United States Office of
Environmental Protection Research and Development
Agency Washington, DC 20460
EPA-600/R-93-214
November 1993
* EPA Conceptual Designs For A
New Highway Vehicle
Emissions Estimation
Methodology
Prepared for Office of Air Quality Planning and Standards
Prepared by Air and Energy Engineering Research Laboratory
-------
EPA REVIEW NOTICE
This report has been reviewed by the U.S. Environmental Protection Agency, and
approved for publication. Approval does not signify that the contents necessarily
reflect the views and policy of the Agency, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/R-93-214
November 1993
CONCEPTUAL DESIGNS FOR A NEW HIGHWAY VEHICLE EMISSIONS
ESTIMATION METHODOLOGY
by
John T. Ripberger
(A Participant in the Earth Team Soil Conservation Service Volunteer
Program Assisting the U.S. Environmental Protection Agency)
EPA Project Officer: Carl T. Ripberger
Emissions and Modeling Branch
Air and Energy Engineering Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Prepared for:
U.S. Environmental Protection Agency
Office of Research and Development
Washington, D.C. 20460
-------
ABSTRACT
This report discusses six conceptual designs for a new highway vehicle emissions
estimation methodology and summarizes the recommendations of each design for improving the
emissions and activity factors in the emissions estimation process. The complete design reports are
included as appendices. EPA asked six contractors to assist in developing ideas for a potential
methodology to estimate highway vehicle emissions that could be developed for use in 5 to 10
years. They were selected because of their experience in working with mobile source emissions
inventories.
In general, the contractors suggest developing new modules within the emission estimation
process to provide users with more detailed information on the causes of vehicle emissions. The
essence of these concepts is the need for more comprehensive integration of data between the
transportation planning model and the emission factor model.
The individual reports did reinforce comments voiced by other experts for modal data and
more responsive transportation models, though they did not identify totally new concepts not
already being considered by EPA research programs.
This report covers a period from March 1991 to July 1991. The work was completed as of
September 1991.
11
-------
CONTENTS
Page
Abstract ii
Figures. iv
Acknowledgments v
1. Introduction 1
2. Recommendations for Improving the Emission Factor Model 4
3. Recommendations for Improving the Activity Factor Model 6
4. Improvements in Emission Inventory Modeling and Integration Techniques 8
References 22
Appendices
A. Alliance Technologies Corporation's Conceptual Design A-L
B. Desert Research Institute's Conceptual Design B-i
C. E.H. Pechan & Associates' / COMSIS Corporations' Conceptual Design C-i
D. Radian Corporation's Conceptual Design D-4
E. Sierra Research's Conceptual Design E-i
F. Systems Applications International's Conceptual Design F—i
-------
FIGURES
Number Page
1 Conventional Urban Travel Forecasting Process 9
2 PECHAN/COMSIS Conceptual Design 10
3 Example Daily Emission Profiles 10
4 Sierra Proposed Modeling System 11
5 DRI Emissions Modeling System 13
6 Radian Motor Vehicle Source Model 15
7 Driving Cycle Matrix 16
8 Metamodel 17
9 Base Case Emission Estimate 18
10 Development of Relationships for Synthesizing and Forecasting Vehicle
Operating Mode Distributions 20
11 Development of Future Year Emission Forecasts 21
IV
-------
ACKNOWLEDGMENTS
We acknowledge the contributions of the following authors of the reports provided in the
appendices. The EPA project officer providing technical direction for these procurements was Carl
T. Ripberger.
Mark Smith of TRC (formerly Alliance Technologies Corporation), for the concepts and
designs provided under EPA Contract # 68-D9-0173, work assignment # 1/129.
Alan W. Gertler, John G. Watson, and William R. Pierson of Desert Research Institute,
for the concepts and designs provided under EPA Contract # 68-D9-0168, work assignment # 24.
James H. Wilson of E. H. Pechan & Associates, for the concepts and designs provided
under EPA Contract # 68-D9-0168, work assignment # 24.
David Levinsohn, J. Richard Kuzmyak, and Richard Pratt of COMSIS Corporation, for
the concepts and designs provided under EPA Contract # 68-D9-0168, work assignment # 24.
L. Bruckman, E. L. Dickson, and W. R. Oliver of Radian Corporation, for the concepts
and designs provided under EPA Purchase Order # 1D1866NALX.
Bob Dulla of Sierra Research, Inc., for the concepts and designs provided under EPA
Purchase Order # 1D1345NASA.
R. G. Ireson of Systems Applications Internationa] and J. P. Nordin, for the concepts and
designs provided under EPA Purchase Order # 1D1519NASA.
-------
SECTION 1
INTRODUCTION
In April, 1991 six tasks were awarded to look for new concepts that could potentially provide
improved emission inventory estimates from highway vehicles in 5 to 10 years. This project was a
follow-on to the two highway vehicle workshops held in the summer of 1990 to solicit research
recommendations from emission inventory experts. (Wilson, 1991). The contractors,
• Alliance Technologies Corporation (TRC)
• Desert Research Institute
• E.H. Pechan & Associates, Inc.
• Radian Corporation
• Sierra Research, Inc.
• Systems Applications International
were selected because of their experience in using and developing mobile source emission
inventories for EPA and California. E.H. Pechan & Associates teamed with COMSIS Corporation
to address transportation modeling recommendations. Guidelines provided to the contractors in
March 1991 included:
Problem Definition
Emission inventory estimates are used by EPA and states in models to
determine the effectiveness of various control strategies for achieving air quality
standards Methods available today for estimating emissions from highway
vehicles may not be adequate for the new fuels, technologies, and transportation
control methods available 5 to 10 years from now. This project is intended to
begin the process of looking for an optimum methodology.
Background
Analyses of the SCAQS [South California Air Quality Study] tunnel study
data indicated that emission inventory estimates for urban areas using the EPA
MOBILE(3-4) or EMFAC7(C-E) emission factor models could be much lower
than actual. All highway vehicle emission estimates of NMHC, NOx, and CO for
ozone and CO SIP non-attainment areas used these models through 1987
[MOBILES and EMFAC7F have been available since 1991]. Since CO, NMHC,
and NOx from highway vehicles are already such a large portion of an urban
areas estimated emissions, a large potential increase from this source could force
-------
significant changes to control strategies that were designed to achieve ambient
standards.
EPA conducted two workshops in 1990 for various experts to brainstorm
about state-of-art improvements that could be made in estimating emissions from
highway vehicles....Some design concerns were noted:
• Any new procedures must be validated against a "ground truth."...
• The uncertainty of data estimates must be defined and reduced.
• New procedures must be affordable by all SIP areas...
• New procedures must be sensitive to changes in vehicles, fuels, and
transportation control measure (TCM's) approaches as well as meeting data
requirements of atmospheric photochemical models.
• Current emission estimation methods rely on parameters (e.g.; VMT,
speeds, I&M effectiveness) that are derived from models and cannot be
satisfactorily validated. [The AWMA paper by] Leonard Seitz ... proposed
that emissions are nearly constant for varied speeds over time. If this is a
reasonable assumption, our future model might concentrate on those
conditions causing emissions to exceed this norm rather than emphasizing
understanding emissions vs. speed.
• High tech scientific developments such as GIS and satellite (or other remote
sensing) data should be considered in new designs.
Studies are being planned and conducted by California and EPA to evaluate
vehicle utilization, driving patterns, emissions as related to vehicle load and speed,
I&M program effectiveness, and VMT projections. Plans have been made to
improve data through modal analyses and computer simulations of emissions as
related to vehicle load and operation.
Program Objective
A focal point is desireable to orient and coordinate research efforts to ensure
that optimum procedures are available in 5 to 10 years. The results of this work
assignment will provide (in the opinion of the contractor) the optimum method for
estimating emissions from highway vehicles. The time frame for implementing
the new method is 5 -10 years.
Because of the special strengths resident with the different contractors, it is
expected that the ultimate optimum approach will be a combination of ideas from
each of the different designs. EPA will establish a working group of Federal,
State, and private experts to evaluate these conceptual designs and suggest a
consolidated methodology design that best meets the needs of aS users.
This report summarizes the conceptual designs from the six contractors. Their complete
reports are included in the Appendices. In general, the contractors believe that the current model
(MOBILE4) provides overly aggregated results and they suggest developing new modules within
the emission estimation process to provide users with more detailed information on the causes of
vehicle emissions. The essence of these conceptual designs is the need for more comprehensive
integration of data between the transportation planning model and the emission factor model. To
-------
present the contractors' recommendations clearly and to avoid redundancy, this report combines
their concepts and organizes them according to topic, rather than simply summarizing each
contractor's conceptual design in a list. First, concepts for improving emissions factors will be
considered, followed by recommendations for improving activity factors, and finally ideas for
organizing and coordinating the new emission and transportation models.
The current approach used by EPA to provide emission estimation procedures is to
continually incorporate new findings as incremental changes to the existing methodology. A
weakness of this approach is that with the advent of major technology changes in pollution
controls, polluting equipment, measurement devices, and analytical tools EPA could miss cost
effective improvement opportunities by considering only the incremental approach. The Joint
Emissions Oversight Group (JEIOG), established by EPA in 1991, provided guidance to the
Office of Research and Development to conduct research to find and develop potential quantum
leap improvements to the emission inventory process that could be implemented in the future.
As EPA develops its research program to prepare emission inventory models for the future,
it is attempting to make sure that no new promising concepts are overlooked. The request to the
six contractors for concepts was an initial step in this process. For the most part, the reports did
not identify any totally new concepts not already being considering by EPA research programs.
Most of the ideas were linked to existing procedures, requiring extensive data collection concerning
vehicle emissions and operations on highway networks. The reports did reinforce comments
voiced by other experts for modal data and more responsive transportation models. Where
concepts, such as modal modeling, can be shown to be affordable and provide more accurate
estimates, future models will attempt to incorporate them. EPA research to evaluate the most
promising concepts and validate algorithms for them is planned through 1996.
-------
SECTION 2
RECOMMENDATIONS FOR IMPROVING THE EMISSION FACTOR MODEL
The primary improvement to the emission factor model which all the contractors
recommend is to begin measuring the emissions produced during the four modes of vehicle
operation: acceleration, deceleration, cruise, and idle. This would be in addition to the emission
measurements for cold and warm starts and for the different vehicle speeds. Vehicles emit
significantly different levels of pollutants during each mode of operation and, as a result, this
additional information is essential for accurate emissions estimates. The quantification of such data
will in turn require transportation models to include information as to where these modes occur
within the travel area. A more complete discussion of this will be covered later in the report.
The collection of modal data (i.e. data concerning various engine operating modes such as
idle, cruise, or acceleration) dictates that, for emissions inventory improvements, modifications
must be made in the current FTP cycle to incorporate the new information. Indeed, most of the
contractors suggest discarding the FTP cycle for emission factor development in favor of some
alternative "in-use" test. For instance, if the nominal FTP driving cycle is replaced with vehicle
operating mode distributions, Systems Applications International (SAI) believes that factors
specific to vehicle operating mode will be needed in order to unambiguously establish the mode a
vehicle is in and how the emissions vary by mode. Consequently, driving-cycle-based emission
factors would be obsolete. The primary input for this model would be factors such as: (1) basic
mode (e.g., cruise or creep); (2) speed; (3) acceleration; (4) engine state; and (5) canister loading.
While the contractors' suggestions for gathering the data differ, their idea that information
concerning vehicle operating mode must be included in the emission factor model remains
constant.
Most of the contractors also recommend including local information in the emission factor
model such as super-emitting vehicles, alternative fuels, and fluctuations in temperature from
morning to evening. Sierra Research's (hereafter Sierra) conceptual design calls for the
development of "engine maps" to represent the mix of vehicles operating in an area. Engine maps
would characterize emissions as a function of speed and load for a set of vehicles including factors
for input such as: (1) the mix of fuel and emission control devices used in previous vehicle models
-------
along with a projection of the mix for future models; (2) the default distribution of vehicle
categories and model year registrations within each category; and (3) the effect local I/M programs
have on the mechanical condition of vehicles and, in rum, the effect they have on emissions
reductions. Desert Research Institute (DRI) advocates the input of diurnal temperature variations
into the calculation of emission rates and a flexible system of allocating emissions to locations.
The contractors also recommend determining the percentage of vehicles emitting in the
super, very high/high, and normal modes and considering their effect on emissions. Sierra
suggests that the sampling for performing in-use emission tests be designed to effectively capture
emission rates, the fraction of vehicles and their travel in each of the emitting modes, and the
reasons for emission control failures when they occur. The contractor concludes that this type of
data collection makes it possible to more accurately estimate base year emissions, as well as
estimate the effect of future control regimes on future year emissions.
Several contractors also suggest including more extensive data concerning stationary
vehicle evaporative emissions in the calculation of emission rates. These include diurnal, hot soak
and resting evaporative losses as well as cold and hot start exhaust emissions. Radian
Corporation's methodology, named EVAP 2.0, calculates evaporative emissions on an hour-by-
hour basis, as opposed to the procedures used in existing models which calculate emissions on a
daily basis and assume emissions are evenly speed throughout the day. Radian believes that the
methods used in MOBILE4 do not permit proper spatial allocation of these emissions, since they
are calculated on a grams per mile basis and spread over the road network. Radian concludes that
their method will permit proper spatial allocation of both exhaust and evaporative emissions since
they will each be calculated separately. In this way, it will be possible to allocate the emissions
closer to where they occur.
DRI suggests a rather different methodology for the emissions measuring process. DRI
suggests deriving empirical on-road estimates of emissions using remote sensing; rather than
extrapolating data from dynamometer tests. The method includes freeze-frame video images of
vehicle license plates. The ratio of visible light transmission to CC>2 can then be related to the
grams of pollutant emitted per gallon of fuel consumed. DRI believes that with reasonable
assumptions about vehicle mileage, this can relate emissions to VMT. In this way, an empirical
model of pollutant emissions per VMT can be determined for different types of roadways, driving
conditions, and times of day.
-------
SECTION 3
RECOMMENDATIONS FOR IMPROVING THE ACTIVITY FACTOR MODEL
As mentioned earlier, the collection of modal data for the emission factor model will in turn
require the development of more comprehensive activity factors. The contractors deal with this
problem either by expanding the current transportation model to include not only VMT but
information on roadway type and time of day or by eliminating VMT altogether in favor of listing
vehicle operating mode distributions by area and time. PECHAN/COMSIS's conceptual design
represents the first of these two views. PECHAN/COMSIS recommends adding three new models
to the travel demand modeling process. First, a model to estimate the percentage of vehicle travel
occurring in the various operating modes. The model would provide percentage estimates of daily
VMT in each of the four operating modes of acceleration, deceleration, cruise, and idle. The model
would be calibrated based on new data that need to be collected relating vehicle operating mode to
variables such as: (1) the types of facilities (e.g., freeway, arterial) over which each vehicle trip is
made; (2) the volume/capacity ratio of the facility; (3) the time of day (peak and off-peak); and (4)
information as to traffic control density (e.g., X traffic signals per mile). PECHAN/COMSIS
believes that these data can be collected through instrumented vehicle trips over a variety of urban
areas. The second model seeks to estimate the fraction of vehicle trips which have cold start and
hot soak emissions listed as a function of trip purpose and trip duration. The output of this model
would be the number of vehicle trip ends in each traffic analysis zone (TAZ) which are estimated to
have cold starts and hot soaks for both peak and off-peak periods. The third model would estimate
the trip reduction effects of the various transportation control measures (TCMs) which urban areas
are likely to implement. The output of this model would be the number of work purpose person
trips for each TAZ-to-TAZ movement which are forecast to utilize mass transit, carpools, or drive
alone measures. PECHAN/COMSIS also recommends expanding the travel demand model to
cover vehicle trips and VMT for areas outside the urban cordon.
SAI represents the opposing view where VMT is abandoned in favor of empirical vehicle
operating mode distributions. Under their conceptual design the actual effects of local factors and
conditions on vehicle operating modes are addressed. S AI's method calls for the collection of
comprehensive data on vehicle operations within an urban area. Principal reliance would be placed
on the use of time-lapse aerial photography and computer image processing and analysis to
-------
determine, by roadway or sub region, such parameters as: (1) the number of vehicles in operation;
(2) the distribution of vehicle speeds by road type; (3) the distribution of acceleration by speed and
road type; (4) the number of vehicles in non-free flow conditions (queued and creeping, and
idling); and (5) the number of trips started and ended. SAI believes that by basing the distribution
of vehicle operating modes on data for the study area, the potential bias introduced by the
nonrepresentational FTP driving cycle can be prevented. Similarly, by using actual observed
speed distributions on links, SAI contends that the arbitrary assignment of single average speed to
each link is avoided, as is the reliance on overly simplistic speed versus volume/capacity ratio
relationships such as the BPR curve and the simplistic treatment of off-network travel as having a
constant speed. Some examples of the factors which effect vehicle operating mode are: lane width,
road surface condition, distractions (road signs, commercial driveways and parking, on street
parking, etc.), hills, and the level of aggressiveness of drivers. SAI concludes that empirical
mathematical descriptions can be generated to show the variations in vehicle operating mode for an
area and time. According to SAI, even though VMT is left out of the transportation picture in
terms of the initial emissions estimations, the information gathered by it will later be necessary for
forecasting purposes (this will be described in more detail later).
Radian Corporation also recommends abandoning VMT in its travel demand model since
transportation modeling systems which incorporate VMT cannot efficiently handle variations in
link-level traffic operations during peak congested conditions. Instead of using VMT, their model
computes vehicle hours traveled (VHT) for each link. Correspondingly, the emission factor model
will measure emission rates on a grams/hr basis, rather than grams/mi.
-------
SECTION 4
IMPROVEMENTS IN EMISSION INVENTORY MODELING AND
INTEGRATION TECHNIQUES
While the recommendations for improving the emission and activity factor models through
increased collection and analysis of empirical data will be important and useful for developing more
precise emission estimates in the future, the primary goal for the contractors was to present a
conceptual design for a new highway vehicle emissions estimation methodology. With one
exception, the contractors attempt to achieve this goal by supplementing the general emission
inventory process (Figure 1) with pertinent new models to facilitate more accurate, efficient and
integrated emission estimates. Generally, these contractors attempt to rectify the limitations in the
contemporary four step travel forecasting process by expanding the activity factor database and
adjusting the current models to account for the factors which produce these limitations. Likewise,
the contractors employ similar means to improve the emissions factor model. The conceptual
designs differ, however, on the type of additional data which will produce better estimates and the
manner with which to integrate that data into the emission and activity factor models.
PECHAN/COMSIS's conceptual design (Figure 2), for example, mimics the current
inventory estimation process almost exactly; except for the addition of three new models to
supplement the travel forecasting process. Two of the models, a "Vehicle Operating Mode Model"
and a "Cold Start/Hot Soak Model", would be calculated off-line incorporating data from the
current travel demand model and then integrated into the emission factor model. The "Vehicle
Operating Mode Model" would be combined with the collection and analysis of emission data by
mode in the emission factor model to produce daily emission profiles (Figure 3). The objective is
to create the smallest number of profiles possible to accurately cover the variety of trip-making that
occurs in an area. The third model, a "Transportation Control Measure (TCM) Impact Model",
would be directly integrated into the travel demand model (TDM), or more specifically into the
mode choice module of the current TDM, rather than using it off-line as it currently is.
PECHAN/COMSIS suggests that one of the advantages of this model design is its ability to
encompass the full range of vehicle operating conditions, in addition to its ability to present data at
the level of detail necessary to validate and verify results. They add that until remote sensing
8
-------
proves to be a practical tool for validating emission estimates, the best tool for emission modeling
in a three to five year time frame would be improving I/M program performance and taking
advantage of the annual data tracking possibilities. Lastly, they suggest using satellite photos for
noting whether episode day conditions differ significantly from "average daily
traffic."
Figure 1. Conventional Urban Travel Forecasting Process (Appendix C, page C-7)
-------
Figure 2. COMSIS Conceptual Design
(Appendix C, page C-17)
( miv.innr.
(I IIIW)
CO. 11C. NO«
Hoi So;ik
(MOW)
(,il Iriporxlr,)
DimrKil
(IIHW)
N
-------
Sierra Research's modeling system (Figure 4) also closely resembles the current method
employed, except they abandon the use of a single drive cycle with correction factors to account for
the effects of local driving conditions. Sierra's conceptual design calls for the construction of
separate emission factor estimates for each link on unique trip routes. They believe that this could
be achieved by modifying current transportation planning models to supply additional information
on: (1) trip length; (2) mix of roads (i.e. facility types); and (3) the congestion encountered on
routes between each origin and destination (OD) zone. Also, by using local transportation surveys
or other improvements, similar information would be developed for the travel that occurs within
each traffic analysis zone (TAZ).
Local Transportation
Survey Data
Figure 4. Sierra's Suggested Modeling System (Appendix E, page E-5)
11
-------
According to Sierra, a TCM module would receive the preliminary outputs of the
transportation planning module and local trip surveys to evaluate the effects of alternative TCM's
on the travel-related determinants of vehicle emissions (i.e. the relationship between vehicle trips,
trip length and congestion levels). This information would then be fed back to the transportation
planning module to reevaluate the level of travel produced.
Sierra's transportation planning module would provide two outputs. The first output
includes information on trip type, trip length, mix of roads, and congestion encountered .on routes
between each OD zone and within each TAZ. All this would then be input for the modal emissions
module (MEM). The MEM would include this route-specific travel information, EPA drive cycle
studies (includes operating conditions like speed and acceleration found on similar roads under
similar congestion levels), and data from engine maps. The MEM output would be a specific
emission factor for each link encountered on each trip. The second transportation planning module
output would be the actual number of vehicle trips between zones and within zones. This
information is later input for a grid cell emissions (GCE) module. The GCE module integrates the
data from the road network and TAZ's with the link-specific emission factor estimates for each OD
pair and the number of trips between each OD pair. A similar computation would be used to
estimate the emissions within each TAZ. Finally, the GCE module aggregates the link and TAZ
emission estimates into the grid cells a user specifies.
Sierra suggests collecting an extensive number of ground counts for an urban area that has
a state-of-the-art transportation planning model with a 1990 or later base year travel estimate for the
validation of travel estimates. The travel model can then be updated to reflect the road network in
place at the time the ground counts were collected. The model can be validated with the normal
level of ground count information. The results can then be compared with estimates that are based
on the more extensive ground count survey to determine the level of accuracy the model has in
representing travel throughout the urban area. Sierra additionally suggests the validation of MEM
estimates through dynamometer-based measurements of emissions on the same drive cycle. Also,
the information EPA and ARE develop through VEHSME and EMFAC/BURDEN may provide a
basis to evaluate the accuracy of the MEM.
Sierra notes that additional funds will be needed to: (1) develop methods to quantify the
travel occurring within TAZ's; (2) modify or replace the existing models to incorporate the new
methodologies; (3) expand the network to include more information about local roads; (4)
determine the mix of engine maps needed to characterize new vehicle emissions; (5) translate
existing I/M program modeling efforts from an FTP-based emission factor methodology to one that
12
-------
uses modal emissions; and (6) analyze the data collected in the EPA and ARE chase car and
instrumented vehicle surveys.
Desert Research Institute's conceptual design consists of an Emissions Modeling System
(EMS) capable of accessing activity data bases from a multitude of information gathering agencies
(Figure 5). Such an EMS then combines this information with meteorological data and validated
emissions/activity relationships (i.e. emissions factors) to provide emissions rates for any selection
of location, time, or sources.
Species specific
enrnuom rales
Figure 5. DRI Emissions Modeling System (Appendix B, page B-13)
13
-------
According to DRI, today's standard definition of an air quality emissions inventory as a
spatially and temporally averaged listing of presumed emissions is outdated. Consequently, they
believe that an EMS consisting of special-purpose modules which can be updated with new
information and new science when they are available and activity levels and emissions/activity
relationships specific to the area being modeled will be more helpful. An EMS would also contain
error propagation algorithms to provide precision estimates on outputs and use independent activity
data bases and emissions/activity relationships of equivalent quality to estimate accuracy.
Similarly, it would adjust the emissions/activity relationship in response to environmental
variables, especially meteorology, and allow the addition, subtraction, or modification of
emissions for specific days. Finally, it would retain the ability to trace the information to allow
quality auditing and provide output displays, statistics, and data bases which could be used for
modeling, data analysis, and quality assurance. For an EMS which contains all these attributes,
DRI suggests referring to the one being developed by the San Joaquin Valley Air Quality Study
and Atmospheric Utility Signatures, Predictions and Experiments (SJVAQS/AUSPEX) in
California.
DRI offers a two-step method for validating the emissions factor model. The first involves
an operational evaluation of the model including a vehicle fleet test which the experimenter
specifies and controls. The user would apply the model to calculate absolute emissions factors,
and stated uncertainties therein, and compare them with observations of known accuracy,
precision, and validity. DRI believes that this first step will permit reformulation of the model
inputs, given the knowledge of the distribution of modes, speeds, grades, and other operating
conditions, to accurately predict observed on-road results. Step 2 involves testing the adjusted
model in a real-world setting where a user can validate it against actual driving conditions. Here,
remote sensing would be used to tell where model inputs may be off and to yield information
necessary for determining actual instantaneous vehicle emissions rates.
Among the conceptual designs which basically follow the current methodology, Radian
Corporation's conceptual design provides perhaps the most detailed approach to vehicle emission
modeling. Radian believes that its motor vehicle source (MV) model (Figure 6) will accurately
generate hour-by-hour link specific emission estimates for urban areas, based upon the
development of travel demand estimates obtained from a transportation modeling system (TMS)
designed specifically for air quality-related analysis (i.e., on-network travel demand estimates).
According to the contractor, the MV model will also accommodate transportation estimates
developed for rural areas through off-network transportation models. Radian bases its MV model
upon the development of emission factors, on a grams/hr basis, for a pre-determined number of
14
-------
driving cycles or a driving cycle/emission factor matrix for different combinations of roadway
types (including minor collectors and residential streets) and area types (e.g., central business
district, rural, entrance/exit ramps, etc.). This matrix (Figure 7) will be stratified by peak and off-
peak travel periods. Each link in a given transportation network will be classified according to its
use and matched to a particular driving cycle and corresponding emission factor. Emission factors
will be developed for each driving cycle on a per-hour basis, based upon the use of a separate
modal analysis model. Once the composite facility type/area type driving cycle emission factors are
calculated, the emission factors are combined with the vehicle activity data developed from the
transportation models for each link and node. The activity data will include: (1) the total vehicle
hours (VHT) traveled on a link; (2) the distribution of VHT by vehicle type and fuel type for each
link; (3) the number of cold and hot starts for each node; and (4) the number of park hours for each
node (diurnal and hot soak emissions). Radian concludes that the product of the emission factors
and the vehicle activity data will be exhaust and evaporative emission estimates by individual link
and node for each hour of the day.
Figure 6. Radian Motor Vehicle Source Model (Appendix D, page D- 2-2)
15
-------
Central
Business
Urban
Residential
Commercial
Suburban
Rural
Facility Type _^___^^_
Freeway
11*
21
31
41
51
61
Expressway
12
22
32
42
52
62
Major
Arterial
13
23
33
43
S3
63
Minor
Arterial
14
24
34
44
54
64
Major
Collector
15
25
35
45
55
65
Minor
Collector
16
26
36
46
56
66
Residential
Streets
17
27
37
47
57
67
Entrance/
Exit Ramps
18
28
38
48
58
68
' ij matrix cell address for: Driving cycle
Driving cycle (off-peak)//
Figure 7. Driving Cycle Matrix (Appendix D, page D- 3-5)
Radian suggests using tunnel studies, remote sensing and source apportionment techniques
to verify the emission estimates generated by the MV model; and data from fuel sales, traffic count
programs, and satellite imagery to verify its transportation estimates.
Alliance Technologies Corporation's (hereafter Alliance) conceptual design consists of two
basic levels. First, a large modeling system, or "Metamodel" (Figure 8), which consists of a broad
range of integrated computerized methods, procedures, and data sources as well as other types of
components, information, and activities related to vehicle emission estimation. The Metamodel's
purpose is to provide an overall structure within which vehicle inventory methods can be
developed and tested and which data can be processed and stored. Alliance believes that this will
allow for the updating of new methods and data as they develop. The Metamodel consists of a set
of modules that would include computerized methods and data or ongoing research and analysis
16
-------
that would be the basis for potential methods. Alliance concludes that the development and testing
of multiple data sources and alternative methodologies within the Metamodel will permit internal
validation within Metamodel modules and in the combined application of the modules. The
Metamodel would also require a flexible set of methods allowing interaction between modules, and
a set of statistical tools for analysis of Metamodel methods and data.
METAMODEL
MODULES
- Research/Evaluation
Data Development
- Methodology Testing
- Internal Validation/
Verification
ANALYSIS / ABSTRACTION METHODS FOR
APPLICATION MODEL DEVELOPMENT
APPLICATIONS MODELS
SCALE
Global/International
National/U.S
State/Urban Area
Modeling Grid/
Small Scale
MODELS I
- Software Packages
- Specialized
- Simpliliea
- 'Consumer' and
Research/Analysis
EXTERNAL I REAL-WORLD
VALIDATION EXERCISES
Figure 8. Metamodel (Appendix A, page A-18)
The "Applications Models" comprise the second level of Alliance's conceptual design.
These models would include individual software packages developed for specific vehicle inventory
cases. Alliance states that these will be more specialized than EPA's MOBILE4 and relatively
simple in structure and data and therefore easier to understand and apply. The Applications Models
contain data and methods relevant to a specific type of use, with simplifications or default values
17
-------
where the use allows. These would be designed for a specific type of application, but will also be
adaptable by reprogramming them using the Metamodel as source material.
Alliance suggests that in order to have parallel structure between the Metamodel and the
Applications Models, the Applications Models should contain all or most of the same modules as
the Metamodel, simplified to suit the specific application. The modules can be arranged along a
continuum from emissions testing to transportation data, with control strategies and information
sources functioning more as auxiliary components that apply across the rest of the Metamodel.
Alliance says that the modules are not strictly defined as computerized components in a modeling
system, but are more like "activity areas" or ways of dividing the overall field of vehicle emission
estimation into a manageable number of logical pieces which can also correspond to parts of a
modeling system.
Of the six contractors' conceptual designs, Systems Applications International's (SAI)
conceptual design deviates most significantly from the current methodology being used. SAI
suggests a three step interrelated process for more accurately determining vehicle emissions
inventories. First, determine the base case emissions (Figure 9) for an area using an activity factor
model (based not on VMT, but on vehicle operating mode distributions by area and time) and a
modal emissions factor model. Second, for forecasting purposes, synthesize the distributions of
the base case emissions inventory by using transportation planning models and small scale models.
Third, each step in the above process should be subjected to independent verification of its
performance under actual conditions.
Vehicle operating
mode distributions by
area and time
Modal emission
factor model
Emission calculation
Figure 9. Base Case Emission Estimate (Appendix F, page F-12)
18
-------
SAI believes that reliance on overly aggregated representations of both vehicle operations
and emissions is the primary cause of the limitations in the current model. SATs main approach
for fixing these problems is by using substantially increased reliance on empirical inputs in their
proposal. The basic emission calculation is the numerical integration of emission rates, within
designated areas, of a vehicle operating mode distribution and a modal emission factor. The
emission rate for vehicles within area / for hour h is given by: J e(v)dFih(v), where
V
v = a vector describing the state and operating mode of a vehicle (it can consist of basic
operating modes, speed, acceleration rate, canister state, fleet characteristics,
environmental conditions, trip end, etc.)
F,/i( v) = the distribution function of vehicle operating modes within an area i for hour h
evaluated at vehicle state vector v (to be determined by aerial photography and
computer image processing)
e(v) = the emission rate (mass/time) for a vehicle in state v.
Total emissions are given by: ^ e(v)dFih(v) .
i.A v
SAI contends that since the empirical distributions F(v) are static, they cannot be directly
used to develop future year forecasts. In the course of developing empirical distributions through
multiple hours and days of aerial photograph analysis, however, data will be collected that show
the sensitivity of these distributions to variations in total network loading. SAI believes that these
data showing sensitivity to change provide the basis for forecasting. SAI defines a new
distribution F'(v) to allow the use of transportation planning model outputs for forecasting. This
new distribution is a synthesized distribution based on a parameterization using transportation
model outputs (Figure 10). SAI uses the following additional notation:
w = a vector of parameters that are outputs of the transportation planning model and that
correspond to measures of overall network loading in the data base used to develop
F(v)
Gi*.(vlw) = the observed distribution of vehicle operating modes v in area i for hour h,
given that the network state is w
19
-------
= a mathematically constructed distribution whose parameters are given by w
that has been shown to be essentially identical to Gn,(v\w).
Then F'i*(v) = G'rt(vlwo) , where w<> is the "average network state" for the base case. SAI
believes that the development of mathematical forms for G', the "calibration" of F', and the
verification of agreement between F' and F will be a key aspect of developing forecasting
capability based upon this proposed approach. With future conditions denoted with an asterisk *,
the contractor contends that future year base case emissions will be given by: J e * (v)dF'ih * (v) ,
V
where F',**(v) = G
Observed vehicle
operating mode
distributions
1
Statistical analyses
of distribution
dependencies
Transportation model
nrnulauons of range
of base year conditions
Synthesis and evaluation
of vehicle operating
mode distributions
Verification of emissions
from synthesized distributions
Selection of
model parameters
Figure 10. Development of relationships for synthesizing and forecasting
vehicle operating mode distributions (Appendix F, page F-14)
SATs estimation of control effects closely parallels existing practices (Figure 11). SAI
contends that there are three types of controls: (1) those that affect vehicle emission rates (e.g.,
retrofit technologies, reformulated fuels, etc.); (2) those that affect trip generation (e.g., ride
20
-------
sharing measures, growth management, etc.); and (3) those that affect vehicle operations (e.g.,
flow improvements, elimination of drive-up windows, etc.). Depending on the specific measure,
the simulation of controls could involve either the modal emissions factors (e.g., reformulated
gas), the future year transportation model forecast parameters (e.g., employee based trip
reduction), or the relationship between transportation forecasts and the synthesized distribution of
vehicle operating modes.
TnnsponaDon
model forecast
Control
measure
effects
Modal emission
factor forecast
Synthesis of future
vehicle operating
mode distribution
Emission calculation
• Conmil
only.
Figure 11. Development of future year emission forecasts (Appendix F, page F-16)
SAI believes that each step in the estimation process should be subjected to independent
verification of its performance under actual conditions. SAI suggests several measures for
verifying the aerial photography and image processing incorporated in the vehicle operating mode
distributions. First, by equipping marked vehicles with accelerometers, fuel consumption and
emission measurement instruments, and position recording equipment, and verifying the accuracy
with which aerial photographic techniques estimate mode, speed, acceleration, and emissions. SAI
says that this technique can also be used on a "trip" (as opposed to area) basis to identify spatial
differences in operating mode distributions, and as a check on the area wide distributions of
operating modes. Second, by conducting "audits" using video techniques, loop counters, and
other fixed-location data, to determine how accurately the primary method describes roadway
dynamics. Third, by use of video-based license identification techniques to verify the age
distribution of in-use vehicles, and the variability of this distribution by time and location. SAI
21
-------
suggests that remote sensing techniques in conjunction with registration-based identification of
vehicle age and technology can be used to verify the accuracy of the modal emissions factors.
REFERENCES
Wilson, J. H., Jr. Proceedings of Two Highway Vehicle Emission Inventory Workshops.
EPA-600/9-91-007 (NTIS PB91-168492), U.S. Environmental Protection Agency,
Air and Energy Engineering Research Laboratory, Research Triangle Park, NC, March
1991.
22
-------
APPENDIX A
Technical Memorandum:
Conceptual Design
for a New Highway Vehicle
Emission Estimate Methodology
Prepared by:
Mark Smith
ALLIANCE TECHNOLOGIES CORPORATION
100 Europa Drive
Chapel Hill, North Carolina 27514
EPA Contract No. 68-D9-0173
Work Assignment 1/129
Project Officer:
Carl T. Ripberger
Air and Energy Engineering Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
JULY 1991
A-i
-------
CONTENTS
Page
Introduction A-l
Basic Concepts A-l
The Metamodel A-l
The Applications Models A-2
The Modules A-4
Potential Additional Modules A-13
Metamodel Integration A-14
Analysis/Abstraction Methods A-14
Potential Applications Models A-15
Validation Methods A-15
Implementation of the Metamodel/Applications Model Concept A-16
Attachment A: Selected References for Highway Vehicle Emission Estimation A - A-l
Figures:
1. Proposed framework for highway vehicle emission inventory
methodology development A-18
2. Metamodel modules A-19
3. Potential Applications Models for highway vehicle emission inventories A-20
4.a. Validation approaches for highway vehicle emission inventory methods —
within modules A-21
4.b. Validation approaches for highway vehicle emission inventory methods —
multi-module internal validation A-22
4.c. Validation approaches for highway vehicle emission inventory methods —
external vali dation exercises A-23
Tables:
1. Characteristics of the Metamodel and Applications Models A-3
A-ii
-------
INTRODUCTION
In response to the EPA request for a "Conceptual Design for a New Highway Vehicle Emission
Estimate Methodology,' this paper outlines an overall approach and structure that can be used in
long-term planning for highway vehicle emission inventory methodology improvement The first
section describes the basic concepts of the proposed approach. The next section describes the
modules that would initially comprise the model set, including interactions between them, potential
inputs to and outputs from each module. Following sections address potential additional modules,
integration of modules and development of specific Applications Models, validation methods, and
implementation of this conceptual design.
BASIC CONCEPTS
The conceptual design proposed here consists of two basic levels. One level is the Metamodel, and
the other consists of a number of Applications Models. These levels are depicted in Figure 1, and are
outlined briefly below. (Figure 1 is a schematic; more details are provided in later figures and text.)
Table 1 provides a comparison of the essential characteristics proposed for the Metamodel and the
Applications Models. Further details are provided in following sections.
The Metamodel
The Metamodel is a large modeling system which consists of a broad range of integrated
computerized methods, procedures and data sources as well as other types of components,
information and activities related to highway vehicle emission estimation. The prefix 'meta-1 was
chosen because one of its definitions is "more comprehensive: transcending.1
The purpose of the Metamodel is to provide an overall structure within which highway vehicle
inventory methods can be developed, tested and evaluated, and in which data sources related to
highway vehicle inventory development can be assessed, processed and stored. It could be
considered a clearinghouse or repository for data and research related to highway vehicle inventories
as well as a framework or tool for research, analysis and creation of inventories and inventory
methods.
The Metamodel structure provides the opportunity to maintain and update methods and data for
highway vehicle inventories as new developments in mobile source emissions and the state of the art
of understanding of how these and other aspects affect highway vehicle emissions. The Metamodel
would be structured and planned to allow incorporation of additional data sources as new data
become available over time, or existing data become relevant, due to changes in mobile source
inventory needs, transportation systems, vehicles, fuels or other factors.
The Metamodel would consist of a set of modules that could each include components ranging from
highly articulated computerized methods and data to ongoing practical and theoretical
research/analysis that would provide the basis for potential methods. At full development, it would
include the results of all relevant research as well as documentation of the derivation/basis for the
computerized portions of the Metamodel.
Within the Metamodel, multiple data sources and alternative methodologies for highway vehicle
inventories can be developed, tested, compared and documented. To the extent that information and
methods allow, this would permit internal validation/verification as well as sensitivity analyses of
A-l
-------
potential options both within individual Metamodel modules and in the combined application of these
modules.
In addition to the individual modules, the Metamodel would also require:
(1) a highly flexible set of methods and procedures allowing interaction and coordination
between the modules, and
(2) a set of statistical and modeling tools for analysis and abstraction of Metamodel data and
methods, which would also be used in the development of the Applications Models.
Applications Models
Applications Models would be individual software packages developed for specific highway vehicle
inventory cases. They would be somewhat more specialized than the current MOBILE emission factor
model series, which has been adapted to allow use in a variety of different contexts. By being
restricted to more limited uses, the Applications Models can be relatively simple in structure and data
requirements, and can be easier to understand and apply. These Applications Models would include
both 'consumer1 versions for regulatory highway vehicle inventory requirements (such as State
Implementation Plans) as well as 'research and analysis' versions that could be used to provide
mobile source inputs to environmental assessment programs and regulatory development efforts at
the national or international scale. In addition, special-purpose Applications Models could be
developed for model verification/validation exercises, by extracting from the Metamodel the best
available information and methods for the specific situations being used in a given exercise.
The Applications Model concept is somewhat analogous to the number of models currently available
for simulation of ambient air quality for different purposes, at different scales and under different
conditions or levels of data requirements.
A-2
-------
TABLE 1. CHARACTERISTICS OF THE METAMODEL AND APPLICATIONS MODELS
METAMODEL
APPLICATIONS MODELS
Comprehensive - Incorporating all available
state-of-the art knowledge related to highway
vehicle inventory methods and data
Flexible - Capable of accepting new modules,
methods and/or data at any time.
Redundant system design - Multiple methods
and data sources wherever possible, for
comparison and validation.
Mainframe or large minicomputer-based -
Combination of standard programming
language (e.g., FORTRAN, to interface with
MOBILE models) and analytical/data-handling
capability (e.g., SAS).
Application-oriented - Contains data and
methods relevant to a specific type of use, with
simplifications or default values where the use
allows.
Relatively "hard-wired* - Designed for a
specific type of application, but adaptable by
reprogramming using Metamodel as source
material.
Minimal options for methods and data -
Using only what is necessary to provide the
outputs for the desired application; few
alternative methods, alternate data sets or
inputs only as required.
Personal computer-based - User-friendly
model structures based on FORTRAN, or
dBase or spreadsheet formats where feasible.
A-3
-------
THE MODULES
The intent of the Metamodel/Applications Model concept is to have a parallel structure between the
Metamodel and each of the Applications Models. Figure 2 presents an expansion of the Metamodel
section of Figure 1, showing basic modules envisioned as a starting point for all models. While this
Figure is specific to the Metamodel, most Applications Models would involve some data and methods
from each of the Metamodel modules, and thus would have all or most of the same basic modules,
simplified to suit the specific application. As indicated in Figure 2, the model structure and software
would be designed to accommodate additional modules which might be created in the future. While
Figure 2 does not attempt to specify relationships between the modules (see individual module
descriptions for this), it does arrange the modules along a continuum from emissions testing to
transportation data, with control strategies and information sources functioning more as auxiliary or
overlay-type components that apply across the rest of the Metamodel.
The following pages provide a general description of the purposes, outputs, data requirements and
sources, and interactions with other modules that might characterize each module. These
descriptions are not intended to be complete or definitive, but rather to suggest possibilities in the
area each Module might cover. They are based on current understanding of factors and data sources
relevant to highway vehicle emission estimation, and thus would be expected to change with more
detailed examination of system requirements and interactions as well as with the availability of new
data sources in the future. Where possible, potential future data sources have been noted as well as
those currently available.
As indicated earlier, these modules are not strictly defined as computerized components in a modeling
system, but might best be thought of as 'activity areas,' or ways of dividing the overall field of highway
vehicle emissions estimation into a manageable number of logical pieces, which can also correspond
to (or contribute to) parts of a modeling system.
Attachment A lists citations and brief abstacts of many of the data sources mentioned in the individual
module descriptions below.
A-4
-------
MODULE: Emissions Test Data / Emissions Relationships
PURPOSES: To be a primary repository/index of available data from emissions testing programs
(U.S. and worldwide)
Provide a set of alternative structures by which emissions data can be used to develop
relationships between emissions and predictive variables (activities and conditions)
Supply state-of-the-art predictive relationships (emission rates or equations) to the
Applications Models
OUTPUTS: Emission rates, equations or other relationships for the entire universe of vehicle-
related emissions (including, but not limited to: speciated hydrocarbons, NOX, CO,
PM-10, CO2, CFCs, toxics, asbestos and engine fluids)
Assessments of adequacy of existing emissions testing procedures and databases
DATA REQMTS/SOURCES:
Emissions test data from all potential sources, including OMS and
CARB regulatory and research programs, Certification testing for
vehicle and engine families, private/industry sources, State I/M
programs
Additional data on the conditions, techniques and assumptions used
in each set of tests
INTERACTIONS: Extensive interaction with Vehicle Fleet module to direct testing and analysis
toward adequate coverage of current and future fleets
Extensive interaction with Driving Behavior module to improve identification of
relevant variables for emissions estimation, and to isolate and quantify the effects
of these variables in emission testing as well as in development of emission rate
prediction tools
Interaction with the other modules to collect and assess potential activity and
condition data sources, and to develop sensitivity to emissions control strategies in
the emission relationships
COMMENT: Although vehicle emissions testing itself is not considered part of this module, one
function of this module would be to combine available test results with explanatory
data and assess the ways in which current testing procedures and databases do not
meet the needs of the predictive process. Thus part of the analytical work done in this
module would be to identify weaknesses in the current testing procedures and
datasets. This would result in suggestions regarding improvements that can be made
to testing programs and methods.
A-5
-------
MODULE: Vehicle Fleet Characteristics / Technology / Fuels
PURPOSE: To provide a sufficiently-detailed description of the highway vehicle fleet to support
analysis, development and implementation of state-of-the-art emissions relationships.
This would include current and historic databases as well as the ability to create and
apply alternate future fleet scenarios.
OUTPUTS: Fleet profiles for analysis years from 1920 to 2050, broken down by vehicle type, fuel,
technology and county
Data on vehicle histories and use (mileage accumulation, tampering with control
devices, misfueling and other factors that may influence emissions characteristics)
DATA REQMTS/SOURCES: R.L Polk and Co. registration information
Mileages, tampering and other data from state I/M programs
Additional data from MVMA and follow-up activities by vehicle
manufacturers
Vehicle history and use data should begin to be available from on-
board vehicle monitoring systems (data dumps from future 'black
boxes') - could be collected as part of I/M program data acquisition or
through manufacturers' recall/follow-up efforts
INTERACTIONS: Primary interactions with the Emissions module (defining the universe that
emissions relationships need to describe, supplying the data to characterize the
emitting fleet in a specific year and location)
Supplying fleet data to the Driving Behavior module and the Transportation Activity
module to support analyses of what behaviors and activities are relevant to
emissions prediction
Interaction with Emissions Control Strategies in development of
vehicle-type-spectfic control strategies and control strategies that consist of
introduction/encouragement of specific technologies or vehicle types
A-6
-------
MODULE:
Driving Behavior / Modes and Transients
PURPOSES: To facilitate the development of improved understanding of (1) what aspects of driving
behavior can affect emissions rates, (2) how the behavior of individual drivers can
influence emissions, and (3) how driving behavior can be characterized to produce
maximum sensitivity to emissions-determining aspects
To develop and maintain a primary database of driving behavior that can be used in
emissions testing/research and in development of methods to predict emissions
OUTPUTS: Analyses and data on driving behavior that can be used to guide emissions testing
and research concerned with sensitivity to driving behavior and the reaction of various
technologies to driving behavior
Databases and characterizations of driving behavior (e.g. statistically-derived functions
or profiles) for use in Applications Models
DATA REQMTS/SOURCES:
Existing and on-going studies of how drivers use the transportation
network (e.g. chase car studies and other on-road behavior
characterization work) and how different vehicles react to specific
aspects of driving behavior (testing heavy accelerations, low and high
speeds and other possible transient high-emission patterns)
Future programming and retrieval of "black box1 on-board recordings
of actual driving patterns on a large scale
Coordinated studies of areas believed to be characterized by specific
driving behaviors or modes of interest, such as joint monitoring of on-
road speed/acceleration and localized ambient measurements
INTERACTIONS: Primary interaction with Emissions module, supplying behavior data and receiving
emissions test data to analyze for behavior-sensitive aspects
Additional interaction with the Transportation Systems and Activity modules, to
coordinate the development and analysis of transportation data that can be used
to predict important aspects of driving behavior, and with the Vehicle Fleet module,
in assessment of the interaction between technologies and driving behavior.
COMMENT: This module would concentrate on alternative methods and data that can be used to
describe driving behavior, including the current typical behavior Federal Test
Procedure approach, and other methods that allow discrete assessment of individual
driving modes and transient events that are becoming more important with the
improvement of our understanding of emissions, driving styles and vehicle technology
(especially engine control systems and programming).
A-7
-------
MODULE: Transportation Systems - Networks, Data, Planning
PURPOSES: To develop transportation system information to support the other modules
To develop improved methods to simulate and assess system/facility-based
transportation control measures
To promote the integration of transportation planning and emissions estimation and
control (e.g. development of an Applications Model that can be attached to urban
transportation models for use in system planning)
To provide support for gridding/spatial allocation of emissions estimates
OUTPUTS: Characterization of highway systems with respect to emissions-related parameters
(speeds, slopes, lane widths and configurations, intersections/signals or other items
that may be identified as significant in future research, as well as systems-related
transportation control measures)
DATA REQMTS/SOURCES: Transportation system networks and related data developed for
local/corridor areas (project studies), urban/regional simulations
(UTPS-style network models) and national applications (Federal-Aid
network data and Highway Performance Monitoring System county-
level roadway files)
Advanced data acquisition concepts such as (1) use of the U.S.
Geological Survey "GeoData" digitized cartographic information system
to develop special-purpose datasets on area-specific road system
extent and characteristics (e.g. slopes derived from topography - see
attached summary of the GeoData project) and (2) use of remote
sensing combined with Global Positioning System for integrated
collection of highly detailed data on roadway and local conditions
(traffic flow/activity). The latter is a concept in development by FHWA,
which could be expanded to include collection of data on other areas
of concern such as driving behavior, vehicle function and on-board
emissions monitoring.
INTERACTIONS: Primary interactions would be with the Transportation Activity module, providing the
basic system description information for activity modeling, and with the Driving
Behavior module, providing data on specific transportation system characteristics
that affect driving behavior, as well as interacting with both of these modules to
identify data needs and improve the available database.
A-8
-------
MODULE: Transportation Activity Data and Modeling
PURPOSES: To provide an analytical structure for the development and assessment of alternative
activity measures for emission estimation
To develop and maintain an up-to-date set of basic reference statistics and tools for
analysis of activity measures and for support activity data needs of Applications
Models
OUTPUTS: Activity measures for use with Emissions module relationships in Metarhodel exercises
and in Applications models
Assessments of the utility, accuracy and other aspects of individual and combined
activity measures for highway vehicle emission estimation, and alternate methods for
activity measure development and application
DATA REQMTS/SOURCES: "Standard" sources of transportation activity data (VMT, trip ends and
traffic/congestion data), including urban transportation modeling and
traffic counts in individual urban areas, Highway Performance
Monitoring System (HPMS) data as reported in Highway Statistics and
HPMS data files, and periodic State/local transportation agency studies
and statistical summaries
A number of potential sources of data which currently appear to be of
marginal utility but which could be utilized with adequate analysis
and/or some changes or additions to the focus of the data-gathering
efforts. These include the Bureau of Census Urban Transportation
Planning Package (a sample of the Decennial Census), the Department
of Energy Residential Transportation Energy Consumption Survey, and
the Federal Highway Administration National Personal Transportation
Study series (See second COMMENT, below.) These sources are
concerned mainly with trip-making behavior, fuel use/refueling
behavior, travel time/distance and associated demographic and
economic variables.
Highway fuel use data, available at the state level and occasionally at
the county level from state tax authorities
Population, income, and employment data from the decennial Census
of Population, at levels as fine as Census blocks
Vehicle population data, available from individual state Departments of
Motor Vehicles but more readily in a national database from R.L Polk
and Company.
Vehicle usage data - potential sources include development of
mileage accumulation databases from odometer readings collected by
I/M programs and "black box* on-board data storage systems
A-9
-------
INTERACTIONS: Primary interaction would be with the Emissions module, in the use of activity
measures with emissions relationships to estimate emissions, and with the
Transportation Systems module, to integrate the available activity data into the
physical landscape of the city or nation for analytical or allocation purposes
The Transportation Activity module data and methods would also have to be
coordinated with each of the other modules, since activity measures would have to
be related to descriptions of driving behavior, the specific vehicle types involved
and emissions control strategies designed to affect activity levels.
COMMENTS: Since activity levels are, along with emissions rates or relationships, the key
parameters used in calculating emissions, it is especially important not to restrict the
types of activity levels that could be applied in future emission estimation methods, or
the research that will lead to these methods and data sources. In particular, there are
numerous logical possibilities for the development of Applications Models that depend
on different activity variables, or that use multiple variables to predict different types of
emissions. One current case in point is the use of miles travelled to estimate exhaust
and running loss emissions and trip-related data to estimate start and diurnal
evaporative emissions. In particular, estimation of 'unconventional' emissions such as
CFCs from air conditioning systems or fluid losses from other engine components
might be much less dependent on vehicle use than other factors such as weather or
maintenance practices. Especially important will be the development of activity indices
that can be coordinated with future data on highly specific transient or modal
emissions phenomena
In all the cases of existing surveys mentioned above, the agencies collecting the data
are amenable to addition of items to the survey instruments and expansion of the
samples or areas of coverage (for a price). Significant lead time is necessary for such
efforts, and surveys are notoriously expensive and subject to major concerns of
sampling and validity at different geographic scales. However, the time horizon and
possible scale of this research effort indicates that each of these possibilities should
be explored in depth.
A-10
-------
MODULE:
Emissions Control Strategies
PURPOSES: To allow simulation of the potential effects of highway vehicle-related control strategies
on emissions from the local to national and global scales
Control strategies of concern would include the traditional' I/M and anti-tampering
programs and activity- and facility-related transportation control measures, but would
also include fuel-based strategies (oxygenated, reformulated and alternative fuels),
vehicle- and technology-based strategies and economic or market-based incentive or
penalty programs
OUTPUTS: Descriptions of potential control strategies and their effects on activity or emission
levels, including 'mini-models' of the effects of specific program parameters on overall
effectiveness or penetration of the program
DATA REQMTS/SOURCES:
Technical studies of individual and combined emission control
programs and strategies, as well as application of these data within
the proposed Metamodel framework to generalize from the results of
specific studies
INTERACTIONS:
This module would be used to develop descriptions of the predicted effects of
emissions reduction/control strategies that can be applied in any of the preceding
modules, depending on the type of strategy. Potential control strategies may cover
the entire range of modules, but individual control measures will typically fall mainly
in the area covered by one or two modules. For example, I/M programs would be
described by changes in Vehicle Fleet Characteristics, whereas integrated
transportation control measure programs would involve Transportation Systems
and Activity and possibly Driving Behavior.
A-ll
-------
MODULE: External Conditions / Information Sources
PURPOSES: Provide input data required by other modules and by 'consumer-level' Applications
Models
Interface with original sources of data
Acquire and process data on 'standing order"
Maintain current and historic data files
Provide analyses of specific data sources to reduce data requirements
OUTPUTS: Data for other modules and models (see below)
DATA REQUIREMENTS/SOURCES: As identified in the development of other modules and models
COMMENTS:
This module would serve as a pre-processor and data acquisition utility for periodic or irregular data
sources that have been defined in other modules and models. It would include a tracking and
ordering system which would track and verify when revised versions of current module datasets
become available and then obtain the updated data and process it for Input to the relevant modules.
It is anticipated that the types and amounts of data that can be accessed and processed in this way
would increase as the level of activity in motor vehicle emission estimation increases and as existing
data sources are transformed into more useful tools in this area One particular function of this
module would be to assemble and make available data from numerous sources required for
ozone/CO nonattainment area mobile source inventories that would otherwise have to be collected
individually by each state or local agency.
Examples of current data sources that would be handled in this way include:
. Annual data sets from the Highway Performance Monitoring System, Highway
Statistics, MVMA annual U.S and World Motor Vehicle Data, and other sources.
Other periodic or irregular data sets such as surveys of travel behavior by U.S. Bureau
of Census (the Transportation Planning Package), FHWA and DOE.
Additional data resource/processing opportunities include collection of ambient
monitoring data from the EPA National Air Data Branch and climatic data from the
National Weather Service Climatic Data Center, and joint processing of these data to
provide statistical abstracts which describe the temperature and other conditions on
days with high ambient pollutant levels. This could be a generalization of the current
requirement for development of temperature data for high-level ozone days for MOBILE
model applications.
A-12
-------
POTENTIAL ADDITIONAL MODULES
The conceptual and data-processing structure of the Metamodel would allow addition of modules as
the need arises. For example, entirely new vehicle or fuel technologies may require other capabilities
or raise new concerns as to environmental impacts that have not been addressed in the modules
discussed above. New data-development possibilities will exist in the future that cannot be predicted
today. Two specific additional modules are discussed briefly below.
One additional module that could be implemented in the near term is a separate "Fuel Use/Fuel
Economy1 module. While the Transportation Activity and Vehicle Fleet Characteristics modules would
include these types of data and analyses, the mission of the overall modeling structure could easily be
expanded to include modeling and projection of fuel use and fuel economy. This will become more
important as the number of alternative fuels and the levels of their usage increase in the next ten
years. This module could be used to do independent analyses of fuel-related issues, using the
supporting data and methods already available in the rest of the modeling structure. The draft
Alliance report 'Methods and Data for Developing Vehicle Miles Traveled and Speed for Highway
Mobile Source Emission Inventories' describes and reviews a collection of references which include
methods and data relevant to this area, which are listed in Attachment A.
Another potential additional module which would relate to a Fuel Use/Fuel Economy module would be
a Fuel Production and Distribution module. In the current context of an oil-dominated transportation
system, this module would integrate desired aspects of the oil production, refining and distribution into
the modeling structure, and have the capability to simulate the effects of transportation, vehicle and
fuel-related policies and technologies. This module could provide a tool for looking at the full picture'
of secondary and tertiary emissions that are related to the present transportation system, including
(1) fuel storage, transportation and distribution; (2) exploration, extraction and refining; (3) construction
and maintenance of the transportation network itself. This module could take advantage of the
numerous models and extensive databases that already exist in each of these areas, focusing on the
ways in which those models can be used to address emissions-producing activities. In addition, this
module could take advantage of available and ongoing EPA and DOE research on the entire
fuel-related economic sector, including investigations of air emissions and other environmental effects.
In addition, a Fuel Production and Distribution module would provide the structure and
transportation/energy demand data to model similar aspects of alternative fuels. In the case of
alternative liquid/gas fuels, some of the oil-based system could be used, with additions to account for
different methods of production or delivery. In the case of electric vehicles, fuel cells, photovoltaic
hydrogen, various hybrid vehicles and other developmental technologies not as tied to the oil-based
economy, the ability to simulate effects on utility demands and other energy sources becomes more
important. Developing a module that can address this level of analysis is obviously a major
undertaking, potentially the size of the rest of this model structure. On the other hand, the fact that
future changes in the transportation fleet/system in individual areas or across the nation may direct
the major environmental impacts of the system away from on-road emissions and towards fuel/energy
production and distribution systems is highly relevant to the overall modeling concept being discussed
here.
A-13
-------
METAMODEL INTEGRATION
Integration of the modules and individual components of the Metamodel can be thought of on at least
two levels:
• Direction and facilitation of the research and analysis efforts that comprise the front
end' of the Metamodel, developing and applying an overall perspective as to the
missions and functions of the entire modeling system (Metamodel and Applications
Models), of related research and regulatory efforts, and of appropriate interfaces
between external activities and modeling system development.
• A highly-flexible set of analytical methods and data-processing procedures allowing
interaction and coordination between the computerized portions of the modules.
Although there may be some individual areas in which specialized software or
database management tools could be applied (e.g. existing modeling systems,
relational databases), it would be highly appropriate to establish a common language
(e.g. FORTRAN) and documentation procedures as standards across the entire
Metamodel/Applications Models system. This would be especially critical if different
components are to be created by groups in a number of different locations and
organizations.
In addition to providing for the basic functioning of the Metamodel at the data-
processing level, this integrative structure would include a capability for conducting
sensitivity and uncertainty analyses across all Metamodel modules, which can be used
on an ongoing basis to identify areas most in need of support and assessment
ANALYSIS/ABSTRACTION METHODS FOR APPLICATIONS MODEL DEVELOPMENT
Creation of Applications Models from Metamodel components will require a similarly flexible (yet
uniform) set of statistical and modeling tools for analysis and abstraction of Metamodel data and
methods. For simplicity of analysis, use of one basic package of analytical/statistical software such as
SAS should be encouraged wherever possible. In some cases, the appropriate tools might be
special-purpose versions of software already developed within the Metamodel. In either context, the
Applications Model concept would be implemented by:
• developing criteria for the proposed application
identifying Metamodel modules and sub-components that are relevant
• assessing the levels of sensitivity of the proposed application to the relevant
Metamodel methods and data
identifying ways in which full-blown Metamodel components can be abstracted or
simplified for the particular application
• developing and testing an Applications Model based on the above information
A-14
-------
POTENTIAL APPLICATIONS MODELS
Figure 3 shows the lower portion of Figure 1, expanded to show a number of candidate Applications
Models that this concept could support. The specific models shown here represent the current range
of applications of highway vehicle emission estimation techniques, with a few other developmental or
anticipated applications. These include two general types of models:
'Consumer-level1 models such as those designed for development of ozone and CO
State Implementation Plan inventories, for the integration of emissions estimates into
urban transportation planning, or for most other smaller-scale applications. The major
exception is specialized Applications Models created for use in local-scale studies for
emissions methodology validation.
Larger-scale models which are generally oriented toward research or policy/regulatory
analysis goals.
Determination of the need for specific Applications Models would be a function of the eventual
organization and support available for activity in this area. Initial development of any given type of
Applications Model would be followed up by tracking of Metamodel method and data changes that
affect existing 'active1 Applications Models, and periodic updating of relevant parts of those models or
'regeneration1 of the models or data sets as appropriate.
VALIDATION METHODS
The proposed Metamodel/Applications Models structure would allow a variety of opportunities for
validation/verification efforts regarding highway vehicle emissions estimates and the supporting
models and data sources. Figures 4a, b and c are schematic depictions of three general types of
validation exercises that are possible, as follows:
Within Module - Figure 4a shows a situation in which multiple methods or data
sources can be used for assessments or calibrations of methodologies within a single
module. This example might involve comparisons of alternate trip-generation
equations or travel-time estimation procedures.
«
• Multi-Module Internal Validation - Figure 4b shows a situation in which two or more
modules are used to produce alternative emissions or activity estimates using multiple
methods and/or data sources. This example might involve assessment of the
applicability or sensitivity of different transportation system models for the analysis of
possible transportation control measures. Other modules might be involved in this
type of analysis, but they would be held static while the three indicated modules
(Control Strategies, Transportation Systems and Transportation Activity) would be
exercised as data and methods would allow.
• Extemal/'Real World' Validation Exercises - Figure 4c shows the potential flows for a
generic external or Teal world* validation exercise. Such an exercise would typically
include one or more of the following:
(1) a set of direct measurements (on-site tailpipe or full-scale laboratory) or
remote measurements of vehicle emissions,
A-15
-------
(2) ambient measurements (e.g., in a tunnel or garage, along a roadway or
around an indirect source), or
(3) measurements/collection of 'ground-truth" data on some critical element in
the emission estimation process (e.g., actual acceleration rates, vehicle or
speed distributions, trip-making behavior, trip reduction program
participation).
The three paths in Figure 4c would apply to validation efforts that use the types of measurements
listed above.
Few 'consumer-level1 Applications Models would produce data that could be used as direct
comparisons to measurements which are actually possible in the field. Most validation efforts have
essentially been 'point estimates,' performed in only a few specific locations which have very specific
characteristics. Even the modeling systems which would be developed for the most localized
Applications Models would necessarily be based on aggregate descriptions of driver behavior,
transportation system characteristics, etc. This is why a separate type of Applications Model for local-
scale validation studies is proposed. Such a model (or alternate models) would be developed from
the full array of data and techniques available in the Metamodel which apply to the specific situation in
the validation exercise. Additional existing Applications Models could be run in parallel with the site-
specific model for comparison purposes. Since the site-specific model might not have parallel
methods, data or structure to the other Applications Models, one of the most important stages of this
type of validation exercise would be determining how to generalize the conclusions to the rest of the
modeling system. This would be done by determining how the validation exercise reflects on
individual elements of the Metamodel, identifying improvements that can be made to Metamodel
methods and data, and then examining all existing Applications Models to determine the utility of
changes to 'consumer-level* models.
IMPLEMENTATION OF THE METAMODEL/APPUCATIONS MODEL CONCEPT
A number of possibilities exist for implementation of the concept described in this proposal. As a
practical matter, the development of the Metamodel and Applications Models would best be performed
by one organization which could also perform and coordinate related research as well as providing
support functions for users of both model sets. Either the EPA Office of Mobile Sources or Office of
Research and Development could undertake this effort. As an alternative, EPA funding and funding
from industry could be provided to an independent organization which could be created or enlisted for
this role. A Center for Research in Transportation and the Environment could be created as a free-
standing entity or could be affiliated with a group such as the U.S. DOT Transportation Systems
Center in Cambridge or the Institute for Transportation Research and Education (a multi-state institute
affiliated with the University of North Carolina).
The basic mission described in this conceptual design could involve a staff of at least ten to twenty
technical and data processing professionals beyond those currently engaged in related work at OMS,
ORD, the California Air Resources Board and other agencies. If this staff were also involved in
research and other functions, additional personnel would be required. The most critical area in
personnel and activity requirements would be the level of support provided by OMS and how the
current emissions data analysis and inventory-related functions could be divided if the models were
not developed and maintained directly by OMS.
While efforts at the level described here may appear to be a major increment to the current activities
in this area, there are some specific aspects that should be taken into account:
A-16
-------
Research and regulation in this area are currently experiencing a major increase in
investment and activity due to Clean Air Act initiatives and technology-forcing
provisions. The effective coordination, assessment and dissemination of research
results will require similar increases in effort, and the framework proposed here would
be an ideal way to accommodate such rapid growth in a highly specialized technical
area
Specific aspects of the proposed Metamodel framework, such as development of
simpler and better-integrated Applications Models and maintenance of databases for
access by researchers and Applications Model users, will actually reduce the time and
effort required to perform the more conventional and widely performed types of
highway vehicle inventories.
A-17
-------
FIGURE 1. PROPOSED FRAMEWORK FOR
HIGHWAY VEHICLE EMISSION INVENTORY
METHODOLOGY DEVELOPMENT
METAMODEL
MODULES:
00
ANALYSIS/ABSTRACTION METHODS FOR
APPLICATION MODEL DEVELOPMENT
APPLICATIONS MODELS
SCALE:
Global/International
I I
Natlonal/U.S.
State/Urban Area
i r
I r
- Research/Evaluation
- Data Development
- Methodology Testing
- Internal Validation/
Verification
Modeling Grid/
Small Scale
MODELS:
I I I I 1 I I I I I I I I I I I I I
EXTERNAL / REAL-WORLD
VALIDATION EXERCISES
- Software Packages
- Specialized
- Simplified
- "Consumer* and
Research/Analysis
-------
FIGURE 2. METAMODEL MODULES
Mt IMMUUtL
|
Emissions
Test Data/
Emissions
Relationships
FUTURE
MODULES
Vehicle Fleet
Characteristics/
Technology/
Fuels
Driving Transportation
Behavior/ Systems —
Modes and Networks, Data,
Transients Planning
Transportation
Activity
Data and
Modeling
Emissions
Control
Strategies
External
Conditions/ Add as needed....
Information
Sources
ANALYSIS / ABSTRACTION METHODS FOR
APPLICATION MODEL DEVELOPMENT
-------
FIGURE 3. POTENTIAL APPLICATIONS MODELS
FOR HIGHWAY VEHICLE
EMISSION INVENTORIES
POTENTIAL
APPLICATIONS MODELS
SCALES:
Global
Global Foreign-
Warming Country
(CO2, CO. Simulations
CFCs) (All Types)
Fuels/
Energy
Policy
Studies
/ International
Integrated
Economic/
Environmental
Analyses
Add Rain
Studies,
Policy.
Controls
Natlonal(U.S.) / Regional
National
Vehicle
and Fuels
Policies
Regional
Oxldant
Modeling
& Control
Multi-State/
Trans-
boundary
Issua*
EXTERNAL 1 REAL-WORLD
VALID A HOW EXERCISES
Ozone and CO
NAAQS Non-
Attainment
& Strategies
State /.Urban Area
Integrated
Air Toxics
Studies/
Strategies
Urban
Transportation
System
Planning
i
Support
to G ridded
Air Quality
Models
Modeling Grid / Small Scale
1
Intersection/
indirect
Source
Modeling
Local-Scale
Ambient
Validation
Studies
-------
FIGURE 4a. VALIDATION APPROACHES FOR
HIGHWAY VEHICLE EMISSION
INVENTORY METHODS —
Within Modules
METAMODEL
MODULES:
~
Transp.
Activity
Comparison of
Alternative Methods/Data
ANALYSIS/ABSTRACTION METHODS FOR
APPLICATION MODEL DEVELOPMENT
APPLICATIONS MODELS
SCALE:
Global/International
Natlonal/U.S.
State/Urban Area
I r
I r
Modeling Grid/
Small Scale
MODELS:
11 i i i ii r i L_J
11 11
1 I I I
I I
1
EXTERNAL I REAL-WORLD
VALIDATION EXERCISES
-------
FIGURE 4b. VALIDATION APPROACHES FOR
HIGHWAY VEHICLE EMISSION
INVENTORY METHODS —
Multi-Module Internal Validation
METAMOD
MODULES:
L
EL
V V \
Transp.
Systems
f
Transp. Control
Activity Strategies
V V \
JOINT EXERCISE
f
ANALYSIS /ABSTRACTION METHODS FOR
APPLICATION MODEL DEVELOPMENT
APPLICATIONS MODELS
sc
MODELS:
ALE:
Global/International
I i i
1
1 1
I 1
1
Natlonal/U.S. State/Urban Area Modeling Grid/
Small Scale
1 1
1
1 1
i i
1 1
\ \ \
]
I 1
i i i
EXTERNAL / REAL-WORLD
VALIDATION EXERCISES
-------
FIGURE 4c. VALIDATION APPROACHES FOR
HIGHWAY VEHICLE EMISSION
INVENTORY METHODS —
External Validation Exercises
METAMODEL'
MODULES:
Emission
Rates
Transp.
Activity
(Measurements
vs. Predictions
ANALYSIS/ABSTRACTION METHODS FOR
APPLICATION MODEL DEVELOPMENT
APPLICATIONS MODELS
SCALE:
Global/International
I I
National/U.S.
State/Urban Area
Modeling Grid/
Small Scale
I i
\
MODELS:
III I
I I II T I I I
I I
J I
J I
EXTERNAL / REAL-WORLD
VALIDATION EXERCISES
Emission Predictions
vs. Ambient Levels
Activity Level
Predictions
vs. Local Data
Feedback lo Entire Metamodel
-------
A-24
-------
ATTACHMENT A:
SELECTED REFERENCES FOR
HIGHWAY VEHICLE
EMISSION ESTIMATION
A-A-1
-------
TRANSPORTATION PLANNING AND NETWORK
MODELING
1. Characteristics of Urban Transportation Systems. Federal Highway Administration, U.S.
Department of Transportation, Washington, D.C., October 1985.
The document provides a source of data on the most important characteristics of urban
transportation systems. The document assesses .only the supply or performance
characteristics of urban transportation systems. Chapter 4 contains a set of quantitative
values for selected supply parameters used to characterize automobile-highway systems
(i.e. auto and truck traffic): speed, capacity, operating cost, energy consumption, pollutant
emissions, capital cost, and accident frequency.
2. Characteristics of Urban Transportation Demand, An Update. Report prepared by Charles
River Associates, Inc. for Urban Mass Transportation Administration, Contract No. DOT-T-
88-18. Washington, D.C., July 1988.
The report presents a selection of updated data on a wide variety of statistics pertaining to
urban travel demand. The information supplements earlier data contained in the UMTA
handbook, Characteristics of Urban Transportation Demand - A Handbook for Transit
Planners. The report is designed to be used by transportation analysts as a source of
data to check the validity of local forecasts developed from traditional planning studies or
as a cross-check on the similarity of travel statistics from one locality to another. Certain
data also may be used as default values for modeling purposes, when such information is
not available locally or would require new or extensive data collection efforts. Much of the
information presented was obtained from reports, documents, and memoranda produced
by or for each study area contacted. A main criterion of the study was that the
information collected be based on surveys, measurements, counts, etc., and not be
synthesized results from analytical modeling efforts.
3. UTPS Highway Network Development Guide. Report prepared by Cosmis Corporation for
the Federal Highway Administration, Washington, D.C. January 1983.
The purpose of the report is to aid the transportation planner in highway network analysis.
The tools for analysis described here are the UTPS programs HNET and UROAD. The
report is divided into five sections. The first section describes the link and node
components of a network and the characteristics of links and nodes. The UTPS guide is
both a user's guide manual to UTPS software and an overview of network development,
planning, and analysis techniques. Chapter 2 describes how UTPS highway network data
may be applied to short-term and long-term travel forecasting, traffic assignment, and
related highway system evaluation. Chapter 3 examines specific assessment methods. A
discussion of impedance follows with comments on speed-capacity options. Data cover
default values of speed and capacity by area type and facility type. The final section
describes methods of network evaluation.
A-A-2
-------
4. Highway Capacity Manual: Special Report 209. Transportation Research Board, National
Research Council, Washington, D.C., 1985
The document is a guide to the design and operational analysis of highway facilities. The
14 chapters represent revision and updates of material contained in earlier editions, and
new material reflecting the many changes in the characteristics of travel and in the
information needed to conduct highway capacity analyses. Chapter 2, traffic
characteristics, presents and discusses values observed throughout North America for
many of the parameters and variables introduced herein. Chapters 3 to 8 cover
uninterrupted flow facilities. Chapter 9 to 14 focus on interrupted flow facilities and their
components.
5. Highway Performance Monitoring System, Analytical Process: Volume It-Version 2.1
Technical Manual. U.S. Department of Transportation, Federal Highway Administration,
Washington, D.C., December, 1987.
The report was developed by FHWA to analyze data furnished by the individual states.
These analyses cover the arterial and collector functional systems, and the HPMS data are
obtained from sample highway sections representing those systems. The results of these
analyses are used by the FHWA for many purposes, including policy development and the
biennial reports to the Congress on the status and performance of the Nation's highways.
The Manual covers the technical aspects of the Analytical Process including an
explanation of the methodology used in the models that comprise this process, the
formulas used, and the values of various tables included in the programs such as
minimum tolerable conditions, design standards, and improvement costs. Actual use of
this process is explained in Volume III - User's Guide.
The section on impact analysis provides a comparison of vehicle performance measures
under various scenarios. Among those scenarios are average overall travel speed.
Overall average travel speed is obtained from initial speed adjusted for different factors.
Appendix D includes tables on vehicle fuel consumption for 7 vehicle types stratified by
speed and grade. Appendix H includes tables on excess fuel consumption incurred
because of speed change cycles, for 7 vehicle types.
6. Highway Performance Monitoring System Analytical Process: Volume III, Version 2.1 User's
Guide. Federal Highway Administration, Washington, D.C., December 1987.
The manual provides guidance for reporting HPMS data and to establish updating
procedures for the annual submission. The manual outlines procedures for reporting 3
types of data: areawide, sample, and universe mileage data Areawide data is a data
base of information reported annually for rural, total small urban, and individual urbanized
areas. It consists of totals for mileage, local system data, travel activity by vehicle type,
and other statistics. Sample data include specific inventory, condition and operational
data obtained for the sample panel of highway sections. These data will be expanded to
represent the universe of highway mileage, permitting the evaluation of the performance of
various highway systems. Universe mileage data include a complete inventory of mileage
classified by system, jurisdiction and selected operational characteristics.
A-A-3
-------
7. Traffic Monitoring Guide. Federal Highway Administration, U.S. Department of
Transportation, Washington, D.C., June 1985.
Section 2 of the guide discusses the structure of traffic characteristics monitoring; how
traffic counting, vehicle classification, and truck weighting relate to the overall monitoring
effort. In Section 3, traffic counting is discussed. The chapter focuses on the development
of the statistical scheme. The sample is critical to both the reliability of information on
traffic volumes and to the later development of samples for vehicle classification and
vehicle weighting. Chapter 4 covers vehicle classification. This section includes aspects of
sample development pertinent to obtaining vehicle classification information along with a
description of the fhwa vehicle categories. Section 5 covers truck weighting and contains
information on collection and summarizing data obtained at truck weight sites.
8. Guide to Urban Traffic Volume Counting. Federal Highway Administration, U.S. Department
of Transportation, Washington, D.C., September 1981.
This report presents methods by which urbanized areas can develop and implement
integrated traffic counting programs to serve the volume data needs of all agencies.
Methods for estimating volumes at a single location, volumes across a cordon line, VMT
within a corridor or other small area, and regional VMT are presented. Sound statistical
sampling concepts permit the collection of data at predetermined levels of precision and in
a cost-effective manner.
9. Guide for Estimating Urban Vehicle Classification and Occupancy. Federal Highway
Administration, U.S. Department of Transportation, Washington, D.C., September 1980.
The manual provides sampling and data collection procedures for field surveys that
estimate vehicle classification and occupancy and (when combined with estimates of vmt
from parallel mechanical traffic counting programs) that estimate travel by vehicle type and
personal travel. Because sound statistical techniques are used, these surveys can provide
valid estimates at predetermined levels of precision and in a cost-effective manner.
10. Highway Functional Classification: A Management Tool. Federal Highway Administration,
U.S. Department of Transportation, Washington, D.C., November 1982.
This report is concerned with functional classification as a highway system management
tool utilized by Federal, State, and local governments. It includes a brief history, the
concepts, system characteristics, and federal and state uses of functional classification.
The report mentions the experiences of several states to present a broad range of
applications of functional classification in state planning efforts, with detailed review in the
cases of Washington and Wyoming.
A-A-4
-------
TRANSPORTATION DATA SOURCES
1. Rudman, M. Leon. Vehicle Miles Traveled: An Evaluation of Existing Data Sources. U.S.
Department of Transportation, Energy Program Division, presented to the Transportation
Research Board, Washington, D.C., January 1979.
This relatively old report evaluates data sources for VMT and gasoline fuel consumption.
Collection, reporting, and estimation procedures are addressed. Since direct
measurement of VMT has never been made, the available information consists of indirect
estimates based on various sets of assumptions. • Since the 1973 energy crisis, FHWA has
requested the states to estimate VMT based on average fuel efficiency rates for different
vehicle classifications. The FHWA has not developed and selected one methodology to
estimate VMT. The paper discusses NPTS data, the Claffey model, and the RDTRAV
computer program, as well as data in Highway Statistics.
2. Highway Statistics. Federal Highway Administration, U.S. Department of Transportation,
Washington, D.C., Annual Series.
This publication brings together annual series of selected statistical tabulation relating to
highway transportation in the areas of highway use, finance, and performance. Of special
interest are tables on state vehicle registration, annual VMT by highway category, vehicle
types, functional system, fuel, and speed data
3. Census Data and Urban Transportation Planning in the 1980s. Transportation Research
Record 981, Transportation Research Board, Washington, D.C., 1984.
This issue of the Transportation Research Record includes a number of papers that deal
with several aspects of the transportation planning process: 1) the use of the 1980 Census
for transportation purposes; 2) the Urban Transportation Planning Package, which is a
special product of the 1980 Census specifically tailored for transportation planning; 3) an
outline of several methods for using the UTPP, such as processing its output on
microcomputers; 4) applying and supplementing census data for transportation planning;
and 5) experience with the Urban Transportation Planning process in the Delaware valley
region, Colorado, and New York.
4. Proceedings of the National Conference on Decennial Census Data for Transportation
Planning. Special Report 206, Transportation Research Board, Washington, D.C.,
December 1984.
5. Journey-to-Work Trends: Based on 1960, 1970, and 1980 Decennial Censuses. Report
prepared by the Office of Highway Information Management, Federal Highway
Administration, U.S. Department of Transportation, Washington, D.C., July 1986.
This report identifies the changes which have occurred in population, journey-to-work
patterns, mode of travel to work, and vehicle availability, at the household level in the
largest metropolitan areas of the U.S. between 1970 and 1980. The report is based on
U.S. Bureau of Census data, and it includes some limited analysis of changes since 1960.
Selected census information for 1960, 1970, and 1980 was summarized from available
census computer data sets, and, to a very limited extent from census publications, to a
A-A-5
-------
geographic base which was compatible for the three decennial years. Chapter 1 contains
the purpose, scope, and limitations of the study. Chapter 2 summarizes population and
other census data for the nation as a whole. Chapter 3 looks at population data for the
large metropolitan areas while Chapters 4 and 5 focus on workers. Chapter 6 identifies
changes in mode of travel to work, and Chapter 7 addresses vehicle availability. Chapter
8 completes the report by referencing the latest available population estimates to evaluate
what has occurred since 1980.
6. Nationwide Personal Transportation Study, Volume II. Federal Highway Administration, U.S.
Department of Transportation, Washington, D.C., November 1986.
The Nationwide Personal Transportation Study (NPTS) is an investigation of the
characteristics and personal travel patterns of the U.S. population. The NPTS derives its
information from a nationwide home interview survey of households. The NPTS is a
unique source of information on personal travel, both as a reference on key travel
measures such as trip rates and vehicle occupancy levels, as well as being a source for
linking the characteristics of households with their travel by all modes of transportation.
This is the Volume II of a two-part report which presents findings from the 1983-1984
NPTS survey. The survey obtained data from a national sample of 6,438 households.
Previous NPTS surveys were conducted in 1977 and 1969. Comparing results from these
three surveys provides a good picture of how the country's population and travel habits
have changed over time.
This report examines drivers and their travel, vehicle trips and travel, person trips and
travel, workers and their journey to work, vehicle occupancy, characteristics of travel
period trips and travel and the use of vehicle safety devices. The various subject matters
are explored relative to household location, income, vehicle ownership, and other socio-
demographic attributes of the household. In addition, trends are evaluated over time.
7. Household Vehicles Energy Consumption, 1988. U. S. Department of Energy, Report No.
DOE/EIA-0464(88), Washington, D.C., February 1990.
This report is a descriptive profile of personal vehicle transportation in the residential
sector. It is based on data from the 1988 Residential Transportation Energy Consumption
Survey (RTECS). The survey was designed to collect data on actual VMT for each vehicle
in the household, information on vehicle characteristics, number and type, and vehicle fuel
efficiencies, consumption and expenditures.
8. National Travel Survey: Travel During 1977. Census of Transportation, U.S. Department of
Commerce, Bureau of the Census, Washington, D.C., October 1979, and 1977 NTS Public-
Use Tape.
9. National Vehicle Population Profile Data Tape. R.L Polk and Company, Detroit, Ml.
(Annual)
Polk data tapes provide county-level data (for the fifty states) on vehicle registration for
eight different weight classes. The tapes also provide data on vehicle age and engine
size for the eight vehicle classes.
A-A-6
-------
10. Truck Inventory and Use Survey. Bureau of the Census, U.S. Department of Commerce,
public use tape modified and expanded by Systems Design Concepts, Inc., for Federal
Highway Administration, Washington, D.C., May 198Z
A-A-7
-------
REVIEW OF DATA AVAILABLE FROM
SELECTED TRANSPORTATION DATA SOURCES
1. Highway Statistics (Annual). Federal Highway Administration, U.S. Department of
Transportation, Washington, D.C.
Highway Statistics includes statistics for the following:
•State motor vehicles registrations: total, automobiles, buses, truck tractors, light trucks, trailers,
semi-trailers
•Total state public road and street mileage classified by functional system
•National Network Mileage for trucks
•Annual vehicle miles of travel by state and functional system
• Nationwide fuel consumption by vehicle type, average annual vehicle miles of travel,
average miles traveled per gallon of fuel, and average fuel consumption in gallons for each
vehicle type
2. 7987 Census of Transportation: Transportation Inventory and Use Survey (Quinquennial).
Bureau of the Census, U.S. Department of Commerce, Washington, D.C.
The Transportation Inventory and Use Survey (DUS) is a sample survey of private trucks registered
in each state, stratified by truck type. Truck types for sample stratification include pickup, van,
single-unit light, single-unit heavy, and truck tractors.
The TIUS includes data on the following:
• Operating characteristics including annual miles, range and base, miles travelled
outside state of registrations
• Vehicle characteristics including size, average weight, engine size
• Energy use including type of fuel used, and miles per gallon
3. Motorcycle Statistical (Annual). Motorcycle Industry Council, Research and Statistics
Department, Irvine, CA.
Motorcycle Statistical is an annual report on motorcycle market and usage. It includes
registrations, penetration by region, and mileage by state (on- and off-highways).
A-A-8
-------
4. National Vehicle Population Profile Data Tapes (Annual). Ft. L Polk and Company,
Detroit, Ml.
Polk data tapes provide county-level data (for the fifty states) on vehicle registration for eight
different weight classes. The tapes also provide data on vehicle age and engine size for the eight
vehicle classes. These data are collected from registration data maintained by state Motor Vehicle
Departments. Availability of data directly from state sources varies.
5. The Urban Transportation Planning (UTP) Process
The purpose of the UTP process is to provide guidance for the development, evaluation, and
implementation of alternative transportation planning proposals. It consists of developing a zone
system, selecting links, trip generation, trip distribution, modal split, and traffic assignment. Output
from UTP-type computer models include estimation of the number of trips to and from activity
centers in the area of analysis, orientation and magnitude of traffic volumes, determination of the
magnitude of travel by mode, VMT, and speed.
There are important limitations associated with existing UTP computer models. For example, VMT
generated by non-coded facilities (i.e. residential streets) are not accounted for. In addition, UTP-
type models only simulate daily and peak period volumes with no consideration for seasonal
variation in volumes. Finally, speeds developed through UTP models serve more as a means of
allocating trips to balance the network than a reliable output.
6. Highway Performance Monitoring System Analytical Process. Federal Highway
Administration, U.S. Department of Transportation, Washington, D.C., December 1987.
The purpose of the Highway Performance Monitoring System (HPMS) is to provide a procedure by
which the nation's functional system of highways can be analyzed based on data annually
sampled by all states. HPMS is composed of two major components: data collection and
analytical process. The analytical process analyzes data for each highway section and expands
the results to represent each functional system. Data reported include of highway mileage, daily
vehicle miles of travel, land area, population, and travel activity by vehicle type.
HPMS includes:
Estimates of daily vehicle miles of travel (DVMT) by functional system are to be
prepared for rural, small urban, and individual urbanized areas of the State on an
annual basis. HPMS disaggregates VMT to twelve discrete daily units based on
levels of congestion.
Development of HPMS estimates of highway travel by functional system are to be
derived using count-based traffic data
HPMS determines an overall average travel speed for different congestion levels.
An initial running speed is estimated which is then adjusted for pavement
conditions, curves grades, stop cycles, and idling time.
HPMS does not include a module that could estimate local road operating
characteristics.
A-A-9
-------
7. Urban Traffic Volume Counting
Vehicle count procedures assume that the VMT in a state during a year can be estimated by
counting the traffic on representative sections of roadways during short periods of time and
expanding these results to statewide totals.
Traffic counting programs are expensive to implement. As such, they tend often to focus on
particular needs of an agency which makes estimates of regional travel often incomplete. In
addition, vehicle speeds and vehicle classifications are not addressed in a counting program.
8. Nationwide Personal Transportation Study. Federal Highway Administration, U.S.
Department of Transportation, Washington, D.C., November 1986
The Nationwide Personal Transportation Study (NPTS) is an investigation of the characteristics and
personal travel patterns of the U.S. population. The NPTS emphasizes the relationship between
demographic and economic characteristics and automobile travel. The relevant factors in
estimating VMT include number of cars per household, trip origins and destinations, urban versus
rural travel, and vehicle age.
NPTS tabulates the distributions of VMT by trip purpose and household vehicle ownership, trip
distance, number of occupants, day of week and time of day. In the 'Worker and Work-to-
Joumey" component of NPTS, average commute length and commute time by urbanized area size
are reported. Those may be used as a very coarse representation of 'average* commute speed.
Commute time is measured "door to door" which includes walk access time in addition to in-
vehicle time.
The NPTS has limitations in estimating VMT. First, it is not known how accurate individuals are in
estimating their VMT. Second, no data are collected on existing fuel prices, which prevents
accurate estimates of cost of travel. Third, the published NPTS report does not reveal geographic
locations of the respondents, therefore it is impossible to relate annual VMT per household to the
special characteristics of the region of residence . Fourth, NPTS is representative of personal
travel at the national level, and therefore, is of limited use at the local and state levels. Fifth, NPTS
does not evaluate operating characteristics such as average annual daily traffic, volume, speeds,
or capacity.
9. Household Vehicles Energy Consumption, 1988. U.S. Department of Energy, DOE/EIA-
0464(88), Washington, D.C., February 1990.
The 7988 Household Vehicles Energy Consumption report is a descriptive profile of personal
vehicle transportation in the residential sector. It is based on data from the 1988 Residential
Transportation Energy Consumption Survey (RTECS). The survey was designed to collect data
on actual VMT for each vehicle in the household, information on vehicle characteristics, number
and type, and vehicle fuel efficiencies, consumption and expenditures.
In the RTECS, annual VMT for a vehicle were either calculated using two odometer readings or
imputed using a regression estimate. For a sample vehicle, beginning-of-year and end-of-year
A-A-10
-------
odometer readings were collected. If two odometer readings were available, the difference
between the two readings (adjusted to a standard annualized VMT) equaled the VMT. In the
second step, the annualized VMT was adjusted to correspond to the time period that a vehicle
was in possession by the sample household.
For vehicles with less than two odometer readings, a multiple regression analysis was used to
estimate the annual mileage, where the number of drivers, household income, age of household
head, type of vehicle, and use of vehicle on the job were all included as explanatory variables.
The regression model resulted in estimates of annualized VMT. These VMT are then adjusted to
correspond with the time the vehicle was in possession of the household.
10. The Urban Transportation Planning Package fTransportation-Related Data from the
Decennial Census)
The Urban Transportation Planning Package (UTPP) of the Census contains information on the
joumey-to-work. The information was collected from responses to the long-form census
questionnaire sent to one sixth of the households. However, only half of these were actually
coded resulting in a sample of about 8.33 percent of all households.
The survey collected data on the distance from home to work, travel time, departure time, and the
principal mean of transportation to work. Questions assumed direct trips from residence to the
work place and did not request information about indirect work trips or trips to second jobs.
Travel times reported are for the total number of minutes that usually takes the respondent to get
from home to work including time spent waiting for public transportation or picking up passengers
on the way to work. Time-of-day travel data were not collected but obtained by converting the
census data into estimates of peak-hour travel.
The census collected data on only one mode of travel, the one used for most of the trip. No
consideration is given for trips in which a combination of modes is used. In addition, the survey
does not include trucks weighting more than one ton as a means of transportation.
11. State Transportation Reports. Various State transportation agencies and metropolitan
planning organizations.
Most States and some metropolitan areas publish annual reports of vehicle miles travelled, traffic
flows and counts and other data developed for transportation planning and engineering. These
data are often the basis for HPMS and Highway Statistics, and are usually more disaggregated
and better suited for emission inventory applications. More detailed data may be available directly
from these agencies.
A-A-11
-------
REFERENCES ON
FUEL USE AND FUEL ECONOMY
1. Vyas, A.D., and C. L Saricks. The Transportation Energy and Emissions Modeling System
(TEEMS): Configuration for Forecasting Transportation-Source Emissions for the 1985 Test
Runs. Argonne National Laboratory Report No. ANL/EES-TM-321, October 1986.
2. Saricks, C.L The Transportation Energy and Emissions Modeling System (TEEMS):
Selection Process, Structure, and Capabilities. Argonne National Laboratory Report No.
ANL/EES-TM-295, November 1985.
3. Wolcott, Mark A. and Dennis F. Kahlbaum, MOBILES Fuel Consumption Model, Prepared
for the Office of Mobile Sources, U.S. Environmental Protection Agency, Ann
Arbor.Michigan, EPA-AA-TEB-EF-85-2, February 1985.
4. Modification of the Highway Fuel Consumption Model. Report prepared by Energy and
Environmental Analysis Inc., for the U.S. Department of Energy, August, 1982.
5. The Highway Fuel Consumption Model, 10th Quarterly Report. Report prepared by Energy
and Environmental Analysis Inc., for the U.S. Department of Energy, Contract No DE-
AC01-80EI-11972, Arlington, VA, 1983.
6. Greene, D.L and A. Rathi. Alternative Motor Fuel Use Model: Model Theory and Design
and User's Guide. ORNL/TM-11448, Oak Ridge National Laboratory, March 1990.
7. Calculations for Fuel Efficiencies. U.S. Environmental Protection Agency, Office of Air
Quality Planning and Standards, National Air Data Branch, Research Triangle Park, NC.
8. Millar, M. et al., Baseline Projections of Transportation Energy Consumption by Mode: 1981
Update. Argonne National Laboratory Report ANL/CNSV-28. April 1983.
9. Fuel Economies of Heavy Duty Vehicles. U.S. Environmental Protection Agency, Office of
Mobile Sources, Ann Arbor, Ml, July 1984.
10. Motor Vehicle MPG and Market Shares Report: Market Share Data System Model Year
1983-First 6 Months. Report prepared by Oak Ridge National Laboratories, for the U.S.
Department of Energy, July 1983.
11. The Technology/Cost Segment Model for Post-1985 Automotive Fuel Economy Analysis.
Report prepared by Energy and Environmental Analysis Inc., for the U.S. Department of
Energy, November 1981.
12. Assessment of Current and Projected Future Trends in Light-Duty Vehicle Fuel Switching.
Report prepared by Energy and Environmental Analysis Inc., for the U.S. Department of
Energy, June 1982.
13. Kulp G. and M.C. Holcomb. Transportation Energy Data Book, Edition 6, ORNL-5883, p.67,
Oak Ridge National Laboratory, Oak Ridge, TN, 1982.
A-A-12
-------
14. Lax, D. Feasibility of Estimating In-Use Vehicle Fuel Efficiency from Household Survey Data.
Research performed under contract for Oak Ridge National Laboratory. Energy and
Environmental Analysis Inc., Arlington, VA. 1987.
15. Evans L, R. Herman, and T. N. Lam. Gasoline Consumption in Urban Traffic. General
Motors Research Laboratories, Warren, Ml, Research Publication GMR-1949, 1976.
16. Hellman, K. H., and J.D. Murrell. Development of Adjustment Factors for the EPA City and
Highway MPG Values. SAE Paper No. 840496, prepared for the U.S. Environmental
Protection Agency, Detroit, Michigan, February, 1984.
17. Hellman, K. H., and J.D. Murrell. Why Vehicles Don't Achieve the EPA MPG on the Road
and How That Shortfall Can Be Accounted For. SAE Paper No. 820791, prepared for the
U.S. Environmental Protection Agency, Troy, Michigan, June, 1982.
18. Hellman, K. H., and J.D. Murrell. On the Stability of the EPA MPG Adjustment Factors. SAE
Paper No. 851216, prepared for the U.S. Environmental Protection Agency, Ann Arbor,
Michigan, 1984.
19. Heavenrich, B. M., and J.D. Murrell. Light-Duty Automotive Technology and Fuel Economy
Trends Through 1990. Technical report prepared for the Office of Mobile Sources, U.S.
Environmental Protection Agency, Ann Arbor, Ml, June 1990.
20. Development of a Methodology to Allocate Liquid Fossil Fuel Consumption by County.
EPA-450/3-74-021, Prepared by Walden Research Corporation for U.S. Environmental
Protection Agency, Office of Air Quality Planning and Standards, Research Triangle Park,
NC, March 1974.
21. Raus, J. A Method for Estimating Fuel Consumption and Vehicle Emissions on Urban
Arterials and Networks. Report No FHWA-TS-81-210, prepared for Federal Highway
Administration, April 1981.
22. Fuel Efficiency of Passenger Cars: An IEA Study. Organisation for Economic Co-operation
and Development, 1984.
A-A-13
-------
U.S. "GeoData"
Digital Cartographic and Geographical Data
The U.S. Geological Survey is in the process of preparing digital data for the entire
country, but coverage is currently very sparse for data corresponding to smaller-scale
maps. Catalogs, indices to available data and price lists are available from USGS
(703/860-6045).
Four basic types of data are available:
. Digital Elevation Models (basic topographic information)
• Digital Line Graphs (planimetric [line map] information for USGS maps). These
include transportation systems, as detailed below.
• Geographic Names Information System
• Land Use and Land Cover
Planimetric Digital Data consists of digitized line graphs (DGLs) at the 1:24,000 (7.5
minute) and larger scales. It is digitized and sold in four thematic layers:
• Boundaries
• Transportation, including (1) roads and trails, (2) railroads and (3) pipelines and
transmission lines
. Hydrography
• U.S. Public Land Survey System
Land Use and Land Cover maps are available at the 1:100,000 and 1:250,000 scales.
Data can be formatted either as a set of polygons which consist of lines enclosing areas
with specific characteristics (e.g. a particular land use) or as in a grid cell format which
uses 10-acre grid cells as the area! unit. Categories of data are:
• Urban or built-up land
• Agricultural land
• Rangeland
• Forest land
• Water areas
• Wetlands
• Barren land
• Tundra
• Perennial snow or ice
Each major class is divided into further categories.
A-A-14
-------
Suitable for use in GIS systems for spatial analyses and geographic studies
Can be used for calculations such as total lengths of highways in an area or in specific
sub-areas
Can be combined with other types of digital data, such as U.S. Census Bureau population
files
Available in compact disc-read only memory (CD-ROM) as well as magnetic tape.
A-A-15
-------
A-A-16
-------
APPENDIX B
A CONCEPTUAL DESIGN FOR A NEW HIGHWAY
VEHICLE EMISSIONS ESTIMATION METHODOLOGY
June 14, 1991
Submitted to:
Carl T. Ripberger
U.S. EPA
Mail Drop 62
Research Triangle Park, NC 27511
Prepared by:
Alan W. Gertler
John G. Watson
William R. Pierson
Desert Research Institute
University of Nevada System
Reno, Nevada 89506
EPA Contract No. 68-D9-0168
Work Assignment 24
B-i
-------
ABSTRACT
Mobile source emissions estimates used throughout the U.S. for developing air pollution
abatement strategies may underestimated by as much as a factor of two or more for both
hydrocarbons (HC) and carbon monoxide (CO). Since motor vehicles are the dominant source
of these species in many non-attainment areas, such large errors make determination of
appropriate abatement strategies difficult and prone to error. In addition, current mobile source
emissions factor models are not user-friendly or designed to evaluate the impact of reformulated
fuels, new technologies, and new control strategies.
The perspective taken in this document to address these deficiencies is based on the
experiences of the authors in evaluating and applying motor vehicle emissions models in non-
attainment areas. This perspective is somewhat limited since the primary emphasis has been on
the emissions rate sub-model of the modeling system used in constructing the mobile source
contribution to the inventory.
Weaknesses in the components used to construct motor vehicle emissions models are
described and potential improvements identified. In particular, limitations in the emissions rate
sub-model are addressed and areas for improvement in future models were noted. An
experimental protocol for performing on-road validation of the model is described.
By concentrating on improving model inputs rather than developing new models, costs
can be minimized. Periodic on-road validation tests would, however, be required to verify
assumptions and data acquisition methods.
B-li
-------
CONTENTS
Page
Abstract B-ii
1.0 Introduction B-l
1.1 Background B-l
1.2 Objectives B-2
1.3 Guide to report B-2
2.0 Components of Motor Vehicle Emissions Modeling B-4
2.1 Travel demand and vehicle miles traveled (VMT) B-5
2.2 Emissions rates B-5
2.3 Spatial and temporal variations B-6
2.4 Speciation B-6
3.0 Emissions Rate Sub-Model B-8
3.1 Limitations of current models B-8
3.2 Maintenance/deterioration B-8
3.3 Emitter level categories B-9
3.4 Speed correction factors B-9
4.0 Recommendations and Conclusions B-l 1
4.1 Recommendations for the year 2000 B-ll
4.2 Conceptual model for motor vehicle emissions B-12
4.3 Availability of data B-14
4.4 Evaluation and reconciliation B-14
4.5 Estimated costs of development and evaluation B-15
4.6 Summary B-15
5.0 References B-17
Figures:
4-1. Emissions modeling system block diagram B-13
B-iii
-------
B-iv
-------
1.0 INTRODUCTION
Mobile source emissions inventory estimates used by both the U.S. EPA and state and
local agencies for developing air pollution abatement strategies may, as described in the work
assignment for this report, be underestimated by as much as a factor of two or more for both
hydrocarbons (HC) and carbon monoxide (CO). Since motor vehicles are the dominant source
of these species in many non-attainment areas, such large errors make determination of
appropriate abatement strategies difficult and prone to error. In addition, current mobile source
emission factor models were not originally designed to evaluate the impact of reformulated fuels,
new technologies, and new control strategies on HC, CO, NOX and other species (e.g., PM10 and
speciated organics), thus hampering projection capabilities over the next 5 to 10 years.
The perspective taken in this document to address these deficiencies is based on the
experiences of the authors in evaluating (Pierson, 1989; Pierson et al., 1990; Gertler et al., 1991;
Gertler and Pierson, 1991) and applying (Watson et al., 1990; Chow et al., 1991) motor vehicle
emissions models in non-attainment areas. This perspective is somewhat limited since the
primary emphasis has been on the emissions rate sub-model of the modeling system used in
constructing the mobile source contribution to the inventory. It is our belief, however, that this
component is crucial if we are to develop a motor vehicle emissions model for the Year 2000.
1.1 Background
In order to calculate automotive emissions factors during the Southern California Air
Quality Study (SCAQS) (Lawson, 1990), the Southwest Research Institute (SwRI) conducted a
study, sponsored by the Coordinating Research Council (CRC), designed to obtain emissions
factors in an on-road setting (Ingalls et al., 1989; Ingalls and Smith, 1990). The experimental
approach was to measure air pollutant concentrations into and out of a roadway tunnel located
in Van Nuys, CA. Air displacement measurements were taken to calculate, using in-out
concentration differences and air flow, the amount of each pollutant emitted. Division by the
vehicle-miles logged by traffic in the tunnel during the measurement yielded the traffic-average
emissions rate per vehicle (grams per vehicle-mile) during the measurement.
The carbon monoxide (CO) and hydrocarbon (HC) emissions factors derived by this in-
situ method were far higher than those predicted by the California Air Resources Board's
(ARB's) emissions factor model EMFAC. The average ratios of tunnel emissions factors to
EMFAC7C (version 1C of EMFAC) emissions factors were: 1) 2.7 ± 0.7 for CO; 2) 4.0 ± 1.5
for HC; and 3) 1.02 ± 0.21 for NO, (nitrogen oxides) (Pierson et al., 1990). These differences
between measured and modeled emissions rates raised substantial concern regarding the validity
of the in-situ measurement, the vehicle emissions modeling procedure, the model inputs, and the
current vehicle emissions inventories, and automotive pollutant abatement strategies.
In response to these questions, the Desert Research Institute (DRI) conducted a
preliminary evaluation of the tunnel experiment at the request of CRC (Pierson, 1989). The
evaluation did not identify any deficiencies sufficient to invalidate the tunnel experiment or to
account for the discrepancies with the model predictions, especially the discrepancies between
experimental and predicted CO/NO, ratios or HC/NO, ratios.
Further work was then undertaken at DRI, under CRC sponsorship, to examine the general
nature of discrepancies between measurements and models using the SCAQS Tunnel Study as
a focal point for this examination. As part of this work, a comparison was made between the
B-l
-------
SCAQS Tunnel Study and other on-road studies (Pierson et aL, 1990). The conclusion of this
phase of the study was that the SCAQS Tunnel Study results were consistent with previous on-
road experiments throughout the United States showing CO/NO, and HC/NOX ratios higher than
dynamometer and model predictions.
An additional evaluation was made of motor vehicle emissions modeling issues (Gertler
and Pierson, 1991) and it was concluded that the differences in the ratios of emitted species was
due less to limitations in EMFAC and MOBILE than to limitations in the data base used to
construct the model input [As mentioned above, a second, more complex, series of issues not
addressed in this work is the limitations in transportation models which use as their input the
results from EMFAC or MOBILE.] While these are not the only questions, they appear to be
at the heart of the model underproductions.
These concerns and those of other scientists were subsequently aired at the CRC-APRAC
(Air Pollution Research Advisory Committee) Vehicle Emissions Modeling Workshop held at the
end of October 1990 in Newport Beach, CA (Cadle, 1991).
1.2 Objectives
The objectives of this report are to:
• Identify the needs for motor vehicle emissions modeling.
Included in this are: 1) identification of the types of pollutants of interest and an
explanation of why and where they are important; and 2) identification of
important species not currently addressed by the model (PM,0 and organic
speciation).
• Specify improvements needed in current motor vehicle modeling technology.
Included are transportation modeling, emissions factor modeling, grid allocation
modeling, temporal distribution modeling, and speciation modeling.
• Describe experimental studies and quality assurance activities needed to provide
model input data and evaluate model results.
Included are dynamometer tests, test cycles, smog certification test data,
acquisition of test fleets, roadway tests, tunnel tests, roadway verification, and
remote exhaust sensing.
Traffic counts and classification, emissions estimates from independent data bases,
uncertainty estimates, and reconciliation methods with on-road and ambient air
measurements are also discussed.
1.3 Guide to Report
This first section details the nature of the problem to be addressed and outlines the
objectives of the report Section 2.0 provides a brief description of the components of motor
vehicle emissions modeling systems, what data they use, and how they relate to each other, along
B-2
-------
with identifying which components will be dealt with in detail. The limitations in emissions rate
sub-models are covered in Section 3.0. A conceptual model for a state-of-the-art emissions rate
sub-model along with conclusions and additional recommendations are presented in Section 4.0.
Section 5.0 contains references to this document
B-3
-------
2.0 COMPONENTS OF MOTOR VEHICLE EMISSIONS MODELING
New emissions modeling systems must be designed with a clear understanding of the
purposes to which they will be applied. Motor vehicle emissions inventories are used for several
purposes. First, they identify actual and potentially important emitters for regulation and control.
Second, they provide a base level against which regulated emissions reductions and projected
increases due to population growth can be evaluated. Finally, they are used in combination with
air quality models to evaluate the anticipated effects of emissions reductions.
The first purpose of identifying emitters requires the least rigor. Nevertheless, current
motor vehicle emissions models do not provide accurate estimates of the relative amounts emitted
by different vehicle types such as diesel, super-high emitters, and new-model passenger cars. As
evidenced in the recent letter by Rubenstein (1991), to estimate the impact of high-emitters
required use of a FORTRAN debugger to pull the information out of EMFAC. This is beyond
the capabilities of most users.
The second purpose of providing base emissions levels is not met by the current inventory
methods. Pierson et al. (1990) demonstrate major differences between on-road ratios of CO and
HC to NOX concentrations in areas dominated by motor vehicle emissions and those estimated
from inventory methods. Chow et al. (1991) applied EPA's MOBILE model to transportation
information in Phoenix, AZ, and calculated that lead emissions constitute 9% of paniculate motor
vehicle emissions. This abundance was 50 to 100 times the abundance measured in the exhaust
from Arizona's I&M testing facility and in paniculate samples taken near heavily traveled roads.
Watson et al. (1991) also found major discrepancies between the average wintertime source
apportionments of PM10 and the ratio of motor vehicle paniculate emissions to emissions from
other PM,0 sources. The Phoenix inventory showed geological material contributing over 90%
of PM10, while the source apportionment modeling found that its real contribution to the highest
PM10 was 45 to 50%. The inventory showed motor vehicle exhaust contributing less than 4%
of PM10, while source apportionment modeling showed contributions of 35 to 40%. Recent
source apportionment studies for hydrocarbons in the South Coast Air Basin (Fujita, 1991) show
similar discrepancies between source contribution and emissions rate ratios.
The final purpose of evaluating potential control strategies imposes the most rigorous
constraints on the accuracy and precision of an inventory and, if this purpose is attained, the
resulting inventory is more than adequate for the first two purposes. Current inventories do not
have the necessary spatial, temporal, and chemical resolutions to target the locations, times, and
reactivities of emissions reductions. The inclusion of accurate organic speciation is especially
important for the evaluation of alternate formulations of transportation fuels and their delivery
systems.
To attain the third purpose, current motor vehicle emissions models need to be
conceptually improved with respect to: 1) travel demand and vehicle miles traveled estimates
(VMT); 2) emissions factors by vehicle type, operating conditions, and meteorology; 3) spatial
and temporal allocation of emissions; and 4) chemical speciation for VOCs and PM10. In order
to implement these conceptual improvements, more efficient and flexible hardware and software
platforms are needed.
B-4
-------
2.1 Travel Demand and Vehicle Miles Traveled (VMT)
Travel demand modeling was developed by highway engineers to simulate traffic flows
for the evaluation of highway design, speed limits, access controls, and traffic light
synchronization. While it serves these purposes well, this type of modeling was never intended
for the estimation of air quality emissions. These models divide a complex highway network into
segments within which different numbers of trips of different types are generated. These are
derived from census data such as: 1) numbers of single-and multiple-family households with 0,
1, or 2+ vehicles; 2) total employment; 3) retail employment; 4) service employment and
education; 5) government employment; 6) non-retail employment; 7) other employment; and 8)
population (JHK and Associates, 1991). These empirical relationships are generally established
by surveys of representative populations and are verified against traffic counts. Depending on
the complexity of the model, VMT can be generated by time of day, day of week, and time of
day. For example, the Bay Area Metropolitan Transportation Commission, which includes the
San Francisco area, makes special provisions for greater tourism in the summertime.
The VMT estimates from these models are usually quite accurate in the aggregate, but
they become less accurate as higher resolution is required for spatial and temporal allocation.
The VMT are very seldom allocated to vehicle types, which is extremely important for air quality
modeling. A state-wide or county-wide distribution of vehicles among light-, medium-, and
heavy-duty vehicles is applied to all roadways, at all times of day, for all types of trips. This
is clearly inadequate for an accurate estimate of vehicle emissions, since these vary substantially
among different vehicle types.
Ideally, traffic demand models would be able to provide VMT for defined vehicle classes
along each roadway segment as a function of time. At a minimum, VMT needs to be specific
to roadway type. With this, a separate distribution of vehicles can be estimated for each roadway
type and a separate emissions factor can be calculated for that type. The distribution of heavy-
duty diesels and light-duty passenger cars, and the speeds at which they travel, are much different
for: 1) expressways; 2) freeways; 3) arterials; 4) arterials-2; 5) collectors; and 6) centroid
connectors. The VMT on each of these should be kept separately within the EMS so that a
specific vehicle exhaust emissions factor can be applied to each one.
2.2 Emissions Rates
For motor vehicle exhaust, either California's EMFAC7, EPA's MOBILE3, or EPA's
MOBILE4 (Seitz, 1989, describes the similarities and differences of these models) have been
used to estimate motor vehicle exhaust These models group vehicles by weight, fuel type (diesel
or gasoline), and model year. They then assign grams/mile (EPA models) or grams/second
(California models) emissions rates for different vehicle speeds. Options are included to modify
these emissions with respect to control device tampering and the effectiveness of inspection and
maintenance programs. These models use area-wide (county or state) vehicle registrations to
determine the distribution of model year, fuel use, and weight class. Emissions from each
category of vehicle are derived from laboratory dynamometer tests of vehicles within each class
operating according to the Federal Test Procedure (FTP).
Ingalls (1989) showed on-road emissions of hydrocarbons and carbon monoxide (CO)
which were significantly larger than those calculated by these emissions models. Pierson et al.
(1990) showed that these results were consistent with observations from other on-road studies,
and that they could not be dismissed as an indicator that models might be wrong. Further
B-5
-------
evidence using remote sensing of individual tailpipe emissions for CO (Stedman, 1989; Bishop
et al., 1989; Bishop and Stedman, 1990) has shown that approximately 20% of the on-rpad
vehicles emit nearly 80% of the CO. Remote sensing methods take the ratio of CO in vehicle
exhaust to the increment in carbon dioxide (COj) above unpolluted air. This ratio is proportional
to the amount of CO emitted per quantity of fuel consumed. A pilot study by Lawson et al.
(1990) in Southern California indicates that the high- and super-emitters cannot necessarily be
identified by model year or by the time which has elapsed since their most recent inspection and
maintenance.
While these studies are specific to CO, they call into question the modeling assumptions
which affect all pollutant emissions rates estimated. For example, Hansen and Rosen (1990)
report the only individual vehicle measurements of particulate emissions, and their results apply
only to the ratio of light-absorbing carbon to CO2. They found a factor of 250 between the
highest and lowest ratio of light-absorbing carbon to CO2 for 60 gasoline-fueled vehicles. They
did not categorize vehicles by weight and age.
Continued use of the current "standardized" motor vehicle exhaust emissions models will
not improve emissions estimates. Section 3 describes the limitations in these models in greater
detail. One alternative to be considered is to derive empirical on-road estimates of particulate
emissions using remote sensing. Stedman (1991) has modified the remote CO sensor which he
has developed to measure the attenuation of light at wavelengths which are specific to CO and
hydrocarbons. It is also possible to measure light extinction across a broad bandwidth which can
be related to particulate matter emissions. Stedman's method includes freeze-frame video images
of vehicle license plates. The ratio of visible light transmission to CO2 can then be related to the
grams of pollutant emitted per gallon of fuel consumed. With reasonable assumptions about
vehicle mileage, this can relate emissions to VMT. In this way, an empirical model of pollutant
emissions per VMT can be determined for different types of roadways, driving conditions, and
times of day. With appropriate analysis of the types of vehicles tested, it might be possible to
identify which are the major emitters of VOCs, CO, and PMIO and to focus control measures on
these vehicles.
2.3 Spatial and Temporal Variations
Spatial and temporal variations in emissions are needed to evaluate different work
schedules and land-use planning strategies for reducing air pollution concentrations. These
variations can be partially addressed by the transportation demand models, but provision for them
must be specifically programmed into those models. Temporal variations in emissions also result
from meteorological conditions. For example, the evaporation rate of VOCs increases as
temperatures rise, while CO emissions are higher upon start-up in cold weather than they are in
hot weather. Provision needs to be made for the input of diurnal temperature variations into the
calculation of emissions rates, and a flexible system of allocating emissions to locations is also
required.
2.4 Speciation
Chemical speciation is important for volatile organic compounds (VOCs) which take place
in ozone reactions and for PM10. This speciation is needed both the verify emissions estimates
against ambient data (using receptor modeling) and to estimate the effects of emissions on
secondary pollutants such as ozone, sulfate, nitrate, and organic particles.
B-6
-------
The most complex speciation is needed for VOCs. This speciation is needed for fuels
which might be used in the future as well as those which are used today so that the effects of
these fuels on ozone can be evaluated. The species need to be identified so that they can be
efficiently combined into the groupings which are used in current air quality chemical modules.
The Auto/Oil testing data set provides a rich source of information which can serve as
a starting point for this speciation. More tests are needed, however, to confirm that these
dynamometer test data actually represent the hydrocarbon species found to be emitted from
vehicles operated on the road.
B-7
-------
3.0 EMISSIONS RATE SUB-MODEL
As described in the previous section, a number of distinct components are combined to
yield the motor vehicle emissions modeling systems used in constructing current emissions
inventories. Weaknesses in the components were described and potential improvements
identified. In this section, we further describe limitations in the emissions rate sub-model and
areas for improvement in future models. The SCAQS Tunnel Study and California's EMFAC
provide the context with which these deficiencies are described; however, the criticisms also
apply to EPA's MOBILE model.
3.1 Limitations of Current Models
As part of the review of the SCAQS Tunnel Study (Gertler et al., 1991), the study was
examined to determine whether or not experimental differences might account for the
discrepancies found with respect to the EMFAC7C emissions model. The experimental
differences included: 1) biases in the chemical measurements; 2) inaccuracies in the tunnel flow
measurements; 3) backflow at the tunnel exit; 4) correlations between tunnel flow and emissions;
5) calculational errors and uncertainties; 6) differences in vehicle operating conditions; 7)
differences in vehicle distributions; and 8) representativeness of the samples taken. It was
determined that none of these (or even all taken together) were sufficient to account for the
observed differences. Furthermore, the impact of any experimental errors becomes irrelevant
when one considers concentration ratios or emissions-rate ratios. These too, however, were found
to differ by factors of close to 3 or more. The evaluation of the differences in the ratios of
emitted species implied that the limitations in EMFAC and MOBILE owe less to deficiencies in
the models than to limitations in the data base used to construct the emissions rate model input
(Gertler et al., 1991).
Based on review of the models and their development and comparison with historical data,
there are a number of groupings of questions about the input assumptions. These include:
• Representativeness of the volunteer fleets for estimating maintenance/deterioration;
• Representativeness of the volunteer fleets for estimating emitter level distributions;
• Speed correction factors (treatment of engine load).
Other questions such as the fraction of vehicles operating in a cold-start mode and impact of
evaporative emissions can be posed; however, they cannot explain the differences in the ratios.
For this reason, we believe the three groupings listed above to be at the heart of the model
underpredictions in both the SCAQS Tunnel Study and other on-road applications.
3.2 Maintenance/Deterioration
In order to establish maintenance/deterioration rates of the vehicle distributions used in
EMFAC and MOBILE, volunteer fleets were solicited. The response to the EMFAC mailing was
on the order of 2 to 5% (Susnowitz, 1990). While this was checked and scaled based on motor
vehicle records to ensure that the age and model distributions were appropriate (see paper by
Carlock in CRC, 1990), there are questions as to whether high emitting vehicles, tampering, and
poor maintenance were adequately represented. Could there have been any "filtering"? (That
is, could owners who allowed testing have had generally better-maintained vehicles than those
who declined?) In addition, what would happen if retesting were extended back beyond the 8-
B-8
-------
year period used to construct the input data base - would the growth-rate assumptions used to
transfer vehicles to higher-emitting categories remain valid? Considering the paucity of
surveillance data for vehicles more than 10 years old, it is difficult to answer these questions.
The tampering factor in EMFAC7C may also be low. Cars that have been tampered with
or "arguably tampered" with constitute 31% of the automobiles and light-duty trucks in the
U.S.(EPA, 1989). The tampering factor is adjusted based on a comparison of the volunteer fleet
data with inspection and maintenance (I/M) and surveillance data; however, this still may not be
satisfactory since the average g/mi from the tampered-with vehicles in the these fleets may not
be the same as (may be less than) the average g/mi from the tampered-with vehicles at large.
The impact of unregistered vehicles is unclear. For example, an estimated 6% of the
vehicles in the Lynwood area of L.A. are not even registered (Lawson et al., 1990) and are
presumably less likely to be properly maintained than the rest of the vehicles. Deterioration-
factor estimates may also be too low, and there appears (Lawson et al., 1990) to be a large jump
beyond 80-100 thousand km and/or for vehicles older than 25 years.
3.3 Emitter Level Categories
The input to EMFAC is divided into a series of categories by level of emissions (super
to low) for each vehicle type. As with the vehicle maintenance/deterioration rates, the
distribution of vehicles within emitter level categories is determined from the volunteer fleet.
On-road remote-sensing CO data compared to CO distributions in volunteer test fleets suggest
that the prevalence and emissions rates of the highest emitting vehicles arc also underestimated.
The remote sensing results of Bishop and Stedman (1990) Lawson et al. (1990) and Stephens and
Cadle (1991) indicate a higher fraction of high CO emitters than previously believed, including
even some vehicles emitting more CO than CO2, and demonstrate the limitations in I&M
programs in reducing emissions from this category. The results also point to a small fraction of
each model year distribution (including the most recent vehicles) as falling into the highest
emitting category. All of the remote sensing results indicate that less than 10% of the vehicles
are responsible for 50% of the CO. Recent results for aerosol black carbon have also indicated
a similar trend (Hansen and Rosen, 1990).
3.4 Speed Correction Factors
The speed correction factors used as the model input are derived from nine transient tests
(not steady-speed tests) including the light-duty-vehicle FTP. The tests span a series of average
speeds up to 65 mi/h. Running the nine cycles and scaling them to construct the speed correction
curves may not accurately mimic real-world driving conditions. Inherent in this derivation of
the speed correction curves is the assumption that averages in the skewed distributions
representing the range of emissions at measured speeds can be validly combined to yield
emissions factors for other (non-measured) speeds. On cycles other than the FTP, emissions are
less well known since far fewer vehicles are tested on the other cycles. Actual emissions at a
given speed depend on load (e.g., acceleration); hence, for example, the fluctuations about a
given average speed - their amplitudes, time constants, and frequency of occurrence - will greatly
affect the emissions accompanying that average speed. Real-world conditions may, in some
cases, exceed the valid range of the test cycles; for example, real-world speeds often exceed 65
mi/h, and real-world accelerations commonly exceed the 3.6 mi/h-s maximum in the FTP cycle.
B-9
-------
The importance of both accelerations and super-emitters may also be underestimated by
the models. In a study by Lawson et al. (1990), it was observed that 10% of the vehicles
produced 55% of the CO, while Groblicki (1990) has reported that a single power acceleration
can produce more CO than is emitted in the balance of a trip. This raises doubts over the
validity of the FTP for use in assessing the true impact of accelerations on tail-pipe emissions
and it is apparent that both the effect of accelerations (and hence load) and impact of super-
emitters are underestimated factors in producing emissions. The validity of predicting emissions
on the basis of a speed correction curve that takes no account of the amount of acceleration in
the various tests from which the curve is constructed, and further may underestimate the impact
of super-emitters, is questionable at best.
B-10
-------
4.0 RECOMMENDATIONS AND CONCLUSIONS
4.1 Recommendations for the Year 2000
The primary step in developing a model for the Year 2000 is to recognize and address
the deficiencies in the current methodology. As mentioned previously, the approach taken in this
evaluation is biased towards addressing the emissions rate sub-model. Other parts of the mobile
source emissions model system will also need to be addressed. Recommendations for a revised
model include:
• Developing more efficient and flexible hardware and software platforms for the
system and documenting the combination of travel demand, emissions rate,
temporal distribution, and speciation modules that are developed.
• Developing a user-friendly package that allows the community to evaluate
multiple control scenarios.
• Allowing the user to accurately estimate the relative amounts emitted by different
vehicle types such as diesel, super-high emitters, and new-model passenger cars.
• Providing accurate estimates of emissions factors for calculating base emissions
levels.
• Specifying model precision and accuracy for evaluating potential control strategies.
• Improving spatial, temporal, and chemical resolutions to target the locations,
times, and reactivities of emissions reductions.
• Adding accurate organic speciation (rather than just HC emissions) for the
evaluation of reactivities of emissions from current fuel mixture and alternate
formulations of transportation fuels and their delivery systems. This is particularly
critical for evaluating ozone control strategies.
• Adding PM,0 in addition to CO, HC (or speciated organics), and NO, to the model
output since many areas are or will be in non-attainment with respect to that
category.
• Upgrading traffic demand models to provide VMT for defined vehicle classes
along each roadway segment as a function of time. This is important since the
distribution of heavy-duty diesels and light-duty passenger cars, and the speeds at
which they travel, are much different for 1) expressways; 2) freeways; 3)
arterials; 4) arterials-2; 5) collectors; and 6) centroid connectors.
• Deriving empirical on-road estimates of emissions using remote sensing rather
than extrapolating from dynamometer tests. Stedman (1991) has modified the
remote CO sensor which he has developed to measure the attenuation of light at
wavelengths which are specific to CO and hydrocarbons. Measuring light
extinction across a broad bandwidth which can be related to paniculate matter
emissions is also possible.
B-ll
-------
4.2 Conceptual Model for Motor Vehicle Emissions
The current combination of travel demand, emissions rate, temporal distribution, and
speciation modules are undocumented, untraceable, incompatible, and inflexible. This makes
these models difficult to update and does not allow information to be traced to its sources.
Given today's computational capabilities and information resources, the standard definition
of an air quality emissions inventory as a spatially and temporally averaged listing of presumed
emissions is antiquated. The prevailing concept in scientific circles is that of an Emissions
Modeling System (EMS). An EMS is capable of accessing activity data bases from a multitude
of information-gathering agencies. The EMS then combines this information with meteorological
data and validated emissions/activity relationships to provide emissions rates for any selection
of location, time, or sources. Figure 4-1 shows a block diagram of such an EMS.
An EMS has the following attributes: 1) it is based on documented scientific and
engineering principles; 2) it is composed of special-purpose modules which can be updated with
new information and new science when they are available; 3) its activity levels and
emissions/activity relationships are specific to the area being modeled; 4) it contains error
propagation algorithms to provide precision estimates on outputs; 5) it uses independent activity
data bases and emissions/activity relationships of equivalent quality to estimate accuracy; 6) it
adjusts the emissions/activity relationship in response to environmental variables, especially
meteorology; 7) it allows the addition, subtraction, or modification of emissions for specific days;
8) it retains traceability of all information to allow quality auditing; and 9) it provides output
displays, statistics, and data bases which can be used for modeling, data analysis, and quality
assurance.
Unfortunately, there is no EMS used or even currently available which attains these
attributes, though many who have developed emissions inventory systems claim otherwise.
Attempts have been made to improve the current state of the art with the creation of the Flexible
Response Emissions Data System (FREDS) (Lebowitz et al., 1987) for the National Acid
Precipitation Assessment Program (NAPAP) and the Emissions Preprocessor System (EPS) for
the Urban Airshed Model (SAI, 1990). These are no more than data management systems and
contain nothing pertaining to travel demand or emissions rates. Though these are improvements
over currently used methods, mostly with respect to data handling, they do not attain the
attributes specified above nor are they easily adaptable to emissions in other areas.
An EMS which is intended to possess these attributes is currently under development as
part of the San Joaquin Valley Air Quality Study and Atmospheric Utility Signatures, Predictions
and gXperiments (SJVAQS/AUSPEX) in California. While this development will not be
completed until sometime in 1992, the conceptual structures being developed are directly
applicable to the issues addressed in this document
An ideal EMS consists of several relational data bases which contain different types of
information. At the base level are activity data from which total emissions are derived. These
data are obtained from a number of sources, and wherever possible there will be different activity
data for the same sources and emissions species. For example, PMIO from construction and
demolition might be independently calculated from construction permits and from U.S. Census
TIGER/Line land-use designations and population growth projections. Differences between
activity estimates can be used to quantify the uncertainty of emissions rates.
B-12
-------
Meteorological
data
Vehicle
type
distribution
Socio-economic
data
Travel
demand
model
VMT
Emissions
model
Verify with
fuel sales
Emissions rates
by vehicle type
Spatial and
temporal
allocation
model
Gridded emissions of PM
NOz, CO and VOCs by hour
10'
VOCandPM
speciation
model
10
Species specific
emissions rates
Figure 4-1
B-13
-------
Upon these activity data are imposed emissions/activity relationships (commonly termed
"emissions factors") which may require meteorological inputs such as temperature, relative
humidity, or wind speed. These relationships are easily replaceable and include the ability to
propagate the precisions specified for the input data. The emissions/activity relationships are
specific for different source sub-types. As an example, certain roads may have greater or lesser
proportions of diesel truck traffic, older and newer vehicles, and vehicles which were recently
started and ones which are fully warmed up. In the idealized EMS, each road segment is
assigned a classification for a specific distribution of vehicles, and a separate emissions/activity
relationship is determined for each classification.
An EMS is constructed using commercially available Relational Data Base Management
Systems (RDBMS) and Geographical Information Systems (GIS). Hardware adequate for a
typical urban area (even one the size of the South Coast Air Basin) can consist of a desktop
workstation with a high-resolution color monitor, high capacity (>500 Mbyte) disk drives, optical
storage media, document scanner (for maps), and color plotter/printers. The RDBMS contains
the needed data and performs mathematical operations, while the GIS accesses geographical data
bases, allocates spatial information to grid squares, and drives the display hardware.
An EMS of this type is being constructed as part of SJVAQS/AUSPEX for central
California. This system might serve as a platform for further motor vehicle emissions model
development.
4.3 Availability of Data
Currently available "useful" data are limited. As described previously, on-road and
ambient data imply that the current methodology is inadequate and continued use of the current
"standardized" motor vehicle exhaust emissions models will not improve emissions estimates.
A number of large emissions rate data bases using remote sensing methods have, however,
become available (e.g., Bishop and Stedman, 1990; Lawson et al., 1990; Hansen and Rosen,
1990; Stedman, 1989; Stephens and Cadle, 1991). These could be used as a base-for developing
empirical on-road estimates of emissions.
The Auto/Oil testing data set provides a source of information which can serve as a
starting point for organic speciation. More tests are needed, however, to confirm that these
dynamometer test data actually represent the hydrocarbon species found to be emitted from
vehicles operated on the road.
As detailed in Section 4.2, the approach of developing an EMS with the stated attributes
may be chosen. If this is the case, the conceptual structures being developed as part of
SJVAQS/AUSPEX can be used.
4.4 Evaluation and Reconciliation
In order to evaluate the emissions rate model, a systematic approach must be taken. Step
1 is an operational evaluation of the model. This would involve testing a vehicle fleet that is
specified by, and under the control of, the experimenter. The SCAQS Tunnel Study was not such
an experiment. It was not originally designed to evaluate automotive emissions models, but
merely to provide input for calculation of the SCAQS emissions inventory. In an experiment
with a test fleet, the model would be applied to calculate absolute emissions factors and stated
B-14
-------
uncertainties therein and compare them with observations of known accuracy, precision, and
validity. This would permit reformulation of the model inputs, given the knowledge of the
distribution of modes, speeds, grades, and other operating conditions, to predict accurately the
observed on-road results.
Step 2 would involve testing of the adjusted model in a second experiment, in a real-
world setting, where it can be validated against actual driving situations. In this experiment,
conducted under passive conditions, driver consent is not necessary and remote-sensing would
be required to tell where model inputs may be off and would yield information critical to
determining actual instantaneous vehicle emissions rates. It would also be necessary to verify
the instantaneous results obtained by remote sensing against results obtained over a longer time
period by performing either multiple remote sensing measurements or on-board measurements
on a few vehicles.
A corollary to this approach would be to include uncertainties in all model outputs. If,
as the tunnel experiment and other data imply, automotive emissions are a greater proportion of
the urban hydrocarbon emissions than previously believed, then even relatively small uncertainties
in the measurements (10-20%) can have a major impact on the development of abatement
strategies. On the other hand, if the contribution of vehicle emissions is small compared with
the total, then large uncertainties will not be as critical. Without knowledge of the uncertainty
in model output data, it is difficult even to know which regime one is dealing with and therefore
to know how good the model needs to be.
4.5 Estimated Costs of Development and Evaluation
If the approach outlined of using input derived from empirical on-road estimates of
emissions using remote sensing rather than extrapolating from dynamometer tests is used, then
development costs will be minimal. The physical hardware for obtaining the data currently
exists. The main cost would be in obtaining remote sensing data in the field; however, this
expense would be offset by savings from not performing dynamometer tests.
The expense of developing improved models rather than using improved inputs is difficult
to quantify. Based on the SJVAQS/AUSPEX effort, we estimate a cost of $2M excluding in-kind
services.
We are currently in the process of planning a model validation study for the Southern
Oxidant Study that parallels that outlined in Step 2, above. The total cost (including in-kind
services) will be approximately $1.5M. Costs for the total evaluation and reconciliation effort
would be on the order of two times that level. Periodic re-evaluations should also be budgeted
for.
4.6 Summary
Mobile source emissions estimates used throughout the U.S. for developing air pollution
abatement strategies may be off by as much as a factor of two or more for both hydrocarbons
(HC) and carbon monoxide (CO). Since motor vehicles arc the dominant source of these species
in many non-attainment areas, such large errors make determination of appropriate abatement
strategies difficult and prone to error. In addition, current mobile source emission factor models
B-15
-------
are not user friendly nor are they designed to evaluate the impact of reformulated fuels, new
technologies, and new control strategies.
The perspective taken in this document to address these deficiencies is based on the
experiences of the authors in evaluating (Pierson, 1989; Pierson et ah, 1990; Gertler et ah, 1991;
Gertler and Pierson, 1991) and applying (Chow et ah, 1991; Watson et ah, 1990) motor vehicle
emissions models in non-attainment areas. This perspective is somewhat limited since the
primary emphasis has been on the emissions rate sub-model of the modeling system used in
constructing the mobile source contribution to the inventory.
Weaknesses in the components used to construct motor vehicle emissions models were
described and potential improvements identified. In particular, limitations in the emissions rate
sub-model were addressed and areas for improvement in future models were noted. Experimental
studies and quality assurance activities needed to provide model input data and evaluate model
results were also specified.
By concentrating on improving model inputs rather than developing new models, costs
can be minimized. Periodic on-road validation tests would, however, be required to verify
assumptions and data acquisition methods.
B-16
-------
5.0 REFERENCES
Bishop, G.A., J.R. Starkey, A. Ihlenfeldt, W.J. Williams, and D.H. Stedman (1989). "IR Long-
Path Photometry: A Remote Sensing Tool for Automobile Emissions." Analy. Chem., 61, 671 A.
Bishop, G.A., and D.H. Stedman (1990). "On-Road Carbon Monoxide Emission Measurement
Comparisons for the 1988-1989 Colorado Oxy-Fuels Program." Environ. Sci. TechnoL, 24, 843-
847.
Chow, J.C., J.G. Watson, L.W. Richards, D.L. Haase, C. McDade, D.L. Dietrich, D. Moon
(1991). "The 1989-90 Phoenix PM,0 Study, Volume H: Source Apportionment" Desert Research
Institute Document No. 893L6F1, Reno, NV.
Cadle, S., (1991). "CRC-APRAC Vehicle Emissions Modeling Workshop Proceedings." Newport
Beach, CA, October 30-31, 1990.
Fujita, E. (1991). California Air Resources Board, Sacramento, CA, personal communication.
Gertler, A.W., W.R. Pierson, J.G. Watson, and R.L. Bradow (1991). "Review and Reconciliation
of On-Road Emission Factors in the South Coast Air Basin." Document No. 8401.1F1 from the
Desert Research Institute to the Coordinating Research Council, Atlanta, GA, January 1991.
Gertler, A.W., and W.R. Pierson (1991). "Motor Vehicle Emissions Modeling Issues," Paper no.
91-88.8, presented at the 84th Annual Meeting, Vancouver, BC. Air and Waste Management
Association, Pittsburgh, PA.
Groblicki, PJ. (1990). General Motors Research Laboratories, Warren, MI, personal
communication.
Hansen, A.D.A., and H. Rosen (1990). "Individual Measurements of the Emission Factor of
Aerosol Black Carbon in Automobile Plumes." JAWMA, 40, 1654-1657.
Ingalls, M.N. (1989). "On-Road Vehicle Emission Factors from Measurements in a Los Angeles
Area Tunnel." Presented at 82nd Annual Meeting, Anaheim, CA. Air & Waste Management
Association, Pittsburgh, PA.
Ingalls, M.N., L.R- Smith, and R.E. Kirksey (1989). "Measurement of On-Road Vehicle Emission
Factors in the California South Coast Air Basin — Volume I: Regulated Emissions," Report No.
SwRI-1604 from Southwest Research Institute to the Coordinating Research Council, Atlanta,
GA, June 1989.
Ingalls, M.N., and L.R. Smith (1990). "Measurement of On-Road Vehicle Emission Factors in
the California South Coast Air Basin — Volume II: Unregulated Emissions," Report No. SwRI-
1604 from Southwest Research Institute to the Coordinating Research Council, Atlanta, GA,
October 1990.
JHK and Associates (1991). "San Joaquin Valley Air Quality Study: Final Technical
Memorandum Comparison of Existing Transportation Models in the San Joaquin Valley."
prepared for the California Air Resources Board by JHK & Associates, Emeryville, CA.
B-17
-------
Lawson, D.R. (1990). "The Southern California Air Quality Study," JAWMA 40: 156-165.
Lawson, D.R., PJ. Groblicki, D.H. Stedman, G.A. Bishop, and P.L. Guenther (1990). "Emissions
from In-Use Motor Vehicles in Los Angeles: A Pilot Study of Remote Sensing and the Inspection
and Maintenance Program," JAWMA 40: 1096-1105.
Lebowitz, L.G., and A.S. Ackerman (1987). "Flexible Regional Emissions Data System for the
1980 NAPAP Emissions Inventory." EPA-600/7-87-025a, U.S. Environmental Protection
Agency, Research Triangle Park, NC.
Pierson, W.R. (1989). Letter report to Timothy C. Belian'of CRC, 26 May 1989.
Pierson, W.R., A.W. Gertler, and R.L. Bradow (1990). "Comparison of the SCAQS Tunnel Study
with Other On-Road Vehicle Emission Data." JAWMA 40: 1495-1504.
Rubenstein, G. (1991). Letter W.R Pierson, DRI, 8 May 1991.
Seitz, L.E. (1989). "California Methods for Estimating Air Pollutant Emissions from Motor
Vehicles." Presented at the 82nd Annual Meeting, Anaheim, CA. Air & Waste Management
Association, Pittsburgh, PA.
Stedman, D.H. (1989). "Automobile Carbon Monoxide Emissions." Environ. Sci. Technol., 23,
147-149.
Stedman, D.H. (1991). Brainerd Phillipson Professor of Chemistry, University of Denver,
Denver, CO. Personal communication.
Stephens, R.D., and S.H. Cadle (1991). "Remote Sensing Measurements of Carbon Monoxide
Emissions from On-Road Vehicles," JAWMA 41: 39-46.
*
Susnowitz, R. (1990). California Air Resources Board, El Monte, CA, personal communication.
U.S. Environmental Protection Agency (1989). "Motor Vehicle Tampering Survey - 1988."
Office of Air and Radiation, Washington, DC 20460, May 1989.
Watson, J.G., J.C. Chow, L.C. Pritchett, J.A. Houck, and R.A. Ragazzi (1990). "Chemical Source
Profiles for Paniculate Motor Vehicle Exhaust Under Cold and High Altitude Operating
Conditions." In The Science of the Total Environment. Special Issue: Highway Pollution,
Proceedings of the Third International Symposium, Munich, West Germany, R.S. Hamilton, D.M.
Revitt, and R.M. Harrison, eds. Elsevier Publishing, Vol. 93, p. 183.
Watson, J.G., J.C. Chow, L.C. Pritchett, J.A. Houck, S. Burns, and R.A. Ragazzi (1990).
"Composite Source Profiles for Particulate Motor Vehicle Exhaust Source Apportionment in
Denver, CO." In Transactions, Visibility and Fine Particles, C.V. Mathai, ed. Air & Waste
Management Association, Pittsburgh, PA, p. 422-436.
B-18
-------
APPENDIX C
CONCEPTUAL DESIGN FOR A NEW
HIGHWAY VEHICLE EMISSION ESTIMATION METHOD
by
James H. Wilson Jr.
E.H. Pechan & Associates, Inc.
Springfield, VA 22151
David Levinsohn, J. Richard Kuzmyak, and Richard Pratt
COMSIS Corporation
Silver Spring, MD 20910
EPA Contract No. 68-D9-0168, WA No. 24
EPA Project Officer
Carl T. Ripberger
Emissions and Modeling Branch
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC 27711
Prepared for:
U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF RESEARCH AND DEVELOPMENT
WASHINGTON, DC 20460
C-i
-------
CONTENTS
Page
Figures C-iii
Tables C-iii
1. Introduction and summary C-1
2. Background C-2
History of motor vehicle emission factor models C-2
Travel demand forecasting process C-6
3. Conceptual design C-12
Improvements to emission factor modeling C-12
Improvements to the travel forecasting process for better emissions estimation... C-16
4. Conclusions C-24
Advantages/disadvantages C-24
Costs C-24
Validation C-24
C-ii
-------
FIGURES
Number Page
1. Motor vehicle emission factor modeling Q_4
2. Conventional urban travel forecasting process C-7
3. Example daily emission profiles C-13
4. Revised travel forecasting processes to better
estimate mobile source emissions C-17
TABLES
Number Page
1 . Information interface w/transportation models C-5
2. Example vehicle operating mode model C-19
3. Example cold start/hot soak model C-21
C-iii
-------
CHV
-------
SECTION 1
INTRODUCTION AND SUMMARY
Our conceptual design for highway vehicle emissions focuses
on improving the interaction between what have been termed the
emission factor modeling and transportation modeling portions of
the emissions estimation process. It is our observation from the
workshops and from other related meetings that while
transportation models are not perfectly designed for use in
producing air pollutant emission estimates, the four step travel
forecasting process is the best tool for performing urban scale
analyses and meeting the needs for highway vehicle activity level
information. We see no need for a complete redesign of this
process. As such, we recommend that new modules be added to the
emission and transportation portions of the overall emission
estimation process and that common data sources useful to both
parts be identified and exploited to more fully integrate these
two processes.
Our specific recommendations are: (1) that three new models
be developed within the travel forecasting process to estimate
the percentage of travel occurring in various operating modes, to
identify the cold starts and hot soaks of vehicle trips as a
function of the temporal spacing of these trips, and to quantify
trip reduction effects of transportation control measures; (2)
that a macroscale model of vehicle miles traveled (VMT) and
vehicle trips be developed for areas outside the urban cordon
area; (3) that there be more collection and analysis of modal
emission data and that it be incorporated into the emission
factor model; (4) that EPA take advantage of opportunities for
collecting data at inspection stations; and (5) that onboard
computer recordkeeping be used in conjunction with auto use
surveys.
This short report is organized as follows. We begin with a
short summary of the history of motor vehicle emission factor
models, which is followed by an overview of the four step travel
forecasting process. Our proposed conceptual design for a new
highway vehicle emission estimation procedure follows the summary
information.
C-l
-------
SECTION 2
BACKGROUND
HISTORY OF MOTOR VEHICLE EMISSION FACTOR MODELS
Programs to estimate mobile source emissions and to provide
emission factors for use by State and local agencies have been
ongoing since 1968. Standard EPA publication of such emission
factors began in 1975 in AP-42, which'-provided emission factors
for 1972 and earlier model year vehicles. At that time, mobile
source emission factors were provided in tables, and if
correction factors had to be applied by a user to develop city-
specific emission estimates, calculations were performed by hand.
If an analysis of more than moderate complexity was needed, the
calculations quickly became cumbersome. Thus, some sophisticated
users of the data started to develop their own computer routines
to perform some of the calculations that were possible using
AP-42 data. EPA's Office of Air Quality Planning and Standards
(OAQPS) developed a computer program with the acronym MVEEP
(Motor Vehicle Emissions Estimation Program), which could be used
to estimate mobile source emissions of light-duty vehicles for
the calendar years 1970 through 2000. A program with similar
capabilities was developed by New York City and was titled
BIG AP-42.
To meet the demand for computerized mobile source emission
factors, EPA's Office of Transportation and Land Use Policy hired
a contractor to provide a computerized mobile source emission
factor model. When this model was completed, it was coined
"MOBILE1". This computer program and an accompanying document
were issued in 1978. SIP analyses for CO, ozone, and NO2 that
have been prepared since the release of MOBILE1 have had to
incorporate this program or its successors into any mobile source
emissions analysis. Because States have submitted SIPs and
revisions to those SIPs since 1978, the MOBILE model user
community has evolved into a fairly sophisticated group.
MOBILE1 and MOBILE2 both contained options to produce
California emission factors, but with CARB producing its own
emission factor model (EMFAC) and the confusion that ensued from
having different emission factors for this State produced by the
two models, EPA stopped including the California emission factor
option in its model. MOBILE4 provides high and low altitude 49-
State emission factors.
While the current versions of the motor vehicle emission
factor models that have been produced by EPA and CARB allow users
to model the effect of many more variables now than was possible
10 years ago, the level of uncertainty associated with the
resulting emission factors is great, especially under the extreme
conditions that are often characteristic of air pollution
episodes. About two years ago, the discovery of excess motor
C-2
-------
vehicle evaporative and running loss organic emissions under high
temperature ozone season conditions made it apparent that the
emission factor models of the time underestimated organic
emissions. MOBILE4 incorporated these running loss emissions as
well as excess evaporative emissions, but because of the limited
number of vehicles tested, commitments to release MOBILE4 at an
early date, and technical/policy concerns at EPA about how to
interpret the data, there remains considerable uncertainty in the
organic emission factors. Recent attempts to compare ambient CO,
HC, and NOX measurements with emission factors have also shown
that the motor vehicle emission factor models continue to
underpredict CO and organic emissions. Thus, the need exists to
more accurately determine actual emission rates and correctly
simulate them in models.
Figure 1. provides our assessment of the information flow in
the "current form" motor vehicle emission factor model.
Interactions with transportation data are noted on Table 1.
C-3
-------
o
L
Test a sample of in-use vehicles
• Mostly autos, light trucks
• Range of model years and technologies
• Tests vary speed, temperature, hot/cold starts
[Certification Dataj-
Trips Per Day
and
Miles Per Day
Assumptions
Motor Vehicle
Emission Factors
Emission Factor
by Vehicle Type,
Calendar Year,
HC Component
Inspection Program Effects
• I/M
• Anti-tampering Program
Independent Variables
• I/M Program
• Anti-tampering Program
• Temperature (min,max)
• Operating Mode
• Fuel Volatility
Figure 1. Motor Vehicle Emission Factor Modeling
-------
TABLE 1. INFORMATION INTERFACE WITH TRANSPORTATION MODELS
1 . Speeds vehicles are traveling
- normally determined by roadway functional class
2. Vehicle miles traveled (VMT) by vehicle type
3. Yearly mileage driven by vehicle age
4. Operating Mode Fractions (cold and hot starts)
5. Travel - VMT/trips (by vehicle type)
Local area data in addition to the above
1 . Low or high altitude location
2. Fuel volatility
3. Ambient temperatures
4. Registration distribution by age
C-5
-------
As an introduction to the proposals for improving aspects of
the travel demand forecasting process to better estimate mobile
source emissions, it is useful to briefly summarize the current
state-of-the-practice in urban travel demand forecasting,
indicating where current practice may introduce errors into the
emissions estimation process. The following section provides a
summary of this process.
TRAVEL DEMAND FORECASTING PROCESS
The current travel demand forecasting process as practiced
by most state DOTs and Metropolitan Planning Organizations (MPOs)
consists of the following five major steps as shown in Figure 2:
1. land use/demographic forecasting
2. trip generation
3. trip distribution
4. mode choice
5. trip assignment
This process is applied to the geographic area within each
urban area referred to as the transportation planning "cordon
area". This cordon area is typically defined to include the
observed "commuter shed" around the central city of an urban
area. The cordon area will change over time as the urban area
grows and is typically adjusted once every ten or so years based
upon information from the decennial census. The expansion of the
cordon area is typically a major technical effort as the travel
forecasting models must be adjusted or recalibrated to include
the travel patterns of the newer, expanded suburbs and exurbs of
the urban area. The cordon area for which travel forecasts are
made may,or may not coincide with the geographic area defined as
the urban airshed for an urban area.
Each of the steps in the travel forecasting process will be
briefly summarized in the following subsections.
Land Use/Demographic Forecasting
Person travel and goods movement is a derivative of the
spatial separation of human activities within urban areas. In
order to forecast the amount and location temporal distribution
of travel, forecasts must first be made of the spatial
distribution and characteristics of land use and urban
populations. These forecasts are typically independent of the
characteristics of the urban transportation systems, but are
inherently dependent upon them. These land use and demographic
forecasts are also typically made by staff independent of the
staff performing the travel forecasts for each urban area.
C-6
-------
Figure 2. Conventional Urban Travel Forecasting Process
C-7
-------
Regional and/or county level forecasts are typically made
for population, households, and employment by type (e.g.: retail,
other service, manufacturing, etc.). These forecasts must then
be geographically disaggregated to smaller geographic units,
called traffic analysis zones (TAZs), which form the geographic
basis for the travel forecasting process. Urban areas typically
are divided into approximately 400 to 600 zones per million
population. Most agencies develop TAZ level land use/demographic
forecasts in two steps: 1) from region to district (or
jurisdiction) and 2) from district (or jurisdiction) to TAZ.
Generally, district level allocation methods do not prckJuce all
required inputs (e.g. household car ownership, household income,
household size distribution, etc.) for the travel forecasting
process. Hence a postprocessing procedure is commonly applied.
The methods used to prepare TAZ forecasts are very diverse
and their use in an urban area depends upon local circumstances.
These methods range from simple analytical methods and Delphi
techniques to complex mathematical models. Because inputs
provided by local jurisdictions are crucial to the accuracy of
the TAZ allocation process, these local jurisdictions are often
responsible for the allocation task.
The accuracy of the travel forecasts is in turn a function
of the accuracy of the land use/demographic forecasts. These
demographic forecasts in rapidly growing urban areas are more
likely to be erroneous than those for more slowly growing urban
areas. Hindsight shows that the errors in earlier travel
forecasts are in large part due to the errors in the forecasts of
the magnitude and geographic distribution of population and
employment, as well as missing the advent of major demographic
changes such as the tremendous increase in two-worker households
over the past two decades.
Trip Generation
Trip generation refers to the set of models in the four step
travel forecasting process which estimate the number and purpose
of the person trips generated by the residents and employees of
the urban area. Most trip generation models estimate the number
of trip productions (the home end of each trip) and trip
attractions (the non-home end of each trip) by trip purpose
(typically home-based work, home-based other, non home-based).
The typical variables used to forecast trip productions and
attractions are the number of households per TAZ, persons per
household, household income and/or vehicle ownership and
employment by TAZ stratified by type of employment.
As with all models in the travel forecasting process, the
trip generation models are typically calibrated with travel
survey data collected locally in each urban area relating the
amount of trip making to the variables listed above.
C-8
-------
It is assumed that trip generation rates are stable over the
time period of the forecasts and that the trip production and
attraction rate relationships to the chosen independent variables
reflect stable relationships among the non-modeled contributory
variables.
Trip Distribution
Trip distribution is the model which allocates each of the
trip productions to each of the trip attractions, creating the
geographic linkages between the production and attraction ends of
each trip. The most widely used trip distribution model is the
"gravity model" which forecasts that the probability of a trip
being produced in TAZ x and attracted to TAZ y is directly
proportional to the fraction of the regional trip attractions
which lie in TAZ y and inversely proportional to the travel
impedance required to travel between TAZs x and y. The model is
calibrated and applied to each trip purpose separately. The
model is typically not sensitive to any socio-economic
characteristics of the trip makers (e.g. household income and
employment classification). The model typically uses estimated
highway travel time as the measure of highway impedance between
traffic zones. The model is calibrated to each specific urban
area by estimating a set of "friction factors" which represent
the effects of travel impedance separation between TAZs on the
likelihood of travel between TAZs.
The gravity model, if not stratified by socio-economic
characteristics, can incorrectly estimate the specific linkages
for work trips between households of a particular socio-economic
status and employment locations of a different status. The
gravity model can also wrongly estimate trip distribution
patterns for urban areas with well developed and utilized transit
systems unless the model is calibrated to a multi-modal measure
of travel impedance which includes transit accessibility
measures.
Mode Choice
Mode choice models are common in most of the larger urban
areas with populations over one million. The current state-of-
the-practice mode choice model is the multinomial logit model,
which relates the share of each travel mode (e.g. drive alone,
transit, carpool) to variables which represent the socio-economic
characteristics of the trip-maker, the characteristics of the
ends of the trip (e.g. density of development) and the travel
time and cost of each mode for each TAZ-to-TAZ movement. The
model is calibrated to each urban area by estimating a set of
coefficients which reflect the observed "importance" of each
variable to the share of trip-makers, by trip purpose, who choose
each mode.
Mode choice models can typically do a good job of
replicating region-wide mode shares. They are less accurate in
C-9
-------
estimating mode shares for specific travel corridors or subareas
within urban areas. They also do a better job of estimating
transit shares to the downtown area than within suburban
jurisdictions. Regional mode choice models typically do not
include variables that allow them to accurately estimate many
transportation control measures (TCMs), particularly those that
are employer based.
Trip Assignment
Trip assignment is the process of allocating the TAZ-to-TAZ
trips by mode to the respective networks representing the highway
and transit systems in the urban area. The highway traffic
assignment process allocates auto (and in some urban areas,
truck) trips to the highway network based upon the minimum
impedance paths through the network from each TAZ to each TAZ.
Capacity restraint is typically applied during the assignment
process to "spread" the vehicle trips over a series of paths as a
function of the estimated congestion on each link forming each
path. Most urban areas perform assignments of average daily
weekday traffic resulting in estimated 24 hour traffic volumes by
facility. Estimates of traffic by time period are usually made
by applying diurnal factors derived either from traffic counts or
from travel surveys.
General Improvements to the Travel Forecasting Process
There are several improvements to the travel forecasting
process which, though not specifically targeted at improving
emissions estimates, would improve estimates of urban travel and
hence the accuracy of emissions estimates. While it is not
advocated that EPA take the lead in promoting this research,
these research topics are brought to EPA's attention so that an
EPA awareness develops of what we believe are additional travel
forecasting research needs. These are listed below. Each
research, need is identified as achievable in the near term (1-3
years) or the long term (4-10 years):
• Develop/refine travel forecasting procedures that
explicitly account for the interrelationship between
land-use development patterns and transportation
infrastructure (highway and transit) (long term)
• More fully integrate the components of the conventional
4-step travel forecasting process to form fully policy
sensitive travel forecasting methods; this probably
requires replacing one or more of the conventional
steps with new methods (near and long term)
• Develop models to better represent vehicle usage for
linked trips and non-work travel (near and long term)
• Update and expand information on the diurnal
distribution of vehicle trips (near term)
C-10
-------
• Develop models to estimate heavy truck trip origin-
destination patterns within urban areas and truck
volumes by highway facility (near and long term)
• Improve traffic assignment procedures to be able to
represent the impact of IVHS strategies on traffic flow
(near and long term)
• Integrate transit park-and-ride vehicle trip estimates
into emissions calculations (near term)
• Integrate transit bus trips/VMT into emissions
calculations (near term)
• Incorporate CIS software in order to facilitate the
storage and transfer of travel forecast data into the
emissions/AQ models (near term)
C-11
-------
SECTION 3
CONCEPTUAL DESIGN
Before developing the conceptual design, we first separated
the needs that should be addressed by this design and have sorted
them into primary and secondary needs.
Primary Heeds
• must capture appropriate measures of vehicle activity (VMT,
number of trips, speed, time of travel)
variations in emissions by fuel type and vehicle type
variations in emissions by operating mode
criteria air pollutant and greenhouse gas emissions
time of day variations
spatial variations, including the ability to place emissions
into a square grid network
• ability to perform forecasts
[Estimating total daily or yearly emissions for a
geographic area is more important than spatial or
temporal allocations]
Secondary Needs
• reactivity of organics varies by exhaust, evaporative,
running loss so it is helpful to tie these to the activities
that produce them
• ability to provide county level emission estimates for the
entire country
• fuel consumption estimates to help in tracking CO2 emissions
IMPROVEMENTS TO EMISSION FACTOR MODELING
Our primary recommendation for improving motor vehicle
emission estimates is to begin collecting and analyzing emissions
data by mode. The objective is to be able to isolate exhaust
emissions by the mode of operation, namely acceleration,
deceleration, cruise, and idle. At the same time, a new module
will have to be developed on the transportation side that allows
one to pinpoint time in mode and identify profiles of trip-
making. (This is described in more detail later.) Then, the
above two steps are used together to develop daily emission
profiles. There are a number of options that might be pursued
for developing these profiles. Our initial thought is that daily
profiles for the different emission components could be developed
and would look similar to those shown in Figure 3. There are any
number of ways that the data could be stratified when
constructing these emission profiles. The objective is to have
the smallest number of profiles possible to be able to accurately
cover the variety of trip-making that occurs in an area. The
illustrative profiles shown in Figure 3 are for a home-based work
trip. At a minimum, there would be profiles for other HC
C-12
-------
Exhaust
Emissions
(HBW)
CO, HC. NOx
6AM 12AM 6PM
Time of Day
12PM
Hot Soak
(HBW)
(at trip ends)
6AM 12AM 6PM 12PM
Time of Day
Diurnal
(HBW)
No Driving
Day
6AM 12AM 6PM 12PM
Time of Day
Figure 3. Example Daily Emission Profile
C-13
-------
emission components, including other diurnals (x trips/ y trips,
multi-day), running loss, and resting loss. Presumably, there
would be a variety of home-based work trip profiles to represent
the variety of trip-making patterns within the urban area.
Emission profiles could either be adjusted to represent the
ambient conditions on an episode day, or could be seasonal.
The results of the above three steps could then be used to
develop emission estimates in an integrated
emissions/transportation model. Emissions profiles could be
constructed endogenously, but we see value in being able to
examine the profiles as an intermediate step in the process.
The other important area of motor vehicle emissions control
is determining the percentage of vehicles emitting in the super,
high/very high, and normal modes. Then, designing inspection,
onboard diagnostic (OBD), and recall programs to catch and repair
the highest emitters is important for future emission control
success. For the most part, there are technological solutions to
the "high emitter" problem. Catching 90 percent or more of the
high emitters and repairing them to normal omitting levels can be
achieved by the "best inspection and maintenance (I/M)" program.
This performance requires a significant financial commitment to
the new equipment- and the attendant bureaucracy needed to track
I/M, OBD, and recall efforts.
Our concern in designing emission estimation methods is not
with solving these problems, but in considering their effect on
emissions. Sampling for performing in-use emission tests,
therefore, needs to be designed to effectively capture emission
rates, the fraction of vehicles and their travel in each of the
emitting categories (normal, high, very high, super), and as much
information as possible about the reasons for failures when they
occur. This type of data collection makes it possible to more
accurately estimate base year emissions, and at the same time,
estimate the effect of future control regimes on future year
emissions.
Modal data will be more difficult to collect than Bag 1, 2,
and 3 samples, but with improving computer technology, this
sampling may prove less difficult than it has been in the past.
Statistical techniques will have to be used to determine the
number of vehicles that need to be sampled and the frequency of
testing. While there may be some penetration into the vehicle
fleet of alternative technologies and fuels in the next 10 years,
current evidence points toward reformulated gasoline and advanced
catalysts with fuel injected engines being the vehicle technology
of the future. If this is the case, then, at least in theory,
there would not be significant differences in emission
characteristics within a vehicle class. Emission differences
would then more likely be related to emission control component
failures than to differences in control technologies. (This is
probably already the case today with the TECH IV group as defined
in MOBILE 4.) This may also reduce the need to characterize the
C-14
-------
emissions from each model year's vehicles -- similar model years
can be grouped.
It is clear from the current research on vehicle emissions
that there are still some major unresolved issues. It has been
known for some time that the Federal Test Procedure (FTP) is not
representative of the driving conditions that are likely to lead
to air quality standard violations. Whether the FTP continues to
be cycle used for certifying vehicles, there needs to be an
alternative "in-use" cycle that better characterizes actual
driving. This cycle needs to include the ability to simulate
"real accelerations," and perhaps it needs to be continually
revised to account for changes in driving patterns.
On the other hand, whatever cycle is developed to simulate
"in-use" emissions should be widely applicable. The assumption
is that driving patterns do not vary much among cities. There
may be special cases, such as Manhattan, that are not well served
by the new cycle, but these will be few, and the New York City
cycle already offers a profile of the driving patterns in that
metropolitan area. It is only important that driving with each
mode - acceleration, deceleration, cruise, and idle - be
consistent from area to area, however, and that the cycle capture
the full range of actual driving practices across various roadway
types.
Even with alternative fuels, it is expected that
manufacturers will design vehicles to certify to the applicable
standards, so the variation in exhaust emissions among fuels may
not be great. Separate evaporative HC emission profiles would
have to be developed for alternative fuels if they achieve
significant market penetration. If alternative fuels are
primarily used in fleet vehicles, the different driving patterns
of that type of vehicle will have to be considered.
Soon, there will be onboard diagnostic equipment on all new
cars, and that, coupled with a trend toward (and perhaps an
EPA requirement for) centralized I/M programs, presents a
tremendous new opportunity for compiling historical data on
yearly mileage for each car and light truck (and perhaps each
heavy-duty gas truck), survival rates, and emissions-related
performance. If transient testing is included in I/M programs,
I/M test results should more closely approximate those of the FTP
and can provide another useful data set on per vehicle emissions.
Increased sophistication of onboard computers also provides
opportunities to coordinate auto use survey material for
transportation planning with records of emission control system
performance. Equipping the cars of households which are keeping
travel diaries with diagnostic equipment that can keep track of
throttle position, speed, etc. also provides significant
opportunities for "cross fertilization" of data between
transportation and environmental needs.
C-15
-------
Motor vehicle emission factor models need to be configured
to provide different outputs in order to serve different
purposes. Having modal data available makes it possible to
provide emission estimates for specific roadways or sets of
roadways that are more likely to match with experimental data
(and therefore, more useful for validation). MOBILE4 really
provides data in a form most useful for performing regulatory
analyses. It is not designed for grid-based model inputs either.
In short, we are recommending that the data in the model be used
in different ways to suit different needs. This means at least
three different MOBILE4 output forms are needed; namely, (1)
model, (2) grid-based episode day, and (3) regulatory analysis
compatible.
IMPROVEMENTS TO THE TRAVEL FORECASTING PROCESS FOR BETTER
EMISSIONS ESTIMATION
In order to better estimate vehicular emissions in the
future, the travel forecasting process must provide different
information about the traffic stream than has been provided in
the past. This information will correspond to the input data
requirements of the proposed new emission factor models described
in an earlier section. Three new models within the travel
forecasting process will be required to produce the information
and sensitivities needed for improved emissions estimates. These
models are:
• model to estimate the percentage of vehicle travel
occurring in various operating modes
• model to estimate the cold starts and hot soaks of
vehicle trips as a function of the temporal spacing of
these trips
• model to estimate the trip reduction impacts of
Transportation Control Measures (TCMs)
Each of these models will be discussed in detail below and their
place in the travel forecasting process is illustrated in Figure
4.
Vehicle Mode Model
The purpose of this model is to provide percentage estimates
of daily VMT that are expended in each of the four vehicle
operating modes which are determinants of emissions:
• acceleration
• deceleration
C-16
-------
Figure 4. Revised Travel Forecasting Processes to Better Estimate Mobile Source Emissions
C-17
-------
• cruise
• idling
It is proposed that the model estimate the percent of each
vehicle trip spent in each operating mode as a function of the
types of facilities (e.g. freeway, arterial) over which each
vehicle trip is made, the volume/capacity (V/C) ratio of the
facility, the time of day (peak and off-peak), and perhaps
information about the traffic control device density on arterials
(e.g. x traffic signals/mi.).
The current traffic assignment models can already report the
percent of VMT that occurs on each facility type and in each V/C
ratio range. These assignments typically reflect 24 hour average
weekday traffic volume estimates. Traffic signal density by
roadway link or by area type is a variable that can be maintained
in the highway network data files, but would need to be added to
those files for most urban areas.
The model would be calibrated based upon new data that need
to be collected relating vehicle operating mode to the variables
cited above. These data can be obtained by developing a sample
of instrumented vehicle trips made in several urban areas over a
range of facility types, traffic conditions, and times of day.
The data could then potentially be reduced to a cross-
classification model with the structure shown in Table 2.
s
The stratification of the variables would depend, of course,
upon an analysis of the data collected. Each cell value would
have both a mean and standard deviation computed to present the
variability of the estimate. A statistical analysis of the
potential confidence level of each expected value would help
determine the sample size of the number of trip observations to
be collected.
This model could be simply applied to most travel
forecasting software package outputs using application software
such as FORTRAN or DBASE. The inputs would be: 1) the estimate
of VMT stratified by facility type, 2) the number of weekday
hours under which each facility type is estimated to operate with
peak and off-peak traffic conditions, and 3) the V/C ratio
produced by the traffic assignment model. The output would be VMT
stratified into each vehicle operating mode by time of day and
geographic location (grid cell).
Cold Start/Hot Soak Model
The purpose of this model is to estimate the fraction of
vehicle trips which have cold start and hot soak emissions, which
is a function of the time span between vehicle usage for vehicle
trips. As is discussed in the emission model section of this
report, cold start emissions and hot soak emissions will be a
greater percentage of total'vehicle emissions for each vehicle
C-18
-------
Table 2. Example Vehicle Operating Mode Model
V/C Range
(Based Upon 24 Hr. APT Assignment)
Facility Type
Freeway
Expressway
Primary Arterial
(5+ signals/mi.)
Primary Arterial
(3-5 signals/mi.)
Primary Arterial
(<3 signals/mi.)
Secondary Arterial
(5+ signals/mi.)
Etc.
0-0.25
Peak
% idle
%accel
%decel
% cruise
off-peak
% idle
% accel
%decel
% cruise
0.25-0.5
0.50-0.75
0.75-1.00
1.00
C-19
-------
trip in the future as running emissions are further controlled.
Hence, being able to estimate the percentage of cold starts and
hot soaks based upon the temporal spacing of vehicle usage by
trip is deemed important for improved emissions modelling in the
future.
The model would estimate the percentage of vehicle trip ends
that are cold starts and hot soaks as a function of trip purpose
(home-based work (HBW), home-based other (HBO), and non home-
based (NHB)) and trip duration. This model would be calibrated
from data obtained in a transportation auto use survey in several
urban areas, such as that collected by the Metropolitan
Washington Council of Governments in their 1968 travel survey.
This type of survey asks the respondent to record the usage of
each vehicle owned by the household over the course of the day.
For each vehicle trip, the following types of information are
collected:
time of trip start
location of trip start
time of trip end
location of trip end
purpose of trip
number of persons in vehicle
The model might look like the example in Table 3. The percent of
cold starts and hot soaks would be determined based upon the time
intervals between vehicle usage. After an hour of non-usage, a
catalyst-equiped vehicle will be considered to be in a cold start
status when used next. The hot soak emissions will occur at trip
end or immediately after the vehicle has been shut off.
The model would be applied to the vehicle trip table input
to the traffic assignment model and the output would be the
number of vehicle tripends in each traffic analysis zone which
are estimated to have cold starts and hot soaks, for both peak
and off-peak periods. It could be developed as a stand alone
program using FORTRAN or DBASE or could be a "set-up" inside any
of the trip table matrix manipulation programs found in all
travel forecasting software packages. These tripends would be
input to the emission model to estimate the tripend component of
emissions which could also be allocated to grid cells based upon
the traffic zone in which they lie and using CIS software to
perform the allocation.
TCM Impact Model
The purpose of this model is to estimate the vehicle trip
reduction effects of various transportation control measures
(TCMs) that will likely be implemented by urban areas in an
attempt to reduce vehicle emissions and highway congestion.
C-20
-------
Table 3. Example Cold Start/Hot Soak Model
Trip DunH ion
< 5Min. 5-9 Min. 10-14 Min. 15-20 Min. 20-30 Min. >30 Min.
HBW Peak
% Cold start
% hot soak
Off-peak
% Cold start
%bot soak
HBO Peak
% Cold start
%hot soak
Off-peak
% Cold start
%hot soak
NHB Peak
% Cold start
% hot soak
Off-peak
%Cold start
% hot soak
C-21
-------
These TCMs, also known in the transportation planning profession
as travel demand management (TDM) measures, are designed to
induce commuters to shift to higher occupancy modes for their
work trips. Example TCMs include improved transit or HOV
alternatives, parking management and pricing, flexible work
hours, guaranteed ride home programs, and carpool and vanpool
incentives. Mode choice models and procedures need to be
developed which can more accurately predict the mode shifts
likely to occur when these TCMs are implemented in urban areas.
These TCM/TDM impact models are already under development in
projects being sponsored by agencies such as FHWA, GARB, and
Seattle Metro. However, additional refinement is necessary to
improve the number of strategies for which these models can
predict travel changes. This refinement will come from the
collection of additional travel survey data from sites at which
the various measures have been implemented, and the calibration
of more robust models to these observed data. Additionally,
these models will need to be integrated so that one set of models
containing all relevant mode choice determinants is produced.
These models will also need to be directly integrated into the
four step travel forecasting process described above, rather than
being used "off-line" as they currently are.
As in current regional mode choice models, these models
estimate the commuter's choice of mode as a function of the
socio-economic characteristics of the tripmaker, the
characteristics of the competing modes of travel, and the
characteristics of the employment site. However, these models
are constructed to specifically include more factors at the work
site, particularly the mode choice effects of a variety of
employer-based TCMs and the aggregate mode choice effects of
varying levels of employer participation in the provision of
these measures. The output of these models is the number of work
purpose person trips for each TAZ-to-TAZ movement which are
forecast, to utilize transit, carpool/vanpools, or drive alone
measures.
Macroscale Models of Vehicle Trips and VMT for Areas Outside
Urban Cordon
As previously discussed, most urban area travel forecasting
models have a coverage area, or cordon area, which covers most,
but not necessarily all of the travel shed of the urban area.
The cordon area is usually adjusted once every ten years or so,
based upon information from the decennial census. Urban airshed
regions, however, may be larger than the area covered by the
travel forecasting models, causing a problem with being able to
accurately estimate mobile source emissions generated within the
portion of the urban area that lies within the airshed, but
outside the travel forecasting cordon. Procedures need to be
developed which allow estimates to be made of vehicular travel in
these "unmodeled" portions of the urban area.
C-22
-------
These outlying areas, typically entire counties, could have
travel estimates developed based upon default models calibrated
using national travel data sources such as the National Personal
Travel Survey (NPTS) and the Census Urban Transportation Planning
Package (UTPP). These models would estimate daily VMT and
vehicle trips per household as a function of common socio-
demographic and land use variables such as:
vehicles/household
workers/household
persons/household
average population density
number of households
number of employees
and transportation supply statistics such as:
• average lane-miles of roadway/square mile
• average trip length
The specific variables and their statistical relationships would
be determined based upon statistical analyses of databases such
as NPTS, particularly for exurban counties of urbanized areas.
Model inputs, for example, could be county level estimates of
households, employment, and highway lane miles, and model outputs
would be average weekday VMT and vehicle trips. The
relationships (models) might be able to be stratified by urban
area population, depending upon how robust the databases are.
The outputs of this aggregate model would augment the
results of the four step models for urban areas where the travel
forecasting cordon does not cover the entire airshed.
C-23
-------
SECTION 4
CONCLUSIONS
ADVANTAGES/DISADVANTAGES
One of the advantages of the model design is its ability to
capture the full range of vehicle operating conditions, in
addition to its ability to present data at the level of detail
necessary to validate and verify results. Previous reliance on
the Urban Dynamometer Driving Schedule (UDDS) and Bag 1, 2 and 3
emission values has led us to overlook some of the most severe
(and high polluting) driving conditions. These problems are also
a function of the limitations of the vehicle testing
technologies.
While the proposed design is complex, the issues are also
complex, and any method that is designed to estimate the effects
of future transportation control measures has to be
sophisticated. For example, a decision about whether to spend
$10 million for a ramp metering system to reduce highway
congestion has to be made with full knowledge of the likely
effects. Increasing the number of hard accelerations by making
vehicles stop on the ramp may produce more emissions than are
being reduced by decreasing congestion and increasing speed on
the highway. Evaluations of land use changes and proposed major
transportation projects are difficult without sophisticated
tools.
Another advantage of the design is that it can build on
advances in the state of the art of transportation modeling.
DOT-FHWA has planned a number of new studies to ensure that
improvement in the next 3 to 5 years and the conceptual design
presented here should be able to include such advances.
COSTS
Costs to apply the procedures suggested here are those in
addition to costs that States and MPOs will be incurring to apply
travel demand forecasting procedures to their areas. We estimate
that the incremental cost to apply the vehicle model and cold/hot
soak models to be $10 to 20 thousand per area. Modeling TCM
impacts will be more involved and could realistically cost as
much as $80K per area.
VALIDATION
Our comments on validation are related to city-specific
validation of the methods suggested here. Transportation model
results are normally validated component-by-component. For
example, model split models are validated by travel surveys.
Similarly, checks on VMT are from count programs. These
C-24
-------
include screen line and cordon line checks by sector. Auto
occupancy checks are another form of validation.
On-road vehicle emission tests are not practical with
today's sampling technologies, especially for every urban area
that needs to prepare an emission inventory. Developing such
technologies is a research need. Remote sensing holds some
promise. The recently initiated project in the South Coast (CA),
along with the other evidence collected to date, should provide a
reasonable assessment of its applicability and cost. Tunnel
studies are probably too expensive to perform more than once
every 3 to 5 years. Until remote sensing proves to be a
practical tool for validating motor vehicle emission estimates,
we believe that improving I/M program performance and taking
advantage of the annual data tracking possibilities is the best
validation tool for emission modeling in a 3 to 5 year time
frame. I/M data can be useful in validating emission modeling
results at various points in the process (as is transportation
model validation as described above) including developing
registration distributions, tracking mileage driven by vehicle
age, and having an alternative source of emissions test data.
Satellite photos are a potential validation tool primarily
as a way of noting whether episode day conditions differ
significantly from "average daily traffic."
C-25
-------
C-26
-------
APPENDIX D
CONCEPTUAL DESIGN OF A
NEW MOTOR VEHICLE SOURCE MODEL
Prepared for:
Carl T. Ripberger
U.S. Environmental Protection Agency
Office of Research and Development
Air and Energy Engineering Research Laboratory
Emissions Modeling Branch
MD-62
Research Triangle Park, NC 27711
Prepared by:
L. Bruckman
E.L. Dickson
W.R. Oliver
Radian Corporation
10395 Old Placerville Road
Sacramento, CA 95827
EPA Purchase Order 1D1866NALX
June 14, 1991
D-i
-------
ABSTRACT
This report describes Radian's new motor vehicle source (MV) model.
The MV model includes procedures to develop travel demand estimates, emission factor
modeling and emission estimation. The MV model requires travel demand estimates
that treat trip end effects separate from running emissions. Link-specific vehicle hours
traveled (VHT) are matched with emission factors that are expressed in units of grams
per hour. Each link in the transportation system is characterized by a driving cycle,
which is a function of its facility type and area type and level of congestion (on-peak or
off-peak). Emission factors are computed for each driving cycle. The Radian/CRC
Evaporative Model is used to compute the evaporative emissions component. The MV
model will be written using fourth generation language software and take advantage of
geographic information system software. It will be menu-driven, user-friendly and
modular to allow for easy updating. This model has the following characteristics:
• Fourth generation language software (i.e., SAS® and ARC/INFO®)
form the basis of the motor vehicle source model;
• GIS techniques are used to simulate all roadways, including minor
collector and residential streets, in a region;
• Origin - destination trip tables are used to develop hour-by-hour
link-specific travel demand estimates;
• Transportation modeling procedures are provided to generate travel
demand estimates for both on-network and off-network areas;
• Trip end effects are treated explicitly;
• A driving cycle matrix, by facility type and area type, is used to
characterize each link in the region;
• Vehicle hours traveled (VHT) are matched with gram per hour
emission factors to develop link-specific emission estimates;
• The transportation model is embedded in and is an integral
component of the motor vehicle source model;
Conceptual Design of A New Motor Vehicle Source Model _
-------
PREFACE
This report describes a new motor vehicle source model. The report is
divided into seven sections. Section 1.0 includes some introductory material. Those
readers familiar with the shortcomings of the current methodology used to develop
emission estimates from on-road motor vehicles may want to skip this section and begin
with Section 2.0. Section 2.0 provides a brief overview of the new motor vehicle source
model. Section 3.0 describes the Transportation Modeling System required to generate
the travel demand estimates for this model, including the development of the driving
cycle matrix (by facility type and area type). Section 4.0 describes how emission factors
will be developed for each driving cycle. Section 5.0 discusses how the travel demand
estimates and corresponding emission factors will be combined to develop gridded,
speciated, temporally resolved emission estimates. Section 6.0 presents the data inputs
required to use this new model. Section 7.0 discusses several techniques that can be
used to verify the emission estimates produced by the new model.
D-iii
Conceptual Design of a New Motor Vehicle Source Model
-------
Emission factors, on a grams per hour basis, are developed for each
driving cycle in the driving cycle matrix, based upon mode test data;
Customized driving cycles for specific applications can be developed.
Evaporative emissions are modeled using the Radian/CRC
evaporative model;
Alternative fuels are accommodated; and
Emission estimates can be gridded into a variety of grid structures,
including uniform, variable and nested grids.
The report summarizes the shortcomings in the current methodology used
to develop motor vehicle emission estimates. A list of required model inputs is provided.
Detailed discussions of all of the model components are provided, along with techniques
that can be used to verify the resulting emission estimates.
Conceptual Design of A New Motor Vehicle Source Model
-------
CONTENTS
Page
Abstract D-ii
Preface D-iv
1.0 Introduction D-1-1
1.1 Shortcomings of existing methodologies D-l-1
1.1.1 Travel demand estimates D -1-2
1.1.2 Emission factor models D-l-7
1.1.3 Emission estimation methodology D -1-11
2.0 Brief Overview of a New Motor Vehicle Source Model D - 2-1
3.0 Development of Travel Demand Estimates D-3-1
3.1 Driving cycle matrix D-3-1
3.2 Origin — destination travel surveys D - 3-4
3.3 Transportation modeling D-3-8
3.3.1 On-network transportation modeling D - 3-9
3.3.2 Off-network transportation modeling D - 3-11
4.0 Emission Factor Model D-4-1
4.1 Modal analysis model D-4-2
4.1.1 Evaporative emissions D - 4-5
4.1.2 Trip-end emissions D-4-6
4.1.3 Alternative fuels D-4-7
4.2 Emission factor model D - 4-8
5.0 Development of Emission Estimates D-5-1
6.0 Data Input Requirements D - 6-1
7.0 Motor Vehicle Source Model Validation Procedures D - 7-1
7.1 Tunnel studies D-7-1
7.2 Remote sensing D-7-1
7.3 Source apportionment D - 7-2
7.4 Fuel sales D-7-2
7.5 Traffic count programs D-7-2
7.6 Satellite imagery D-7-3
D-v
-------
CONTENTS
Page
Figures:
2-1. New motor vehicle source model D - 2-2
4-1. Hypothetical driving cycles for urban freeway D-4-9
Tables:
3-1. Driving cycle matrix D-3-5
D-VJ
-------
1.0 INTRODUCTION
The development of cost-effective air pollution control strategies that will
achieve federal and state ambient air quality standards (AAQS) requires accurate
estimates of emissions from applicable source categories. Volatile organic compound
(VOC), nitrogen oxide (NOX) and PM10 emissions from on-road motor vehicles continue
to make-up a major portion of the total emissions inventory from these pollutants in
most areas of the nation, many of which are currently exceeding AAQS. In addition, on-
road motor vehicles emit significant quantities of toxic air pollutants (e.g., benzene).
Consequently, accurately quantifying VOC, NO,, PM10 and various specific toxic
compounds from on-road motor vehicles will play a critical role in the development of
various air pollution control programs. Recent studies indicate that current procedures
for estimating emissions from on-road motor vehicles significantly underpredict observed
values. Furthermore, it is doubtful that existing methodologies will be able to properly
treat alternative fuels, new vehicle technologies and transportation control methods that
will be available in the near future. It is imperative that new procedures be developed
to more accurately estimate emissions from on-road motor vehicles. The following
discussion presents the conceptual design of a new motor vehicle source model to
estimate emissions from this source category.
1.1 Shortcomings of Existing Methodologies
The methodology used to develop emission estimates from on-road motor
vehicles can be expressed in simple terms as the processing of age-composited emission
factors for the various classes of vehicles for the year of interest, along with spatially and
temporally distributed travel demand estimates, to produce emission estimates for a
given region. These estimates may be gridded, speciated and temporally resolved
depending upon their desired use. Each of these components (i.e., emission factor
development, generation of travel demand estimates and emissions modeling) currently
Conceptual Design of a New Motor Vehicle Source Model 0-1-1
-------
has some severe shortcomings. Some of the most important limitations of the current
procedures are described below.
1.1.1 Travel Demand Estimates
Although many different types of transportation models (e.g., MINIUTP,
UTPS, System II, EMME-2) are used to develop travel demand estimates in urban areas,
these models can be divided into the following six general categories:
• Development of socioeconomic/demographic data;
• Trip generation;
• Network development;
• Trip distribution;
• Mode choice; and
• Trip assignment.
Uncertainties and potential errors in the magnitude of the emission
estimates based upon the use of vehicle activity data from current transportation models,
as well as the spatial and temporal disaggregation of these values, may be introduced at
each step of the transportation modeling system (TMS) process. Examples of these
uncertainties are described below.
Socioeconomic/Demographic Data. The development of socioeconomic
and demographic data is an important component of the TMS and actually drives the
process. This information is typically provided by land use forecasts of the modeled
region and the socioeconomic characteristics of its inhabitants (e.g., number of
households by income level, population and employment characteristics of the residents
of the region). Many areas of the nation, (e.g, the South Coast Air Basin [SoCAB] in
Southern California) have been rapidly growing and although forecasts may be made as
often as every five years, this may not be adequate for the purposes of transportation
Conceptual Design of a New Motor Vehicle Source Model D - 1
-------
modeling. In addition, and perhaps more importantly, accurate future year
socioeconomic and demographic forecasts would be even more difficult to make,
especially those that are 20 years into the future.
Trip Generation. Trip generation involves developing estimates of the
rates of trip productions and trip attractions for discrete areas, usually identified as travel
analysis zones (TAZ). Each trip can be visualized as being produced in one TAZ and
attracted to another. A significant portion of the VOC emissions from on-road motor
vehicles (i.e., the evaporative emissions), are directly proportional to the number of starts
(origins) and parks (destinations) in a given TAZ. The starts are further broken down
into cold starts and hot starts, and the emissions during the park period are a function of
the number of hours a vehicle is parked before it is re-started. In order to accurately
quantify and locate these "trip end" emissions in a region, it is necessary to accurately
determine the number of origins and destinations in each TAZ. In addition, errors in
the calculation of trip productions and attractions will lead to errors in the quantification
of vehicle miles traveled (VMTs).
There is reason to believe from recent studies that trip generation rates in
various urban areas, such as the SoCAB, have been changing over time due to the
changing nature of the inhabitants of a given region. While average household size has
been decreasing, trips per household have increased due to the ever increasing number
of two-worker households. Increased vehicle ownership in recent years also contributes
to a greater number of vehicle trips per household. While trip generation models
account for some of these changes in trip generation rates, the data and assumptions that
are used in the models need to be periodically updated. Trip generation models are
based upon a combination of regression analyses of certain demographic and
socioeconomic characteristics that yield trip making units, which are converted into
person trips by applying trip generation rates. The trip generation rates are developed
from survey data and define the number of person trips generated by a particular kind of
housing unit having a given level of vehicle ownership, for a specific trip purpose. A
Conceptual Design of a New Motor Vehicle Source Model
D • I "•*
-------
similar process is used to develop estimates of trip attractions. The use of regression
techniques and outdated survey information (travel surveys in most urban areas have not
been updates since the late 1960's or mid-1970's due to the associated costs of obtaining
and analyzing this information) may introduce considerable uncertainties in the
development of trip productions and attractions. The fact that these rates are typically
assumed to be constant throughout a given county may not properly account for
socioeconomic and demographic differences that may exist across certain counties in a
given region, such as the SoCAB.
Another concern in the development of trip productions and attractions is
how large trip generators, such as airports, are handled in the trip generation model.
The TMS used by most transportation modelers does not appear to treat these large trip
generators separately; therefore, the attractiveness of large trip generators are typically
simulated by their employment characteristics, which may significantly underestimate the
attracted trips for the TAZs containing these large trip generators. Also, episodic travel
events (e.g., concerts, baseball games, etc.), are not properly included in the trip
generation model.
In addition, trip productions and attractions are generated for a typical
weekday and do not reflect differences in travel patterns that occur during various
seasons in different parts of the nation (e.g., winter and fall in New England, summer in
the SoCAB) or on weekends. Several types of trips, such as work and school trips, are
reduced during the summer months, while agricultural and recreational trips increase
during these warmer periods. The use of average weekday productions and attractions
may introduce significant uncertainties in the generation of trip productions and
attractions during other periods.
Trip Distribution. The next step in the TMS process is trip distribution,
which connects zonal productions with attractions. Productions are typically distributed
among all available attractions in a region in proportion to the size of the attractions in
Conceptual Design of a New Motor Vehicle Source Model D - 1 -4
-------
each TAZ and inversely proportional to the distance or travel time between the
production and attraction using the so-called "gravity model." The distance variable in
the trip distribution model is computed using a path-building algorithm that is based
upon travel time. The resulting matrix of directionless trips between TAZs may be
converted into a matrix of origins and destinations by using the ratio of the production to
attraction movement to the origin and destination information obtained from trip
surveys. These origin and destination surveys are badly outdated in most regions (e.g.,
the trip generation model in the SoCAB was last updated with travel survey data
collected in 1976). Therefore, considerable errors may be introduced in properly
locating these trip ends in a given region.
Trip ends are used to compute emissions from cold start/hot start (by
technology type, e.g., catalyst equipped vehicles, etc.) and parks (hot soaks and diurnals)
in California. EMFAC, the emission factor model used in California, treats these trip
end effects separately from running emissions. However, MOBILE4, the emission factor
model developed by EPA and used in the rest of the nation, treats all emissions
proportional to VMT. This is accomplished by assuming an average number of trips per
vehicle per day and miles per vehicle per day for a given region. MOBILE4 contains
default values for these parameters. Although the user can supply region-specific
estimates, this is rarely done. The use of average default values in MOBILE4 has the
potential to introduce significant errors in the calculation of these emission components.
In addition, distributing these emissions on a per mile basis (instead of the grid cell in
which the emissions occur) results in spatial allocation errors.
The development of transit facilities and the encouragement of ridesharing
in many areas may make it difficult to forecast future year travel demand estimates
based upon current travel patterns, particularly for work trips. Consequently, there is
considerable uncertainty in the accuracy of travel demand estimates, particularly for
periods well into the future (i.e., the years 2000 and 2010), and the corresponding on-
road motor vehicle emission estimates that are a function of these travel characteristics.
Conceptual Design of a New Motor Vehicle Source Model D - 1 -5
-------
Trip Assignment. After trips have been distributed between TAZ, they are
assigned to specific links on the highway network. In most urban areas this assignment is
done on a daily basis, although some regions, such as the SoCAB, may make separate
trip assignments for the AM and PM peak periods. Capacity restraint techniques,
particularly for peak periods, are one of several procedures that may be used to assign
trips to specific roadways. Factors which affect capacity restraint assignment are
uncongested link speeds, link volume capacities, and the relationship between volume to
capacity (V/C) ratios and congested speeds. Speed on a given link is a function of the
V/C ratio of the link and the associated free flow speed. This relationship can be
simulated by many different types of functions. This, along with the fact that a single
assignment is performed for the entire day, makes the determination of the average
vehicle speed for a given link (and especially for a given hour of the day) very suspect.
Since on-road motor vehicle emission estimates are sensitive to changes in vehicle
speeds, the fact that hourly link-specific vehicle speeds are not developed is a
shortcoming of the TMS process and may introduce considerable uncertainty in the
development of emission estimates.
TMS Calibration and Verification. Calibration and verification of the trip
assignment model is less straightforward than the other portions of the TMS process due
to a lack of a single variable that can be adjusted to bring predicted and observed data
into agreement. In addition, highway trip assignment includes the errors occurring in
other parts of the TMS process. The indicator commonly used to assess the accuracy of
the highway assignment process is a comparison of estimated and observed link traffic
volumes. VMT is computed by multiplying link volumes or traffic counts by the link
distance and summing over all links. VMT comparisons primarily reveal systemwide
inaccuracies in the TMS process. For example, total regional estimated VMT may differ
from observed VMT totals. This probably indicates an error outside of the highway
assignment model, such as incorrect trip generation values from the trip generation
model, trip lengths obtained from the trip distribution model, or mode shares from the
Conceptual Design of a New Motor Vehicle Source Model D - 1 -6
-------
mode choice model. Improper calibration may introduce additional uncertainties in the
travel demand estimates produced by the TMS.
1.1.2 Emission Factor Models
On-road motor vehicle emission factors, are determined by a complex
process that considers the characteristics of the vehicles involved, as well as the
conditions under which they are operated. Uncertainties exist in the manner in which
the current emission factor model treats variations in speeds, temperatures, super
emitting vehicles, certain evaporative emission effects (e.g., running losses, resting losses),
and off-FTP cycle operations, as well as some of the inputs into the model (e.g.,
registration distributions, mileage accumulation rates).
Registration Distributions and Mileage Accumulation Rates. Application
of MOBILE4 and EMFAC7E requires, as inputs, registration distributions and mileage
accumulation rates by model year for each vehicle type. Currently, while registration
distributions and mileage accumulation rates will vary depending upon the calendar year
being modeled, application of both models (i.e., MOBILE4 and EMFAC7E) typically
involves the use only one set of registration distributions and mileage accumulation rates
for all regions. Variations in the registration distributions by region could impact the
resulting emission rates. This could be particularly important in certain parts of the
nation, such as the SoCAB, where excellent climate allows older, high polluting vehicles
to remain on the road longer. Recent research has shown that the use of alternate
distributions and rates can alter the resulting emission factors by approximately 10
percent for light-duty gasoline vehicles.
Speed Corrections. For emission factor calculations, speeds on individual
links are determined from transportation models as discussed above. These are checked
against roadside survey data to determine the accuracy of the estimates. The accuracy of
these speed estimates are important because both MOBILE4 and EMFAC7E show
Conceptual Design of a New Motor Vehicle Source Model D - 1 -7
-------
considerable sensitivity to variations in average speed, particularly at extreme high and
low speeds.
Temperature Corrections. For forecasting of emission impacts in a given
region, spatially resolved hour-by-hour temperatures are typically unavailable for use in
calculating emission factors. In certain regions of the nation, such as the SoCAB,
temperatures can vary significantly from one hour to another and from one part of the
region to another. Given the non-linear characteristics of the temperature correction
factors in both MOBILE4 and EMFAC7E, this could lead to inaccuracies in the
predicted emission factors, especially if spatially and temporally resolved emission
estimates (required as inputs to photochemical models) are desired.
Running Losses. The current methodology for calculating running losses in
both MOBILE4 and EMFAC7E is based upon testing of 47 vehicles sponsored by EPA.
The 47 vehicles were tested under a variety of temperatures and driving cycles. The
results from these 47 vehicles have been used to cover all model years of light-duty
gasoline automobiles and trucks.
Although EPA uses the same test data for calculation of running losses in
MOBILE4 as ARE does in EMFAC7E, EPA's calculation methodology differs from
ARB. According to ARB's documentation of EMFAC7E, ARE conducted a statistical
analysis and found significant effects of speed and temperature on running losses. EPA
assumes that running losses also vary with temperature, but that they are constant with
speed. ARB's methodology has running losses at a maximum at idle and decreasing
linearly down to zero at 45 miles per hour. At lower speeds and lower temperatures, the
difference between ARB's and EPA's methodologies are not significant. At higher
speeds and higher temperatures, the difference increases and EPA's running losses are
significantly greater than ARB's. For example, at 95°F, the difference between
MOBILE4 and EMFAC7E ranges from 0.05 gram/miles at 2.5 miles per hour up to 1.0
gram/mile at 45 miles per hour and above.
Conceptual Design of a New Motor Vehicle Source Model D - 1 -8
-------
Resting Evaporative Emissions. Resting evaporative emissions are
emissions that escape from the fuel or evaporative control system as a result of
permeation through various non-metallic fuel system sources, or as a result of vapor
migration allowed by other system design features, such as open-bottom canisters.
Resting evaporative emissions represent a potentially significant new class of ROG
emissions. They have not been accounted for in the past because evaporative emissions
tests were too short in length to detect them. The new, longer evaporative emissions test
procedure adopted by ARE will facilitate measurement of resting losses. At the present
time, however, data to fully quantify resting losses are not yet available. Therefore, the
current estimations of their impact are very imprecise. Recent research indicates a range
of possible resting loss emission factors from 0.07 grams/mile to 1.02 grams/mile. A
resting loss emission rate of 1.02 grams/mile would equate to an increase of 37 percent
in VOC emissions for on-road motor vehicles in the SoCAB (using EMFAC7E) based on
gasoline VMTs of 225,709,000 miles per day.
Evaporative Control System Failure Rates. EPA recently released data
from testing of evaporative control systems during a short duration loaded dynamometer
test The tests evaluated the system's ability to hold vapors under pressure and to
properly purge vapors under vacuum conditions. Analysis of the data showed that the
control system failure rates were significantly higher than previously believed.
Data released by EPA showed a minimum 20 percent increase in total
evaporative emissions. The overall effect on emissions in various regions is not clear
given the confounding effect of an I/M program. A rough estimate showed that the
minimum 20 percent increase in evaporative VOC emissions would lead to a 6 percent
increase in overall on-road motor vehicle VOC emissions.
Super Emitters. The emission factor model used by EPA groups vehicles
in emitter categories, from normal emitters that are at or below the emissions standards,
up to high and super emitting vehicles that emit at many times above the standards. The
Conceptual Design of a New Motor Vehicle Source Model D - 1 -9
-------
highest emitting vehicles, super emitters, represent an important group of vehicles for
calculation purposes. They do not represent a large portion of the vehicle population,
but their effect is magnified by their high emission rate, especially for VOC. While VOC
and NOX emission rates all increase as emission control systems deteriorate, VOC also
increases further when engines fail to maintain optimum air/fuel ratios, which can occur
due to vehicle aging or lack of maintenance. NOX emissions will decrease when engines
are operated under non-optimum air/fuel ratios.
Both MOBILE4 and EMFAC7E are quite sensitive to the emission rate
and growth of the super emitting vehicle population. Given a 20 percent increase in the
base super emitter emission rate in EMFAC7E, an increase of 6 percent in VOC is seen
in the overall light-duty automobile (LDA) emission factors. If the super emitter growth
rate is increased by 20 percent, LDA VOC emission factors are increased by 4 percent.
NOX emissions are not as sensitive to the growth in super emitters, since NOX emissions
are promoted by higher combustion temperatures in properly running engines. Analysis
of EMFAC7E showed that a 400 percent increase in the super emitter growth rate would
result in a 50 percent increase in the LDA ROG emission factor and a 30 percent
increase in the NOX emission factor. A 400 percent increase in super emitter growth
rates should be considered a worst-case scenario. Assuming a 400 percent increase in
super emitter growth rate and a 20 percent increase in base super emitter emission rates,
the LDA VOC emission factor would increase by almost 60 percent.
Off-FTP Cycle Operations. The impact of vehicle operation outside of the
FTP75 procedure on VOC emissions could represent a large source of error in the on-
road motor vehicle emissions inventory. While the FTP75 cycle is a transient procedure
that includes many accelerations, and consequently higher engine loads, it has been
noted that the procedure does not include severe acceleration rates often encountered in
normal vehicle operation. Accounting for operation outside of the FTP cycle could be
important to accurately determining in-use vehicle emissions. Knowledge of actual
driving conditions and their relationship to the FTP75 procedure becomes important.
Conceptual Design of a New Motor Vehicle Source Model
D -1 -10
-------
Recent work by ARE confirms that variations in engine speeds and loads outside of
those represented in the FTP75 procedure could significantly impact VOC emissions.
At this point in time, it is not possible to estimate the impact of differences
in off-FTP cycle operation on the overall emissions inventory. Clearly, operation outside
of the FTP75 conditions could significantly affect the accuracy of on-road motor vehicle
emission estimates.
Trip End Effects. As was stated above, MOBILE4 treats trip end effects
(i.e., cold start/hot start emissions, hot soak and diurnal evaporative emissions)
proportional to VMT, while EMFAC handles these emissions on a per trip (i.e., start or
park) basis. The potential errors introduced by the MOBILE4 methodology have already
been discussed above.
1.L3 Emission Estimation Methodology
Currently, the procedures available for combining travel demand estimates
and emission factors into emission estimates are limited, particularly for developing
speciated, temporally and spatially resolved values for input to photochemical models.
California has developed two emissions models for this purpose- the Direct Travel
Impact Model (DTIM) and BURDEN. DTIM was developed by ARB and the
California Department of Transportation (CALTRANS) to develop gridded, speciated,
hour-by-hour emission estimates for input to a photochemical model. DTIM uses travel
demand estimates generated by a TMS (i.e., hourly link-specific VMT and speed
estimates and TAZ trip end estimates) and EMFAC-derived emission factors. Although
DTIM provides the user with a great deal of flexibility, it is written in FORTRAN and is
a very difficult program to use. DTIM would have to be substantially modified to use
emission factors generated by MOBILE4. DTIM is not user-friendly and requires
several model runs to develop hour by hour emission estimates. In addition, DTIM will
not easily handle variable size grid cells or a nested grid structure.
Conceptual Design of a New Motor Vehicle Source Model _ -\-\-\
-------
ARE uses BURDEN to calculate on-road motor vehicle emission estimates
on a regional (e.g., county) basis for those areas that are not covered by TMS.
BURDEN estimates the number of vehicle trips in a given region, along with the hot
start/cold start splits and the associated VMT. Emission factors are obtained from
EMFAC and used to compute regional emission estimates. In this manner, motor
vehicle emission estimates are treated as an area source and spatial surrogates (e.g.,
population, length of road) are required to spatially allocate these emissions to grid cells.
Generic temporal profiles are used to develop hour-by-hour emission estimates.
Models similar to DTIM and BURDEN are typically not available for use
with MOBILE4. Most states use crude procedures to match regional estimates of VMT
with MOBILE4-derived emission factors to calculate on-road motor vehicle emission
estimates. Most areas do not directly use the output of TMS to develop link-specific
emission estimates and, therefore, software to compute these emission estimates in that
manner is not readily available.
Conceptual Design of a New Motor Vehicle Source Model
-------
2.0 BRIEF OVERVIEW OF A NEW MOTOR VEHICLE SOURCE MODEL
The new motor vehicle source (MV) model proposed by Radian
Corporation is based upon the development of a matrix of driving cycles and the
characterization of each link (by facility type and area type) in the transportation system
according to the driving cycle that best fits its particular use. Trip end effects are treated
separately from hot stabilized operation. Emission factors will be developed for each
driving cycle on a per-hour basis, based upon the use of a separate modal analysis
model. Alternative fuels will be accommodated. The modular nature of the MV model
will make it easy to substitute updated emission factors and driving cycles as vehicle
technology and vehicle usage change over time.
The MV model will generate hour-by-hour link specific emission estimates
for urban areas, based upon the development of travel demand estimates obtained from
a TMS designed specifically for air quality-related analysis (i.e., on-network travel
demand estimates). The new TMS will accommodate large trip generators and also be
able to handle micro-scale changes in travel demand estimates resulting from the
implementation of transportation control measures.
The MV model will also accommodate travel demand estimates developed
for rural areas for which the TMS will not be run. These travel demand estimates will
be distributed on the transportation network obtained from the TIGER/Line data and
treated in the same manner as on-network generated estimates.
The MV model will be user-friendly and be able to produce speciated,
gridded, temporally resolved emission estimates for input to a photochemical model.
The MV model will also be designed to facilitate projections of future year travel
patterns and emission estimates and various sensitivity analyses. This new MV model
would require a five (5) to 10 year timeframe for implementation.
A schematic diagram of the new MV model is depicted in Figure 2-1.
D - 2-1
Conceptual Design of a New Motor Vehicle Source Model
-------
1 "i
r ?2- *<•<* •-'-.<' i
'* ARCaNFO*^ 1
SAS» Based Functions
g-.^/neften*-?,;
*L'>"T^ - '"'
/f™*tai F«nr la*4Jp Ttbit\
I (Gnrn«*-T«np«in).br I
V V
lM-tn
-------
3.0 DEVELOPMENT OF TRAVEL DEMAND ESTIMATES
Currently, a diverse range of methodologies are being used to develop
estimates of vehicle miles traveled (VMT) and speeds in urban areas, resulting in a wide
range of accuracies. TMS have been designed to develop regional travel demand
estimates and, accordingly, do not handle variations in link-level traffic operations during
peak congested conditions very well. Emphasis is generally on planning for major
corridor-level projects and on forecasting traffic volumes on major roadways.
Considerably less importance is placed on air quality planning needs such as travel
speeds, particularly for minor collectors and residential streets.
Many urban areas do not even use a TMS but develop regional travel
demand estimates based upon traffic count data, fuel sales or a hybrid approach. Since
MOBILE4 treats trip ends proportional to VMT, no attempt has been made outside of
California to develop accurate estimates of starts and parks and their distribution (both
spatial and temporal) for a given region. In order to produce accurate travel demand
estimates for Radian Corporation's new motor vehicle (MV) source model, a new type of
TMS, which emphasizes generating the travel demand estimates required to drive this
model, must be developed. The following discussion discusses how such a TMS might be
structured.
3.1 Driving Cycle Matrix (by Facility and Area Type)
The Federal Test Procedure (FTP) was developed by EPA to certify that
passenger cars comply with applicable federal emission standards. Current emission
factor models (MOBILE4 and EMFAC) rely on FTP-based emission estimates, corrected
for varying average speeds and temperatures. However, in-use driving patterns can vary
dramatically from the FTP cycle, particularly the number of accelerations/decelerations
associated with a particular driving cycle. Speeds greater than 48 mph are not even
included in the FTP. MOBILE4 only includes speeds from 5 to 55 mph. The model
Conceptual Design of a New Motor Vehicle Source Model D - 3-1
-------
proposed by Radian will be based upon the development of emission factors, on a
grams/hr basis, for a pre-determined number of driving cycles or a driving cvcle -
emission factor matrix for different combinations of roadway facility types (e.g., freeway,
expressway, etc.) and area types (e.g., central business district [CBD], rural, entrance/exit
ramp, etc.). This matrix will be stratified by peak and off-peak travel periods. Each link
in a given transportation network will be classified according to its use and matched to a
particular driving cycle and corresponding emission factor.
The vehicle distribution associated with each facility type/area type
combination in the driving cycle matrix, must also be developed on an hourly basis, along
with the diurnal distribution of weekday (and weekend) trips. This information can be
obtained by instituting a comprehensive and statistically valid traffic count program
designed around the driving cycle matrix. Additional information regarding link-specific
vehicle distribution data will be obtained from satellite imagery. At a scale of 1:400, it is
possible to count traffic and at a scale of 1:100, it is possible to discern passenger cars
from trucks, and it might be possible to actually further resolve vehicles by type (other
than just cars and trucks), using improved computerized data reduction techniques.
The possible facility types that will be part of Radian's proposed TMS
model include the following classifications:
• Freeway;
• Expressway;
• Major arterial;
• Minor arterial;
• Major collector;
• Minor collector;
• Residential streets; and
• Entrance/Exit ramps.
Conceptual Design of a New Motor Vehicle Source Model
D-3-2
-------
The first five (5) classifications are typically coded into the highway
network currently used in TMS. Minor collectors and residential streets are not usually
accounted for in a typical transportation network model, but will be included in Radian's
model. Vehicle activity on these types of links typically account for, 10 to 15 percent of
low speed, congested, stop and go traffic volume in a given urban area. However, the
low speeds and congested nature of travel on these surface streets makes these roadways
significant sources of vehicle emissions. Radian's TMS will take advantage of the
increased storage capacity and processing speed of today's (and tomorrow's) computers
and geographic information system software (GIS) to expand the transportation network
to include vehicle activity on minor collectors and residential streets.
Entrance and exit ramps will also be explicitly included in the TMS due to
the large amount of vehicle emissions generated during accels and decels. This is
especially true of metered entrance ramps. Inclusion of entrance/exit ramps in the
driving cycle matrix will be an important additional facility type that has been typically
excluded from existing network models.
The possible area types that will be part of Radian's proposed TMS model
include the following classifications:
Central Business District (CBD);
• Urban;
• Residential;
• Commercial;
• Suburban; and
Rural.
Comprehensive traffic count programs will be developed to support the
construct of the driving cycle matrix. The traffic counting program will include the use
of permanent embedded counters that are able to transmit data directly to a central
Conceptual Design of a New Motor Vehicle Source Model
»* _ *
-------
computer system for reduction, analysis and storage. These embedded counters will be
supplemented with traditional counters, video recorders of selected roadways (for vehicle
classification identification) and data obtained from satellite imagery.
Driving cycles that best characterize each facility type/area type
combination will be developed. These driving cycles will form the foundation for the
development of emission factors, on a grams/hr basis, as described in Section 4.0. Table
3-1 presents an example of a possible driving cycle matrix by facility type and area type.
3.2 Origin — Destination Travel Surveys
The cornerstone of any TMS is comprehensive and current origin -
destination travel survey data. These data, along with the demographic and
socioeconomic characteristics of the residents in a particular region, form the basis of the
trip generation models that produce the trip generation rates that drive the TMS.
Current trip generation models used in TMS generate trip productions and attractions by
traffic analysis zones (TAZ).
Trip generation estimates (i.e., trip productions and attractions) are
normally made for an average weekday for each TAZ and are independent of trip
direction, length, and duration. The following five trip types are used in the South Coast
Air Basin (SoCAB), for example, and illustrate the type of trips that are typically
considered:
• Home-other;
• Home-work;
• Home-shop;
• Other-work; and
• Other-other.
Conceptual Design of a New Motor Vehicle Source Model Q _ *m
-------
o
o
Table 3-1
o
n
v>
Driving Cycle Matrix
Mode Distributions by Facility Type and Area Type
2
n
o
s
o
o
1
Area Type
Central
Business
District (CBD)
Urban
Residential
Commercial
Suburban
Rural
Facility Type
Freeway
11*
21
31
41
51
61
Expressway
12
22
32
42
52
62
Major
Arterial
13
23
33
43
53
63
Minor
Arterial
14
24
34
44
54
64
Major
Collector
15
25
35
45
55
65
Minor
Collector
16
26
36
46
56
66
Residential
Streets
17
27
37
47
57
. 67
Entrance/
Exit Ramps
18
28
38
48
58
68
<*>
en
* ij matrix cell address for: Driving cycle (peak)//
Driving cycle (off-peak)//
-------
The first three trip types listed above refer to trips that either start or end
at the home, e.g., the home work trip encompasses the trip from home to work in the
morning and the return trip from work to home in the evening. The two trip types
classified as other-work and other-other are trips that neither start nor end at home.
The other-work trip is a trip that either starts or ends at work, but does not involve the
home as a beginning or end point. The other-other trip is a trip that neither involves the
home nor the work location.
A trip production is defined as the home end of a home-based trip or the
origin of a non-home-based trip. Similarly , an attraction is defined as the non-home
end of a home-based trip, i.e., the "work", "shop" or "other" end of a home-based trip.
The "other" end of the other-work trip is defined as the production end and the "work"
end as the attraction. The other-other trips are usually split evenly into productions and
attractions.
Each trip can be visualized as being produced in one TAZ and attracted to
another. A significant portion of the VOC emissions from on-road motor vehicles (e.g.,
evaporative emissions), are directly proportional to the number of starts (origins) and
parks (destinations) in a given TAZ. The starts are further broken down into cold starts
and hot starts, and the emissions during the park period are a function of the number of
hours a vehicle is parked before it is re-started. In order to accurately quantify and
locate these "trip end" emissions in a given region, it is necessary to accurately determine
the number of origins and destinations in each TAZ, not productions and attractions.
The Radian TMS will be based directly on origin - destination data and
will not convert these data into productions and attractions for subsequent distribution
(as is typically done in today's TMS using the so-called "gravity" model). In addition, the
information gathered in these surveys will be modified to include time of day
information, preferred route of travel for each trip, use of alternative fuels, vehicle class
and technology type, and an estimate of the time a vehicle is parked before it is re-
Conceptual Design of a New Motor Vehicle Source Model _
U - 3-o
-------
started. The surveys will be expanded to include information on commercial travel,
which is currently not handled properly in trip generation models. In addition, more
comprehensive data regarding truck traffic will be gathered so a comprehensive trip
generation model can be developed. These data, along with socioeconomic and
demographic data obtained from the TIGER/Line data (TIGER/Line Data contain
cartographic representations of different geographic features, such as census tract
boundaries, which include census statistics), will be used to construct new trip generation
models that produce origin (starts) and destination (parks) data for each TAZ by:
• Time of day;
• Day of week;
• Vehicle class and technology type;
• Park duration; and
• Fuel type.
Information will also be gathered to determine if a significant amount of
"short trips" (e.g., moving one car out the way so another can leave in the morning,
driving around parking lots, moving from one location to another in a shopping mall,
etc.) are typically generated. These trips, which are not included in TMS, may
significantly contribute to motor vehicle emissions. These data will also be used to
develop a trip distribution model that estimates trips between and within a TAZ, by hour
of the day and fuel type, based directly on origin - destination travel survey data.
These comprehensive origin « destination surveys will be conducted every
three to five years, except in high growth areas like the SoCAB, where yearly surveys
may be required.
These surveys could either be included as part of permit requirements for
industry, e.g., both new construction and operating permits and operating permit
renewals (such as those required under Title V of the 1990 Clean Air Act Amendments),
Conceptual Design of a New Motor Vehicle Source Model Q . 3.7
-------
new permit requirements for major trip generators (e.g., shopping malls, airports) or
additional permit fees assessed on existing pollution sources (keyed to the number of
employees). A computerized survey could be designed to facilitate the analysis of these
data. It might be possible to add questions regarding the use of certain consumer
products to gather activity data to produce more accurate area source emission estimates,
which will help justify the cost/effort of conducting these surveys. One of the national
bulletin boards, such as PRODIGY® might be elicited to add an origin ~
destination/fuel use survey to their product line as a public service. Local/national
environmental groups could be used to gather this type of information, especially if
questions regarding the use of public transportation were added. And finally,
computerized software could be developed as part of a national/local educational
campaign and used by students to query their parents and supply the information to their
local air pollution control agency on magnetic media (e.g., floppy disk).
Results of these surveys could be published and, given the current state of
environmental awareness, all who participated in gathering this information would
become part of the solution to local air quality problems. Local business and industry
could be solicited in a cooperative effort to develop the software needed to make this
survey "fun" to complete and cost-effective, a sort of "environmental census." Although
origin -destination surveys are expensive, they are critical to the development of a.TMS
that is more tuned to providing more accurate on-road motor vehicle estimates. There
are numerous ways, some of which have been suggested above, to cost-effectively carry
out these surveys.
3.3 Transportation Modeling
Separate procedures will be developed to treat urbanized areas (i.e., on-
network travel demand estimates) and rural regions (i.e., off-network travel demand
estimates) as described below. The resulting TMS will be embedded in the motor
Conceptual Design of a New Motor Vehicle Source Model _
U- 3-8
-------
vehicle (MV) source model to allow for integration of travel demand estimation,
emission factor modeling and the development of emission estimates.
3.3.1 On-Network Transportation Modeling
Current TMS used in urban areas include the following component models
as described earlier:
• Trip generation;
• Network development;
• Trip distribution;
• Mode choice; and
• Trip assignment.
As was stated above, the updated origin - destination travel survey
information will be used to develop new trip generation and distribution models that are
based directly on origin and destination information. The trip generation and
distribution models will account for trips generated by commercial travel, including
trucks, trips on minor collectors and residential streets, and external trips (trips that exist
outside of the region of interest and either pass through the region or end in that area).
These models will be developed in fourth-generation software, such as SAS®, to facilitate
the analysis of the origin - destination travel data and the development of the algorithms
needed to accurately develop trip generation and distribution rates.
The network configuration will be developed using GIS software, such as
ARC/INFO®, to allow for the modeling of all streets (including minor collectors,
residential streets and on/off ramps). Modern computer technology, with its increased
storage and processing power, will allow for this expanded network configuration. This
will facilitate the subsequent assignment of all trips to links, which will be categorized
according to the driving cycle matrix classification scheme described above.
Conceptual Design of a New Motor Vehicle Source Model « _ 3.9
-------
Trip assignments will be made on an hourly basis, for each fuel type used
in a given region. The updated origin - destination travel surveys will provide the
detailed data required to make hourly assignments, by fuel type. Instead of VMT, the
output of the assignment model, which will also be coded in SAS®, will be vehicle hours
traveled (VHT) on each link (by time of day and fuel type). Travel times instead of
average link speeds will be modeled to calculate vehicle hours on each link. Travel
times are probably more accurate than average link speeds and the emission factor
model that will be required, which is described below, will generate emission rates on a
grams/hr basis instead of grams/mi. Current techniques based upon the use of speed
corrections require link-specific speed estimates. However, the capacity restraint
techniques that are used to manipulate link speeds do not necessarily represent accurate
speed estimates for a given segment, but rather a value that optimizes traffic assignment
over the entire "congested" network. Link-specific travel times are easier to predict and
gram/hr emission rates, "for links by facility type/area type combinations in the driving
cycle matrix, will be more uniform than those based upon grams/mi values.
The use of origin - destination travel data to build the TMS will facilitate
the development of hourly trip end information. The updated survey information will
also provide the data necessary to more accurately determine cold start/hot start splits
and park durations (by time of day and day of week). A separate SAS®-based module
will be developed to manage trip end data. The emission factor model recommended by
Radian will treat trip ends separately from running emissions (by fuel type). Trip end
emissions will be associated with the centroid of the traffic analysis zone (TAZ) in which
the corresponding trip origins and destinations occur. The emissions model described
below will allocate the trip end emissions to the appropriate grid cell(s).
Micro-Scale Transportation Modeling. Radian's proposed TMS will be
constructed to better handle the micro-scale changes in travel demand resulting from the
addition of relatively large trip generators (e.g., shopping malls, civic centers). This same
model must also better accommodate the benefits resulting from instituting various
Conceptual Design of a New Motor Vehicle Source Model — "4 1 n
-------
transportation control measures (TCMs). While techniques currently exist to estimate
the change in travel demand resulting from the implementation of various TCMs, these
micro-scale models are not currently incorporated into current TMS. Radian's TMS will
be developed in a manner that allows these micro-scale transportation techniques to be
directly incorporated into its TMS, using a nested TMS approach. The TMS user will be
able to first run the TMS and then select an area for micro-scale analysis using the
results of the TMS as boundary condition inputs to the micro-scale model. The micro-
scale model will then be run and the resulting link-specific changes in travel demand will
be input into the network configuration that comprises the TMS. This procedure will be
facilitated by incorporating the minor collectors and residential streets in the TMS. This
will allow the air quality/transportation specialist to accurately analyze the effects that
various types of TCMs or trip generators might have on travel demand in the modeled
domain. These changes in travel demand will be matched with the appropriate emission
factors and the emissions processor will calculate emission estimates for comparison
purposes.
3.3.2 Off-Network Transportation Modeling
The TMS described above will be used for most urban areas. However,
there will be considerable portions (the more rural areas) of a given region that will lie
outside of the area that will be covered by the TMS. In most instances, it will not be
cost-effective to develop detailed origin - destination data for these areas as they
contribute a small percentage of the total trips in a given region and the nature of travel
in these rural areas (i.e., free flow) is such that emissions are minimized under most
circumstances. Therefore, a separate type of transportation modeling procedure,
described below, will be developed to estimate vehicle activity for this "off-network"
travel.
Travel demand estimates can be obtained in the off-network areas by
developing a relationship between fuel sales and vehicle activity. This relationship can
Conceptual Design of a New Motor Vehicle Source Model D - 3-11
-------
also be used for quality assurance purposes, particularly in regions where multiple fuels
will be used. For example, gasoline sales are tracked at the state level for tax purposes
and fuel marketing-specific information at the station level may be available from state
Leaking Underground Storage Tank (LUST) programs or from the individual stations,
through the use of the TIGER/Line data and surveys. Statewide transportation
modeling is another potential source of the off-network vehicle activity data. Although
statewide transportation models typically include major roadways (e.g., freeways and
expressways), the amount of travel on minor roadways in rural areas not covered by the
statewide system is of lesser importance. VMT data can be converted to VHT by
assuming an average free flow speed. Trip end information could be obtained from the
results of statewide surveys or extrapolated from results obtained from origin -
destination travel surveys conducted in the more urbanized areas.
Off-network travel demand estimates will then be allocated to the major
roadways in the off-network region using the transportation system included in the
TIGER/Line data and using GIS techniques. The area defined by the on-network
transportation model will be removed from this "pseudo transportation network" and off-
network vehicle activity will be allocated to these "pseudo-links" based upon the ratio of
the link-specific road length to the total length of roads that comprise the pseudo off-
network configuration. Once travel demand has been allocated to links, temporal
profiles will be used to develop hourly estimates and fuel type sales will be used to
develop fuel-specific activity data. These off-network travel demand estimates will then
be treated in the same manner as the on-network generated data.
A separate driving cycle will be developed to characterize driving patterns
on the off-network roadways. Corresponding emission factors will be generated for this
driving cycle and the same emissions processor will be used to develop gridded,
speciated, temporally resolved emission estimates. These off-network procedures could
also be used in urban areas that are unable to conduct the necessary origin - destination
Conceptual Design of a New Motor Vehicle Source Model
D-3-12
-------
travel surveys needed to develop the TMS described earlier and the associated detailed
travel demand estimates.
Conceptual Design of a New Motor Vehicle Source Model _
L/ ~ J~l
-------
D-3-14
-------
4.0 EMISSION FACTOR MODEL
On-road motor vehicle emission factor modeling is currently based on a
database developed primarily for vehicle emissions certification, not emissions modeling.
Consequently, these data have required extensive modifications and additions to allow
the database to cover the wide range of operating conditions experienced by in-use
vehicles. Recent research has pointed towards limitations imposed by the use of this
database and the need for development of additional data devoted to emission factor
estimation.
Some of the areas of emission factor modeling which need improvement
include:
The impact of off-FTP cycle operations and improvement of the
model's ability to handle a wide variety of operating conditions.
Recent research by ARE and others indicates operation outside of
the conditions of the FTP can significantly impact emissions.
Improve procedures within the motor vehicle emissions model to
account for vehicle speeds, and ambient temperatures. Both
MOBILE4 and EMFAC7E show significant sensitivity to the
accuracy of speed and temperature predictions.
The potential impact of super emitting vehicles. Both MOBILE4
and EMFAC7E are sensitive to the population and emission rates of
this small subset of the vehicle fleet.
Further investigation into resting evaporative losses. Research is
needed to quantify the impact of these emissions and to include
them in future emission factor models.
Incorporation of revised evaporative control system failure rates.
Data from testing by EPA show significantly increased evaporative
hydrocarbon emissions due to pressure and purge failures of current
evaporative control systems.
Further development of the database for vehicle exhaust and
evaporative emissions at the higher temperatures (greater than 75
Conceptual Design of New Motor Vehicle Source Model D - 4-1
-------
°F) likely to be encountered during the summer months in many
ozone non-attainment areas, such as the SoCAB. Data for some
emission types, such as running evaporative losses, are based on
limited data for operation over 75°F.
Development of region (county)-specific registration and mileage
accumulation distributions. Much of this data is available from
inspection and maintenance databases, along with more region-
specific information on mileage accumulation rates.
The following discussion describes a new type of emission factor model
which addresses many of the shortcomings of current versions of both MOBILE4 and
EMFAC7.
The key to this model is the use of a modal emissions database to build
emission factors on a grams/hr basis, for the facility and area type driving cycle matrix
described in Section 3.0. These facility type/area type specific emission factors will be
built from a database of vehicle emissions by operating mode. In this case, operating
mode will refer to general vehicle operating conditions, such as idle, cruise, acceleration,
cold starting, etc. rather than the current limited MOBILE4 definition of cold start, hot
start and hot stabilized operation. Combinations of emissions from each of the operating
modes will be aggregated to build a driving cycle which approximates vehicle emissions
for each vehicle type within a facility type/area type driving cycle combination. The
operating mode-specific emission factors will be developed in the Modal Analysis Model,
described in Subsection 4.1. The individual emission factors will be assembled into
emission factors for specific facility and area types in the Emission Factor Model,
described in Subsection 4.2.
4.1 Modal Analysis Model
The Modal Analysis Model will be developed to generate emission factors
for several specific driving, or operating, modes. This is in contrast to the current test
procedure, the FTP which develops emissions for one driving cycle and adjusts the
Conceptual Design of New Motor Vehicle Source Model _ . _
D - 4-2
-------
emissions for different conditions of speed and temperature. Recent research has shown
that certain vehicle operating conditions, such as high loads and accelerations, can
produce significantly higher vehicle emissions than normal cruising. The purpose of the
Modal Analysis Model will be to quantify vehicle emissions by these specific operating
modes.
The design and characteristics of the individual operating modes needed to
fully characterize the range of vehicle operations is yet to be completed. However,
research towards this need is underway, sponsored by the California Air Resources
Board (ARB) and others. For discussion purposes, we estimate that a minimum set of
operating modes would include:
Idle
• Steady state normal (surface street) cruising;
• Steady state high speed (freeway) cruising;
• Moderate acceleration;
• Severe acceleration;
• Deceleration;
• Cold starting; and
• Hot starting.
Previous research by U.S Environmental Protection Agency and others has
shown that vehicle emissions over a variety of operating cycles can be reasonably
approximated using emissions based on individual operating modes. The output of the
Modal Analysis Model would be a set of emission factors describing vehicle emissions
by:
Vehicle Type;
• Operating mode;
• Fuel Type; and
Conceptual Design of New Motor Vehicle Source Model
-------
• Ambient Temperature.
Note the lack of vehicle speed as an output of the model. The Modal
Analysis Model would express vehicle emissions as a function of operating time rather
than distance (grams per hour versus grams per mile). When vehicle emissions are
divided into separate operating modes, the effect of vehicle speed can be accounted for
in the choice and distribution of the individual operating modes. All of the operating
modes listed above are dependent more upon operating time than speed. The exception
to this would be steady state cruising. However, while vehicle emissions do change with
vehicle speed when expressed on a grams per mile basis, they are relatively constant with
vehicle speed when expressed on a grams per hour basis at normal speeds. This trend
changes at high speeds such as freeway cruising. For this reason, a separate operating
mode for high speed cruising is proposed.
Development of the modal emissions database would obviously require a
significant testing effort to include the necessary vehicle types, emission control
technology types, fuel types, model years, and emissions levels (normal, high and super
emitters). This effort could be mitigated to some degree by developing a driving cycle
that includes all of the individual operating modes within it. If emissions were measured
on an on-line basis, rather than with discrete bag measurements in the current FTP, then
all modes could be measured within one driving cycle. This new simulation could be
added to the existing FTP cycle to gather data during the vehicle certification process.
In addition, if the FTP test itself were measured on an on-line basis, the
additional information could be used to supplement the added driving cycle. This would
provide critical information regarding the variations of the emissions measurements. In
turn, this information could be used to help estimate the uncertainty in the emission
factor model, an aspect currently lacking from present models.
Conceptual Design of New Motor Vehicle Source Model
D-4-4
-------
4.1.1 Evaporative Emissions
The Modal Analysis Model is designed primarily to deal with exhaust and
running evaporative vehicle emissions. Stationary vehicle evaporative emissions, such as
diurnal, resting and hot soak emissions should be dealt with separately. The current
methods used in MOBILE4 does not permit resolution of evaporative emissions by hour
of the day. In this motor vehicle source (MV) model, evaporative emissions will be
produced from a distinct calculational methodology such as an expanded version of the
Coordinating Research Council (CRC) Evaporative Model, EVAP 2.0. EVAP 2.0 was
developed by Radian to model vehicle evaporative emissions. The current EVAP model
includes diurnal and hot soak emissions and is being modified to include resting
evaporative losses.
The current methods in MOBILE or EMFAC are not fully able to account
for time and driving cycle effects on diurnal and hot soak emissions. This is due mostly
to a lack of available data at the time the models were developed. For example,
MOBILE4 and EMFAC7E divide diurnal emissions into three time and driving cycle
groups: partial diurnals, full-day diurnals, and multiple-day diurnals. In the development
of EVAP 2.0, vehicle driving pattern information has been analyzed to resolve vehicle
use characteristics by hour of the day.
Sensitivity analyses of the EVAP 2.0 model has shown many parameters to
be significant contributors to evaporative emissions. The analyses also included
parameters that were not currently included in the model to help identify areas for
further development. The important parameters accounted for in the model include:
Fuel Reid Vapor Pressure (RVP);
• Ambient temperature;
• Driving pattern;
• Vehicle evaporative emission control technology; and
Conceptual Design of New Motor Vehicle Source Model
-------
• Older and/or tampered vehicle emissions.
The EVAP 2.0 model also calculates diurnal and hot soak emissions on an
hour-by -hour basis, as opposed to the procedures used in existing models which
calculate emissions on a daily basis and assume emissions are evenly spread throughout
the day. While this distinction does not affect annual inventories, the EVAP 2.0 model is
much more suitable for use in more detailed inventories where emissions are modeled
on an hourly basis.
4.12 Trip-End Emissions
Many vehicle emissions occur when the vehicle is stationary rather than in
running operation. These include diurnal, hot soak and resting evaporative losses as well
as cold and hot start exhaust emissions. The current methods used in MOBILE4 do not
permit proper spatial allocation of these emissions, since they are calculated on a grams
per mile basis and spread over the road network. For regional emission inventories, this
is adequate, but as the need for more accurate emission estimates increases, it will be
difficult to achieve the proper spatial resolution with the current method.
This MV model will permit proper spatial allocation of both exhaust and
evaporative emission since they will each be calculated separately. In this way, it will be
possible to allocate the emissions closer to where they occur. In addition, by separating
emissions into individual operating modes, it will be possible to construct operating
cycles specific to certain areas, such as freeway entrance and exit ramps. This should
further improve the spatial resolution capabilities of the model.
Conceptual Design of New Motor Vehicle Source Model Q .
-------
4.13 Alternative Fuels
The potential growth in the number of automotive fuels presents a
significant logistical problem for a future motor vehicle source model. The problems
that alternative fuel use presents include:
• Determining the emission characteristics of vehicles operating on
different fuels;
• Quantifying the use of different fuels by vehicle type;
• Spatially resolving the emissions from alternative fuel use; and
• Dealing with the variable emissions of flexible-fuelled vehicles.
If the fuels under consideration had enough similarities, it might be
possible to group fuel by their emission characteristics. However, the significant
differences in emission characteristics of the fuels currently being considered require the
development of individual databases for each fuel.
In this MV model, modal exhaust and evaporative emissions will be
determined by fuel type for each affected vehicle type. It is proposed that the MV
model develop emission factors, rather than corrections to the existing gasoline-based
emission factors. One advantage if the fuels are dealt with separately is the capability to
add new data if more fuel types become available. The Modal Analysis Model will
output emission factors by fuel type for each operating mode.
The difficulty of dealing with flexible-fueled vehicles presents a unique
problem, since their emission characteristics vary with the fuel being used. At this time,
it does not appear that flexible-fueled vehicles operating on a wide range of fuels will be
common enough to warrant special treatment. Most likely, vehicles with flexible fuel
capability will be operated primarily on specific types of fuel, enabling the use of specific
emission factors. Although survey data should be gathered, the time spent operating on
Conceptual Design of New Motor Vehicle Source Model
_ *_j
-------
non-standard fuel blends may be small enough to be treated through small adjustments
to the standard fuel emission factors.
42 Emission Factor Model
The Emission Factor Model takes the individual operating mode emission
factors from the Modal Analysis Model and combines them to form driving cycle
emission factors specific to the elements of the Driving Cycle Matrix (see Table 3-1).
Separate emission factors will be constructed to reflect peak (congested) and off-peak
(uncongested) travel.
The driving cycle will be a composite emission factor that reflects the
percentage of time the fleet spends in each of the individual operating modes. To build
a composite emission factor for a specific facility type and area type, survey data are
needed to determine the amount of time spent in each of the individual operating
modes. This time distribution is then used to develop a single facility type/area type
emission factor from the composite of the individual emission factors. The time
distribution concept is demonstrated in a hypothetical example shown in Figure 4-1. The
example shows two distributions, both for an urban (area type) freeway (facility type),
but for on-peak and off-peak travel.
Operating mode time distributions will be developed for each facility type,
area type and congestion level (on-peak or off-peak) through survey data. It will be
possible to embed default time distribution values into the model for applications where
region-specific data are not available. If other data are available, then the user can have
the opportunity to input new operating mode distributions for specific facility and area
types.
One immediate advantage to the proposed system is the ability to use
region-specific information regarding operating mode distributions. Additionally, it is
Conceptual Design of New Motor Vehicle Source Model p .
-------
Q
LLJ
_J
LU
1
W
cc
Z>
o
HI
_J
o
X
LU
U_
O
I-
LU
O
OC
HI
DL
HK^
H^ss -
S^BKVN
^ip^
:
:%
URBAN
FREEWAY
ON-PEAK
1
1
I
'•mji
: 3^/8888 ,
ESS
IDLE
STREET CRUISE
S§3
FWY CRUISE
MODERATE ACCEL.
SEVERE ACCEL.
DECEL
URBAN
FREEWAY
OFF-PEAK
Figure 4-1. Hypothetical Driving Cycles for Urban Freeway-Operating Mode Distribution.
-------
also possible to easily create additional facility or area types with unique operating mode
distributions. In this way, the model can be truly customized for local applications.
Once the composite facility type/area type driving cycle emission factors
are calculated, the emission factors are combined with the vehicle activity data developed
from the transportation models for each link and node. The activity data will include:
• Total vehicle hours traveled (VHT) on a link;
• Distribution of VHT by vehicle type and fuel for each link;
• Number of cold starts, hot starts for each node;
• Number of park hours for each node (diurnal and hot soak
emissions);
The product of the emission factors and the vehicle activity data will be exhaust and
evaporative emission estimates by individual link and node for each hour of the day.
The development of emission estimates is described in more detail in Section 5.0.
Conceptual Design of New Motor Vehicle Source Model D . 4_-j Q
-------
5.0 DEVELOPMENT OF EMISSION ESTIMATES
Travel demand estimates on a link-specific basis need to be combined with
corresponding emission factors to develop emission estimates. The methodology that is
used to perform these calculations should be able to handle regional variations in
temperature on an hourly basis. In addition, it is desirable to be able to grid and
speciate the resulting emission estimates for use in photochemical models. Currently, the
only software available to calculate link-specific emission estimates and to grid and
speciate those values is the Direct Impact Travel Model (DTIM) developed
cooperatively by the California Air Resources Board (ARE) and the California
Department of Transportation (CALTRANS). DTIM reads CALTRANS formal trip
assignment and network description files and files of emission rates generated by
EMFAC7. DTIM uses these travel demand estimates and emission factors to develop
gridded emission estimates. DTIM permits travel demand estimates represented on the
assignment file to be allocated to any hour of the day with a corresponding temperature.
The user can also provided gridded temperature values on an hourly basis. The
CALTRANS trip assignment program and its output file allow trips to be stratified by
time of day or trip purpose, each with separate emission factors. DTIM resolves both
link and zonal emission estimates into user-specified grid cells. However, DTIM is
written in FORTRAN and is difficult to use. The Radian motor vehicle model will use a
more user-friendly model to compute link-specific emission estimates that is based on the
use of fourth generation software (i.e., SAS® and ARC/INFO®).
Radian's motor vehicle source (MV) model (see Figure 2-1) will accept as
input the ARC/INFO® transportation network configuration from the TMS (for
urbanized areas) and the off-network transportation system (for rural areas) as described
in Section 3.0. These coverages will then be merged into a single coverage of links and
nodes (including TAZ centroid nodes) for the entire region (i.e., both on-network and
off^network domains). The links that comprise these coverages will contain the hourly
vehicle hours traveled (VHT) (for use with the grams/hr emission factors) and an
Conceptual Design of New Motor Vehicle Source Model D . 5--J
-------
identifier which matches a given link with a driving cycle (from the driving cycle matrix,
based upon the facility type/area type combination that best characterizes a given link).
TAZ nodes will contain relevant trip end data.
A coverage of gridded temperatures will be developed based upon the use
of observed values or the output of a meteorological model, and the temperature will be
added to each link/TAZ node. A separate coverage will be maintained for each hour of
the day.
These coverages will be converted into SAS® data sets and passed through
a temporal and spatial adjustment processor which will allow the modeler to modify
temporal profiles or link-specific vehicle hours for a user-defined area (useful for
sensitivity analyses and forecasting purposes).
A link-specific age-composited emission factor, for each vehicle class, will
be obtained from the emission factors that have been developed for each driving cycle.
An emission factor look-up table based upon temperature will be developed for
evaporative, cold start/hot start and diurnal emissions. Each link will contain the vehicle
mix (i.e., fleet mix), which is obtained from its respective facility type/area type
classification. The emission factor for that link/node, for each vehicle class, will be
obtained from the look-up table, fleet composited, and added to the file.
Emission estimates for each link/node will be computed in SAS® by
multiplying the fleet composited emission factor by the VHT trip end activity data
(assigned to each TAZ node as described earlier). This computation will be performed
for each link and TAZ node, for each hour of the day. The SAS® data set containing
the link/nodes and emission estimates will be converted back into ARC/INFO®. A
series of hourly coverages, which now includes the appropriate emission estimates, will
be gridded in ARC/INFO® according to a user-defined grid structure (uniform grids,
variable grids, nested grids, etc.).
Conceptual Design of New Motor Vehicle Source Model
D-5-2
-------
The gridded, hourly coverages will be converted back into SAS® and
speciated according to user-specified speciation profiles. The speciated VOC emissions
can be used to generate emission reports or combined into the user-defined VOC-species
groupings required by the chemical mechanism in a desired photochemical model.
Conceptual Design of New Motor Vehicle Source Model p . 5.3
-------
D-5-4
-------
6.0 DATA INPUT REQUIREMENTS
The following provides a list of the types of data inputs that will be
required to run Radian's new motor vehicle model:
• Socioeconomic/Demographic data;
• Origin/Destination travel surveys;
• Origin/Destination trip tables, by hour of the day;
• Traffic count data;
• Transportation system roadway characteristics;
• Tiger/Line data;
• Fuel sales;
• Gridded temperatures for region under analysis;
• Vehicle mode test data;
• Region-specific vehicle data (e.g., registration distributions, I/M
programs, etc.);
• Engine maps;
• Driving cycle matrix (by facility type and area type);
• Grid definition; and
• Speciation profiles.
The resources required to gather these data will be substantially greater
than those needed to use current procedures. The cost of obtaining the detailed origin -
destination travel survey information necessary to develop the travel demand estimates
for Radian's motor vehicle source model will be significant. However, several options
for minimizing the amount of resources needed to conduct these surveys were presented
in Section 3.0. In addition, the mode test data needed to develop the driving cycle
matrix and associated emission factors will also be costly. On the other hand, motor
vehicles are the most important source category in most urban areas and a strong
argument can be made that more resources should be devoted to improving the
methodology used to characterize motor vehicle emission estimates.
Conceptual Design of a New Mntnr Vehicle Source Model D - 6-1
-------
D-6-2
-------
7.0 MOTOR VEHICLE SOURCE MODEL VALIDATION PROCEDURES
One of the most important elements of model development is the
procedures that will be used to verify the resulting emission estimates and travel demand
data used to drive the model. The following discussions describe different techniques
that will be used to verify Radian Corporation's motor vehicle source (MV) model.
7.1 Tunnel Studies
Perhaps the best known tunnel study is the 1987 Van Nuys tunnel study
conducted by Southwest Research Institute for the Coordinating Research Council
(CRC) as part of the Southern California Air Quality Study (SCAQS). In this work,
carbon monoxide (CO), volatile organic compounds (VOC) and nitrogen oxide (NOX)
concentrations were measured by bag sampling of the air in the tunnel and at the tunnel
entrance and exit. These measurements were used, along with measurements of air flow
rate into the tunnel and examination of the vehicle flow characteristics, to determine
overall emission rates for the vehicle population traveling through the tunnel. Emission
rates were compared with those predicted by EMFAC7. VOC and CO emissions were
considerably higher than those predicted by EMFAC7, while NOX emission rates
compared favorably. Additional tunnel studies could be conducted to verify emission
estimates generated by the new MV model.
12 Remote Sensing
Remote sensing instrumentation has recently been developed to monitor
CO levels and has been used to identify CO super emitters. This type of equipment will
soon be expanded to remotely monitor VOC emissions from vehicles. These
measurements could be used to verify emission rates for in-service vehicles operating on
links that have been categorized as following a particular driving cycle and then
compared to motor vehicle model emission rates.
Conceptual Design of a New Motor Vehicle Source Model D - 7-1
-------
73 Source Apportionment
Researchers are developing source apportionment techniques to develop
emission estimates. These techniques could be applied in areas where emissions are
dominated by vehicles, and the resulting emission estimates could be compared to those
generated using the MV model. This technique might be especially useful in verifying
vehicle speciation profiles that will be incorporated into the MV model.
7.4 Fuel Sales
Gridded vehicle fuel (e.g., gasoline) sales could be obtained for a particular
region and using average vehicle fuel economy (miles per gallon), vehicle miles traveled
(VMT) estimates could be generated. Average speed estimates could be used to convert
VMT to vehicle hours traveled (VHT) and emission estimates could be computed.
These gridded emission estimates could be overlaid on gridded emission estimates
produced by the MV model, using GIS techniques, and used to check the reasonableness
of model predictions.
7.5 Traffic Count Programs
Traffic count programs could be established to verify the traffic volumes
predicted by the TMS on various links in the driving cycle matrix. Many researchers
consider traffic counters the equivalent of ambient air quality monitors. Such traffic
count programs could also be used to verify the temporal profiles used in the driving
cycle matrix and hourly trip rates used in the trip assignment model of the TMS. In
addition, information regarding weekday/weekend travel pattern variations and seasonal
changes could also be obtained from traffic count programs and used to verify profiles
and adjustment factors included in the MV model.
D - 7
Conceptual Design of a New Motor Vehicle Source Model " " ' "
-------
7.6 Satellite Imagery
Satellite imagery, in the future, might be used to check traffic volumes on
selected links and compared to those generated by the TMS. It might even be possible
to use these techniques to derive estimates of vehicle distributions and check those
against profiles used in the MV model.
Conceptual Design of a New Motor Vehicle Source Model D - 7-3
-------
D-7-4
-------
APPENDIX E
CONCEPTUAL DESIGN OF A NEW HIGHWAY
VEHICLE EMISSIONS ESTIMATION METHODOLOGY
prepared for:
Emissions and Modeling Branch
Air and Energy Engineering Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
July 1991
prepared by:
Bob Dulla
Sierra Research, Inc.
1521 I Street
Sacramento, California 95814
(916) 444-6666
EPA Purchase Order 1D1345NASA
E-i
-------
CONTENTS
Page
1. Introduction E-l
1.1 Understanding the problem E-l
1.2 Project objectives E-3
1.3 Report organization E-3
2. Model Description E-4
2.1 Model logic E-4
2.2 Module descriptions E-7
3. Cost Implications E-15
3.1 Transportation activity E-15
3.2 Emission factor development E-16
3.3 Model integration E-17
4. Validation E-18
4.1 Travel estimates E-18
4.2 Emission factor estimates E-19
Figures:
2-1. Proposed motor vehicle emission methodology E-5
2-2. Illustration of zone/node concept E-8
Tables:
2-1. Baseline travel characteristics E-i0
E-ii
-------
1. INTRODUCTION
1.1 Understanding the Problem
As the stringency of the federal motor vehicle emissions control program
has increased, the relationship between the standards vehicles are
required to meet and the actual performance of vehicles in customer
service is becoming more of an issue. Based on the official
certification test results for cars built during the last decade,
exhaust emissions of hydrocarbons (HC) and carbon monoxide (CO) should
have been reduced by over 90% and oxides of nitrogen (NOx) emissions by
over 75%, evaporative hydrocarbon emissions should have been reduced by
over 90%, and crankcase emissions should have been eliminated. Despite
these findings, numerous communities across the country have failed to
attain the National Ambient Air Quality Standards (NAAQS) for either CO
or ozone. Concerns about the contradiction between laboratory
measurements and on-road performance have been further enhanced by
"tunnel" studies which indicate that current estimates of motor vehicle
emissions produced by EPA and the California Air Resources Board (ARB)
seriously underestimate the levels of HC and CO being emitted under
in-use conditions. There are many reasons for the problems with motor
vehicle emission estimates. Some are well understood and steps are
being taken to resolve them, others are just beginning to be explored.
Lack of proper maintenance in customer service is known to be a major
factor that the motor vehicle inspection/maintenance (I/M) programs are
attempting to address. Previous studies by Sierra Research have shown
that emissions could be reduced by over 50% if all vehicles were
properly maintained. Current programs, however, are achieving limited
benefits. For example, Sierra's evaluation of the California Smog Check
program in 1987 showed that HC and CO reductions were only in the range
of 10%. Based on the use of a computer model developed by Sierra, the
latest enhancements to the California Smog Check program are projected
to yield net emission reductions (for vehicles subject to the program)
of only 20-25% for HC and CO. Assuming that these estimates are not
unrepresentative of programs operating in the rest of the country,
substantial program enhancements are necessary to improve the quality of
maintenance being received by vehicles in customer service.
Inadequate control of evaporative emissions is a problem that has been
recognized more recently and is soon to be addressed through the use of
lower volatility gasoline and revised evaporative emissions test
procedures. Under the evaporative emission test procedures that have
been used for the last twenty years, there has been no requirement to
measure or control emissions that occur during multiple days of
inactivity, during extremely high temperatures, or during periods of
time when the engine is running (running losses). The new evaporative
test procedures will force the redesign of evaporative control systems
to deal more effectively with the full range of conditions that occur in
E-1
-------
customer service; however, cars already in customer service will be
unaffected. Data for such vehicles are needed to accurately estimate
the emissions from the motor vehicle fleet.
A problem similar to that with the evaporative emissions test procedure
also exists with the basic test procedure used to measure exhaust
emissions. Under the official test procedure, exhaust emissions are
measured during stop and go driving using a driving cycle that was
developed about twenty years ago. By its very design, the "LA4" driving
cycle was intended to represent a typical commute trip, by a typical
driver. As a result, the cycle does not reflect the full range of
vehicle operation that is known to occur. One of the most obvious
weaknesses of the cycle is that it contains very modest acceleration
rates. This has enabled vehicle manufacturers to disregard emissions
during higher rates of acceleration. Given the stringency of the
current emission standards, there can be a 100-fold difference between
the level of emissions control that must be achieved during the official
test procedure and the level of emissions that occurs when no emissions
control is required. Another perhaps less obvious problem with the
current test procedure is that it is based on the assumption that only
two "cold starts" occur each day and about three other engine starts can
be represented by the emissions from a vehicle that was thoroughly
warmed up and then shut off for exactly ten minutes. Data currently
being obtained by Sierra under contracts with EPA and ARE indicate that
this is a very bad assumption. Based on surveys being conducted in the
Sacramento area, it appears that there may be substantially more engine
starts per day that previously assumed and many of those starts may be
with an engine that isn't fully warmed-up.
For reasons discussed above, and for similar reasons, there is
substantial uncertainty regarding the current estimates and forecasts of
emissions of motor vehicles in customer service. Further work is
required to:
• better characterize vehicle operation in customer
service;
revise emission estimates to reflect the full range of
vehicle operation observed in customer service;
analyze the feasibility of improving the identification
and correction of maintenance-related problems by using
more sophisticated I/M concepts;
develop estimates of evaporative emissions from vehicles
in customer service using more representative test
procedures; and
revise the current approach used to estimate mobile
source emissions to include the information improvements
outlined above.
E-2
-------
1.2 Project Objectives
The RFP specifies that the goal of the effort is to "develop a highway
vehicle emissions estimation conceptual design that could be implemented
nationally within 5-10 years." The model must be capable of tracking
the emissions produced by all expected vehicles, technologies and fuels.
It must also be capable of estimating the benefits of all potential
emission control strategies (e.g., I/M program design changes and a
variety of transportation control measures). In short, the proposed
design must not only replace the existing methodology used by states and
local communities to develop emission inventory estimates, but also
expand the range of tools available to develop these estimates (EPA does
not currently offer a system to estimate the benefits of transportation
control measures).
Emission inventory estimates require information on vehicle activity as
well as the rate at which pollutants are emitted. Therefore, the
proposed system must address not only the approach used to develop
emission factor estimates, as discussed above, but also the methodology
used to develop vehicle activity estimates. At present, communities are
pretty much left to their own devices when it comes to developing
estimates of local travel levels. Recently, EPA and the U.S. Department
of Transportation (DOT) have prepared guidance on the methods that
should be used in developing local estimates of travel that are
incorporated in emission inventory estimates submitted in response to
the 1990 Clean Air Act Amendments.
It appears, however, that little research is being devoted to developing
new tools that improve the accuracy of local travel patterns. It is
unclear how emission inventory estimates can be improved unless there
are similar improvements in local travel estimates. For this reason,
the proposed methodology must identify areas of inadequacy in the
current models that are used to develop local estimates of vehicle
activity and make recommendations for improvements.
1.3 Report Organization
Immediately following this introduction is a discussion of the proposed
model structure in Section 2. It includes a description of each of the
modules contained in the model and available data sources that can be
used to support their development. Section 3 presents a summary of the
cost of developing this model compared with current procedures. Section
4 discusses methods that could be used to validate the results of the
proposed model.
E-3
-------
2. MODEL DESCRIPTION
2.1 Model Logic
The following principles/assumptions guided the development of the
proposed motor vehicle emissions estimation methodology:
Driving patterns in each urban community are believed to
be unique.
• The current driving cycle is not representative of
driving patterns in nonattainment communities. More
importantly, correction factors cannot be used to adjust
any driving cycle to accurately represent the effects of
local driving patterns on motor vehicle emissions.
• Emissions are not "nearly constant for varied speeds
over time."
Accurate estimates of total urban emissions are
insufficient, they must also be temporally and spatially
resolved.
• Limited changes in the accuracy of transportation
planning models are possible. Nevertheless, it is
believed that estimates of heavy-duty vehicle operation
can and will be added to most models. It is also
expected that the accuracy of zone-specific travel
estimates will be substantially improved.
Uniform goals for accuracy will be established for both
travel and emission factor estimates at the beginning of
the project. A system will be established to track the
accuracy of both estimates throughout the model
development process.
Local transportation surveys will collect travel
information that can be used to aid the development of
motor vehicle emission estimates.
• Improvements in computer technology will continue, resulting in
faster processing power and greater data storage capacity.
An overview diagram of the proposed motor vehicle emission methodology
is contained in Figure 2-1. It shows an approach to estimating motor
vehicle emissions that is fundamentally different from the one currently
employed by either EPA or ARB. First and foremost, it scraps the use of
a single drive cycle with correction factors to account for the effects
E-4
-------
Figure 2-1
Proposed Motor Vehicle Emission Methodology
Transportation
Planning Module
Local Transportation
Survey Data
Motor Vehicle
Emissions By
Grid Cell
Grid System
Specifications
E-5
-------
of local driving conditions on emissions estimates. Instead, it is
proposed that separate emission factor estimates be constructed for each
link on unique trip routes. This can be accomplished by modifying
transportation planning models to supply information on the trip length,
mix of roads (i.e., facility types) and congestion encountered on routes
between each origin and destination zone (OD pair). It is also proposed
that similar information be developed for the travel that occurs within
each traffic analysis zone. As will be discussed later, this
information can be developed either through use of local surveys or the
improvement of transportation models to better account for this category
of travel.
The heart of the proposed methodology is a modal emissions model. It
will combine route-specific travel information, including the length of
time between starts by trip type and hour of the day, with (1) a
distilled version of the data collected in the ongoing EPA and ARE drive
cycle studies on operating conditions (e.g., idle time, speeds and
accelerations, etc.) encountered on similar roads (i.e., grade as well
as road type) under similar congestion levels and (2) engine maps that
characterize emissions as a function of speed and load for a
representative set of vehicles. The output of this calculation is a
specific emission factor for each link encountered on each trip.
The distribution of engine maps used to represent the mix of vehicles
operating within an area is governed by a series of relatively familiar
modules. The profile of technologies employed by model year is governed
by the technology module. It incorporates a summary of the mix of fuel
and emission control technologies manufactured in previous vehicle
type/model year groups along with a projection of the mix of
technologies that will be employed within each vehicle category in
future model years. The activity module specifies the default
distribution of vehicle categories by model year and the engine maps
needed to represent the baseline mechanical distribution of vehicles on
the road. The I/M module quantifies the effect of the local I/M program
on the mechanical condition of the vehicle fleet and in turn, the effect
of the I/M program on vehicle emissions.
The outputs of the transportation planning module and local trip surveys
are input to the transportation control measure (TCM) module. That
module evaluates the effects of alternative TCMs on the travel-related
determinants of vehicle emissions: the relationship between vehicle
trips and person trips, trip length and congestion levels. It provides
a feedback loop to the transportation planning model to reevaluate the
level of travel produced when measures substantially alter baseline
travel behavior.
The transportation planning module provides two separate outputs. The
first is information on the trip type (which in fact may come from a
travel survey), trip length, mix of roads (i.e., facility types) and
congestion encountered on routes between each origin and destination
zone and within each traffic analysis zone (TAZ). The first set of
outputs is input to the modal emissions model. The second set of
outputs is the actual number of vehicle trips between zones and within
zones; this is input into the grid cell emissions module.
E-6
-------
The grid cell emissions module integrates data on the characteristics of
the road network and TAZs with link-specific emission factor estimates
for each OD pair and the number of trips between each OD pair. A
similar computation is used to estimate the emissions within each TAZ.
The grid cell emission module aggregates the link and TAZ emission
estimates into the grid cells specified by the user.
A complex set of controls is required to operate the proposed modeling
system. Not shown, but implicit in the system, is a set of
preprocessing steps that are needed to translate local information
(e.g., travel survey, node coordinates, etc.) into a format that is
compatible with the proposed modeling structure. A more substantial
effort is required to organize the information collected in the ongoing
chase car/instrumented vehicle surveys into a range of common
facility/congestion and grade specific travel profiles.
The decision to use this approach was largely motivated by the
requirement to accurately resolve emissions estimates, both temporally
and spatially. Motor vehicle emissions are disproportionately high
during cold start operation and under high acceleration rates. This
means that emissions are high during the time before catalyst light-off
(and to a lessor extent, before the engine fully warms up) and on
certain types of roads where high levels of acceleration occur.
Extremely high levels of emissions occur when hard acceleration rates
are experienced before the catalyst lights-off. The emissions from
these events cannot be spatially resolved with any appreciable accuracy
using average values.
2.2 Module Descriptions
Transportation Planning Module - This module requires several categories
of outputs from transportation planning models. Figure 2-2 provides an
illustration of the zone/node structure employed in transportation
planning models. It is presented to help the reader visualize the
different categories of information discussed below. The first category
covers interzonal travel and includes the trip generation matrix (i.e.,
number of trips between origin and destination zones) for peak periods
of operation; the distribution of trip types (e.g., home-based work,
home-based other, etc.) for each OD pair by period of operation; and the
specific links included in a representative trip path between OD pairs.
The network description must include a specification not only of
distance between node points, but the grade as well. With the above
information, it is possible to compute the trip length and the fraction
of the trip that occurs on each link and facility type. As discussed
earlier, it is assumed that most transportation planning models will be
upgraded, within the time horizon envisioned for this effort, to include
both commercial vehicle operation and through traffic.
E-7
-------
Figure 2-2
Illustration of Zone/Node Concept Used
In Transportation Demand Modeling
Roadway Links
Facility Type Key
1 • Local/Collector
2 - Arterial
3 - Highway
4 - Freeway
Area Type Key
- Urban/CBO
I | - Suburban/Rural
Zone Boundaries
Zone Centroid
The second category of information addresses intrazonal travel. At
present, this is the number of trips estimated for equal OD pair zones
(e.g., 1-1, 2-2, etc.) in the trip generation matrix. Conversations
with transportation planners, university researchers and consultants in
the field all agree, however, that these estimates are quite weak. No
information on the intrazonal road structure is included in the road
network, and trip lengths within the zone are frequently approximated on
the basis of the diameter of the zone. These simplifying assumptions
are used despite the documented shift in shopping trips and recreational
travel toward local/regional shopping and recreation centers. As
discussed later, one of the assumptions driving the development of this
approach is that money will become available from several different
sources to improve the state of the art in estimating the level of
travel within zones. It is also assumed that because .of this research,
travel estimates on a par with those developed for interzonal travel
will become available for use in the development of motor vehicle
emission inventories.
Several of the outputs specified above are not routinely produced by
existing models. For example, it is possible to prepare code that
E-8
-------
tracks the links in assigned trips between OD pairs; however, it is not
a routine product. Similarly, information on grade is not routinely
coded for the road network. It should also be obvious that most of the
suggested improvements noted above do not currently exist.
Transportation Surveys - Surveys are infrequently conducted by local
transportation planning agencies to quantify the travel behavior of a
representative sample of the population. The data collected in the
surveys are used to specify many of the travel profiles that
characterize behavior in the transportation planning models. These
profiles (e.g., average commute trip length, average vehicle occupancy,
etc.) are also needed to support the evaluation of TCMs.
The travel surveys can be expanded to collect information that can be
used to either support the development of the intrazonal travel
estimates discussed above or to provide the basis for developing
exogenous estimates of the distribution of trips and trip lengths
occurring within each of the TAZs. The surveys can also be used to
estimate the fraction of hot and cold starts that occur within local
communities. This can be accomplished through the use of home
interviews or diaries that track the time of each trip in a typical day.
By analyzing the time between starts and stops, it is possible to
determine the condition of the engine and catalyst at the start of each
trip. The alternative to this information is the use of defaults that
will be developed from the planned instrumented vehicle surveys.
It should also be noted that the transportation control measure module
will require a substantial amount of information about the cost of local
travel (e.g., cost per gallon of fuel, average out-of-pocket commute
trip expense, etc.). Information on the mix of vehicles operating on
typical roadways will be required by both the transportation planning
models (to evaluate the benefits of goods movement) as well as the modal
emissions model to correctly specify the mix of vehicles operating on
facility types (heavy-duty trucks are much more prevalent on freeways
than collectors). This information will not be supplied by either the
transportation planning model or the travel surveys discussed above.
This information is developed from a variety of sources and related but
separate surveys.
Transportation Control Measures - A wide variety of TCMs are currently
being evaluated as nonattainment areas search for alternative control
strategies to reach attainment. Table 2-1 provides a summary of the
travel characteristics needed by a TCM analysis model developed for use
by air pollution control districts in California. That model, developed
by Sierra and its subcontractor JHK, quantifies the effects of 25
separate TCMs on the parameters that current emission factor models use
to determine emissions (i.e., trips, VMT and speed). The proposed
methodology will not require information on speed, but will require
information on trips and trip length. Therefore, it is assumed that the
basic travel parameters outlined in Table 2-1 should be representative
of the information that will be needed to support the quantification of
TCM benefits.
E-9
-------
Table 2-1
** *** a***********************************************************************
BASEUNE TRAVEL CHARACTERISTICS
General
Total person trips 1
Total commute person trips .1
Total commute vehicle trips 1
Total non-commute vehicle trips 1
Total peak period VMT 1
Total off-peak period VMT 1
Drive-alone share of commute person-trips 79.7%
Percent of all trips in peak period 39.6%
Percent of all trips that are commute trips 33.8%
Percent of all trips that are non-commute trips 66.2%
Percent of commute trips in peak period 60.8%
Percent of non - commute trips in peak period 28.8%
Percent of peak trips that are commute trips 51.9%
Percent of off-peak trips that are commute 21.9%
trips
Average commute trip length 10.7
Average non-commute trip length 6.0
TCM Specific
Demand Management
Average daily commute out-of-pocket $5.00
costs per vehicle
Average number of telecommuters per day 0.0
Transit Improvements:
Percent of all trips that are transit 1.0%
Commute trip share of transit 5ZO%
Total transit vehicle miles 0
Bicycle:
Percent of commute trips less than 6 miles 25.0%
Percent of non -commute trips less than 5 25.0%
miles
Pricing:
Average cost of gas per gallon $1.44
Average cost per mile to drive $0.300
Average commute out-of-pocket costs per $3.00
vehicle per trip
Average non-commute out-of-pocket costs per $1.00
vehicle per trip
Freeway Management/Goods Movement
Percent of VMT on freeways 55.8%
Average trip length for trucks 10.7
E-10
-------
Current TCM models compute the effect of individual TCMs on aggregate
levels of travel. The model structure outlined in Figure 2-1 provides
for a feedback loop between the TCM module and the transportation
planning model. This feedback can be used to evaluate the effects of
significant changes in trip generation rates (as will be discussed in
Section 4. on the order of 5 percent or more). This approach also
allows for a more sophisticated appraisal of the effects of alternative
TCMs (e.g., land use controls which shorten future trip lengths) on
future travel patterns.
Facility Type Travel Profiles - Ongoing EPA and ARB drive cycle studies
are collecting information on vehicle operating conditions (e.g., idle
time, grade, speed and acceleration, etc.) using two primary data
collection methods: chase cars which follow traffic over representative
road routes, and instrumented vehicles which record operating data from
a representative set of vehicles. These studies will be conducted in a
minimum of four communities across the country. The data collected will
be used to develop a "representative" driving cycle by EPA that could be
used to certify new vehicles. ARB plans to develop three separate
driving cycles that are representative of the travel that occurs during
each of the peak operating conditions of the day (i.e., a.m. peak, p.m.
peak and off-peak). These cycles will be used to improve the accuracy
of existing emission inventory estimates.
The data collected in the chase car studies will provide the basis to
quantify typical driving profiles that occur on different categories of
roads (i.e., facility types) as a function of congestion (time of day)
and grade. Similarly, the data collected in the instrumented vehicle
study will provide the information needed to quantify the distribution
of short trips (i.e., less than a tenth of a mile) that frequently occur
off public roads in shopping centers and driveways as well as the soak
time between trips by trip type. The results of these analyses can be
organized into data files that quantify the driving profiles of vehicles
over all possible roads within urban nonattainment areas.
Modal Emissions - Both ARB and EPA are supporting the development of a
modal emission model entitled "Vehicle Simulation-Emissions" (VEHSIME).
It determines the power demands required for a specific engine/vehicle
combination to achieve instantaneous speed and acceleration rates on a
second-by-second basis for any specified time-speed trace. Based on the
engine-transmission combination, second-by-second power values are
translated to engine speed load values. Fuel consumption and emissions
are interpolated from a steady-state map of emissions and fuel economy
for the full range of the engine.
As shown in Figure 2-1, the proposed modal emissions model requires
three primary inputs: travel profile of the trip of interest; the mix of
possible time/speed profiles that can be encountered on links within the
route of interest; and a set of engine maps that are representative of
vehicles operating within the community of interest. A critical input
also required is the average soak time between trips.
The travel profile must specify the length and grade of each node that
makes up a trip route. The facility type travel profiles must specify
the time speed trace on a second-by-second basis for individual roads as
E-11
-------
a function of congestion and grade. The engine maps require
measurements of fuel consumption and emissions for specific engine speed
and load points.
Even after the current work of adding a cold start algorithm and
increasing the number of available engine maps is completed, a
substantial amount of work to be required to allow the model to operate
in the manner envisioned in this module. Software would be required to
integrate trip requirements with the available facility profiles and the
soak time. Multiple runs of the model would be required to integrate
the effects of alternative engine maps representing different model year
groups and technologies. Software would also need to be developed to
quantify the link specific emission factors.
Technology Profile - The technology profile specifies the mix of fuel
types (gasoline, Diesel, etc.), fuel delivery systems (e.g., multipoint
versus throttle-body fuel injection) and emission control technologies
(e.g., three-way closed loop versus oxidation catalysts) manufactured in
previous vehicle type/model year groups. It also provides a forecast of
the mix of technologies that will make up future vehicle category/model
year groups. Current emission factor models track up to five separate
technology categories for the 25 model years that make up the vehicle
fleet in a single calendar year. As discussed below in the I/M module,
each technology group has multiple emission regimes. The disaggregation
of the vehicle fleet into more than 250 subcategories of emission
performance has significant implications for the information
requirements of the engine map module.
Engine Maps - Engine maps consist of formatted data files that contain
emissions and fuel consumption rates over a range of engine speeds and
loads. At each speed/load point, supplemental information regarding
throttle angle (in degrees) and manifold vacuum values (in inches Hg)
may also be included. At present, VEHSIME can only accept engine maps
with up to 20 different speeds and 20 different loads at each speed. It
is possible, however, to expand the size of the matrix to cover a
greater number of points over the range of potential engine operation.
I/M Program Benefits - I/M programs are designed to reduce the level of
pollutants emitted by vehicles operating under in-use conditions.
Differences in program designs can have a significant impact on their
effectiveness in reducing emissions. For example, the date at which an
I/M program is first implemented, or the date when significant program
changes take effect, has implications for the benefits of the program.
EPA and ARB I/M models currently segregate the vehicle population into
different emission regimes (e.g., normals, moderates, highs, supers,
etc.). The California I/M model CALIMFAC simulates deterioration by
modeling the migration of vehicles from one regime to another with
increasing mileage. Emission factors for specific technology/model year
groups are developed from test data representing the mix of emission
regimes. The benefits of I/M are quantified by altering the baseline
mix of emission regimes.
The proposed methodology would require the development of engine maps of
fuel consumption and emissions for a mix of vehicles with implanted
defects designed to simulate the range of possible emission regimes for
E-12
-------
each model year and technology group subject to an I/M program.
Baseline fleet average emissions would be based on a mix of engine maps
representing not only the appropriate model year technology groups which
make up the vehicle fleet, but also the range of defects found in that
fleet. The benefits of I/M would be quantified by altering the mix of
defects found in that fleet.
Activity Profile - The activity profile specifies the default
distribution of vehicle categories and model year registrations within
each vehicle category. This information must be combined with the
distribution of technologies that make up each model year/vehicle
category to specify the mix of engine maps needed to represent the in-
use vehicle fleet. Information on the registration status of vehicles
is also tracked to support the quantification of I/M program benefits
(unregistered vehicles are not likely to be inspected).
The level of detail included in this profile could be expanded to
provide default estimates of the mix of vehicles operating on different
classes of roads at different times of day. Current inventories assume
that the mix of registered vehicles is uniformly operated over the
entire road network. Data collected in ARB's San Joaquin Valley Air
Quality Study have shown that there are substantial variations in the
mix of vehicles operated on individual road categories.
Emission Factors - A specific emission rate for each link in a specified
trip route is required. It is unclear how the effects of ambient
temperature would be accounted for in the proposed modeling structure.
One method would be to adjust the cold start profile used in the modal
emissions model to account for the effects of alternative ambient
temperatures on catalyst and engine warm-up. Another approach would be
to adjust the emission estimates produced by the modal model. If the
latter approach were selected, the emission factor module would be the
appropriate location for the correction factor to be applied.
Grid System Specifications - This set of inputs defines the boundaries
of cells within the modeling domain. Zonal (TAZ) boundaries must also
be identified so that intrazonal emission estimates can be properly
allocated to grid cells. It is expected that a standardized coordinate
system would be used to define grid cell and zone boundaries. A CIS
(Geographic Information System) could be used to map irregularly shaped
TAZs into rectangular grid cells.
Grid Cell Emissions - Trip route and link-specific emission factors are
combined with trip volume estimates from the TCM module to estimate
link-specific pollutant emissions for each trip route for OD pairs
(interzonal travel). A similar methodology would be used to compute
intrazonal emission estimates. Route emissions can be summed to produce
link-specific emission estimates. These emissions can then be either
directly allocated to grid cells or apportioned on the basis of link
length. The approach used to allocate zone-specific emissions to grid
cells will depend on the level of intrazonal network definition that is
available. Emissions within the zone, however, would be computed in a
manner similar to that employed for interzonal travel.
E-13
-------
Program Controls - A complex set of controls would be required to
operate the proposed modeling system. Not shown, but implicit in the
system, is a set of preprocessing steps that are needed to translate
local information (e.g., travel survey, node coordinates, etc.) into a
format that is compatible with the proposed modeling structure. The
actual operation of the system would have to broken down into a series
of steps that match the logical flow of information required to
characterize local conditions (e.g., I/M program specifications, local
registration data, travel parameters) and minimize computer system
requirements.
E-14
-------
3. COST IMPLICATIONS
This section presents a qualitative discussion of the incremental costs
that could be incurred in developing the proposed modeling system. The
critical elements to be considered when evaluating the costs of the
proposed modeling system are as follows:
• Trends in existing and planned research that will
generate the information needed to support the
development of the proposed model. What information is
already being developed, what information is not?
New information requirements (e.g., testing programs)
that will need to be produced on an annual basis that
are not currently produced and therefore not covered in
existing budgets.
• Staff requirements for developing and then maintaining
the proposed modeling system. Can it be assumed that
current staff used to maintain and upgrade existing
emission factor models will be used in the development
of the new modeling system?
• Any new system will require a substantial effort to
educate local users. As the information requirements of
the proposed system increase, so do the support costs
for local users.
The modeling system recommended for development in this report is quite
sophisticated and therefore will require substantial amounts of new
information as well as a substantial development effort. In order to
simplify the discussion of cost implications, the proposed modeling
system is divided into three aggregate functional categories:
transportation activity; emission factor development; and model
integration. A brief review of the cost implications associated with
each of these areas is presented below.
3.1 Transportation Activity
At the outset, it should be stated that this category has the greatest
level of cost uncertainty. The primary reason is that the source of the
money for transportation system model improvements is unclear. At
present, the Department of Transportation has eliminated support for the
Urban Transportation Planning System (UTPS) model. A new class of
transportation planning models (e.g., MinUTP, System II, etc.) that can
be operated on a wider range of computers has been developed by several
consulting firms. These models are widely used by transportation
E-15
-------
planning agencies in communities across the country. Unfortunately, none
of these models appears to substantially improve the state-of-the-art in
terms of the accuracy of intrazonal travel estimates.
Up until recently, there has been no specific demand for detailed
estimates of this category of travel. Demand for this data, however, is
growing as travel within local areas expands and communities struggle
with the internal allocation of maintenance and road construction funds.
Funds will be required to develop methods to quantify the travel
occurring within TAZs. Additional funds will be required to modify or
replace the existing models to incorporate these new methodologies.
Funds will also be required to expand the network to include more
information about local roads. While there are many potential sources
of funds for this work, the key is that this is a new and at present
unfunded area of effort.
Effort will also be required to develop the software that takes data
from transportation planning models and organizes it into a consistent
format that can be used in the proposed modeling system. It is unclear
as to whether this effort can be satisfied with existing EPA staff or if
it falls into a category of new expenditures.
The high cost of travel surveys limits the frequency at which they can
be conducted. Many areas of the country conduct them on a 10-year cycle
that may or may not: be related to the Census. Some areas conduct them
on 15-year cycles. The incremental cost of adding specific questions to
collect data on soak times should not be particularly onerous. The
level of effort required to improve the accuracy of intrazonal travel
estimates would be more substantial. The key is where communities are
in their current survey cycles. It is assumed that most agencies are
already collecting the information that would be required by the TCM
analysis module.
No TCM analysis module currently exists at the federal level. It
appears that EPA does not plan to develop a module that can be
integrated with the MOBILE series of emission factor models. Instead, a
series of guidance documents is being prepared to help communities
evaluate the benefits of alternative measures. As discussed earlier,
California currently has an integrated TCM evaluation model. It,
however, employs the emission factors produced in EMFAC7E. Because of
the limited state of knowledge about the ability of alternative measures
to influence trip behavior, the state-of-the-art should be described as
in its infancy. A substantial effort will be required to track the
state of knowledge as it evolves from the experiences of communities
that experiment with alternative measures. It appears, however, that
EPA is staffing up to track this information. Because it appears that
EPA staff will be tracking this area of information, the costs of
developing a TCM analysis module should not be particularly significant.
3.2 Emission Factor Development
This category of data and model development is expected to be the most
expensive of the proposed modeling system. That is because (1) it is
expected that EPA will continue to use either the FTP or some
E-16
-------
modification of it (e.g., the addition of a bag 4 to represent high load
conditions) to certify new vehicles and track their in-use performance
and (2) a substantial testing effort will be required to characterize
the engine maps of the on-road vehicle fleet. The collection of engine
maps will thus not replace existing FTP-related activities, it will be
in addition to them. This means that new staff/contractor support would
be required to develop this information.
A substantial analytical effort will be required to determine the mix of
engine maps needed to characterize new vehicle emissions, mileage
deterioration and malperformance effects. A great deal of effort would
also be required to translate existing I/M program modeling efforts from
an FTP-based emission factor methodology to one that uses modal
emissions. The translation of the activity and technology modules into
the proposed modeling system should be trivial.
As discussed earlier, a substantial programming effort will be required
to modify VEHSIME to quantify the emissions for a representative sample
of engine maps for each link in unique routes. There is room for a
great deal of creative thinking in how to minimize the programming
requirements of the proposed system.
Finally, it should be pointed out that a large analytical effort will be
required to analyze the data collected in the EPA and ARE chase car and
instrumented vehicle surveys. This analysis will be required to produce
the link-specific driving profiles that are a function of grade and
congestion level. The above programming tasks can clearly be satisfied
by existing EPA staff members. The cost of the effort will depend on
whether the appropriate staff will be assigned the responsibility for
the development of this software.
3.3 Model Integration
All of the remaining programming tasks (i.e., grid cell specifications
and emissions, and overall system programming) will require a
significant effort in order to make the modeling system available to a
wide range of users. Again, it is unclear who will have the
responsibility for the development of the overall modeling system. It
appears, however, that this capability is resident within EPA.
E-17
-------
4. VALIDATION
The proposed modeling methodology is based on the assumption that the
accuracy of the travel estimates from the transportation planning models
will be roughly equivalent to the accuracy of the emission factors
produced by the modal emissions model. From the perspective of the
emissions inventory, there is no reason for one estimate to have any
more accuracy than the other. Because it is quite difficult, if not
impossible, to use ambient air quality measurements to validate the
results of the proposed modeling system (tunnel studies cannot be used
to validate urban airsheds), it is proposed that effort be devoted to
validate the results of the two principle components of the system:
travel estimates and emission factor estimates. A brief description of
the process that could be used to validate each component is presented
below.
4.1 Travel Estimates
It is very difficult to validate a variable that cannot be measured.
The problem of validating VMT estimates is not new. Compounding the
problem is the fact that transportation planning models are not designed
to predict VMT, they are designed to predict volume. A recent AWMA
paper entitled "Travel Demand Forecasting Limitations For Evaluating
TCM's" suggests that
"travel demand forecasting models are calibrated and expected to be
accurate enough so that the model's inaccuracies in the volume
forecast will not affect the number of lanes when designing a
highway project."
The paper goes on to indicate that typical error standards for
regionwide ground counts should be less than 5 percent. The suggested
error limits for functional classifications, however, range from less
than 7 percent for freeways to less than 25 percent for collectors and
frontage roads. The paper suggests that similar levels of error should
be seen in VMT estimates. It cautions, however, that these estimates
only cover the roads included in the network. The error for roads not
included in the network but nevertheless included in an estimate of
overall travel levels can be substantially higher.
The implication of this paper is that it is unreasonable to expect
transportation planning models to evaluate the benefits of TCMs when
the measure is expected to have an impact on travel that is lower than
the accepted accuracy of the modeling system (i.e., 5 percent). This is
an important perspective to maintain when evaluating proposed emission
modeling systems. It is unclear what effect the suggested improvements
in intrazonal travel will have on the overall accuracy of the
transportation planning models themselves.
E-18
-------
Sierra is aware of many approaches that have been used to independently
validate local, state and national travel forecasts. These have
included the use of fuel consumption estimates to back out travel
activity, because fuel consumption is based on tax receipts. Another
approach has extrapolated diary-based odometer readings for specific
vehicle classes to represent travel for the entire vehicle population.
Another approach relied on the use of annual changes in odometer
readings from vehicles participating in I/M programs to represent travel
as a function of vehicle age. All of these methods have had problems
accounting for either fuel or travel that did not occur within the study
domain. They have also had problems accounting for vehicles traveling
through the study domain that are not locally registered. Therefore,
the first step in the validation process of these models should be a
thorough review of the literature.
The next step should be the collection of an extensive number of ground
counts for an urban area that has a state-of-the-art transportation
planning model with a 1990 or later base year travel estimate. The
travel estimate should be based on a recently completed travel survey.
The model can then be updated to reflect the road network in place at
the time the ground counts were collected. The model can be validated
with the normal level of ground count information. The results can then
be compared with estimates that are based on the more extensive ground
count survey to determine the level of accuracy the model has in
representing travel throughout the urban area. This process can be used
to develop an estimate of the accuracy of the model before and after the
intrazonal travel estimates are improved.
4.2 Emission Factor Estimates
Of the two components, the emission factor estimates are the easiest to
validate. The use of a modal emissions model allows estimates of
emissions for any particular drive cycle to be validated through
dynamometer-based measurements of emissions on the same cycle. There
are, however, many reasons to expect that dynamometer-based emission
estimates for a particular vehicle will not equal the values predicted
by the modal emissions model. These can include differences between the
selected vehicle and the technology represented by the emission map, the
condition of the vehicle, and the accuracy of the selected start
algorithm.
Both EPA and ARB are well aware of the fact that the modal emissions
model cannot account for all elements of vehicle-specific emission
control system performance. For this reason, EPA is planning to use
VEHSIME as a screening tool to provide a rough indication of the
emissions performance that a representative sample of vehicles would
produce under alternative driving conditions (time/speed traces). On
the basis of those preliminary indications, EPA will then test a
representative sample of vehicles on dynamometers on the cycles that are
determined to be of interest. The information developed in this process
will provide the basis for a careful appraisal of the accuracy of the
VEHSIME model.
E-19
-------
ARB has plans to integrate a modal emissions model into its emission
factor system (EMFAC/BURDEN). They also plan to use VEHSIME to evaluate
the information developed in the ongoing chase car study. As stated
earlier, the goal of that effort is to develop representative driving
cycles for three peak periods of operation. One element of that study
is a comparison between modal-based emissions estimates for those cycles
and dynamometer measurements of vehicles operating on the same cycles.
The information developed in both the EPA and ARE efforts should provide
the basis to evaluate the accuracy of the proposed modal emissions
model. It is also expected that if these efforts discover significant
discrepancies between the two measurements, further effort will be
devoted to determine the source of the error and develop a methodology
to correct it.
E-20
-------
APPENDIX F
Technical Report
CONCEPTUAL DESIGN FOR A NEW
HIGHWAY VEHICLE EMISSION
ESTIMATION METHODOLOGY
SYSAPP-91/083
June 18, 1991
Prepared for
Carl T. Ripberger
U.S. Environmental Protection Agency
Air and Energy Engineering Research Laboratory
MD-62
Research Triangle Park, NC 27713
Prepared by
R. G. Ireson
Systems Applications International
101 Lucas Valley Road
San Rafael, California 94903
and
J. P. Nordin
West Union, Iowa 52175
EPA Purchase Order 1D1519NASA
F-i
-------
CONTENTS
Page
1 Conceptual Design for a New Highway Vehicle Emission Estimation Methodology F-1
2 A Brief Overview of Recognized Limitations in Current Procedures F-3
3 Rationale and Overview of Proposed Approach F-5
4 Model Concept Description F-11
5 Verification of Inputs, Relationships, and Results F-18
6 Research Issues F-21
7 Summary F-23
References F-24
Appendix A: Collection of Detailed Vehicles In-use Information F - A-l
Appendix B: Small-scale Traffic Simulations F-B-1
Appendix C: The Urban Transportation Planning System F-C-1
Figures:
1. Speed distributions over time along a freeway segment F-7
2. Tune-distance diagram showing operating mode trajectories for three vehicles F-8
3. Time-distance diagram for a vehicle approaching a stop-sign-controlled
intersection F-9
4. Base case emission estimation F-12
5. Development of relationships for synthesizing and forecasting vehicle
operating mode distributions F-14
6. Development of future year emission forecasts F-16
F-ii
-------
1 CONCEPTUAL DESIGN FOR A NEW HIGHWAY VEHICLE
EMISSION ESTIMATION METHODOLOGY
BACKGROUND AND OBJECTIVES
Procedures that are currently used to estimate on-road motor vehicle emissions exhibit
several significant limitations. These limitations compromise the accuracy of base year
estimates as well as future year forecasts. Because the potential errors in emission estimates
affect the predicted relative impact of mobile source emissions on ambient pollutant
concentrations (especially for ozone), improving the accuracy of these estimates is a critical
need. In addition, because of a reliance on a number of broad assumptions, existing
emission estimation techniques can produce inaccurate predictions of the effects of specific
controls, such as transportation control measures (TCMs) and alternative fuels. Many of the
limitations of existing procedures arise from data, analyses, and models that are deeply
ingrained in the mobile source emission estimation procedures; eliminating these problems
requires a willingness to rethink each component of the process.
This report develops a conceptual design for a next generation emission modeling system for
highway vehicles. The underlying objective for this specific conceptual design is to provide
adequate support for air quality planning: source attribution; forecasting; control strategy
evaluation; and verification of control strategy effectiveness. Meeting this objective requires:
Accurate base year emission inventories
Emission inventories that are resolved to levels that are appropriate for air quality
analyses (i.e., to at least a few kilometers spatially, and one hour temporally)
Emission inventory forecasts that are accurate (within the ability to predict growth)
relative to base year emissions
Control scenario forecasts that are accurate relative to the future year base case
forecast.
Failure to achieve any one of these can lead to a systematic bias in the attribution of ambient
concentrations to sources, and (depending on the selected controls) strategies that are
unnecessarily costly, ineffective or even counterproductive.
91078-2 F-1
-------
In developing the conceptual design presented in this report, we reviewed recognized
problems in current procedures, and possible approaches for addressing each. Subsequent
sections of this report present:
An overview of recognized limitations of current approaches
The rationale for, and overview of our proposed approach
A description of the conceptual design of the emission model
Methods and issues in verification of inputs, relationships, and results
Anticipated research issues in model development and application
A summary of our recommendations
Three appendixes are included with this report:
A discussion of alternatives and capabilities for collecting information regarding in-
use vehicles
A review of small-scale traffic simulation modeling alternatives and capabilities
A review of the methods and limitations of urban transportation planning models as
related to the modeling of mobile source emissions.
91078-2
F-2
-------
2 A BRIEF OVERVIEW OF RECOGNIZED LIMITATIONS
IN CURRENT PROCEDURES
The heart of the emission estimation process is the identification of an appropriate "vehicle
term" and "emission rate term" for use in the equation
Emissions = (Activity Level) x (Emission Factor).
For a single vehicle, there are several different causes of emissions, and therefore a number
of different possible emission factor/activity level pairs that could be used. For example,
hot-soak evaporative emissions can be estimated by multiplying an estimate of the number of
grams of emissions per hot-soak by the number of trips that the vehicle makes. Tailpipe
emissions can be estimated based on gram/mile, gram/minute, or gram/gallon of fuel,
multiplied by the corresponding activity level for the vehicle. Similar options exist for
running losses, excess cold-start and hot-start emissions, and breathing losses.
Both MOBILE4 and EMFAC7 (the U.S. EPA and California Air Resources Board emission
factor models) treat these different emission sources, with EMFAC7 providing separate
outputs for "trip-end" (hot-soak, cold-starts and hot-starts). However, emission factors from
both models are aggregated to some extent (e.g., assuming constant speed for all vehicles).
The activity level parameters that are currently used to estimate emissions are derived
primarily from data and models used in transportation planning. In their simplest form,
these parameters are vehicle-miles-travelled (VMT) estimates for entire urban areas; more
sophisticated emission estimates have been generated using transportation model outputs for
traffic volumes and speeds on individual road segments, or "links." Like the emission
factors, however, such estimates are still aggregated to some extent, such as by treating
"average" peak hour or weekday average traffic volumes, and by estimating a single speed
for each link.
Historically, the development of a fleet-average composite emission factor was necessary to
develop emission inventories from limited available data. Recent studies suggest, however,
that a much more explicit treatment of individual vehicles and operating conditions is now
needed to accurately estimate emission rates. Several specific factors should be noted:
Current emission control technologies such as closed-loop catalyst systems perform
very well under moderate driving conditions (mild acceleration and deceleration, idle,
cruise) but less well under hard acceleration.
91078-2 F-3
-------
Such control technologies appear to be optimized for the federal test procedure (FTP)
driving cycle used in vehicle emission certification programs. Thus, departures from
this cycle result in excess emissions.
A relatively small number of vehicles have been found in field studies to be
contributing the majority of total emissions.
At the present time, then, the "average" composite emissions for the fleet appear to be made
up of a relatively large, cumulative VMT for vehicles that are quite "clean," and a relatively
small VMT for vehicles that are "dirty" either because of their condition or their current
operating mode. Aside from questions regarding the actual accuracy of the estimates of
"average" vehicle emissions, "average" link volumes, and "average" link speeds, we believe
that a major problem with existing techniques is their failure to explicitly address those
vehicles and operating modes that contribute most to total emissions. In particular, we
believe that the proportion of atypically high emitters in different cities depends strongly on
local factors. For example, the fraction of miles accumulated under heavy acceleration can
be influenced by hills and highway geometry.
In summary, we believe that improvement in the accuracy of emission estimates hinges on
disaggregation of both the "vehicle term" and the "emission rate term" in ways that allow
explicit treatment of those conditions leading to high per-vehicle emission rates. This
approach, of necessity, requires substantially more detailed information as input to the
emission calculation, and the conceptual design of the proposed methodology is based on the
objective of treating individual vehicles as explicitly as possible.
91078-2
F-4
-------
3 RATIONALE AND OVERVIEW OF PROPOSED APPROACH
The basis for recommending the proposed methodology is: (1) that current practices cannot
reliably treat the range of important local factors that affect vehicle use; (2) that current
practices cannot accurately describe emission rates for vehicles as used; and (3) the principal
cause of these limitations is the reliance on overly aggregated representations of both vehicle
operation and emissions. The primary approach for rectifying these problems in the
proposed methodology is substantially increased reliance on empirical inputs, especially for
vehicle operations.
Need for a Detailed Empirical Basis for Vehicle Operations
In order for such a method to succeed, comprehensive data on individual vehicle operations
within an urban area are needed. We believe that a combination of techniques can meet this
need. Principal reliance would be placed on the use of time-lapse aerial photography and
computer image processing and analysis to determine, by roadway or subregion, such
parameters as:
Number of vehicles in operation
Distribution of vehicle speeds by road type
Distribution of acceleration by speed and road type
Number of vehicles in non-freeflow conditions (queued and creeping, idling)
Number of trips started and ended.
Aerial photography and video data analysis for traffic flow have been recently explored in a
number of contexts by Makigami et al. (1985), Becker (1989), Persaud and Hurdle (1988),
and Taylor et al. (1989). Related studies include the vehicle classification and identification
schemes reported by Pfannerstill (1989), Pursula and Kosonen (1989), and Williams et al.
(1989).
Calibration and verification of this technique can be accomplished through the use of traffic
counters and instrumented vehicles that can be identified in the photographs. The
development of large-scale data bases of vehicle operations under a variety of conditions
would provide a robust basis for operating mode distributions.
91078-2
-------
Value of Empirical Distributions of Vehicle Operating Modes
By basing the distribution of vehicle operating modes on data for the study area, the potential
bias introduced by nonrepresentativeness of the generic FTP driving cycle can be prevented.
Similarly, by using actual observed speed distributions on links, the arbitrary assignment of a
single average speed to each link is avoided, as is the reliance on overly simplistic speed
versus volume-to-capacity ratio relationships such as the BPR curve and the simplistic
treatment of off-network travel (centroid connectors) as having a constant speed. Instead, the
actual effects of local factors and conditions on vehicle operating modes are addressed. Such
factors include lane width, road surface condition, distractions (road signs, commercial
driveways and parking, on street parking, etc.), hills, and the level of aggressiveness of
drivers. Such conditions are addressed in transportation models by attempting to calibrate
the model's predictions to a limited set of observations.
At the urban modeling scale, much of the natural variability in vehicle operating modes is
ignored on the assumption that driving cycle approximations provide a reasonable basis for
describing operating modes. This simplification is not necessary. Empirical and
mathematical descriptions can be generated to show these distributions. Figure 1, for
example, shows speed distributions by time and location for a six-mile segment of a Los
Angeles freeway. Matzoros (1990) has explored vehicle operating modes at intersections
(Figures 2 and 3) for incorporation into a simple emission and air quality dispersion model
based on modal emission factors and vehicle dynamics. The time-distance diagrams shown
provide a potentially useful way of classifying vehicle operating modes from aerial
photography under conditions where the transition deceleration to queue to acceleration may
be ambiguous.
Modal Emission Rate Modeling
By replacing a nominal driving cycle with vehicle operating mode distributions, we eliminate
the option of using driving-cycle-based composite emission factors. Instead, factors specific
to vehicle operating modes are needed. Some work has been carried out in this area, and
more will be needed in order to determine how to unambiguously establish the operating
mode of a vehicle, and how emissions vary by mode. In this report, we suggest that modal
emission factor models be developed whose primary input is a vehicle state vector consisting
of basic mode (e.g., cruise or creep), speed, acceleration, and other relevant factors such as
engine state (time since start) and canister loading.
Forecasting-Transportation Modeling Outputs as Parameters of Vehicle Operating
Mode Distributions
To this point, the suggested approach is appropriate for base year estimates, but is difficult to
use for forecasting due to the reliance on empirical vehicle operating distributions. Growth
and congestion effects, as well as the effects of capital improvements to the roadway network
91078-2
-------
Speed Contour Plots
Southbound Hwy 101, Los Angeles
Morning
Data Dated: 05/13/85-Upper Plot 05/14/85 - Lower Plot
MM
MM
Ifl
MM
2 2JO POSt
PM5.7 PU44 PUX2 PM24 "*
Santa Monica . Vermont Rampart Gtcndak*
SPEED LEGEND (mph)
••10
10-XO
40-M
SO
Direction of travel —»
Scale: H: 1" * 1 mlla, V: 1" • 1 hour
Reprinted from Report No. FHWA/CA/TL-69/12, October 1089, Office of Transportation
Materials and Research, California Department of Transportation by J. Palen with permission.
FIGURE 1. Speed distributions over time along a freeway segment.
F-7
-------
Tl
-------
•n
-------
require the use of some form of model. However, it is not necessary to directly use the link-
level transportation model outputs as the "vehicle term" for emissions. If sufficient data on
vehicle operating distributions exist over a range of network loadings (e.g., over a range of
times of day), relationships can be empirically derived between those distributions and
transportation model outputs.
Objective: To Model Vehicle Operations or Roadway Performance?
The preservation of vehicle operating mode distributions, is a major departure from traditional
transportation modeling in several respects. The primary objective of transportation
modeling systems is to simulate the operation of the highway network and assess its ability to
handle travel demand. That is, the emphasis is on modeling of the facilities and systems.
The transportation modeler has little need for knowledge of small scale fluctuations and
variability-it is how the network behaves on the average that is important. In contrast, as
discussed earlier, emissions may vary greatly for a specific average speed, depending on the
distribution of speeds and accelerations that yield the average. Therefore, our objective is to
describe individual vehicle operations, as represented by a multivariate distribution. The
transportation model becomes an "instrument" for measuring the relative change in
parameters that describe these distributions. In this way, the transportation model can be
used as the basis for forecasting the effect of population and employment growth and network
improvements on distributional parameters.
91078-2 F_1 o
-------
4 MODEL CONCEPT DESCRIPTION
Basic Emission Calculation
The computational structure of the proposed mobile source emission model is the numerical
integration of emission rates, within designated areas, of a vehicle operating mode
distribution and modal emission factor. This process is shown schematically in Figure 4.
We use the following notation:
v A vector describing the state and operating mode of a vehicle
Fft(v) The distribution function of vehicle operating modes within area i for
hour h evaluated at vehicle state vector v.
e(v) The emission rate (mass per unit time) for a vehicle in state v
The vehicle state vector consists of an identification of basic operating mode (cruise, idle,
accelerating, creeping/queued), as well as continuous variables for key factors (speed,
acceleration rate, canister state, engine/catalyst temperatures, etc.). For simplicity of
notation, we will also assume that the vehicle state vector includes those terms usually used
in defining fleet characteristics (model year, technology, mileage, etc.) as well as those
describing current environmental conditions (temperature, relative humidity).
The emission rate for vehicles within area i for hour h is given by
e(v)dFJ(v) .
Total emissions are given by the summation over areas (i) and hours (h):
e(v)
91078-2 F-11
-------
Vehicle operating
mode distributions by
area and time
Modal emission
factor model
Emission calculation
FIGURE 4. Base case emission estimation.
eee/rgiOOl
F-12
-------
Base Case Emissions
The base case emissions for an area are based to the extent possible on empirical
information. In particular, the vehicle operating mode distribution F(v) is derived from
detailed aerial photographic data. Time lapse photography (e.g., three frames at three
second intervals) or video techniques can be used to determine the speed and acceleration of
every vehicle within the photograph. Computer analysis techniques, coupled with geographic
information systems (CIS) including roadway information, can be used to identify and extract
these data. Although potentially time consuming, a relatively large and robust data base can
be developed for any area, without the need for labor-intensive traffic observations or
capital-intensive traffic monitoring. Since aH vehicles are counted and their states determined
at every point along roadways, questions of the representativeness of traffic count data are
avoided.
This data base will also identify trip ends, in that newly parked or departed vehicles can be
identified. This information on trip initiation/termination will be incorporated in the vehicle
state vector distribution in a manner appropriate to the emission factor treatment. To avoid
unnecessary repetition, we will not present here the emission rate equations for running
losses, trip end emissions, or breathing losses; these are essentially identical equations to that
already shown, differing only in the definition of the vehicle term v.
Development of Forecasting Relationships
Empirical distributions F(v) are static, and cannot be directly used to develop future year
forecasts. Transportation planning models can be used to forecast general traffic trends, but
not distributions of key variables. However, in the course of developing empirical
distributions through multiple hours and days of aerial photograph analysis, data will be
collected that show the sensitivity of these distributions to variation in total network loading.
For example, normal variations in day-to-day traffic volumes allow development of speed
and acceleration distributions as a function of (for example) the total number of vehicles
currently operating within area i, or the total VMT accumulation rate within area i. These
data showing sensitivity to change provide the basis for forecasting.
To allow the use of transportation planning model outputs for forecasting, we define a new
distribution F'(v). This new distribution is a synthesized distribution based on a
parameterization using transportation model outputs. Figure 5 shows the process for
developing these distributions. We use the following additional notation:
w A vector of parameters that are outputs of the transportation planning
model and that correspond to measures of overall network loading in
the data base used to develop F(v)
Gft(v| w) The observed distribution of vehicle operating modes v in area i for
hour h, given that the network state is w
91078-2 F-13
-------
Observed vehicle
operating mode
distributions
Statistical analyses
of distribution
dependencies
Transportation model
simulations of range
of base year conditions
Selection of
model parameters
Synthesis and evaluation
of vehicle operating
mode distributions
Verification of emissions
from synthesized distributions
FIGURE 5. Development of relationships for synthesizing and forecasting vehicle
operating mode distributions.
eee/rgi003
F-14
-------
A mathematically constructed distribution whose parameters are given
by w that has been shown to be essentially identical to
Then
where w0 is the "average network state" for the base case. That is, F' is simply G' evaluated
under average base case conditions. It is clear that F' must be shown to give comparable
vehicle distributions and emissions to those of F. It is also clear that the mathematical
development of G' may be difficult, but can be simplified by limiting the dimensionality of
w. In the absence of data bases of the type described for developing F(v), the best approach
for the development of G' is uncertain. The development of mathematical forms for G', the
"calibration" of F', and the verification of agreement between F' and F will be a key aspect
of developing forecasting capability based on this proposed approach. We do not explore
specific methods here, but note that there are small-scale models (i.e., models that address
individual vehicle operations within small areas) and model development efforts that are
potentially useful in developing the needed relationships. For example, Wong (1990) and
Rathi and Santiago (1990) describe the capabilities and structure of TRAF-NETSIM, a Monte
Carlo simulation model of individual vehicle behavior within small networks.
Forecasting
The use of the above techniques to generate future year forecasts is straightforward, as
shown in Figure 6. Denoting future conditions with an asterisk ", future year base case
emissions are given by
[
dF'm'(v)
where
Control Scenarios
The estimation of control effects under this proposed approach closely parallels existing
practices. As shown by the dotted lines in Figure 6, there are three basic types of controls:
those that affect vehicle emission rates (e.g., retrofit technologies, reformulated fuels); those
that affect trip generation (e.g., ride sharing measures, growth management strategies); and
91078-2 F-1 5
-------
Transportation
model forecast
Synthesis of future
vehicle operating
mode distribution
Control
measure
effects
Modal emission
factor forecast
Emission calculation
* Control scenarios only.
FIGURE 6. Development of future year emission forecasts.
eee/rgi002
F-16
-------
those that affect vehicle operations (e.g., flow improvements, elimination of driveup
windows). Depending on the specific measure, simulation of controls could involve either
the modal emission factors (e.g., for reformulated gasoline), the future year transportation
model forecast parameters w (e.g., for employer-based trip reduction); or the relationship
between transportation forecasts and the synthesized distribution of vehicle operating modes,
G'(v\w).
The disaggregate nature of emission calculation via numerical integration allows explicit
treatment of factors whose effects might otherwise be treated in overly simplistic ways. For
example, modal emissions for vehicle idling can be directly treated in a modeling system that
explicitly treats the number of vehicles idling at traffic lights. The principal limitations of
providing this opportunity arise from both the computational intensiveness of any such
changes, and the need to generate reliable estimates of vehicle operating mode distributions.
Small-scale transportation models were previously mentioned as having the potential for
addressing the effects of, for example, travel demand or signal timing on distributions of
acceleration, idle, creep, and cruise modes. Such models become important here for two
reasons: (1) the initial data on which G' will be based will be of relatively short duration and
will not be able to statistically support estimation of the effects of all types of flow
improvements; and (2) the vehicle operating mode distribution F' must provide a reasonable
representation of changes arising from system changes that expedite or impede flow.
91078-2 F-17
-------
5 VERIFICATION OF INPUTS, RELATIONSHIPS, AND RESULTS
The historical failure of emission estimation procedures to provide accurate values is
cautionary. Each step in the estimation process should be subjected to independent
verification of its performance under actual conditions. The proposed approach offers
opportunities to verify individual steps and components of the emission calculation. We will
consider the individual parts of the emission estimation process independently.
Vehicle Operating Mode Distributions
Photographic and image processing techniques for establishing vehicle operating mode
distributions have been explored, but they cannot be considered to be standardized.
However, a number of opportunities exist for evaluating the performance of such techniques.
These include
Equipping marked vehicles with accelerometers, fuel consumption and emission
measurement instruments, and position recording equipment, and verifying the
accuracy with which aerial photographic techniques estimate mode, speed,
acceleration, and emissions. This technique can also be used on a "trip" (as opposed
to area) basis to identify spatial differences in operating mode distributions, and as a
check on the areawide distributions of operating modes.
Conducting "audits" using video techniques, loop counters, and other fixed-location
data, to determine how accurately the primary method describes roadway dynamics.
Use of video-based license identification techniques (Williams et al., 1989) to verify
the age distribution of in-use vehicles, and the variability of this distribution by time
and location.
Note that although these methods are useful to test and verify the accuracy of the primary
technique, it is not necessary that they be found to be "representative11 of typical conditions
in order for them to be of value for verifying performance of the emission estimation
method. This cannot be said of such techniques if they are ultimately directed toward a
nominal driving cycle such as that of the FTP.
91078-2 F-18
-------
Modal Emission Factor Model
The development of emission factors has been a perpetually troublesome step, and will
remain so until reliable methods for in-use vehicle emission measurements have been
developed. Nevertheless, the use of modal emission factors allows more direct comparison
of vehicle-specific emission rate measurements with predicted values. The Stedman and
Bishop (1990) remote sensing technique shows promise, especially in conjunction with
registration-based identification of vehicle age and technology. As noted in the preceding
paragraph, test vehicles can also be used to verify the accuracy of model-based emission
estimates of single vehicle emissions.
Synthesized Distributions of Vehicle Operating Modes
As noted earlier, this is perhaps the area where the most new development work is needed.
Small-scale models of roadway dynamics will need testing and verification. In addition, the
ability to synthesize operating mode distributions is a severely multidimensional problem. It
cannot be arbitrarily assumed that traffic dynamics are independent of factors that are
primarily of concern for emission rates (e.g., vehicle size and age). Early work is needed to
determine the extent to which independence of traffic dynamics and vehicle or area
characteristics can be established. For example, if it can be shown that systematic bias in the
joint distribution of speed and acceleration on freeways (given fixed volume) is independent
of time of day and area, then the development of synthesized distributions is greatly
simplified.
As part of this effort, parameterization based on transportation planning models is needed,
along with verification that the use of outputs from such models as inputs to smaller-scale
models yields appropriate distributions. This is perhaps the weakest area of any forecasting
technique, in that the "vehicle term" that is generated is a traffic modeling forecast of a
traffic modeling forecast of a socioeconomic forecast. Errors in such forecasts are to be
expected. Appropriate checks are needed to verify that the procedures followed produce an
ultimate product (the synthesized distribution F*) that is properly sensitive to the key inputs:
travel demand and network status. Maximum use should be made of the natural (day-to-day
and seasonal) variation in travel volumes to establish operating mode distributions over as
wide a range of network loadings as possible. These data can then be used to test the
performance of model-synthesized distributions for heavier loading and roadway
improvements. Similarly, planned road maintenance closures can be treated as experiments
to determine the sensitivity of operating mode distributions (both empirical and synthesized)
to changes in overall network capacity, and other planned experiments (e.g., signal timing)
can be conducted.
Finally, although there may well be errors in absolute forecasts, the opportunity exists (with
appropriate data collection) to verify the basic procedures. On a year-to-year basis, it can be
determined whether the procedures for forecasting vehicle operating mode distributions
91078-2 F-19
-------
would have reproduced those that were actually observed, if the area-wide forecast itself had
been correct.
91078-2 p.20
-------
6 RESEARCH ISSUES
Much of the work needed to develop this proposed emission modeling concept has been
recently explored, or is currently underway. This is not to say, however, that an integrated
program can be immediately assembled from existing data bases, tools, and programs.
Research strategies for developing modal emission factor models for in-use vehicles are not
yet integrated with traffic engineering efforts to describe vehicle dynamics. Similarly,
transportation planning models have been used to provide inputs for smaller scale models,
but treatment of distributions of vehicle operating modes has been avoided. Where the
stochastic nature of traffic flows has been treated, such as by Daly and Ortuzar (1990) and
Chang and Kanaan (1990), the emphasis has been on the effect of sample size, level of
aggregation, and model complexity on error bounds and confidence limits for model-
predicted transportation parameters.
An integrated effort will be needed to develop the proposed emission estimation approach.
Common definitions of terms and variables is needed, as is careful review of the capabilities
and limitations of each of the data collection and modeling techniques. An important step is
the development of an operational definition of vehicle operating mode that is appropriate
from the perspective of traffic data collection, traffic modeling, and emission modeling. We
have suggested by way of example that four basic modes (cruise, acceleration, idle, and
creep) be used along with other variables to define vehicle operating modes. Other ways of
dimensioning the vehicle mode state space may be more readily integrated with traffic data
and modeling.
Defining characteristics or boundaries for subareas will also require careful attention.
Arbitrary gridding of an urban area is feasible, but may lead to difficulties if an individual
areas' traffic behavior interacts too closely with its neighbors, or if an area is too diverse to
exhibit consistent patterns. Definition based on roadway characteristics is not an option,
unfortunately, in that there is a clear dependence of traffic flows on different types of roads
within an area. This interaction (e.g., spillover of freeway traffic onto arterials during
congested periods) makes it impossible to separately treat adjacent roadways.
Many other research areas arise from the suggested approach, including the prospect of
developing weekend emission simulations (unreliable to date), and exploring the "normal"
variability in traffic volumes and emissions from day-to-day. For example,
weekday/weekend effects in ozone trends are well established; how significant in determining
maximum ozone concentrations is the coincidence of adverse meteorology and abnormally
congested traffic networks? Palen (1989) shows that at least at the roadway level, substantial
91078-2 p-21
-------
variation from day to day is possible. Development of emission factor models based on the
physics and chemistry of combustion and controls is another potentially fruitful area.
Comparability of vehicle operating mode distributions within and across cities will be of
particular interest as well. Although descriptions of human driving behavior cannot be
developed from physical principles, the large number of vehicles, and the basic physical
constraints of cars and roadways may make it possible to develop mathematical forms that
describe the stochastic nature of vehicle operations in ways that are applicable in all areas.
91078-2
-------
7 SUMMARY
In this report, we have suggested a conceptual design for a next-generation mobile emission
modeling system. Although it draws from current work, the proposed system places
substantial reliance on detailed empirical distributions of vehicle operating modes as the
primary vehicle-related input to the calculation. The intent is to simulate individual vehicle
operations, rather than VMT accumulation. In this way, potential unrepresentativeness of
arbitrary driving cycles does not influence emission calculations.
This approach requires the development of a modal emission rate model whose input
parameters are consistent with those that define vehicle operating modes. Given the
distribution of vehicle modes and the modal emission model, it is possible to numerically
integrate over the range of possible vehicle states to develop emission estimates. If properly
defined, and with adequate data, such estimates provide explicit treatment of many of the
factors that are not currently treated in mobile emission forecasts.
Forecasting under the proposed system requires the use of transportation planning models and
small-scale models to synthesize distributions of the base case emission inventory (for
purposes of verification). This hybrid approach (i.e., model-based synthesis of empirically
developed distributions) may require substantial testing and development, but has several
advantages. Most notably, since distributions of vehicle operations are preserved, overly
simplistic treatment (from an emission modeling perspective) of vehicle operating conditions
is avoided.
The proposed approach is both data and computationally intensive due to the size of the data
bases involved. However, advances in computing speed, and the ability to automate much of
the data reduction and analysis (e.g., through computerized analysis of video images) should
substantially reduce costs, as well as enhancing data comparability between cities.
Substantial reliance is also placed on aerial photography for data collection. While high-
resolution (and costly) mapping and surveillance techniques could be used, adequate results
should be achievable with substantially less expensive platforms. The development of base
case emission inventories should be relatively inexpensive in that no transportation modeling
is required. Additional costs (e.g., use of instrumented vehicles and remote sensing to verify
in-use emissions are those that should be anticipated, regardless of the design chosen for the
emission estimation system. The costs of forecasting are likely to be substantially higher for
the first city studied, due to the need to explore the ability to synthesize vehicle operating
mode distributions from traffic model outputs. Such efforts should be transferable to other
cities, however, resulting in an overall emission forecasting system whose costs are
comparable to those expended in transportation modeling.
91078-2 F-23
-------
ICES
All Safwat, K. Nabil, and Thomas Magnanti. 1988. "A-Combined Trip Generation, Trip
Distribution, Modal split, and Trip Assignment Model," Transportation Science.
v22:l, Feb. 1988, pp. 14-21.
Antonisse, Robert, Andrew Daly, and Moshe Ben-Akiva. 1989. "Highway Assignment
Methods Base on Behavioral Models," Transportation Research Record. #1220,
1989.
Becker, U. J., 1989. "Automatic Traffic Data Collection Using Aerial Video Images,"
Second International Conference on Road Traffic Monitoring.
Ben-Akiva, Moshe, and Steven Lennan. 1985. Discrete Choice Analysis: Theory and
Application to Travel Demand,. Cambridge, MTT Press.
Dial, Robert. 1976. "Urban Transportation Planning System: Philosophy and Function,"
Transportation Research Record #599, 1976, pp. 60-64.
Dial, Robert. 1976. "Urban Transportation Planning System: Philosophy and Function,"
Transportation Research Record #619, 1976, pp. 43-48.
Hagerty, Bradley. 1987. "User's Guide for TRAP H NETSIM," Michigan Department of
Transportation.
Kollo, H. P. H., and C. L. Purvis, "Regional Travel Forecasting Model System for the San
Francisco Bay Area." Transportation Research Record. #1220, pp. 58-65.
Leonard, D. R., P. Gowen, and N. B. Taylor. 1989. "CONTRAM: Structure of the Model"
TRRL Research Report, 1989, no. 179.
Makigami, Yasuji, Hamao Sakamoto, and Masachika Hayashi. 1985. "An Analytical
Method of Traffic Flow Using Aerial Photographs', ITE Journal. v.lll:4, July 1985,
pp. 377-394.
Matzoros, Athanasios. 1990. "Results from a model of road traffic air pollution
featuring junction effects and vehicle operating modes." Traffic Engineering
+ Control. January. 1990, pp. 24-34.
91078-2 r F-24
-------
Palen, J. 1989. 'Freeway Speed Profiles and Fuel Consumption Relationships,"
FHWA/CA/TL-89/12, Office of Transportation Materials and Research, California
Department of Transportation, October 1989.
Persuad, Bhagwanti, and V. O. Hurdle. 1988. "Freeway Traffic Data from Time-Lapse
" Transportation Sdet^, v.22:4, November 1988, pp. 251-258.
Pfennerstill, E. 1989. "Automatic Monitoring of Traffic Conditions by Re-Identification of
Vehicles," Second International Conference on Road Traffic Monitoring.
Pursula, M., 1989. "Microprocessor and PC-Based Vehicle Classification Equipments Using
Induction Loops", Second International Conference on Road Traffic Monitoring.
Rathi, Ajay. K., and Alberto Santiago. 1990. "The New NETSIM simulation program,"
Traffic Engineering and Control, May 1990, pp. 317-320.
Stedman, D. H., and G. H. Bishop. 1990. "An Analysis of On-Road Remote Sensing as a
Tool for Automobile Emissions Control," I1ENR/RE-AQ-90/05, Illinois Department
of Energy and Natural Resources, 1990.
Taylor, M. A. P., W. Young, and R. G. Thompson. 1989. "Headway and Speed Data
Acquisition Using Video," Transportation Research Record. # 1225, 1989., pp. 130-
139.
Taylor, N. B. 1990. "CONTRAM 5: An Enhanced Traffic Assignment Model," TRRL
Research Report, No. 249.
Urban Mass Transit Administration. 1990. "Urban Transportation Planning System: Utown
Case Study," July 1986.
Weiner, Edward. 1988. "Urban Transportation Planning in the U. S.: An Historical
Overview", 3rd ed., revised, DOT-T-88-26, Sept 1988.
Williams, P. G., H. R. Kirby, F. O. Montgomery, and R. D. Boyle. 1989. " Evaluation of
Video-Recognition Equipment for Number-Plate matching," Second International
Conference on Road Traffic Monitoring.
Wong, Shui-Ying. 1990. 'TRAF-NETSIM: How it Works, What It Does," JTE Journal,
April 1990, pp. 22-27.
Zarean, M., and Z. A. Nemeth. 1988. 'WEAVSIM: A Microscopic Simulation Model of
Freeway Weaving Sections." Transportation Research Record. #1194, 1988, pp. 48-
54.
91078-2T F-25
-------
F-26
-------
APPENDIX A
COLLECTION OF DETAILED VEHICLE EM-USE INFORMATION
INTRODUCTION
Given the significance of driving cycle information to the emissions modeling efforts,
significant data collection efforts over the next five years should be devoted to ascertaining
real world driving cycles under a variety of conditions. Certain developments in the traffic
engineering environment may provide a useful base for our efforts.
POSSIBLE STUDIES
Monitoring of Vehicles by On-vehicle Sensors
The objective of this study is to extend current technology as required to facilitate the placing
of a "black box" in either a series of test vehicles who would drive standard courses
throughout the city, or, with more advanced technology, place recording devices on a
random sample of vehicles in several urban areas. Devices should record time and speed
information for periods of at least a week.
Several projects reported in the literature have done related activities. Palen (1989) reports a
massive study where test vehicles were instrumented to measure fuel consumption on a
variety of test runs over a carefully monitored section of highway. Link volume data was
collected by various passive count devices. The study derived various relationships of
vehicle performance to fuel consumption and evaluated other models. The data set is
available to other researchers.
Monitoring of Selected Urban Links
The objective of this activity is to develop monitoring techniques of actual vehicles to record
and assess the dynamics of interest for emissions. Vehicle monitoring has been conducted
for decades. There are four principal methods:
Manual counting
Fixed position sensors in or on the roadway
Fixed position camera
91078-2 F - A-1
-------
Aerial photography, including helicopters, airplanes, and blimps.
Manual counting.
This requires the least technology, but really records only count data. If sufficiently detailed
(turning movements, vehicle types, number in queue) it might support other studies, but is
not sufficient in itself.
In-road sensors.
While this is primarily a way of obtaining counts, some research has been done (see Pursula
1989) at using the analog signal of a loop detector to measure vehicle class.
Photographic techniques.
It seems clear that for the assessment of the vehicle dynamics of acceleration, deceleration
and idling, more detailed observations are required. Filming traffic movements for later data
reduction is a technique that goes back to the 70s, or further. The cost of film and the
difficulties of reducing data to a computer processable record have prevented widespread use
of this technique. Both limitations are being eased. The reduced cost of video equipment
and the advent of computerized methods of deriving data from video are spurring current
research.
Persaud (1988) reviews the.history and feasibility of this approach, and reports a project to
obtain data on variations in speed as a function of flow and location and discusses methods of
extracting data from film. Taylor (1989) reports on acquiring headway and speed
information via video and describes computerized techniques for automatic data reduction
from the video record. The specific possibilities of aerial photography are explored, in
among other places, Becker (1989) and Makigami (1985).
One development that could greatly expand the information available to analysts is the
possibility of automatic number-plate recognition and matching. With several video cameras,
and linking to registration databases, it is possible to both determine the actual routes and
elapsed time on routes for vehicles and the exact mix of vehicle classes traversing the routes.
Work in this area is described by Williams (1989) and by Pfannerstill (1989).
Given the current ferment in research on image processing and the declining cost of video
equipment, there would be a likely opportunity for significant advances in this area with
funding. Note in particular that work is already focusing on obtaining some dimensions of
vehicle dynamics necessary for the emissions modeling effort.
QUESTIONS FOR DATA EVALUATION
A research task could be formulated to use and generate monitoring data focusing on the
needs of emissions analysis. Some specific questions to guide these data collection tasks are
given here.
91078-2
F-A-2
-------
What is the extent and nature of temporal and spatial variability on the pattern of
driving cycles?
What is the significance of vehicle class on driving cycles? Specifically, how many
divisions of vehicle class are necessary due to different driving dynamics and how
does that relate to the vehicle classes used in emissions analysis?
Is roadway type a significant indicator of driving cycle? This would be particularly
useful if true, since roadway type can be obtained from map data.
Does link congestion dominate other variables in the assessment of driving cycles? In
other words, can we use link volumes as proxies for driving dynamics?
F-A-3
-------
F-A-4
-------
APPENDIX B
SMALL-SCALE TRAFFIC SIMULATIONS
The UTPS and similar transportation planning modeling systems are oriented toward urban
scale modeling. They cannot deal with the dynamics of individual vehicles, nor is that their
purpose. There are many other models of traffic operations that provide better platforms for
analyzing the detailed dynamics that play such a significant role in emissions. We review
these for their suitability for providing a platform for our efforts.
SPECIAL PURPOSE SIMULATIONS
There are myriad special purpose simulations that model a particular roadway situation. As
an example, Zarean (1988) reports the development of WEAVSIM, a model of freeway
weaving sections. These might be useful for specific studies, but are not general enough and
are not supported and documented at a level necessary to be used by the wider community.
GENERAL SIMULATIONS
More general are those simulations that can be configured to describe a section of road
network and are flexible enough to deal with small scale issues like turning lanes, delay from
parking vehicles, and so forth. Key among these are CONTRAM, developed in the U. K.
(see Leonard, 1989, and Taylor 1990) and NETSIM, now TRAP H NETSIM in this country
(see Rathi, 1990, Wong, 1990, Michigan Dept Highways, 1987).
These models are primarily intended for the modeling of changes to the transport system to
investigate impacts on capacity and delay. One very significant class of changes of concern
to traffic engineers has to do with traffic signal light timing.
As NETSIM is supported by FHWA, we focus our assessment on this model. NETSIM was
developed in the early 1970s, and has been applied to a variety of problems throughout the
U.S. Individual vehicles are modeled on a 1 second timestep. Various classes of vehicles
are supported. The model captures various detailed movements such as turning across
opposing traffic, reacting to obstructed lanes, yield and stop signs and pedestrian delay of
vehicles. Reflecting its traffic engineering perspective, there is a very detailed ability to
model traffic signals. Some emission modeling capability is included.
91078-2
F -B-1
-------
SUITABILITY OF TRAF H NETSIM
A preliminary assessment of descriptions of NETSIM show that it has potential to be
modified to serve as a platform for supporting emission modeling. NETSIM does model
individual automobiles, and allows for variation in driver dynamics, albeit at a simple level.
Approach and departure from intersections is governed by a simple deceleration and
acceleration formula, giving access to these dynamics. The modeling of fuel consumption
involves calculations that have some similarity to what we expect for emissions. It would
certainly support modeling of a variety of TCM measures relevant to emission controls.
However, certain extensions would need to be made to NETSIM, based on our current
understanding of the model. More complex deceleration and acceleration curves may be
needed, and changes in simulation outputs would be required.
91078-2
-------
APPENDIX C
THE URBAN TRANSPORTATION PLANNING SYSTEM
INTRODUCTION
Given the intimate connection of estimating the spatial and temporal emissions of automobiles
with the transportation modeling problem, it is essential for emission modeling
to understand the dynamics of the transportation problem.
After thirty years of effort, the current state of urban vehicular modeling can be summarized
by saying that there is a default modeling framework, and intensive investigation of
alternatives to it
The default system is known as UTPS, the Urban Transportation Planning System. This is a
set of related computer programs that implement the default modeling techniques. The
history of UTPS is covered in the two reports of Dial (1976) and Weiner (1988). A case
study using UTPS was published as UMTS (1986).
As we review this system, we will evaluate its suitability for the emission modeling effort,
and review various alternative methods currently undergoing research.
FUNDAMENTAL ASSUMPTIONS
Inputs
The default modeling approach takes as inputs various socioeconomic data, a description of
the road network at the individual street level, and an empirically derived relationship
between travel time and the likelihood of a trip being made. There are four primary data
sets required:
Transport Network. A link and node description of the highway and transit
network. Primary link characteristics are: end nodes, length, maximum speed,
capacity per lane, lanes. Primary node characteristics are x, y coordinates.
Note that artificial nodes for zone centroids and links from zone centroids to a
real node are required.
91078-2 F - C-1
-------
Zone Data. These consist of information that will be used to predict the
attractions (number of trips ended) and productions (number of trips started) in
each zone. They are typically derived from a regression analysis of various
factors. Productions may be modeled as a function of population, income,
number of autos, household size and other factors. Attractions may be
modeled as a function of square footage in the zone devoted to retail,
manufacturing, etc. Trips may be modeled by type (home-based work, home-
based shop, non-home-based, etc.). Note that zones are seldom, if ever, a
regular X, Y grid, and they do not correspond to grid cells of air pollution
models.
Friction Factors or Travel Time Factors. These are a set of points that define an
empirical curve that describes the way travel time affects the attractiveness of a trip.
In other words, using the metaphor of the gravity model, these curves define how the
"force of gravity" falls off as travel time increases.
Travel Survey Data. Various travel survey data for the base case or previous
years are needed. These may be used, prior to forecasting applications, to
derive the friction factors and to calibrate base case results.
Outputs
The primary outputs of the UTPS system are link volumes and travel times for a single point
in time of equilibrium flow. From the perspective of emission modeling, this is at once too
little and too much. It is too little because emission modeling requires temporal
disaggregation into time of day, and because of the importance that acceleration and
deceleration play in the emission process. Further, if the primary acceleration, deceleration,
and idling times are associated with stopping for intersections, then the focus of UTPS on
links may not be modeling the particular points of concern.
On the other hand, it is too much because we are really not interested in link volumes as
much as flows in a grid cell.
BASIC MODELING STEPS
The standard modeling process implemented in UTPS involves a classic four-stage process:
trip generation, trip distribution, model choice and route choice. The stages are sequential,
so the results of a downstream stage do not influence the upstream stages. We consider each
stage in more detail.
91078-2 F - C-2
-------
Preliminary Activities
Before the model can begin, certain preliminary efforts are required which also have
implications for emission modeling. First is the preparation of the network. For a major
urban area this could be enormously complex, even if residential areas were not completely
modeled. However, this has been done for urban areas that have been using UTPS for some
time. Further, the development of computerized mapping techniques and the use of GISs
simplifies this process.
Of more significance, however, is the construction of zones. This immediately defines the
level of aggregation of data. Each zone has a centroid, which is connected to the real
network by an artificial link. These artificial links are the sources and destinations for trips
within the zones, and are used to describe all travel that does not occur on identified real
links.
Trip Generation
How many trips are there in an urban area? This stage generates origins (productions) and
destinations (attractions) by zone. Typically, a multiple-regression analysis is used to form
an equation relating productions and attractions to various socioeconomic criteria.
Developing this model may depend on some sort of traffic survey or more general household
survey. These surveys are so complex and expensive that they are performed only every
twenty years, or have limited resurveys every ten years.
The zonal approach to trip generation is a fundamental assumption which has several
consequences on the shape of modeling. First, it should be noted that this is not a behavioral
model. The "zone" is not an entity which decides to make certain numbers of trips. The
number of trips is an aggregate of various individual decisions which one hopes, with some
expectation, will result in correlations at the zonal level. Also, the condition of the transport
network itself, fuel availability and price, have no impact on the number of trips, something
which is clearly true only in the short run.
Trip generation models become quite complex, since each type of trip (home-based work,
recreation, non-home based, etc.) generally has a separate model. For an example of this
complexity, see Kollo (1989).
Trip Distribution
Trip distribution is the linking of origins and destinations. Each production in a zone gets
assigned to an attraction in another zone. The classic technique here is the gravity model,
first described by Alan Voorhees in 1955. The gravity model calculates the trips between
each zone pair using the "attractiveness" of the destination zone divided by the sum of
"attractiveness" of all the destination zones for the origin zone. "Attractiveness" is a
91078-2 F - C-3
-------
function of the actual attractions in the destination zone, the friction factors, the minimum
travel time on the minimum path between the zones and a further factor. This last factor is
calibrated on the base case data to make the model converge to observed data.
There are a number of obvious issues here. While the choice to make a trip and the choice
of where to go may be separable without difficulty for work trips, it makes less sense for
non-work trips, where the choice of destinations and the ease of getting there can
dramatically affect the number of trips themselves.
Travel Mode Split
The third stage is to split trips by mode (e.g., single occupant vehicle, transit) as a function
of time and cost of travel, although the user may specify the exact form of the model.
Trip Assignment
This step distributes trips by link according to the following process:
Assign trips to best paths using minimum travel time
Calculate link volumes
Adjust link travel times based on congestion
If the change in travel times has converged to an acceptable level, quit, or else go
back to step 1 (assignment of trips), using the new link travel times.
As with the other stages, the assignment process has been critiqued. Note that each trip
loads each link from origin to destination simultaneously rather than being an individual
vehicle which flows over the network. Further the artificiality of the loading of the network
and searching for equilibrium has resulted in various algorithms being proposed for how the
network is to be loaded. More behaviorally based models of traffic assignment have been
proposed (see Antonisse, 1989).
NEW APPROACHES
Two directions of alternatives to UTPS may be briefly noted. First, models aimed at
combining the various stages are common, for example see All Safwat (1988) and the
literature cited there.
A second approach is to abandon the zonal framework altogether and make the household the
basic unit of modeling. Aggregate statistics are then based on sampling over the urban
91078-2 F - C-4
-------
region. This approach first arose in the context of sketch planning. See Ben-Akiva (1985)
and the literature cited there.
GENERAL ISSUES FOR UTPS
There are some additional issues, and key working assumptions of UTPS which are well
known by practitioners:
Multi-Segment Trips
Trips of more than one destination (x to y to z) are modeled only by breaking them into
single segment trips. The chaining of trips can be significant (e.g., as a reaction to higher
gas prices or scarce availability); this change must be handled externally rather than being
modeled.
Varieties of Vehicle Types
Travel is classified by trip type, but vehicle type is not addressed.
Information Availability
The ability of drivers to know the current state of congestion is not modeled. One proposed
TCM is to improve driver information to permit a reduction of idling and congestion.
Time Dynamics
This is a steady state model, and not one where temporal issues are considered (e.g. leaving
early or late to avoid the rush). The trips happen all at once. However, note that peak
versus off-peak periods may be modeled separately.
SUITABILITY FOR EMISSION MODELING
While UTPS, when used by knowledgeable and skilled analysts, produces better results than
might be supposed based on the preceding discussion, the model framework is of limited
value for emission modeling.
If link volumes, the primary output of UTPS, are to be used as a basis of emission modeling,
then a significant analytical effort must be developed to understand the connection between
an average link volume and the complex of vehicle operating dynamics that determine
emissions.
F-C-5
91078-2
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/R-93-214
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Conceptual Designs for a New Highway Vehicle
Emissions Estimation Methodology
5. REPORT DATE
November 1993
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
John T. Ripberger*
8. PERFORMING ORGANIZATION REPORT NO
9. PERFORMING ORGANIZATION NAME AND ADDRESS
See Block 12
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
NA (Inhouse)
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final; 3/91-5/93
14. SPONSORING AGENCY CODE
EPA/600/13
15.SUPPLEMENTARY NOTES AEERL project officer is CarlT> Ripberger. Mail Drop 62, 919/
541-2924. (*) A participant in the Earth Team Soil Conservation Service Volunteer
Program, assisting the U. S. EPA.
is. ABSTRACT
repOrj- discusses six conceptual designs for a new highway vehicle
emissions estimation methodology and summarizes the recommendations of each
design for improving the emissions and activity factors in the emissions estimation
process. The complete design reports are included as appendices. EPA asked six
contractors to assist in developing ideas for a potential methodology to estimate
highway vehicle emissions that could be developed for use in 5-10 years. They were
selected because of their experience in working with mobile source emissions inven-
tories. In general, the contractors suggest developing new modules within the emis-
sion estimation process to provide users with more detailed information on the cau-
ses of vehicle emissions. The essence of these concepts is the need for more com-
prehensive integration of data between the transportation planning model and the
emission factor model. The individual reports did reinforce comments voiced by
other experts for modal data and more responsive transportation models, although
they did not identify totally new concepts not already being considered by EPA re-
search programs.
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COS AT I Field/Group
Pollution
Motor Vehicles
Emission
Estimating
Mathematical Models
Pollution Control
Stationary Sources
13 B
13F
14G
12 A
8. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (ThisReport)
Unclassified
21. NO. OF PAGES
237
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
F-C-6
------- |