT date for Equations 3-23 and 3-24 are unavailable, an approximate value can be
obtained from the vehicle VMT fraction data from the MOBILE model output. The
VMT vehicle fraction would then be substituted into Equation 3-23 in place of the VMT
data. Note that the MOMLE vehicle VMT fractions are a function of calendar year. As
noted in Step 1, in 1990 MOBILE4.1 reports a LDGV VMT fraction of 0.626 (i.e.,
62.6% of the total fleet VMT is from LDGVs) and a LDGT1 VMT fraction of 0.171
(national average default values). The approximate value of >VMT can ^ obtained from
substituting these values into Equations 3-23 and 3-24 to yield *rvMT»LDGV = 0*785 and
Determine Hot-Stabilized Exhaust Emission Chanes
A significant portion of total emission changes are exhaust emission reductions due to
reduced VMT (through fewer trips and through reduced trip length). This section
explains how to calculate this change in hot-stabilized exhaust due to TCM related VMT
changes. Hot-stabilized exhaust emissions are the exhaust emissions after the vehicle has
warmed-up and are calculated in grams per mile by MOBILE using 100% hot-stabilized
operating mode. The hot-stabilized emission factors calculated by MOBILE vary by the
STEP 2 348 Emission Ana* w of VMT Changes
-------
vehicle speed specified by the user. In this analysis, hot-stabilized emission factors
should be determined from the vehicle speeds observed prior to TCM implementation.
Emission changes resulting from the change in speed (before and after TCM
implementation) are evaluated separately in Step 3 of this chapter.
The following equations can be used to determine the peak and off-peak period changes
in hot-stabilized emissions:
*7VMTJJ)GV* STBLDGVtHC,Op] + (3-26)
IANETVMTOP * JVMTJJ>GTI * ^uxm ,HC,OP)
= (&NETVMTp *1VMTJJ3GV*STBLDGV,CQ,P} * (3-2?)
p (3-28)
(AA!£rV!MrOp * yvMTjMGTi * S^IDGTI ,CO,OP)
(ANETVMTp * JVMTJDGTI * STBLDGT1 ,NQ*,p]
l&NETVMTOp * JVMTJJ3GT1 * ^^LDGTl ,NOx,Op)
where ANETVMT is the change in total VMT in the units of total miles for the
subscripted period (P = peak period, OP = off-peak period) determined in Step 8 of
Chapter 2; STB is the hot-stabilized exhaust emission factor in the units of grams per
mile for the subscripted vehicle class, pollutant, and period; and YVMT is the vemcle
VMT fraction for the subscripted vehicle class and is defined in Equations 3-23 and 3-24.
The hot-stabilized emission factors (STB) used in Equations 3-25 through 3-30 are
determined from MOBILE evaluated at the operating mode of 100% hot-stabilized. The
peak and off-peak period subscripts on STB are used to distinguish peak and off-peak
period speeds which are generally different resulting in different emission factors for peak
and off-peak periods. An example application of the calculation of hot-stabilized
emissions changes is given in Example 3-5.
STEP 2 3-19 Emission Analysis of VMT Changes
-------
Determine VMT-Related Evaporative Emissions
The VMT-related evaporative emissions consist of the VMT-dependent, non-exhaust
categories of running loss, crankcase, and refueling emissions. Running loss and
crankcase emissions, expressed as gram-per-mile emission factors, occur while the
vehicle is in operation and are therefore affected by any change in VMT. Refueling
emissions, expressed in grams per gallon of fuel, occur while the vehicle is refueling;
however, the grams per gallon emission factor can be converted to grams per mile using
fuel economy data (miles per gallon). MOBILE reports refueling emission factors in
both grams per gallon and grams per mile, the latter of which is used in this analysis.
The following equations can be used to determine peak and off-peak period VMT-related
evaporative emission changes:
= (ANETVMTP*'YvMTjD0v* VEVPLDOV) + (3.31)
(ANETVMTp
***CVEVP,OP = OPyWTJI>GV LDGV (3.32)
(ANETVMT0p * JVMTJDGTI * VEVPijJGTl )
where ANETVMTp, and ANETVMTOP are the peak and off-peak change in total VMT
determined in Step 8 of Chapter 2, VEVP is the VMT-related evaporative emission factor
for the subscripted vehicle class determined from the sum of the gram per mile nmning
loss, crankcase and refueling emission factors reported by MOBILE, and 7vMT *s *^e
vehicle VMT fraction for the subscripted vehicle class and is defined in Equations 3-23
and 3-24.
Peak and off-peak VMT-related emission factors are used in Equations 3-31 and 3-32
because running loss emissions are influenced by vehicle speed changes resulting in
different emission factors for peak and off-peak periods. An example of the calculation
of VMT-related evaporative emissions is given in Example 3-6.
Due to VMT Changes
Summing the emission changes of the of peak and off-peak hot-stabilized and running
evaporative emission categories into one net emission change, the following equations can
be used to determine the total HC, CO and NOx emissions changes due to VMT changes
resulting from TCM implementation:
3_20 Emission Anaiysis of VMT Changes
-------
0-34)
p (3-35)
where the values of AHCSTB, AHCVEVP, ACOSTB, ANOxgyg are defined and calculated
in Equations 3-25 through 3-32. The resulting values of the total emission changes due
to VMT changes, AHCy^, ACCy^, and ANOxyj^j, determined in Equations 3-33
through 3-35 are required later in Step 4 for the calculation of the total emission change.
An example of the determination of AHCyj^j, ACOy^, and ANOxyj^j is presented
in Example 3-7.
STEP 2 3-21 Emission Analysis of VMT Changes
-------
EXAMPLE 3-5: Hot-Stabilized Exhaust Emission Changes
Example MOBILE*. 1 Emission Factor Data
National Default Fleet, 75°F Ambient Temperature, 9.0 psi, 1990 Calendar Year,
No 1/M Program
Vehicle Operating
Mode, Speed
100% Hot Stab., 20 mph
100* Hot Svib,, 35 mph
LDGV Emission
Factor (grams/mile)
•HG
1.23
mm
CO
14.62
7.79
NOX
1.15
IM
LDGT1 Emission
Factor (grams/mile)
HC
1.77
0.94
CO
18.05
9.46
NOX
1.41
1.39
Using Equations 3^25 through 3-30 to dfitermmehofestabilizedexhmist emission
withhotr«tabili2»demissioii:lactors (STB) taken from tneidattabowesfassumingia^peak
period speed of 20 mpkand an; off-peak of 35 mpli), with...ANETVMT(calculated in Step 8
of Chapter 2) taken from Example 2-9, and with ^y^p determined from ;MOBILE VMT
distribution data:
(gnuas) ."'.••.'."; :b
= ••••>£ 78*057 * 0. 785 * 0:69) ^•(i7
x 10* (grams) . • . ,, ^ '
'
' '
illustrative purposes only. xl-*.::':.::j^:-:;:^::':'i:.: ':-^^:'i'I:;-.!:Sl^:' !f^^':/ ' '•;
3-22
Analysis of WMT Changes
-------
EXAMPLE 3-6: VMT-Related Evaporative Emission Changes
Example MOBILE4.1 Emission Factor Data
National Default Fleet, 75 "F Ambient Temperature, 9.0 psi,
1990 Calendar Year, No I/M Program
Emission Category
(Speed)
Running Loss (20 mph)
Running Loss (35;mph)
Crankcase (all speeds)
Refueling (all speeds)
LDGV Emission
Factor (grams/mile)
0:22
0.12
0:03
0;19
LDGT1 Emission
Factor (grams/mile)
0.22
0.13
0.06
0.25
(1) Calculating the WMT-related evaporative emission factor as the sum of running loss,
crankcase and refueling emission factors (assuming a peak period speed of 20 mph and an
off-peak ;:period of 35 mph):
= 0.22 + 0.03 •+ 0.19 = 0.44 (gram/mile)
= 0.12 + 0.03 + 0.19 - 0.34 (grams/mile)
= 0.22 +0.06 +0.25 = 0.33 (grams/mile)
= 0.13 + 0,06 + 0,25 = 0.44 (grams/mile)
(2)sl3sing Equations 3-3 1 and 3 *32 to determine the jVMT-related evaporative emission
changes with VEVP emission factors determined in (1), with AVMrTQp (calculated in Step
8 bf;Chapter:;2)! taken ifrom Example 2-X, and with r determined from MOBILE VMT
distribution data:
V = 0-785
DGH —0*
ANETVMTp = -118^989 (mUes)
= -78;057 (miles)
V03 (grans) < • '•.' "•''<:•
* 0885-* 0.34^+J- 78,057
= -38:2 xlO3 {grams}
Discussion
0.44)
Using Equations 3-31 and 3-32, the VMT-related evaporative emission changes can be
detennmed from the VMT (^^ Emission
SIKP2
3-23
Emisaon Analysis of VMT Changes
-------
EXAMPLE 3-7: Total Emission Changes Dae to VMT Changes
Using Equations 3-33 through 3-35 to sum all componeots of the VMT-related emission
changes with the data of the components taken from Examples 3-5 and 3-6:
p = -160 x 103 (grams)
C^BQp = -58;l x 103 (pams)
. = -1J83 x 10s (grams)
AGOgr^op = -636 x 1Q3 (grams)
p = -143' x 103 (grams)
AN°S«TB,OP = -88.3 x I03 (grams)
-54,6 x 103 (grams)
:= -28:2 x 1Q3 (grams)
••/- 16Q - 58.1 - 54.7 -2&2) xl « -301 x 1 (grams)
ACOVMT = -I'SSxltf-eStivcKf*^ -2J7x 10? (grams)
'= (-143 - 88.3) x iti* = -231 x JO* (g rams)
This example illustrates; &e total ^VM^^ated emission changes, Jn comparison to me total
trip-related emission c&anges of i:Example 3-4, ;me \*MF-rela^ ;«inissions dianges are
isigmficantly larger;; ; : : .• ••...^^'.:::,:,,--.M,:...,.^^<:.^:;,x^S-2^.-..-
STEP2 3-24 Emissum Analysis of VMT Oteutge*
-------
STEP 3: Emission Analysis of Fleet Speed Change*
This step evaluates emission changes due to the changes in vehicle speeds. The emission
categories influenced by this evaluation are hot-stabilized exhaust and running loss
emissions. This step differs from Steps 1 and 2 in that all vehicle classes are affected by
speed changes. It is important to note again that the methodology used here considers
regional average speeds and will not capture the complexities implied by the fact that
vehicles are traveling at different speeds in different parts of the region. The parameters
required to complete the speed change emissions analysis are:
« SPEEDp BASE - speed for peak period (P) prior to TCM implementation
(indicated by the subscript BASE).
• SPEEDOPsBASE - off-peak period (OP) base speed.
« SPEEDpjTCM - peak period speed after TCM implementation (indicated by the
subscript'TCM).
* SPEED0p .JQ^ - off-peak period speed after TCM implementation.
* VMTpjTCM - total peak period VMT for modeling region after TCM
implementation.
* VMTQp TCM - total off-peak period VMT for modeling region after TCM
implementation.
Of these parameters, the base speeds are region dependent and should be known prior to
this analysis. The TCM speeds can be determined from;
SPEEDPTCM = SPEEDp MSE+&SPDP (3-36)
(3-37)
where ASPDp and ASPDop were determined in Step 9 of Chapter 2. The values of
VMTpjTCM and VMTOP TCM can also be determined from the parameters used in Step 9
of Chapter 2:
VMTPtTCM = VMTp+WMTp (3-38)
VMTOP,TCM = VMTOP^AVMTOP (3-39)
where the parameters of Equations 3-38 and 3-39, AVMTP, AVMTOP, VMTP, and
VMTOP, were also identified in Step 9 of Chapter 2.
STEP 3 3-25 Emission Analysis of Fleet Speed Changes
-------
The emission change due to a change in speed is determined from the difference in
emission factors (hot-stabilized exhaust and running loss) evaluated at the speed prior to
TCM implementation and at the speed subsequent to TCM implementation. This is
expressed in the following equations which can be used to determine the net emission
change due to an overall peak period fleet speed change:
P ~ VMTTCM,P*t(SJBFLTJIC,P,TCM't'RNLFLT,P,TCM) (3-4Q)
- VMTJCfAp *
(3-41)
- VMTTCM Op
ACOSPDP
ANOxSPDP = VMTjcMp* (STB^TNQKpT(^-SWfLT^0^pfBASE) (3-44)
where the subscript SPD (i.e. AHCSTD) indicates speed-related changes of the indicated
pollutant, STB and RNL arc the hot-stabilized and running loss emission factors for the
subscripted pollutant, and the subscript FLT indicates fleet emission factors. The
subscripts OP and P, indicating off-peak and peak period, and BASE and TCM,
indicating base speed and TCM-related speed, are used to identify the correct speed used
in the evaluation of the emission factor.
The overall emission change is the combined changes observed in the peak and off-peak
period and is calculated by the following equations:
AHCSPD . ASCmj, - MtCSPDfOP (3-46)
ACOSPD - *COSPDfP + *COSPDtOP (3-47)
ANQxSPD - *NOXSPDtP * MfOxop (3-48)
The values of AHCSPD, ACOgpjj, and ANOx^p determined in equations (3-41) through
(3-43) are required later in Step 4 for the calculation of the total emission change. An
example application of determination of emission changes due to speed changes is given
in Example 3-8.
STEP3 3-26 EmtsvoHAMobBs of Fie* Sp~d Changes
-------
EXAMPLE 3-8: Total Emissions Changes Due to Speed Changes
Example MOBILE4.1 Emission Factor Data
National Default Fleet, 75°F Ambient Temperature, 9.0 psi, 1990 Calendar Year,
No I/M Program
Scenario (speed)
BASE, Peak Period (20 mpfa)
BASE, Off-peak Period (35 mph)
TCM, Peak Period (22 mph)
TCM, OSfepeak Period (36 mph)
Fleet Emission. Factors
(grains/mile)
HC
1,676
1.938
1.524
O06
CO
:8;421
9.916
6i670
9,614
NOX
'2,739;
2328
2.674
2:532
Run. Loss
0.212
0.120
0.195
0.416
(1) ;Usin£ Equations 3-40:flirough 3-45 to determine hot*stabttizedierfiaust emission changes
with hot-stabilized (S3!B) and the running loss (KNL) emission factors taken from the data
above at ;tbe -indicated speeds, and iwfth Vl^
= 33,4 x 106 (miles)
' ,
'-;:'x'';"vr5::';:^:::::'J'' ""
m
•
'
^^s^^s^^^v»:^^^^^^^
(2) Using Equations 3-46 ;to 3-48 to sum flie;peakandbrlripeak:c6mponents:
i^ww^^'.';::.'!'':i;"^.;^:-.!
,=-f 2*m0^X^=^^ . ''.' V,,
AGO
This application of emission changes from speed CfeM
only. In the emission analysis of the rideshare program which has been use^ •
examples undl mis one, the speed change analysis given in Example 2-1^
isjjeedj changes. Thus the emission change of mat particular rideshare program
sa^^
rrpm:-me^ '• •,,
- speedssof m :iwiiii|^^ ;-§eaeratf::3afge?«mMka;:;«ianj|K::; • -: :ffiMKe«er,.;it;:is:;>.;:: i-:
important to note mat sr^eed changes i^izedb^ ^ . y
X^^ \ t||Q'||;| Iffi^:''; '^;S:A:fZf,,:-
STEPS
3-27
Emission Analysis office* Speed Changes
-------
STEP 4: Total Emission Changes Due to TCM Implementation
The final emission change realized by a region due to the implementation of a TCM is
the sum of the emission changes determined in Steps 1 through 3, and can be calculated
in the following equations:
AHC = AHC+AHC+AHC (3-49)
ACO = ACOnup+ACOyMT+ACOspz) (3-50)
ANOx = ANOx+ANOX+ANOx (3-51)
where AHC, ACO, and ANOx are the final emission changes which combine emission
changes due to trip changes determined in Step 1, VMT changes in Step 2, and speed
changes in Step 3. An example application of Equations 3-49 through 3-51 is given in
Example 3-9.
3_2g Total E
-------
EXAMPLE 3-9: Total Emissions Changes Due to Speed Changes
Using Equations 3-49 through 3-52, the total emission changes are the sum of the trip-
related,VMT-related, and speed-related emission changes determined in Steps 1,2 and 3
respectively:
= -301 x 103 (grams)
» -53.9 x 103 (grams)
AHCSPD = 0
= -554 x 103 (grams)
= -2.47 x 106 (grams)
= 0
= -21.9 x 103 Cpanis)
= -231 x 1# (grams)
223P••+ Q)tx tl^ — -253*tiff (grams)
..:. .-.-''. . . ' Discussion •
AHC, ACOV and ANOx are the final emission changes realized by TCMimpleme^
Note the disoission of Example 3-8 for the evaluation of emission changes due to speed
changes. Useful conversion facers for me conversion of emissions in grams to tons (1.10 x
STEP 4 3-29 Total
-------
4 TCM INTERACTIONS AND MODE CHOICE
DEPENDENCE ON MULTIPLE ATTRIBUTES
INTRODUCTION
Few agencies arc likely to implement only one TCM at a time. More often, several
TCMs will be implemented together. However, TCMs are rarely independent of one
another, thus separate analyses of individual TCMs in a package may be misleading.
Two issues should be considered when conducting an analysis of a package of TCMs:
(1) measures overlap target audiences and it is possible to double count the effectiveness
of TCMs lacking consideration of this overlap. For example, one person cannot both ride
the bus and carpool to work. Similarly, some measures may be effective but may attract
participants from other, preexisting programs. For instance, a rideshare participant may
switch to transit if transit passes are offered); and, (2) the implementation of some
measures either improves or diminishes the chances for successful implementation of
other TCMs, These synergies need to be recognized while analyzing the effectiveness of
a given TCM (one example: parking pricing strategies improve the success rate of other
programs such as rideshare).
This chapter presents a method for evaluating packages of TCMs rather than individual
measures. By conducting the analyses presented below, analysts will be able to predict
changes in the mode split1 of a target population in response to different packages of
TCMs. Once the TCM participation rates are known (e.g., an additional 5 percent of all
peak period trips will be rideshare trips, or an additional 5 percent of works trips made
by people working for a major employer will be made using transit), analysts can
evaluate measures individually to determine their travel and emission impacts. The
packaging methodology in this discussion builds upon analytical approaches developed by
the authors under separate sponsorship (Austin et al., 1991; Eisinger et al., 1991).
The relationship between TCMs and mode choice is a subject in its infancy of
development. Mode choice behavior is difficult to predict for several reasons. One
reason is that mode choice is a behavioral response to a variety of factors, many of which
cannot easily be quantified. For example, a person may choose to take public transit one
day just because he/she is simply not in the mood to sit in gridlock traffic on the freeway
that day, and the same person may choose to drive alone the following day because of an
1 Traditional mode split generally refers to the distribution of travel among SOVs, ridesharing, and
transit. Mode split here refers to the distribution among all possible modes, including modes that may
be introduced by TCMs such as telecommuting or compressed work weeks.
92093.02 4-1
-------
uncomfortable experience on the bus. Such factors, which can sometimes be arbitrary,
add a large random element to a person's mode choice. Adding to the difficulty of
predicting human behavior with respect to mode choice, TCMs most often are not
independent of one another. For example, the introduction of flextime to a population
that already has a successful rideshare program may encourage SOV drivers to drive
during off-peak hours and may also disrupt carpools. The methodology presented here
provides an analytical framework for identifying and roughly quantifying the impact of
such relationships.
There exist transportation models that may be used to study the change in mode splits
resulting from the implementation of packages of TCMs. A good example includes
TRIPS (Harvey, 1991), which is based on 5,000 households in Los Angeles. The model
uses empirical equations that have been statistically calibrated from a particular database
(from the Los Angeles area). Often the data used to calibrate such models is outdated or
only representative of the region from which it was collected. These models require
extensive data to be supplied by the user (typically collected in surveys and individual trip
diaries), such as average household disposable income, the number of workers in a
household, in-veMcle travel time, and number of autos in the household. If a recently
calibrated transportation model is available and the user has the extensive input data
required, more accurate estimates of TCM package effects may be made in this manner
than with the more approximate approach suggested here.
Other approaches requiring less data are also available. A good example is the pivot
point technique (CSI, 1979), which calculates incremental mode shifts as a function of the
utility of each mode. The model, still in use, applies a multinomial logit formulation and
coefficients from a 1968 Washington, D,C. travel survey. A limited comparison of the
coefficients with 1967 Los Angeles and 1963 New Bedford coefficients found them to be
quite similar. The pivot point model does not address modes other than SOV,
carpooling, and transit, and does not address nonwork trips. Pivot point, with the
original coefficients, is used in a California Air Resources Board model for TCMs called
AQAT (Randall and Diamond, 1990). AQAT is described in Appendk A together with a
number of other California methodologies.
The TCM packaging methodology presented here is not an empirical model statistically
calibrated from an extensive database. Rather, it is based on principle attributes of travel
modes and the comparative values of the different modes for each attribute. It provides a
framework for the analyst to think through the transportation alternatives available to the
population (or region) being studied, and is a means to identify the strengths and
weaknesses of different TCMs combinations, and an opportunity to better understand the
factors influencing the current mode split and what it would take to alter the mode split
significantly. The analyst selects every coefficient in the model to represent the
population under study. The only data required for the approach are current mode choice
splits, costs, and travel times. The minimal data requirements make it inexpensive to
use, and its calibration flexibility makes it transportable to different regions and
populations. This approach is recommended for use when the resources required to run a
detailed transportation model accurately are not available. However, because this model
is largely conceptual, the analyst should be prepared to demonstrate the reasonableness of
92093.02 4-2
-------
the coefficients chosen to represent the population of study. The selection of coefficients
should be guided by empirical data from regions with similar demographics. The
predicted mode splits should not be viewed as exact, but as approximations.
The rest of this chapter includes five discussions: (1) a brief overview (the "big picture")
of how the packaging methodology works, (2) a discussion outlining key mode choice
considerations, (3) a step-by-step discussion that describes in detail how to conduct a
packaging analysis, (4) sample applications of the methodology using empirical data from
San Francisco and the Phoenix metropolitan area, and (5) example applications of a
TCM.
OVERVIEW OF THE PACKAGING METHODOLOGY
The packaging methodology is founded upon the premise that individuals choose their
travel mode based upon the attributes of each mode choice option. Example attributes
influencing mode choice include the convenience of SOV use, the lower costs associated
with rideshare, or the convenience and cost savings of not having to make a trip when
telecommuting. The packaging methodology focuses on valuing individual TCMs based
upon how well each TCM rates in consideration of several important mode choice
attributes2, TCM participation rates are quantitatively estimated based upon the total
value of a specific measure in comparison to the value of other mode choice options
available to the trip maker.
Interdependent TCMs are easily addressed using this methodology. The overlap among
TCMs is accounted for by comparing the relative value of each TCM based upon mode
choice attributes. More valuable measures are assumed to attract a greater percentage of
a given target audience (such as individuals making work trips). Synergies are accounted
for by the way in which TCMs alter the value of a mode's rating for an individual
attribute. As an example, if a parking management program combined with a ride
matching program increases the costs of using a SOV while decreasing the inconvenience
of ridesharing, then the total value of SOV travel diminishes and the total value of
ridesharing increases in the methodology's ranking system,
A key component of the methodology involves validating the approach using actual mode
split data. As an initial step in evaluating packages, one sets up TCM rankings so that
the methodology replicates existing mode splits. Once this is done, one can then alter
variables to estimate the participation rates among new or enhanced TCMs.
The methodology is most easily implemented using a spreadsheet or a simple FORTRAN
program. The example applications included in this chapter were developed using a
FORTRAN program. First the selected coefficients were validated by illustrating model
agreement with actual mode splits and then the inputs were altered to reflect the
2 As discussed in more detail later in this section, the literature on mode choice decisions points to
four key attributes that individuals weigh: time, cost, reliability, and convenience of travel
92093.02 4-3
-------
hypothesized implementation of a set of TCMs. The program enables the user to easily
modify data inputs and to quickly adjust the validated inputs to reflect the TCMs.
MODE CHOICE CONSIDERATIONS
The foundation of the packaging methodology is the ability to understand and rank the
key factors individuals consider when determining mode choice. Specific TCMs alter trip
making behavior if the measures affect the factors that people consider in deciding upon
the nature and frequency of their trip making. To analyze the effects of a package of
measures it is important to understand how each measure (both individually and in
concert with other measures) affects the key variables that people weigh when deciding
upon mode choice.
Individuals directly and indirectly consider numerous attributes of the mode choice
opportunities they have when deciding what trips to make and how to make them. The
four most important factors are travel costs, time, reliability, and convenience.
Cost and Travel Time
For most mode choice decisions, travel time and cost are two of the most important
factors and have been the key variables used in mode choice models for some time
(HutcMnson, 1974). As Wachs (1990) states in a paper summarizing implications of
behavioral research on transportation demand, "Applications of behavioral science to
transportation planning give greatest emphasis to travel time and travel cost as the
characteristics of travel modes most likely to influence choices made by commuters."
Recent findings still support this relationship. For example, Willson, et. al. (1989)
summarizes existing studies of the relationship between poking subsidies and SOY users
in a number of areas.
Certain costs are more important than others. Analyses of specific TCMs support the
idea that day-to-day costs are the most important cost considerations. Feeney (1989) says
that parking costs are weighted more heavily than mileage related or car maintenance
costs. Wachs states that "Many studies of commuters9 willingness to carpool have shown
that commuters consider the out-of-pocket costs of carpooling versus driving alone to be
among the two or three most important factors influencing the choice between these
modes, the others being travel time and convenience..." (Wachs, 1990).
Similarly, certain time costs are more important than others. "Excess time" (time spent
other than just driving enroute) has substantially greater disutility than driving time (e.g.,
some studies show that walking time has twice the disutility of in-venicle time; Feeney,
1989). Wachs (1990) states "A variety of studies, conducted in different environments,
involving different trip purposes and different modes, have shown that people
psychologically weight 'out-of-vehicle time' somewhere between two and three times as
heavily as they weight 'line-haul' time or moving time in their travel decisions."
92093.02 4-4
-------
Other Attributes Weighed By Trip Makers
While the four factors listed above are key, others can also influence TCM choices.
Analysts should be sensitive to whether these factors may be operative in their study
region, and incorporate them into the analysis to the extent possible. It is recommended
that this be accomplished through adjustments to the weights assigned to the four primary
factors. The discussion presented below is intended to introduce some of these factors.
Regarding transit, for example, Wachs (1990) observes that the following variables (listed
in order of importance, beginning with most important) determine whether transit is
taken: cost, reliability, travel time savings, comfort (climate control, exposure to rain,
snow and hot sun), space for packages. Among the range of important considerations
identified in other behavioral studies are the following:
Ridesharing: There is a strong correlation between commute distance and mode choice;
the longer the commute distance, the higher the ride sharing rates (Grain, 1984; DOT,
1985). A major deterrent to carpooling is incompatibility of people's schedules. Valdez
and Arce (1990) found in a study on ridesharing that about 42 percent of respondents .
"believed that depending on others was not worth the money carpooling would save,
higher proportions (55 to 58 percent) believed that achieving time savings in commuter
lanes or fulfilling the requirement for pooling to obtain a guaranteed parking space at
work would be worth depending on others or leaving work at a fixed tune each day."
Childcare issues also serve to deter potential ridesharing; people want to be able to
respond to emergencies if necessary (Grain, 1984; DOT, 1984), and "parents who need
to leave children at child care facilities, or are concerned about their ability to react to
emergency situations involving their children are reluctant to rideshare" (Valdez and
Arce, 1990). Part-time carpooling is viewed more positively than regular (i.e., daily)
carpooling. In one study, cne group expressed a willingness to rideshare on a part-time
basis, for example 2 or 3 times per week. The other days they either have specific
obligations that require car availability or they simply wish to have their car for the
freedom it offers. Some factors considered by people when deciding whether or not to
rideshare or take transit involve the other individuals they would travel with; personal
characteristics such as smoking can either encourage or discourage potential ridesharers,
depending upon their feelings about these habits (Grain, 1984; DOT, 1984).
Transit Use: The principle factors influencing choice transit riders are* "...the relative
service properties of competing transport modes such as in-vehicle travel times, excess
travel times, out-of-pocket costs and the overall convenience of travel" (Hutchinson,
1974). A main transit problem (as cited by SOV commuters) is the length of time
necessary to take the bus to work. Surveys reveal a perception that driving takes
substantially less time than a transit trip. Another perception is that bus service may be
unreliable (Grain, 1984; DOT, 1984), Also, perceived difficulties in deciphering bus and
other transit schedules can deter inexperienced transit users from considering transit as a
viable alternative to driving (Grain, 1984). Some commuters, especially women, feel that
their personal safety is at ride while waiting at or walking to bus stops, especially in the
dark (Grain, 1984).
92093.02 4-5
-------
Other Factors Influencing Mode Choice; Childcare issues can affect commuter mode
choice. Many individuals cite the need to respond to child health emergencies and
childcare needs as reasons for driving to work rather than taking transit. Many non-work
related trips during the day are generated by parents shuttling children to daycare, doctors
appointments, and other commitments. This trend is more prevalent with women in the
woric force than men (Grain, 1984; Raux et al., 1986; DOT, 1985). Also, commuters
often underestimate or are not aware of their true commute costs; for example, one
woman estimated her commute costs, including insurance, gas, maintenance, etc. to be
around four cents per mile--in actuality, the total cost was around 18 cents per mile.
Summary of Important Attributes
The major attributes of travel choices can be broken down into four broad categories: (1)
cost—including long- and short-term costs; (2) time—including direct travel time enroute
plus excess time from the origin and destination to the mode choice; (3) convenience--
including comfort, safety, and flexibility of the mode choice; and (4) reliability-focusing
on the predictability of the mode choice's ability to reliably deliver the rider to her or his
destination.
STEP-BY-STEP APPROACH TO CONDUCTING A TCM PACKAGE ANALYSIS
TCM packages can be evaluated by defining how, when packaged with the other
measures, each individual measure compares across the four major attributes determining
travel characteristics. For example, when two TCMs overlap a common audience, trip
makers will choose whether or not to participate in one or another of the TCMs based on
each TCM's ability to offer cost, time, convenience, or reliability improvements to the
traveller's driving conditions. Figure 4-1 illustrates how a mode choice can be described
as a function of these attributes; the figure also describes units of measure for each
attribute. The basis for the packaging methodology is frequently referred to as a "multi-
attribute analysis." It is based on conceptual analytical methodologies described in
decision analysis literature (see Stokey and Zeckhauser, 1978),
Conceptually, the idea behind the packaging methodology is simply to value each TCM
using the attributes of cost, time, convenience, and reliability as a framework for
assigning a total value to any one measure. Since each of these four attributes has a
different unit of measure (dollars, minutes, etc.), a mathematical framework is
constructed to translate these attributes into a common unit of measure, and then they are
summed across the attributes and estimate a total value or "utility" for a given measure.
Analysts can use the methodology to compare measures that overlap target populations in
terms of their total value and to define mode choice preferences. The concept of overlap
is key; measures which do not overlap can be analyzed individually. Most important,
analysts can use the methodology to consider the synergies among measures. For
example, assume that SOV users are not participating in a rideshare program (i.e.,
assume that the total value of SOV use in terms of cost, time, convenience, and reliability
is greater than the total value of rideshare). To encourage ridesharing, an employer
92093.02 4-5
-------
Attributes
Total
Choice Value
Reliability
Time
Total Travel Time
Total Travel Distance
Convenience
Comfort
Safety
Flexibility
Access to Transport
(handicap, walk distance)
During Travel
(seating, music, companions)
Ability to Complete
Olher (mid-day) trips
Im mediate-Emergency
Travel Availabtlty
Start-End Times
Cost
Long-Term (months-years)
Immediate
Weekly-Monthly
Daily
FIGURE 4-1. Attributes of travel choices.
aaa9t085rl
Units of Measure
Percent of time transport is available;
predictability of start/end times
Minutes or hours
Miles
Ease of and lime to access
Aesthetic value
Risk of injury
Ability to complete other trips/chores.
Access to immediate transportation
Range of (and number of options for) starl~end limes
Dollars
Dollars
Dollars
-------
begins to charge for SOV parking; thus the employer's parking management program
now interacts synergistically with the rideshare program. Using the analytical approach
presented in this chapter, this synergy can be explicitly captured—the total value of a trip
made by SOV now drops relative to the total value of a rideshare trip because the "cost"
attribute associated with SOV use is now more expensive.
Three broad steps need to be completed to conduct a packaging analysis: (A) gathering
of travel data and establishing base-case conditions; (B) establishing a validated base case;
and (C) conducting the TCM package analyses (an eight-step process described below).
A. Collect Travel Data, Establish Base Case Conditions
A full discussion on the collection of data is included in Chapter 2. Data requirements
for individual and TCM packages are similar. We recommend dividing the population
(or region) of interest into smaller segments to study whether characteristics vary
substantially across segments. Factors to address when collecting relevant data include
the following:
(1) Establish the preexisting conditions in the areas to be analyzed. This includes
to what extent TCMs have already been implemented, what AVO (average vehicle
occupancy) levels have been achieved during peak and off-peak periods, what
types of trips occur in the region, and what mode choices are available.
Determine the percentages of people who travel via each of these modes (i.e., the
mode split).
(2) Determine the relative costs of the different modes. The costs to consider
consist of out-of-pocket expenses. For example, the cost for driving a SOV may
include the cost of gasoline, parking, and tolls. The cost of taking transit consists
of the fare. Ridesharers may or may not have to pay the cost of parking or tolls,
depending on the programs affecting the population. Telecommuters pay nothing
on the days they stay home, but they may pay the same as SOVs on the days they
travel to work. Thus average travel costs over the entire work week are lower.
(3) Determine the relative travel times of the different modes. How long does the
average trip take driving a SOV, taking transit, or ridesharing? If these travel
times are not directly available, they may be estimated by dividing trip distances
by average speeds. For a TCM such as telecommuting, we recommend averaging
the travel times over all days in the week.
(4) Estimate the relative convenience of the different modes. For example, to what
extent are the different travel options available within the region of interest? How
widely dispersed are transit services? How available is transit during peak and
off-peak hours? Are there express services? What park and ride lots are available
and what percent of their parking spaces are vacant? What kind of immediate
access is available, and how easy is it to complete mid-day trips (i.e., can you
come and go as you please)? How safe and comfortable are the different travel
92093.02 4-8
-------
options? Note that these estimates are "guessed at" by the analyst. They are to
be used in Step 1 below. If they are slightly off, the calibration procedure will
identify this.
(5) Determine the relative reliability of the different modes. How often is transit
on time? How often are carpools on time? Do people often get held up in traffic
jams? These estimates are to be used in Step 1 below.
B. Establish "Validated" Base Case
Analysts will need to use the eight-step packaging methodology twice: once to establish
base case conditions, and a second time to establish potential participation rates given
new or enhanced TCMs. Replicating base case conditions validates the approaches and in
essence creates a model of the region's mode choice decision making. The validation
process runs through the eight-step approach using variables that reflect existing
conditions. These variables include the values of the different attributes of the modes of
travel (i.e., the cost, convenience, travel time, and reliability) to the target population of
each of these attributes. A set of reasonable variables must be determined that accurately
characterizes the current travel patterns of the target population.
C. Eight Steps to Conduct Packaging Analysis
The eight steps described below establish base case conditions that calibrate the approach,
and then adjust the base case inputs to reflect new or enhanced TCMs.
Step 1: Determine the cost, time, convenience, and reliability for each mode of travel;
i.e., determine the "attribute profile" of each mode. For example, the time associated
with an average transit work trip in the modeling region is 1 hour.
The convenience and reliability measures are determined based on the factors discussed
above. They represent the study population's perception of how convenient and reliable
each mode is. This perception is dependent on factors such as levels of service, safety,
and timeliness.
Step 2: For a given trip type (work or non-work trips) determine the best and worst
limits of the attribute profile; i.e., determine the best and worst cost, convenience, travel
time, and reliability that are possible in the region being analyzed for the TCM package.
For example, the worst cost could be $4.00 per trip, and the best possible cost could be
$0.85 per trip. The best possible convenience could be immediate access, and the worst
possible convenience could be only trip-end access.
A general guideline to follow when selecting the best and worst limits for attribute values
is that the median value of all the modes should be close to 0.5. This ensures that the
mode values are not all clustered toward 0.0 or toward 1.0. If the mode values cluster
toward 1.0 for a particular attribute, then that attribute is artificially given more weight in
92093,02
-------
the final calculation, and similarly, if they cluster toward 0.0, then the attribute is
artificially weighted less,
Step 3: Scale the cost, convenience, travel time, and reliability associated with each
mode of travel on the "best to worst" scale for that trip type. Figure 4-2 provides an
example. This should be done separately for work and non-work trips,
The convenience and reliability of a mode of travel must be given a value between 0 and
1 , where 0 is the worst possible value and 1 is the best possible value. Travel time and
cost values may be calculated directly as follows:
= (worst cost - COSTk)/(worst cost - best cost) (4-2)
= (worst time - TIMEk:)/(worst time - best time) (4-3)
where
k = the travel mode (if there are 3 modes being analyzed, then k will range from 1 to 3),
COSTk = the cost of travel of mode k,
= the travel time of mode k,
= the cost value of mode k, and
= the time value of mode k.
Step 4: Determine a weight profile (or set of coefficients)
(>wACOnvenience>\ime>\eliability)> for the population in the region.
When measuring the potential change that would occur from a set of TCMs, the
importance the target audience places on the four attributes must be estimated. For
example, if a traveler values cost far more than reliability, time, and convenience, then
increasing their cost of travel will more likely alter their behavior than decreasing the
reliability. This step involves assigning the relative importance of each attribute to the
average member of the study population. This relative importance is called the "weight
profile". Note that if the mode values for a particular attribute cluster toward 1.0 or 0.0,
as discussed in step 2 above, then the choice of the weight profile will be influenced by
the artificial weighting created by the choice of best and worst values.
The X's are assigned the percentage of importance of each of the attributes. They must
sum to 1.
Based on the mode choice literature, a reasonable weighting of each attribute's
importance for work trips might be: 0.3 for cost, 0.3 for travel time, 0.2 for
convenience, and 0.2 for reliability (summing to 1.0). Analysts are encouraged to select
weights appropriate to their specific areas and trip types.
Step 5: Calculate the total utilities of each mode of travel. The total utilities will reflect
the relative weight coefficient of each attribute (from Step 4), and each measure's scaled
value (between 0.0 and 1.0) for that attribute (from Step 3). See Figure 4-3.
92093.02 4-10
-------
Cost
$ .I0/mile
Convenience
Immediate Access
Best
Worst
$ .85/milc
Only trip-end
Time
0.75 hours
Reliability
Nearly 100%
1.5 hours
85%
• = Rideshare =$.20/mile
A = Transit = $.14/mile
Emergency Access
Limited mostly to trip-end
1.0 hour
I.I hours
90%
90%
FIGURE 4-2. Sample hypothetical attribute profile for two measures, rideshare and transit.
aaailOOSrl
-------
K)
Cost
1.0
Best
Worst
.95
.85
0
Convenience
1.0
,3
.2
Indicates "value function" of each measure's attributes.
• = Rideshare
A = Transit
FIGURE 4-3. Value function for each attribute for rideshare and transit.
aaa91085rl
-------
The calculation of the total utility for a particular travel mode, k, and weight profile,
TUk = COSTVAL^Xj + CONWALk*X2 + TIMEVALk*X3 + RELTVALk*X4
(4-4)
where
TUk = the total utility of mode k,
COSTVAL},. = the cost value of mode k,
CONWALjj = the convenience value of mode k,
the time value of mode k,
the reliability value of mode k,
Xj = the relative weight of cost,
X2 = the relative weight of convenience,
X3 = the relative weight of time, and
X4 = the relative weight of reliability with respect to the other attributes.
Step 6: Calculate the estimated percentage of time a person from the target population
group with weight profile (X1,X2,X3,X4) will travel via mode k relative to the travel
modes considered.
Presented is a mathematical representation of how to estimate the percent of the target
audience that will use the individual travel modes. Suppose there are N different modes
of travel that are available to the travelers that are targeted for the TCM package (for
example, N might be 4, representing SOVs, transit, rideshare, and telecommuting). For
each person, there is a set of values TUk for k = 1 through N, where TUk represents the
total utility of TCM "k." The probability that a person will travel using mode k will
depend upon the TU of k in relation to the TU of their remaining mode choice options.
For example, if a person has a total mode utility value of 0.9 for single occupant travel
and only 0.1 for public transit, then it is unlikely that this person will travel on public
transit. However, if a person has a TU of 0.41 for single occupant travel and 0,40 for
public transit, then it is only slightly more likely that this person will ride in a single
occupant vehicle instead of taking public transit. In precise terms, the packaging
methodology must consider that the percentage of people that will travel via mode k is
dependent upon the differences in magnitude between TUk and all the other TU numbers
for that population.
When extending a travel mode's total utility to an entire target market of trip makers, the
total utility serves as a surrogate for the degree to which a specific measure will be
utilized. To relate total utility to percent of time a TCM is utilized, it is important to
represent that measures of little value are not likely to be utilized, while measures of
greater value are likely to be substantially utilized. As a concrete example, consider one
of the packages just discussed: a person has a total mode utility value of 0,9 for single
occupant travel and only 0.1 for public transit. With such a large disparity in utility
92093.0Z 4-13
-------
between these two travel options, it is unlikely that transit will be utilized. It is therefore
inappropriate to linearly relate target market for a TCM to its total utility. In other
words, it is unlikely that 10 percent of the target market would utilize transit. A better
mathematical approach relating total utility to target market use of a TCM is to relate the
percent of time that the target audience will utilize a given TCM as a function of e6-5TU,
where "e" is the exponential function, TU is the total utility of that mode choice. The
constant 6.5 was determined empirically to provide the most reliable results.
Mathematically, using the exponential function accentuates the differences between
measures that have widely different utilities, while maintaining a closer balance between
measures that have similar utilities. Note that the methodology poses this mathematical
relationship as a model to estimate travel behavior; the relationship is not developed from
a large sample of empirical data.
Equation 4-5 illustrates how to calculate percentages of mode travel for a person with TU
values TUj through TUN:
£(6.S)TUk
Pk = N ,(6.5)717,. <4-5>
where
Pk = the probability that a person from the target group (or the
percentage of the target group) will travel via mode k relative to the
N modes examined.
Note that if all possible modes available to the target population are not examined, this
probability (or percentage) captures the percent of people who will travel via mode k out
of the total population traveling only via modes 1 through N.
It is important to note the methodology assumes that the relationship between the
percentage of time a person travels via mode k does not depend linearly on TUk.
When validating a base case, if the analyst is not reaching reasonable agreement with
actual mode splits, then we recommend first reconsidering the choice of X's (because of
their greater degree of uncertainty). If after adjusting the X's, the analyst still has
difficulty getting the methodology to agree with actual mode splits, we then recommend
reexamining the best and worst limits chosen in step 2 (keeping in mind that the mode
values should not cluster toward 1.0 or 0.0).
Step 7: Determine the new mode values (cost, time, convenience, and reliability) that
reflect the implementation of a TCM package. With these new values and the weight
profile determined from the base case, repeat steps 2-6 above to determine the new
percentages of time people will travel via the different N modes of travel (once the TCMs
are implemented). The new percentages are the results of the "control case" and the old
percentages are the results of the "base case".
92093.02 4-14
-------
At this point in the analysis, it is possible to introduce a new travel mode that was not
previously available to the target audience (such as telecommuting). In this case,
depending on the total utility of the new mode, one might expect a decrease in all
previously available modes from members of the target population choosing to use the
new mode.
Step 8: After repeating Steps 2-6 with new model values reflecting the TCM package,
compare the differences in the percentage of people that will travel via each mode after
the implementation of a TCM package. This is the difference between the control case
and the base case, and reveals the potential impact of each measure in the TCM package.
EXAMPLE APPLICATION OF PACKAGING METHODOLOGY
The first task in applying the methodology involves calibrating the model against base-
case values using "real world" data. Using mode split and travel time information for
work trips to the city of San Francisco as our target values (MTC, 1991), a FORTRAN
program was used to calculate total utility and percent use using the formulas presented
earlier in this chapter. The analyst may find that using a spreadsheet or other program
will make it easier to quickly evaluate the impact of altering attribute values on the total
value and percentage use of each TCM.
The second task in applying the methodology involves estimating the changes that a TCM
package would have on the attribute values of the different modes of transportation. For
this task, we assumed a 1 % increase in the cost of transit and then compared the results
to the elasticity of price with respect to demand calculated in MTC's three-mode work
mode choice model (see Figure 4-8).
Assumptions and Calculations for the Base Case Simulation
of the City of San Francisco
To establish base case conditions and "validate" the model approach, this example
simulates work trips to the city of San Francisco using 1987 statistics of the available
modes of travel, their travel times, and the percentages of mode splits (MTC, 1991).
This data established base conditions for three possible mode choices: SOVs, public
transit, and ridesnaring.
The travel times are calculated as follows: the average SOV trip length to the city is
11.33 miles, and the average peak period speed is 24.63 mph (MTC, 1991). This
corresponds to 0,46 hours travel time per trip. Assuming that carpoolers travel an
additional 2 miles per trip, their average travel time is assigned 0.54 hours. The average
work trip vk transit is 18.24 minutes (MTC, 1991). Adding an additional 8.76 minutes
for walking and waiting time (27 minutes altogether), the average transit time assigned is
0.45 hours.
92093.02 4-15
-------
The out-of-pocket travel costs for a SOV are calculated from the following: the 1987
cost of gasoline is assumed to be SLOG per gallon (1987 prices, MTC, 1991), and the
average fuel economy of light duty vehicles is 23,4 miles per gallon (from MOBILE4.1
data). The average parking price (calculated by averaging parking prices in different city
zones from MTC, 1991) is assumed to be about $.37 per hour, or $3.00 per day. The
average bridge toll is assumed to be $1.00. Then, the cost per trip is calculated as:
(distance)*($ per gallon) / (fuel economy) + parking fee + bridge toll
= (11.33)*(1.00/23.4) + 3.00 + 1.00
= $4.48 per trip.
The out-of-pocket travel costs for ridesharers is calculated as follows: the bridge toll is
assumed to be waived, but the parking price is not waived. The average vehicle
occupancy of carpools is 2,28 (MTC, 1991), and the costs are assumed to be equally split
among all members of the carpool. So the cost is calculated as:
[(distance)*($ per gallon) / (fuel economy) + parking fee] / (vehicle occupancy)
= [(13.33)*(1.00/23.4) + 3.00J/2.28
- $1.56 per trip
The out-of-pocket travel costs for transit riders is calculated by taking a weighted average
of the transit fees from different origins into the city. The fees are weighted by the
population traveling into the city from different origins (given by MTC, 1991), This
weighted average is $1.03 per trip. (Note that most of the work trips to San Francisco
originate in San Francisco, and so the transit cost is less than from outlying areas.)
The convenience of SOVs is rated high. It is not set equal to 1 but 0.8 because driving
in the city is congested and parking can be hard to find. The convenience of transit is
rated slightly lower than 0.5. The frequency of transit and the availability is reasonably
high in the city, but is less so in outlying areas. In addition, the transit rider may have to
walk to transit, which may be inconvenient during odd hours of the day or night,
lowering the convenience, so it is assigned the value 0.4. The convenience of ridesharing
is set equal to 0.3, lower than transit because ridesharers are restricted to specific travel
times set by the carpool. Immediate access is not readily available, and members of the
carpool may not have flexible schedules.
The actual 1987 mode splits (from MTC, 1991) for work trips into San Francisco are
39.8% SOV, 40.7% transit, and 19.5% rideshare. Figure 4-4 shows an example input
file for a simple FORTRAN program that applies the packaging methodology for this
base case simulation. The first approximation of the weight profile is chosen to be 0.3,
0,3, 0.2, and 0.2 for cost, time, convenience, and reliability, respectively.
Figure 4-5 shows the program output with different weight profiles. Rideshare, in the
first approximation, is assigned too large a percentage of the population. It is rated high
in cost and time, so the next weight profile is lowered in cost and time, but raised in
convenience and reliability to be 0.29, 0.29, 0.21, and 0.21. This reduces the gross
error from 3.8% in the first attempt to 1.08%. The next weight profile again lowers cost
92093.02 4-16
-------
4 # of titles
Base Case Simulation
Illustration of Calibration Technique and
Selection of Reasonable Weight Profiles for
The City of San Francisco.
base.out
3 # of modes
sov
Transit
Ridesiiare
0.00 8.0
dollars)
4.48
1.03
1.56
0.0
0.46
0.45
0.54
0.8
0.4
0.3
0.8
0.6
0.4
0.398
1.0
** BEST cost and WORST cost (in
0.407 0.195
cost of SOV
cost of transit
cost of rideshare
** BEST time and WORST time
time of SOV
time of transit
time of rideshare
convenience of SOV
convenience of transit
convenience of rideshare
reliability of SOV
reliability of transit
reliability of rideshare
(in hours)
FIGURE 4-4. San Francisco base case input file.
22093.03
4-17
-------
Mode Attribute Values:
Cost Time
SOV 0.440 0.540
Transit 0.871 0.550
Rideshare 0.805 0.460
Convenience
0.800
0.400
0.300
Reliability
0.800
0.600
0.400
Weight Profile:
Cost Time
0.300 0.300
Total Utilities:
SOV
Transit
Rideshare
Convenience
0.200
Reliability
0.200
MODE:
SOV 38.098
Transit 41.289
Rideshare 20.613
Mean Error: 1.135
Gross Error: 3.810%
0.614
0.626
0.520
PERCENTAGES:
Calculated Actual Split
39.800
40.700
19.500
%Difference
-4.276
1.448
5.707
Weight Profile:
Cost Time
0.290 0.290
Total Utilities:
SOV
Transit
Rideshare
Convenience
0.210
Reliability
0.210
MODE:
SOV 39.781
Transit 40.292
Rideshare 19.928
Mean Error: 0.285
Gross Error: 1.082%
0.620
0.622
0.514
PERCENTAGES:
Calculated Actual Split
39.800
40.700
19.500
%Difference
-0.048
-1.004
2.193
FIGURE 4-5. Base case simulation illustration of calibration technique and selection of
reasonable weight profiles for the city of San Francisco.
92093.03
4-18
-------
Weight Profile:
Cost Time
0.280 0.280
Total Utilities:
SOV
Transit
Rideshare
Convenience
0.220
Reliability
0.220
MODE:
SOV 41.488
Transit 39.270
Rideshare 19.242
Mean Error: 1.125
Gross Error: 3.025%
0.626
0.618
0.508
PERCENTAGES:
Calculated Actual Split
39.800
40.700
19.500
%Difference
4.240
-3.513
-1.323
Weight Profile:
Cost
0.310
Total Utilities:
SOV
Transit
Rideshare
Time
0.310
0.608
0.631
0.525
Convenience
0.190
Reliability
0.190
MODE:
SOV
Transit
Rideshare
Mean Error:
Gross Error:
PERCENTAGES:
Calculated Actual Split %Difference
36.443
42.261
21.296
39.800
40.700
19.500
-8.435
3.835
9.211
2.238
7.160%
FIGURE 4-5. Concluded.
92093.03
4-19
-------
and time and raises convenience and reliability. Now the gross error increases to 3.02%,
and so the previous weight profile is a better choice. Finally, the last weight profile
entered into the program is 0.31, 0,31, 0.19, and 0.19. The gross error corresponding to
this weight profile is yet larger than the gross error from the first weight profile, 0.3,
0.3, 0.2, 0.2. From this simulation, we determine that 0.29, 0,29, 0.21, 0.21 is the best
weight profile, and the model is calibrated for San Francisco work trips.
Control Simulation for the City of San Francisco
This control simulation reflects the change in mode split corresponding to a 1 % increase
in transit fares. This TCM has been selected so that the results of the simulation can be
compared with elasticity data calculated from a logit model based on a 1980/81 data base
in Harvey, 1989. Figure 4-6 shows an example input file for a FORTRAN program that
applies the packaging methodology for this control scenario, and Figure 4-7 shows the
program output. The last line of Figure 4-6 consists of the calculated mode splits
corresponding to the weight profile 0,29, 0.29, 0.21, 0.21 from the base case simulation.
The result shows a -0.14 percent change in transit use. The calculated elasticity from
Harvey, 1991 for a 40% mode share and a $1.00 base travel cost is -0.21. The elasticity
for a 50% mode share is -0.17. These elasticities are in the same general range,
especially given that -0.3 is a widely used transit fare elasticity (TTE, 1982) for a smaller
mode share. The national average transit ridership is around 7 percent, while San
Francisco's is about 41%.
Assumptions and Calculations for the Base Case Simulation
of the Maricopa County Metropolitan Area, Arizona
The travel characteristics of the Maricopa County metropolitan area come from the
Maricopa Association of Governments Freeway/Expressway Plan, 1987, and from MAG
TPO personnel (Howell, 1992).
The actual work-trip mode shares in the Phoenix area are: 77.4% SOV, 17.4%
rideshare, 1.6% city bus, 1% motorcycle, 1% walk, 1.5% bicycle. People who walk or
bicycle must live close to where they work (less than a couple of miles if they walk).
Because the average person in the metropolitan area commutes 10 miles to work, they do
not have the option of walking or riding a bicycle (given current land use), so these travel
modes are not included in this example analysis. The percentage of people who ride
motorcycles to work is insignificant in comparison to the SOV share and not substantially
different from SOVs, so we have also not designated motorcycles as a separate mode
from SOVs.
The actual mode shares have been renonnalized to represent the fraction of people who
travel by SOV, transit, or rideshare out of the number of people who previously travel by
SOV, transit, and rideshare. These new fractions are: 80.3% SOV, 18.04% rideshare,
and 1.66% transit. Note that these three fractions sum to 100%.
92093.02 4-20
-------
2 # of titles
Control Simulation
1.0% Increase in Transit Fares in San Francisco
control.out
3 # of modes
sov
Transit
Rideshare
0.00 8.0
4.48
1.04
1.56
0.0
0.46
0.
0,
1.0
0,
0,
.45
,54
,8
,4
0.3
0.8
0.6
0.4
.39781
.40292
** BEST cost and WORST cost
cost of SOV
cost of transit
cost of rideshare
** BEST time and WORST time
time of SOV
time of transit
time of rideshare
convenience of SOV
convenience of transit
convenience of rideshare
reliability of SOV
reliability of transit
reliability of rideshare
.19928
(in dollars)
[in hours;
FIGURE 4-6. Control scenario input file for the city of San Francisco.
92093.03
4-21
-------
Mode Attribute Values:
Cost Time
SOV 0.440 0.540
Transit 0.870 0.550
Rideshare 0.805 0.460
Convenience
0.800
0.400
0.300
Reliability
0.800
0.600
0.400
Weight Profile:
Cost
0.290
Total Utilities:
SOV
Transit
Rideshare
Time
0.290
0.620
0.622
0.514
Convenience
0.210
Reliability
0.210
MODE:
SOV
Transit
Rideshare
Mean Difference:
Gross Difference:
PERCENTAGES:
Calculated
39.819
40.235
19.947
0.038
0.110%
Base Case
39.781
40.292
19.928
%Difference
0.094
-0.142
0.093
FIGURE 4-7. Control simulation 1.0% increase in transit fares in San Francisco.
92093.03
4-22
-------
Demand Elasticities with Respect to Travel Cost
Base Mode Share
Bese travel 0.10 0.20 0.30 0,40 0.50 0.60 0.70 0.10 0.90
cost (cents)
50
100
150
200
250
300
350
400
450
500
-0.11
-0.31
-0.4?
-0.63
•0.71
-0.94
-1.10
•1.25
•1.41
-1.57
-0
-0
-0
-0
-0
-0
-0
"1
-1
"1
.14
.28
.42
.56
,70
.S3
.97
.11
.25
.39
-0.12
•0.24
•0.37
-0.49
-0.61
-0.73
-0.15
-0.57
-1.10
-1.22
-0.10
-0.21
-0.31
-0.42
-0.52
-0.63
-0.73
•0.33
-0.94
-1.04
•0.09
•0.17
•0.26
-0.35
-0.43
-0.52
-0.61
•0.70
-0.71
-0.17
-0
-0
-0
-0
-0
-0
•0
•0
-0
-0
.07
.14
.21
.21
.35
.42
.49
.16
.63
.70
•0.05
-0.10
-0.16
•0.21
-0.26
•0.31
•0.37
-0.42
-0.47
•0.52
•0.03
-0.07
-0.10
-0.14
-0.17
-0.21
-0.24
•0.21
-0.31
-0.35
-0.02
-0.03
-0.05
-0.07
-0.09
-0.10
-0.12
-0.14
•0.16
-0.17
FIGURE 4-8. Demand elasticities with respect to travel cost,
(Source: Creig Harvey, "Screening of Transportation Control
Measures for the San Francisco Bay Area," working paper for
the Metropolitan Transportation Commission, 1989).
4-23
-------
The travel times for the three modes are calculated as follows: the average commute
distance for SOVs is 10 miles, and the average speed is 27 mph. This corresponds to an
average commute time of 0.37 hours. Assuming that ridesharers travel an additional 3
miles (to pick up the other members of the carpool), for a total of 13 miles, and that their
average speed is also 27 mph, we calculate an average commute time of 0.48 hours. The
actual average transit trip is 6 miles in the area. However, in order to compare the travel
time per mile of transit to the other modes, we consider the travel time of a transit trip
over 10 miles. Assuming that the average speed of a city bus is 18.3 mph (from San
Francisco Bay Area data, MTC, 1991), the travel time for 10 miles is 33 minutes.
Because city buses do not run frequently throughout Phoenix, an additional 20 minutes
was added to the 33 minutes to account for waiting for the bus and walking to the bus
stop. This is a total of 53 minutes, or 0.88 hours.
The out-of-pocket costs of traveling using a SOV are calculated from the following: the
cost of gasoline is currently about $1.23 per gallon. The average fuel economy is
assumed to be 23.4 mph (MOBILE4.1 data). Parking costs are rare in the metropolitan
area, and parking is abundant. There are no bridges, and hence, no tolls. The cost of a
commute trip, is then:
(distance) *($ per gallon)/(fuel economy)
= 10*(1.23/23.4)
= $0.52
The only difference in cost for ridesharers is that the distance is slightly longer, and that
the members of the carpool split the cost. Assuming the vehicle occupancy of a carpool
is 2, the cost of a rideshare trip is:
[(distance)*($ per gallon)/(fuel economy)] / (vehicle occupancy)
= 13*(1 -23/23.4)7 2.0
= $0.34
The cost of transit is assumed to be the cost of a bus fare, $0.85.
Figure 4-9 shows an example input file for a FORTRAN program that applies the
packaging methodology for the base case scenario, and Figure 4-10 shows the output
from the simulation. In this simulation, the error is minimized when the weight profile is
0.28, 0.28, 0.22, and 0.22 for cost, time, convenience, and reliability, respectively. The
gross error consists mostly of the error in trying to simulate the low transit share. Even
though the model calculates only 2.4 percent transit ridership, the percentage difference
between 2.4 and 1.66 is 44.6 percent because 1.66 is a small number. Note that the
mean error is only 0.494, which is extremely small. Given that the mean error is this
small, the weight profile 0.28, 0.28, 0.22, and 0.22 is satisfactory for the base case
calibration.
92093.02 4-24
-------
2 # of titles
Base Case Simulation
Maricopa County Metropolitan Area
baseM&G.out
3 # of modes
SOV
Rideshare
Transit
0,00 1,00 ** BIST cost and WORST cost
0.52 cost of SOV
0.34 cost of Rideshare
0.85 cost of transit
0.0 1.5 ** BEST time and WORST time
0.37 time Of SOV
0.48 time of rideshare
0.88 time of transit
i. 0 convenience of SOV
0.4 convenience of rideshare
0,1 convenience of transit
0.9 reliability of SOV
0.3 reliability of rideshare
0.2 reliability of transit
0.803 0.1804 0.0166
;in dollars!
(in hours;
FIGURE 4-9. Base Case Input File for the Maricopa County Base Case Simulation
92093.03
4-25
-------
Mode Attribute Values:
Cost Time
SOV 0.480 0,753
Rideshare 0.660 0.680
Transit 0.150 0.413
Convenience
1.000
0.400
0.100
Reliability
0.900
0.300
0.200
Weight Profile:
Cost
0.300
Total Utilities:
SOV
Rideshare
Transit
Time
0.300
0.750
0.542
0.229
Convenience
0.200
Reliability
0.200
MODE:
SOV
Rideshare
Transit
Mean Error:
Gross Error:
PERCENTAGES:
Calculated Actual Split
77.366 80.300
20.016 18.040
2.617 1.660
1.956
24.088%
tDifference
-3.653
10.956
57.655
Weight Profile:
Cost
0.290
Total Utilities:
SOV
Rideshare
Transit
MODE:
SOV
Rideshare
Transit
Mean Error: 1.019
Gross Error: 18.926%
Time
0.290
Convenience
0.210
Reliability
0.210
0.757
0.536
0.226
PERCENTAGES:
Calculated Actual Split %Difference
78.771 80.300 -1.904
18.720 18.040 3.772
2.508 1.660 51.101
FIGURE 4-10. Base case simulation for Maricopa County metropolitan area.
92093.03
4-26
-------
Weight Profile:
Cost
0.280
Total Utilities:
SOV
Rideshare
Transit
MODE:
SOV
Rideshare
Transit
Mean Error: 0.494
Gross Error: 15.983%
Time
0.280
Convenience
0.220
Reliability
0.220
0.763
0.529
0.224
PERCENTAGES:
Calculated Actual Split %Difference
80.110 80.300 -0.236
17.488 18.040 -3.058
2.401 1.660 44.65S
Weight Profile:
Cost
0.270
Total Utilities:
SOV
Rideshare
Transit
Time
0.270
0.770
0.523
0.221
Convenience
0.230
Reliability
0.230
MODE:
SOV
Rideshare
Transit
Mean Error:
Gross Error:
PERCENTAGES:
Calculated Actual Split %Difference
81.384
16.320
2.296
80.300
18.040
1.660
1.350
-9.536
38.335
1.147
16.407%
FIGURE 4-10. Concluded.
92093.03
4-27
-------
Control Simulation for the Maricopa County Metropolitan Area, Arizona
Figures 4-11 and 4-12 show the input file and output files for the control scenario. A 1 %
increase in transit fares is assumed. The calculated mode split is taken from the base
case results with the weight profile (0.28, 0.28, 0.22, 0.22). The percent change in
transit ridership is -1.75. The elasticity from Harvey, 1989, for a 10% mode share and
$1.00 base travel cost is -0.31. The elasticities in the Harvey report increase, however,
both with respect to an increase base cost and with respect to a decrease in base mode
share. In this scenario, the base mode share is substantially less than 10%. In addition,
the elasticities in the Harvey report do not take into account the costs of other travel
modes available to the study population relative to transit. Considering the substantial
difference in the perceived cost of transit compared to ridesharing and SOV, the predicted
-1.75% change in transit ridership is reasonable.
Summary
Overall, the model reflects the general trends one would expect from the increase in
transit fares in two very different cities. The modal shares in San Francisco and Phoenix
could hardly be more different, yet the model replicates each city fairly well. The
predicted elasticity in San Francisco is small, in accordance with the large base case
mode share, and the relatively similar cost of transit and rideshare. The predicted
elasticity in Maricopa County is quite large, reflecting the large difference in the
perceived (i.e., out-of-pocket) base case costs, and the substantially lower total utility
measure of transit with respect to SOV and rideshare.
As a final note, it should be stressed that this approach is very new and has not been
extensively tested for other urban areas or for different sets of TCMs. It is likely that the
model will evolve over time as it is applied in more situations. However, the analytical
framework it provides is expected to prove a useful tool for TCM evaluation.
92093.02 4-28
-------
2 # of titles
Control Scenario
1.0% Transit Fare increase in the Maricopa County Metropolitan Area
cntlM&G.out
3 # of modes
sov
Rideshare
Transit
0.00 1.00
0.52
0.34
0.86
0.0 1.5
0.37
0.48
0,88
1.0
0.4
0.1
0.9
0.3
0.2
.80110 .17488
** BSST cost and WORST cost
cost of SOV
cost of Rideshare
cost of transit
** BEST time and WORST time
time of SOV
time of rideshare
time of transit
convenience of SOV
convenience of rideshare
convenience of transit
reliability of SOV
reliability of rideshare
reliability of transit
.02401
I in dollars)
[in hours)
FIGURE 4-11. Control scenario input file for the Maricopa County metropolitan area.
92093,03
4-29
-------
Mode Attribute Values:
Cost Time
SOV 0.480 0.753
Rideshare 0.660 0.680
Transit 0.140 0.413
Convenience
1.000
0.400
0.100
Reliability
0.900
0.300
0.200
Weight Profile:
Cost
0.280
Total Utilities:
SOV
Rideshare
Transit
MODI:
SOV
Rideshare
Transit
Mean Difference:
Gross Difference:
Time
0.280
0.763
0.529
0.221
PERCENTAGES;
Calculated
80.145
17.496
2.359
0.028
0,613%
Convenience Reliability
0.220 0.220
Base Case
80.110
17.488
2.401
%Difference
0.044
0.046
-1.749
FIGURE 4-12. Control scenario 1.0% transit fare increase in the Maricopa County
metropolitan area.
92093,03
4-30
-------
References
Austin, B. A., D. Eisinger, L. Duvall, S. Shepard, J. Heiken, and J. Fieber, 1991.
"Transportation Control Measure Analysis Procedures," Final Report, Prepared for the
California Air Resources Board, Mobile Sources Division by Systems Applications
International and K,T. Analytics, Inc., 25 November 1991.
CAAA. 1990. (Clean Air Act Amendments of 1990) 42 U.S.C. 7401, PL 101-549, November
15, 1990.
Grain, J., et al. 1984. "Santa Clara County Solo Driver Commuters: A Market Study."
Prepared for Santo Clara County Transportation Agency and Santa Clara County Goals
Review Committee by Grain and Associates, Inc.
CSI. 1979. Transportation Air Quality Analysis Sketch Planning Methods. Volume 1
Analysis Methods. Prepared for Environmental Protection Agency, Office of
Transportation and Land Use Policy by Cambridge Systematics, Inc. (NTIS PB80-
158702), December 1979.
DOT. 1981. Traveler Response to Transportation System Changes, Second Edition, prepared
for the U.S. Department of Transportation (Contract DOT-FH-11-9579) by Barton-
Aschman Associates, Inc., R.H. Pratt and Co. Division. My 1981.
DOT. 1985. "Personal, Social, Psychological and Other Factors in Ridesharing Programs."
Final Report. Prepared by Morgan State University, Center for Transportation Studies,
Baltimore, Maryland. January 1984.
Eisinger, D. S., S. B. Shepard, L. L. Duvall, and B. S. Austin. 1991. "Analyzing
Transportation Control Measure Packages." Presented to the Air and Waste Management
Association/U.S. EPA Specialty Conference; Emission Inventory Issues of the 1990s.
Durham, North Carolina, September 9-12, 1991.
EPA. 1981. Procedures For Emission Inventory Preparation, Volume IV: Mobile Sources.
U.S. Environmental Protection Agency (EPA-450/4-81-026d), July 1989 (Revised version
pending).
EPA. 1991. User's Guide to MOBILE4.1 (Mobile Source Emission Factor Model), U.S.
Environmental Protection Agency (EPA-AA-TEB-91-01).
92093.09
-------
Feeney, B. P. 1989. A review of the impact of parking policy measures on travel demand.
Transportation Planning and Technology, 13:229-244.
Fieber, J. L., L. Duval, and D. Eisinger. 1992. Systems Applications International, E,
Gimnzow, J. Coxey, Urban Analysis Group. "Approaches to Improving Travel Demand
Modeling for Air Quality Analysis." January 14,1992 (SYSAPP 92/024). Prepared for
EPA OAQPS and OMS.
Harvey, G. 1989. "Screening of Transportation Control Measures for the San Francisco Bay
Area - Part 1: Methodology." Submitted to Metropolitan Transportation Commission,
Oakland, California. November 1989.
Harvey, G., and B. Deakin. 199L Draft Report "Toward Improved Modeling Practice,"
Prepared for National Association of Regional Councils. November 1991.
Hutebinson, B. G. 1974. "Principles of Urban Transport Systems Planning", Scripts Book
Company.
Ismail, D. 1991. "Travel Demand Forecasting Limitations for Evaluating TCMs," Presented
at 1991 Air and Waste Management Annual Meeting (Paper 91-87.1).
Kitamura, R., J. M. Niles, P. Conroy, and D. M. Fleming. 1989. "Telecommuting as a
Transportation Control Measure: Initial Results of State of California Pilot Project."
Presented at Transportation Research Board 69th Annual Meeting, 7-11 January 1990,
Washington, D.C.
Maltzman, F. 1987. "Casual Carpooling: An Update." Prepared for RIDES for Bay Area
Commuters, Inc., June 1987.
MTC. 1990. "Bay Area Travel Forecasts: Year 1987 Trips by Mode, Vehicle Miles of Travel
and Vehicle Emissions - Technical Summary." August 1990. Metropolitan
Transportation Commission, Oakland, California
MTC. 1991. "Bay Area Travel Forecasts - Congestion Management Program Databook #1:
Regional Summary." March 1991. Metropolitan Transportation Commission, Oakland,
California
Ornelas, L. 1990. Sacramento Metropolitan Air Quality Management District. "Sacramento
Air Quality Attainment Plan Volume Five: Transportation Control Measures Program."
Randall, P., and A. Diamond. 1990, "Air Quality Analysis Tools (AQAT-3): Computer
Models to Determine the Emission Impacts of Various General Development Projects and
Mitigation Measures." California Air Resources Board Stationary Source Division.
Raux, C. 1986. "Employment, Childcare, and Travel Behavior: France, the Netherlands, the
United States." Laboratoire d'Economie des Transport and Sandra Rosenbloom,
University of Texas at Austin. Behavioral Research for Transport Policy.
92093.09 R-2
-------
RIDES. 1990. "RIDES' 1990 Database Survey." D. Bureh, RIDES for Bay Area
Commuters, Inc., April 1990.
Schewe, et al. 1990. Guideline For Modeling Carbon Monoxide From Roadway Intersections.
Prepared by PEI Associates for the U.S. Environmental Protection Agency, October
1990,
Sierra Research. 1991. "Users Guide to the Transportation Module." Prepared by JHK and
Associates for the San Diego Association of Governments and Siena Research, Inc.
March 1991, Sierra Research.
Stokey, E., and R. Zeekhauser. 1978. A Primer for Policy Analysis. W. W. Norton &
Company, New York.
Wachs, M. 1990. Transportation demand management: Policy implications of recent
behavioral research. Submitted to the Journal of Planning Literature. March 1990.
Willson, R., and D, Shoup. 1990, "The Effect of Employer Paid Parking in Downtown Los
Angeles." Prepared for SCAG, May 1990.
WSEO. 1989.
USDOC. 1988. County and City Data Book 1988. U.S. Department of Commerce, Bureau
of the Census.
UMTA. 1985. "National Ridesharing Demonstration Program: Comparative Evaluation
Report." U.S. Department of Transportation, Urban Mass Transportation
Administration, Office of Technical Assistance.
Valdez, R., and C. Aree. 1990. Comparison of Travel Behavior and Attitudes of Ridesharers,
Solo Drivers, and the General Commuter Population." Transportation Research Record
No. 1285. Transportation Forecasting 1990.
92093.09 R_3
-------
(This page intentionally blank for 2-sided printing)
-------
APPENDIX A
SUMMARY OF RECENT TCM METHODOLOGIES DEVELOPED IN CALIFORNIA
-------
APPENDIX A
SUMMARY OF RECENT TCM METHODOLOGIES DEVELOPED IN CALIFORNIA
Spurred by TCM requirements in the California Clean Air Act and the Clean Air Act
Amendments of 1990 as well as continuing nonattainment problems, many California agencies
have developed TCM methodologies for use in air quality planning. This appendix reviews a
number of these methodologies in order for the reader to better understand the range and type
of methodologies that are being developed and used for TCM analysis. The appendix covers
five methodologies developed for (1) the San Diego Association of Governments and the
California Department of Transportation (SANDAG/Caltrans), (2) the Sacramento Metropolitan
Air Quality Management District (SMAQMD), (3) the San Francisco Bay Area Metropolitan
Transportation Commission (MTC), (4) The California Air Resources Board (ARE) Technical
Services Division, and (5) the ARE Mobile Source Division, Each summary presents a brief
description of the methodology, and a preliminary assessment of its strengths and weaknesses.
SANDAG/CALTRANS
Responding to the requirements of the California Clean Air Act (CCAA), the California
Department of Transportation (CALTRANS) provided a grant to the San Diego Association of
Governments (SANDAG) to study the relationship between transportation control measures
(TCMs) and emissions reductions. Guided by a state-wide Steering Committee, Sierra Research
and JHK & Associates (Sierra/JHK, 1991) developed a three-part methodology for quantifying
the travel, emissions, and cost impacts of different TCMs. The entire set of methodologies were
then incorporated into a PC-compatible software package.
The three modules of this methodology include a transportation module, emissions module, and
a cost-effectiveness module. The transportation module is designed to estimate the effect of
selected TCMs on trips, vehicle miles of travel (VMT), and speeds. There are 25 pre-defined
TCMs included in the software and user options to define five additional measures not included
as defaults. Ongoing work slated for completion at the end of 1992 is expected to include
development of three additional non-work related TCM methodologies (Valerio, 1992). Local
estimates of travel activity are combined with assumptions about how travelers will respond to
individual TCMs in a Lotus spreadsheet program. This spreadsheet produces a summary of the
baseline travel characteristics and the effects of each TCM on peak and off-peak period trips,
VMT, and speed. These outputs are used as inputs to both the emissions and cost-effectiveness
modules.
The emissions module consists of a computer program (written in FORTRAN) which combines
the TCM-specific travel impacts (calculated by the transportation module) with the emission
factor data contained in the EMFAC7E and BURDEN models (specific to California) and
selected default parameters (defined by Sierra and JHK) to create a baseline emissions estimate
that includes reactive organic gases (ROG), carbon monoxide (CO), nitrogen oxide (NOJ, and
A-2
-------
participates (PM). This baseline estimate is then used to determine the emissions impacts of
each TCM evaluated. The two output files from this module summarize the pollutant-specific
percentage reductions for each TCM (used by the cost-effectiveness module) and a print file that
contains information summarizing the specific run, the baseline emission estimates, tilt emissions
estimates after each TCM is implemented, and a pollutant-specific percentage reduction for each
TCM (Sierra, 1991).
The cost-effectiveness module is a Lotus spreadsheet program which uses the travel impact
information from the transportation module and the percent emission reductions generated by
the emissions module, in combination with additional user-supplied information, to calculate the
costs and cost-effectiveness of each TCM to be evaluated. The user-supplied information
includes baseline parameters (e.g. year, study area, pollutants of interest and daily emissions
totals for each of these pollutants) and default parameters that include basic cost per unit data
and other parameters. These default parameters were developed by Sierra based upon a survey
of transportation planning and other agencies in California. The user's guides for each of these
three modules all stress the importance of customizing default values to better characterize the
region being studied.
Methodology Application
The following summarizes the procedure for evaluating a TCM, in this case ridesharing, using
the Sierra/JHK methodologies.
Step 1: Transportation Module
There are three types of data required by the spreadsheet program:
Baseline Travel Characteristics - These define the baseline travel patterns for the analysis
year and for the region for which the TCM is being evaluated; for instance, all examples
given in the user's guide for this module use values for the San Diego County area in the
year 2010. These parameters include (for ridesharing) drive alone share of commute
trips; total commute (person) trips; percent of commute trips in peak period; average
commute trip length; and total peak and off-peak VMT. The values in the spreadsheet
are default values: user's have the option to input values more specific to their region
of interest.
TCM Specific Parameters - These factors are supplied by the user and for ridesharing
would include percent increase in non-drive-alone modes; percent of the maximum VMT
realized due to circuity of ridesharing or access to transit; average carpool size; and
percent of employees affected.
Assumptions - For ridesharing, the assumptions embedded in the spreadsheet program
include a value for the elasticity of speed with respect to volume. The default value for
this was developed by Sierra/JHK based upon the San Diego County region.
A-3
-------
The calculations performed by the spreadsheet for ridesharing include:
Reduction in trips (person trips, peak, off-peak, and total)
Reduction in VMT (peak, off-peak, and total)
Percentage change in speeds (peak and off-peak)
Output from this module includes a summary of these travel effects for each TCM evaluated,
including an output file suitable for input to both the emissions and cost-effectiveness modules.
Step 2: Emissions Module
This module is designed to estimate the influence of selected TCMs (for this example,
ridesharing) on mobile source emissions. Inputs to this module include both required and
optional values. Required inputs include air basin of interest (in California), county, year, TCM
data file name (generated by the transportation module), I/M indicator (indicates whether
emission factors are applied with or without I/M credits), and the output file name. Optional
inputs include the file name for an ambient temperature profile and eight sets of input data that
the user can specify that will replace the default values of the program. These default values
include travel information (e.g. speeds, trip fractions, vehicle fleet mix) that can be specific to
the user's region of interest.
After all necessary information is supplied either by the user or the default values, the program
proceeds to calculate the emissions reductions for each TCM of interest. Upon completion of
this step, two output files have been created: a summary report file containing the tabulated
baseline and post-implementation emissions and an emissions reduction file suitable for input into
the cost-effectiveness module.
Step 3: Cost Effectiveness
This spreadsheet program uses values calculated by the other two modules in combination with
user-specified inputs and an extensive set of default values to determine the cost and cost-
effectiveness of each TCM. The reason for the use of the multitude of default values is due to
the nature of these values and the extensive amount of data required for region-specific
applications. Data such as plan preparation cost, administrative costs, O&M costs, etc., make
this module the most data-intensive of the three. User's are advised that "..it is critical that
users identify and document the use of proper cost date and other information for their particular
area."
After the spreadsheet calculations are complete, the user may choose to view the results or
generate printouts of these results.
Strengths and Weaknesses of Methodology
A-4
-------
The Sierra/JHK methodology provides a relatively easy-to-use software package that is based
upon extensive survey information, literature reviews, and professional expertise. It provides
the user with a convenient tool for evaluating the consequences of their assumptions about travel,
operating conditions anJ costs on emissions and cost-effectiveness. It represents a collaborative
effort between the air quality and transportation communities to develop a comprehensive model
of TCM effects. However, there are some weaknesses:
• In order to apply these methods, the user must have access to a PC-compatible computer
with the LOTUS 1-2-3 software installed.
• The extensive use of default values indicates the region-specific nature of these
calculations. Although the developers of the user's guide for the methodology frequently
point out the importance of customizing these default values to better describe the region
of interest, little guidance is given as to the procedure for obtaining or calculating the
necessary values (e.g. elasticities). Many areas may not have access to the extensive
information used by Sierra/JHK to generate the default values. In these cases, use of the
defaults may result in inaccurate estimates of TCM-related emissions reductions and cost-
effectiveness.
• The documentation states that the spreadsheet for the transportation module requires user-
input of information, such as jobs/housing balance or urban density, for only a few
TCMs (due to the area-specific nature of these variables). However, more than half of
the TCMs require user input of TCM effectiveness (i.e. the user enters the percent
increase in non-drive alone modes). The appearance that the system calculates these for
the user (because it translates such inputs into trip or VMT changes) may encourage
misuse of the system.
e The system may incorrectly estimate effects on start emissions. For example, it assumes
that the trip reduction is directly related to the number of new non-drive alones in the
ridesharing example referred to above.
• Calculations of travel and emissions impacts do not sufficiently consider potentially
offsetting effects; for example,
For telecommuting there is no accounting of the potential increase in non-work
related trips by the telecommuter or a member of the telecommuter's households
as a result of increased vehicle availability. The model calculates an average
change in travel activity for a weekday but does not account for the existence of
"favored days" (e.g. more people telecommute on Wednesday than on Friday),
which could make the daily estimates differ. Further, not all people work five
days per week, while others often and/or regularly work more than five days per
week. Commuting to satellite centers is also not covered, which means that the
authors assume that all telecommute days reduce work trips (and therefore cold
starts). Finally, some individuals working in occupations that would be
A-5
-------
appropriate for telecommuting may also be participating in compressed work
week programs.
In ridesharing, a default percentage for the realization of potential VMT
reductions (e.g. 80% of potential reductions will be realized when accounting for
circuity of ridesharing routes or distance to transit), but little guidance is offered
regarding the methods used to arrive at this value for the user to draw upon when
calculating a value specific to the region of interest.
For staggered work hours/flextime, mode changes that could result, such as
decreased or increased ridesharing, are not calculated. Changes in non-work trips
latent demand, or peak spreading (the length of the peak period increases) are not
addressed. The potential for shifting peak trips out of the peak for areas such as
Los Angeles where the peak period is defined as running from approximately 6
a.m. to 10 a.m. is not addressed. Further, these TCMs can function to move
trips only partially out of the peak period rather than completely out of the peak.
This methodology is designed to evaluate the emissions impacts and cost-effectiveness
of individual TCMs. While it addresses numerous public and private costs and their
relationship to individual driving patterns, there are a number of key issues such as latent
demand and effects on non-work travel that are not addressed. Also, although the user's
guide includes guidance for qualitatively evaluating interactions among groups of
measures, the software does not include a consideration of the effects of combining
various measures. User's are cautioned tihe effects of these combinations are not likely
to be additive, but little information is given as to the benefits and shortcomings of these
combinations.
SACRAMENTO METROPOLITAN AIR QUALITY MANAGEMENT DISTRICT
The 1991 Air Quality Attainment Plan for the Sacramento Metropolitan Air Quality Management
District (SMAQMD) contains a program for the evaluation and subsequent inclusion of selected
transportation control measures (TCMs) in the Plan. Volume Five of the Plan (SMAQMD,
1991) describes the methodology used for this TCM evaluation. The work done for this Plan
is characterized as a starting point for a series of predicted improvements to be implemented
when developing future plans.
The first step in TCM selection begins with the identification of possible TCMs. From a series
of "brainstorming" sessions of the Technical Advisory Committee, a list of 317 potential TCMs
was assembled. This list was then initially screened using a two-dimensional matrix that
considered the effectiveness and feasibility of each TCM. The top 30% from this screening
process were immediately added to a list for future evaluation. The bottom 30% were
immediately discarded, and the remaining measures were re-evaluated, with the top measures
A-6
-------
from this evaluation being included on the list for additional evaluation (in this case, 122
measures in all were included on this list)
Subsequent screening for redundancy and duplication resulted in a final list of 38 measures for
consideration. These 38 were subdivided into three implementation "terms": near-term (1991-
1993), mid-term (1994-1996), and long term (1997-2010). In this study, 20 of the TCMs were
designated "near-term" TCMs, with the remaining 18 classified as mid- or long-term measures
for future consideration. It should be noted that the methodology used for this Plan focused only
upon the near-term measures.
Modeling Methodology
The modeling methodology used by the District to evaluate the candidate near-term TCMs
involves three parts: transportation impacts, emissions impacts, and cost-effectiveness. A
combination of computer models, combined with other qualitative analyses, was used to conduct
this evaluation. It should be noted that for the District's application, all candidate near-term
TCMs were evaluated as a package of measures. Each TCM was evaluated for
transportation/air quality impacts based upon the assumption that the total package would be
adopted and implemented according to a pre-determined schedule.
The transportation impacts evaluated for each candidate TCM include cold start trips, hot start
trips, vehicle miles travelled (VMT), idling time, average speed, and time of day. Additionally,
each measure was considered in terms of the number of GRACE travel markets affected. The
term GRACIE is an acronym for six travel markets that TCMs could potentially affect: Goods
Movement, Recreation, Activity Center, Commercial, Institutional, Employment. The
effectiveness of each measure varied with the number of travel markets affected. The first
computer model, TCMARK, determines the scope of TCMs in each of the GRACIE travel
markets. Each of these markets has some unique characteristics that tend themselves to different
sets of TCMs, such as demographics, trip lengths, time, parking price, and mode shift
characteristics (SMAQMD, 1991). Each of the near-term TCMs were run through this model
and assigned a "yes" or "no" rating for each of the six GRACIE markets based upon whether
or not the measure would impact that market.
TCMPACT is the second component in the system. Each candidate TCM is qualitatively ranked
either positive or negative on a scale of 1-6 (each number representing a range of emission
reductions in percent) based on its impact on each of these emissions categories. A negative
ranking indicates that the measure would increase an emission source. It is possible for a TCM
to have negative impact on a particular emission category, but be determined to have a positive
overall impact on emission reductions.
The TRAVDEM component of the modeling methodology is the travel demand forecasting
model used to evaluate the transportation and cost-effectiveness impacts of each TCM. This
model incorporates information regarding trip purposes, modes, and specific figures for number
of trips (both person and vehicle) and VMT to produce estimates of average trip time, average
A-7
-------
trip distance, and mode split for each of four trip purposes for each year between 1989 and
2010. Travel characteristics for each mode are then converted to regional vehicle trips and
VMT by trip purpose. These regional figures are then converted to GRACIE travel markets and
market shares.
EMISSION is the estimation model that produces the planning inventory of on-road mobile
source emissions for the SMAQMD region. This model is based upon data supplied from the
EMFAC/BURDEN models (specific to California). For each of five modeling years
(1987,1991, 1997, 2000, 2010) a total daily planing inventory of emissions of ROG, NOX, and
CO were identified separately and assigned to specific travel aspects: VMT emissions, cold-start
emissions, hot-start emissions, and hot-soak emissions. Values for each of these emission
categories were then interpolated for each year between 1987 and 2010, ultimately yielding
estimates for each emission category. It is unclear from the documentation whether temporal
variations in travel caused by TCMs are addressed.
The final component of the TCM modeling methodology is the calculation of net-present value
for each TCM of interest. Each of the candidate measures are ranked in terms of cost per unit
pollutant reduced ($/ton per day) in 1987 dollars, calculated using a simple spreadsheet program
that incorporates output from the other model modules described above.
In addition to the modeling methodology summarized above, an additional qualitative evaluation
was made to determine the technical feasibility and public acceptance of each candidate measure.
This analysis was based upon the professional judgment of the analysis team, combined with
information regarding TCM implementation in other regions.
Strengths and Weaknesses of the Methodology
The TCM evaluation methodology developed by the District provides a starting point for the
future development of a more comprehensive program. It represents a first-step in determining
which combination of TCMs would be most effective in achieving desired mobile source
emissions reductions in the SMAQMD region. Because of the interim nature of this
methodology, however, there are areas for improvement which can be identified:
• There are numerous places in the methodology where a lack of sufficient, accurate data
are identified. Information or quantitative estimates regarding TCM effectiveness, both
as individual measures and packages, for the SMAQMD region would result in more
accurate estimates of TCM impacts. Additionally, a lack of up-to-date travel activity
information, including the relationships between vehicle technology and trip purpose or
GRACIE travel market, for the six GRACIE travel markets is noted in the TCM
evaluation documentation. Since the GRACIE evaluation criteria are integral to this
methodology, inadequate data could result in inaccurate estimates of TCM effectiveness.
In addition to the GRACIE travel activity information, more precise estimates of idle,
speed, and time-of-day related emissions would clearly improve the model's predictive
abilities.
A-8
-------
This methodology is based upon the evaluation of a complete package of candidate
TCMs. There are no provisions made for evaluating a single TCM for its impact in a
localized area, nor is there a simple way to alter the package of measures evaluated
without repeating the entire modeling procedure. Additionally, the synergistic effects of
combining different TCMs is assumed to be accounted for by the GRACIE modeling
process. As noted above, the information regarding the GRACIE travel markets is
preliminary, so these effects may be somewhat inaccurate. Also, this method of
assigning TCM impacts focuses primarily on shifting between modes by the commuter.
The generation of additional trips resulting from increased vehicle availability (e.g. a
member of a telecommuter's household can now use the vehicle on certain days) is not
addressed.
Many areas of this methodology rely upon qualitative evaluations. The initial ranking
of TCMs to be included in the package; the division of TCMs into near-, mid-, and long-
term measures; and the evaluation of technical feasibility and public acceptance are all
based upon "professional judgment" that could be highly variable. The accuracy of the
model's prediction is likely to improve if, at a minimum, the method for conducting this
"guesswork" is more clearly defined.
This methodology is specific to the SMAQMD region and is largely dependent upon the
availability of computing resources for implementation. While there are some conceptual
contributions that can be made for other agencies, there is little transferability to other
locations. This lack of transferability is also augmented by the subjective nature of many
of the evaluation steps. There are few criteria outlined in the methodology that could be
used as a guide for other air quality/transportation agencies.
AQAT-3: AIR QUALITY ANALYSIS TOOLS
The AQAT package was developed by the California Air Resources Board Stationary Source
Division and links four computer tools for assessing air quality impacts of transportation
programs: URBEMIS, EMFAC, CALINE4, and PIVOT POINT. The package is supplied on
two diskettes and can be used with any IBM compatible microcomputer with 128K of memory,
a color graphics video adaptor, and a disk drive. The four components of the package are
described below under modeling methodology.
Modeling Methodology
URBEMIS can be used to estimate emissions from vehicular traffic associated with new or
modified land uses based on changes in the number of trips associated with a given land use, and
the VMT for each trip type. The user inputs whether the project under analysis changes
residential or commercial trip generators. The user then sets up EMFAC parameters (study
A-9
-------
region or default EMFAC inputs) for vehicle fleet mix, temperatures, trip speeds, trip lengths,
the percent of travel by operating mode (hot and cold starts), and percent of travel by trip type.
EMFAC7PC estimates on-road emission factors (i.e. grams per mile travelled) for a vehicle
fleet. The model has streamlined the fleet characterization and some other aspects of the
mainframe EMFAC (used in place of the MOBILE models in California). Rather than the
detailed model year by model year information contained in EMFAC, EMFAC7PC details the
percent of vehicles by vehicle class (i.e. light duty auto, light duty truck), fuel (leaded,
unleaded, or diesel) and the percent of travel by each vehicle class (and fuel type within each
class).
CALINE4 was developed by Caltrans to calculate pollutant concentrations near roadways, based
on Gaussian algorithms. Users define source strength, site geometry and other cite
characteristics, and meteorology, the model calculates pollutant concentrations for receptors
within 150 meters of the roadway,
PIVOT POINT is a sketch planning methodology for estimating the impact of transportation
control measures on the use of various travel modes (i.e. single occupant vehicle, carpool,
transit). The methodology was originally developed by Cambridge Systematics (CSI, 1979) as
a manual worksheet method. Pivot Point evaluates the change in mode choice based on changes
in travel time or travel costs for specific transportation modes. The model is based on a
mathematical formulation frequently used in transportation mode choice models (multinomial
logit). The model considers that the probability of choosing a given travel mode is a function
of the utility of the mode divided by the sum of the utilities of all possible modes. Pivot Point
calculates revised probabilities based on an existing base mode share and estimated changes in
the utilities (i.e. lower transit costs equal higher transit utilities).
Inputs to Pivot Point include, for each population subgroup analyzed, income, employment, and
auto ownership information, base mode shares, the average carpool size, and average trip lengths
for work and non-work trips. The user then translates each TCM being analyzed into potential
level of service changes. These are entered in units such as changes in round-trip in-vehicle
travel time, round trip out-of-vehiele travel time, or out-of-pocket travel costs.
Strengths and Weaknesses
The AQAT package utilizes a number of commonly used computer software programs and may
be a reasonable screening approach for looking at TCMs. There are a number of weaknesses
as well. A primary one is that the documentation is not sufficient to understand the precise
techniques that are used to calculate changes. Such 'black box' techniques may be problematic
for agencies preparing or reviewing TCM emission estimations. Key for the TCM travel effects
changes is the use of the pivot point model. The model calculates only work trip changes and
it uses regression coefficients developed from a 1968 Washington DC travel survey. The model
itself is a useful way to roughly approximate modal shares resulting from changes in level of
A-10
-------
service. It does not precisely calculate how the changes in modal shares would translate to trip,
VMT, and speed changes. With respect to the emission factors, the effect of the use of
abbreviated inputs for fleet characterization is not discussed although this may significantly affect
the results. Finally, the model is very California specific and would not be easily transportable
to other states.
SYSTEMS APPLICATIONS INTERNATIONAL/CALIFORNIA AIR RESOURCES
BOARD
The California Clean Air Act (CCAA) requires nonattainment areas to adopt TCMs to reduce
vehicle activity levels, and growth in these levels due to population increases. The Federal
Clean Air Act Amendments (CAAA) of 1990 also require TCMs to be adopted in many
nonattainment areas. Both State Implementation Plans and attainment plans required under the
CCAA must include detailed evaluations of the emission reductions associated with the TCMs
proposed. However, no comprehensive methodology for evaluating the effects of TCMs was
available when these provisions were promulgated. The Mobile Source Division of the
California Air Resources Board (ARB) sponsored a study to provide such a methodology.
One of the primary purposes of this methodology was to address what were felt to be numerous
overly simplistic assumptions that had been used in past TCM evaluation efforts. It was felt that
these assumptions produced exaggerated estimates of a TCM's effectiveness, making it difficult
to rely on control strategies which utilize them. For instance, it is often assumed that each new
ridesharer will reduce one trip, or that employees working a four-day work week will reduce
trips by 20 percent. In reality, a ridesharer may drive to a park-and-ride lot (reducing VMT but
not trips), while compressed work week workers may make extra non-work trips on their days
off from work. Other simplifying assumptions have been made regarding TCM packages.
Combinations of TCMs are frequently assumed to be additive in their effects although some may
not be (i.e. one cannot ride a bus and carpool simultaneously) and some may be synergistic.
The methodologies developed by SAI for the ARB provide methods for evaluating both
individual measures and packages of measures, much like the work currently underway for EPA.
The individual methodologies cover a limited number of TCMs: rideshanng, telecommuting,
parking management, flextime/staggered work hours, compressed work weeks, traffic flow
improvements, and traffic signal synchronization. This set was chosen to represent most of the
key analytical problems associated with TCM analysis as well as to include commonly
implemented TCMs. Other TCMs may be assessed by slightly modifying approaches that are
similar. These individual methodologies attempt to quantify both the total effect on overall trips
(and, consequently, emissions) of each measure, taking into account as many variables as could
be quantified and that could potentially affect the overall benefit of the TCM.
The packaging methodology was designed to enable user's to employ a multi-attribute analysis
of groups of TCMs, in order to assess the overall effect of the package and account for
phenomenon such as overlap and synergy between various measures.
A-ll
-------
Methodology Application • Comparison with EPA Methodology
As noted above, the methodologies developed for ARB are similar in nature and function to the
methodologies presented in this document. The key similarities and differences are highlighted
below:
• The ARB methodologies for individual TCMs were designed for a specific set of
measures (listed above). While the user can generally apply the methodologies to other
measures by making slight modifications to one that is provided, the approach for doing
such modifications was not discussed at length. Recognizing that the vast number of
TCMs and potential implementation strategies makes it impossible to develop and present
methodologies to cover every possible situation, the methodologies developed for the
current work for EPA are structured in a manner that allows the analyst to quickly adapt
them to TCMs and situations other than those specifically presented in this document.
An effort to make the evaluation of individual measures much more generalized and
streamlined, focusing on the logic behind the TCM analysis and providing guidance to
the user on modifying a methodology for a specific need. Attention has also been
focused on clarifying the theory and assumptions behind these methodologies to make
them more intuitive to the user. It is intended that these modifications will ultimately
result in methodologies that are more widely applicable than those developed for ARB.
• The calculation of emissions impacts of TCMs in the ARB report were based upon the
EMFAC7E model, which is specific to California. The EPA methodologies utilize the
MOBILE4.1 model, making them applicable on a nation-wide scale.
* The packaging methodology developed for ARB involved the manual calculation of the
impacts of groups of measures based upon the evaluation criteria given. Because the
packaging methodology requires a variety of qualitative decisions be made by the user,
the calculation of a multitude of scenarios could be very cumbersome, confusing, and
time consuming. The packaging methodology included in this report represents a
"second generation" of the ARB methodology. The process has been automated via a
computer software package, so many scenarios using a variety of utilization rates can be
quickly screened. Additionally, more complete guidelines for making decisions regarding
the application of the packaging methodology have been provided, making the entire
model easier to use and more intuitive than that which was developed for ARB.
• The packaging methodology presented in this report utilizes a normalizing function that
is an exponential and not a square. This was changed to more closely resemble the
widely accepted logit model used in the transportation community, as well as to obtain
results that are better than those obtained using the ARB methodology.
Shortcomings of the ARB Methodology
A-12
-------
There are number of potential effects of TCMs that are not addressed by the ARE
methodologies. These include changes in work habits, auto hold/purchase decisions,
employee residence location, employee office location, and overall changes in land use
as a result of altered commute patterns. It is conceivable that these effects could be
addressed in later revisions to the methodology.
There are areas in the methodology that require professional judgment in assigning key
values. An example is the determination of the utility values for cost, convenience, time,
and reliability that are iclied upon by the packaging methodology. These values are
highly subjective, and users of the methodology should be sensitive to this.
Several of the individual TCM methodologies employ elasticity values in order to
calculate the response of the commuter to changes in the transportation system. For
instance, the ridesharing methodology employs the elasticity of peak speed with respect
to volume. Elasticities are by nature very specific to the region being studied. They are
intended to be used as a screening tool only, and should be supplemented with more
complete information. This report includes a discussion of how elasticities are calculated
in order to provide the user with guidance on developing region-specific values for the
necessary elasticities.
As noted above, all of these methodologies are data-intensive. In order to obtain the best
estimate of a TCM's impact in a specific region, the user is encouraged to use as much
local data as is available. This may be a problem for smaller districts that do not have
the resources to have region-specific travel surveys and other forms of data collection.
In these cases, users will be forced to use less accurate data from sources such as the
U.S. Bureau of the Census and Department of Transportation. This is likely to result
in less accurate estimates of a TCM or package TCM's effects on travel behavior and
emissions.
A-13
-------
(This page intentionally blank for 2-sided printing)
-------
APPENDIX B
Methodology to Evaluate Peak Period Trip Shifts of Flextime and Compressed Work
Participants
This appendix is provided as a supplement to Step 4 of Chapter 2 where in peak period
trip shifts were evaluated for flextime and compressed work week participants. The
details of the methodology, presented here for completeness, was separated from the main
report since an understanding of the exact methodology is not required to complete the
TCM activity level assessment of Chapter 2, Flextime is the scheduling of work hours
allowing for a broader period of travel to and from work; compressed work week
scheduling adds one or two working hours to every four days in order to eliminate one or
two days every two weeks from the work schedule. In both cases, the daily changes in
the travel period of the participants results in a fraction of work trips made by the
participants which shift from the peak to the off-peak period.
The evaluation of peak period trip shifts centers on the identification of the fraction of the
total trips which will shift from the peak period to the off-peak period. This parameter,
SPUEX for flextime participants and 6CWW for compressed work week participants will
vary by the length of the peak period and by the change in length of the participant travel
period. For example, the longer the peak period the less likely a flextime or a
compressed work week trip is going to be removed from the peak period. An evaluation
°f ^FLEX and 5cww can be made by establishing a few assumptions in regard to the
expected travel changes and the distribution of work trips. In this methodology, it is
assumed that the distribution of targeted work trips can be predicted by a normal (or
Gaussian) distribution.
The remainder of this appendix is divided into 3 sections:
(1) Discussion of the Gaussian distribution and comparison of the Gaussian
distribution with actual work trip distribution data,
(2) Evaluation of the fraction of the total trips which will be removed from the
peak period for flextime participants
(3) Evaluation of the fraction of the total trips which will be removed from the
peak period for compressed work week participants 0&cww).
The analysis of the two TCMs are handled individually, as the travel characteristics of
each is unique. Flextime participants experience a broadening of travel period due to
flexible work scheduling; compressed work week participants experience earlier travel
during the AM peak period and later travel in the PM peak period due to extended
working hours.
92093.08 B-l
-------
Using a Gaussian Distribution to Simulate the Work Trip Distribution
The Gaussian distribution, also known as the normal distribution or the "bell-shaped
curve" is a symmetrical distribution which represents the distribution of occurrence of
many phenomena (Figure B-l). The probability density function, f(x), of the normal
distribution shown in Figure B-l is defined by the following equation with mean, p, and
variance, a*:
The continuous distribution function, F(XQ), is the area under the probability density
curve, f(x) representing a total number of occurrences for the interval between x=0 and
x=sx0. This function is defined by:
In assuming a normal distribution of work trips, f(x), becomes the trip density function of
the independent variable x, which in this case is time. The standard deviation of the trip
density function, a, will be determined from the peak period length and is illustrated in
the TCM applications that follow this section. F(x0) is then the total number of trips
occurring between time x=0 and x=x0. The total number of trips between an interval
defined as x=Xj and x—x2 is determined from F(x2) - F(XJ).
Since the probability density function, f(x), is difficult to integrate, Equation B-2 is
seldom used to directly evaluate the continuous distribution function, and a standard
normal table is used instead. The standard normal able is based on a normal distribution
with mean, p,, of 0 and a standard deviation,
-------
f(x) = probability density
function
FIGURE B-l. Example illustration of the Gaussian or normal distribution curve showing
f(x), the probability density function and /t, the mean value of the independent variable,
x. In the case of f(x) equaling the trip density function, x represents time, and ft
represents the mean value of the peak period.
B-3
-------
0.14
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
End Hour
FIGURE B-2. Actual hourly work trip distribution data for Phoenix, AZ showing the
bell shaped curves of the AM and PM peak periods.
B-4
-------
the AM and PM peak periods is apparent confirming the assumption that work trip
distributions are similar to the Gaussian distribution and can be modeled as such.
Furthermore in this analysis, the targeted work trips (i.e. the TCM participants) are
assumed to have an equivalent normal distribution of trips.
Evaluation of Peak Period Trip Shifts For Flextime Participants
This analysis determines the net trip shift from the peak period due to flextime work
scheduling. Flextime allows for a broader period of travel to and from work resulting in
a shift of work trips to before or after the normal peak period. Only a fraction of the
total trips made by flextime participants will experience a shift from peak period to the
off-peak period malting it necessary to define and identify fraction of the total trips made
by the participants which will shift from the peak period to the off-peak period. This
fraction is identified by the symbol
The value of %x£X can ^ determined by establishing the average increase in travel
period for flextime participants, by assuming the change in travel time is equally as
probable to occur earlier or later (than before flextime implementation), and by assuming
a normal distribution (Gaussian distribution) of targeted work trips. For example,
examine the trip density curve illustrated in Figure B-3 where the peak period length is
shown by ^. Let u, shown in Figure B-3, be the average shift in time the flextime
participant is willing to travel. All work trips which occur within w hours of the
endpoints of the peak period have the potential to shift out of the peak period. Of those,
it assumed that 1/2 of the people representing these potential trips will choose to travel
earlier and 1/2 will choose to travel later. The 1/2 of the trips identified at the earlier
endpoint which choose to travel earlier and the 1/2 of the trips identified at the later
endpoint which choose to travel later will be removed from the peak period; therefore,
only 1/2 of the trips originally identified in the region which is within w hours of the
endpoints of the peak period will actually experience a trip shift. The assumption that
people are equally likely to travel earlier as later is conservative estimate. Since there is
more incentive to move away from the peak period (due to traffic considerations) it is
likely that more that 1/2 of the people would move away from the peak in the trip
distribution. The value of ^FLEX 1S t^611 calculated to be the fraction of the total trips
which are removed from the peak period, and can be seen to be a function of ^ (peak
period length) and w (average time shift of flextime participants).
If a normal distribution of targeted work trips is assumed, then Sj^gx can be calculated
from the continuous distribution function defined in Equation B-3. Using the Equation B-
3 to determine the value of the standard deviation, a, and assuming a curve that captures
92W3.08 B-5
-------
H
o>
.Q
£
f(x)
f(x) = trip density
function
Area representing
the number of potential
trips which could shift
from the peak period
v- x,
.0/2.
CO
Coordinates of ^ and x
x=
FIGURE B-3. Example Gaussian work trip distribution curve for a peak period length,
0. The shaded areas indicate the fraction of targeted work trips which occur within w
hours of the endpoints of the peak period where u indicates the average amount of time a
flextime participant is willing to shift.
B-6
-------
1.
95 % of the trips within the endpoints1:
1.96 =
_ 4>/2
where 1.96 is the value of z corresponding to a 95% of the total trips determined from a
standard normal table and ^/2 (i.e. 1/2 the length of the peak period) is the value of x2-fi
(x2 equals one of the endpoints shown in Figure B-3). Solving for a yields:
0 -
3.92
(B-4)
All work trips which occur with in w hours of the endpoints is equivalent to two times the
trips which occur between the time interval Xj=^/2-«+/i and x2=^/2+/i (shown in
Figure B-3) where the factor of 2 is to account for both areas at each side of the curve.
Of these, it was indicated above that only 1/2 will fall out of the peak period, so the 2
and the 1/2 cancel each other, and the fraction of trips removed from the peak period can
be determined from the area of the curve between the time interval Xj=^/2-w+/i and
x2=^/2+ft. This area of the curve is determined by evaluating the difference of the
continuous distribution function F(x) at X] and x2:
,
\for Co*-*
for i—
for +M can be translated in to z coordinates using Equation
B-3 with the value of a determined in Equation B-4 yielding:
1 A curve that captures 100% of the trips is ideal, but due to the asymptotic nature of
the normal distribution, identifying a curve that captures 100% of the trips produces
unrealistic results.
92093.08
B-7
-------
1-GJ +
2 = 1.96* 11-^1 (B-6)
3,92
The value of Sfj^ can ** determined for any combination of 4> (peak period length) and
01 (average time shift of flextime participants) using Equation B-5 and a standard normal
table with z identified in Equation B-6. For example, for a peak period length of 3 hours
and an average time shift of 1 hour for flextime participants, z is determined to be 0.653
from Equation B-6. From this value of z, the corresponding continuous function value is
0.242 determined from a standard normal table. Then using Equation B-5 to determine
SJXEX, results in ^xjsx = 0-233. 0.233 is the fraction of trips removed from the peak
period for the conditions of $=3 and «=1. The results of this methodology are given in
Table 2-6 (included in the text of Chapter 2) for a range of ^ and w.
Evaluation of Peak JPeriod Trip Shifts For Compressed Work Week Participants
This analysis determines the net trip shift from the peak period due to compressed work
week scheduling. Compressed work week scheduling adds one or two working hours to
every four days in order to eliminate one or two days every two weeks from the work
schedule. This daily extension of working hours from a compressed work week schedule
results in a fraction of participants experiencing trips shifts outside the peak period as
commuters travel earlier to work and return home later. This fraction of trips removed
from the peak period is identified by the symbol
The value of £CWw can be determined by establishing the average increase in travel
period for compressed work week participants, and by assuming a normal distribution
(Gaussian distribution) of targeted work trips. For example, examine the probability
density curve illustrated in Figure B-4 where the peak period length is shown by ^ and
where this is assumed to be an PM peak period. Let u, shown in Figure B-4, be the
average shift in travel time per peak period experienced by the compressed work week
participant (i.e. a compressed work week participant who works a nine-hour day instead
of an 8-hour day will shift AM travel time 1/2 hour earlier and PM travel time 1/2 hour
kter). Since this figure represents a PM peak period, the participants travel period will
all be shifted to a later time. For compressed work week scheduling, all work trips
which occur within w hours of the earlier end of the AM peak period and within u hours
of the later end of the PM peak period will shift out of the peak period. The value of
5CWW is then calculated to be the fraction of the total trips which are removed from the
peak period, and can be seen to be a function of ^ (peak period length) and « (average
time shift of compressed work week participants).
92093 .OS B-8
-------
CD
.s-
H
(M
O
U
£
s
2;
f(x)
f (x) = trip density
function
Area representing the
number trips which will shift
from the peak period.
CO
Coordinates of Xj and x,
= /2-co
x=
FIGURE B-3. Example Gaussian work trip distribution curve for a PM peak period
length, «£. The shaded areas indicate the traction of targeted work trips which occur
within u hours of the later endpoint of the PM peak period where u indicates the average
amount of time per peak period a compressed work week participant will shift in order to
accommodate longer working hours.
B-9
-------
If a normal distribution of targeted work trips is assumed, then S^^w can k° calculated
from the continuous distribution function defined in Equation B-3. Using the Equation B-
3 to determine the value of a and assuming a curve that captures 95 % of the trips within
the endpoints2:
1.96 =
where 1,96 is the value of z corresponding to a 95% of the total trips determined from a
standard normal table and ^/2 (i.e. 1/2 the length of the peak period) is the value of XQ-H
at the endpoints. Solving for a yields:
(B-4)
3.92
All work trips which occur with in u hours of one of the endpoints is equivalent to the
trips which occur between the time interval x=^/2-«+ji and x=«£/2+ji (shown in Figure
B-4) and the fraction of trips removed from the peak period can be determined from the
area of the curve between the time interval x^^/2-w+ji and x2=^/2+ft. This area of
the curve is determined by evaluating the difference of the continuous distribution
function F(x) at Xj and x2: and the fraction of trips removed from the peak period can be
determined from the following values of the continuous distribution function F(x):
'cww
'cww
for
In this equation, the value of the endpoint F(^/2+^i) is already known to be 0.475 (half
of 95 % occurs between the mean and each endpoint) so these equations can be simplified
to become:
'cww
*cww = 0.475
for &):<—
for «>—
2
(B-5)
6CWW can be evaluated using Equation B-5 where the value of F(^/2-«+ft) can be
evaluated using Equation B-2 or a standard normal table. If a standard normal table is
2 A curve that captures 100% of the trips is ideal, but due to the asymptotic nature of
the normal distribution, identifying a curve that captures 100% of the trips produces
unrealistic results.
92093.08
B-10
-------
used, then the value of x=^/2-u+M can be translated in to z coordinates using Equation
B-3 with the value of results in Scww = 0,323. 0.323 is
the fraction of trips removed from the peak period for the conditions of ^=2.5 and w— 1.
The results of this methodology are given in Table 2-7 (included in the text of Chapter 2)
for a range of and w.
92093.08 B-ll
-------
(This page intentionally blank for 2-sided printing)
------- |