United States
           Environmental Protection
           Agency
                                       June
                                       1992
           Air
wEPA
Highway  Vehicle
Emission  Estimates

-------
   i .   UNITED. STATES ENVIRONMENTAL PROTECTION AGENCY
   /                  ANN ARBOR. MICHIGAN  48105
                             JUL   8 1992                        OFFICE OF
                              ui_   u iw^c                     AIR AND RADIATION

      In  response  to  concerns regarding  potential  underestimation  of
 in-use highway vehicle emissions by  highway  vehicle  emission  factor
 models,  the Office of Mobile  Sources  (OMS)  has  prepared  the enclosed
 paper. "Highway Vehicle Emission  Estimates."   A  number  of studies
 conducted over the last  few  years have  provided indications that
 current  estimates of total  highway  vehicle emissions may  be too  low.
 Among  these  are  tunnel  studies (such as  the 1987  Van  Nuys tunnel
 study),  roadside emission measurements  (including the  use  of  remote
 sensing  devices),  and  ambient  concentration  comparisons.  Each has
 provided  suggestive but  not definitive evidence that  the total
 emissions  contribution  of highway vehicles  is  understated by current
 emission  inventory  development  procedures,  which  include  -the  use
 of emission factor  models (such as MOBILE4.1) to estimate  average
 in-use per vehicle  emission  rates in grams per mile.

      EPA is  aware of a number of areas in  which current practice.
 including  the  collection  of  in-use vehicle  emission data,  could be
 improved.   Some  of these  improvements  will be  reflected  in
 MOBILES, while others  will take  longer to implement.  The EPA
 approach  to estimating  highway vehicle emission factors prior  to
 1990. the  known  limitations  of that approach,  issues  that  have  been
 identified  as  possibly contributing  to  the underestimation  of in-use
emission  levels  by the  model,  and  the  approaches  EPA has
undertaken and  is  planning  for the  future to address  these  issues,
are discussed  in this paper.

      This paper is intended to provide an overview  and to  facilitate
discussion; it  is  not a statement of official  EPA  policy.   Comments on
the paper  and the  issues discussed therein are  welcome, and should
be directed to:

                 Mr. Terry  Newell  (AQAB)
                 L'.  S.  Environmental Protection  Agency
                 National Vehicle  and Fuels  Emission  Laboratory
                 2565  Plymouth Road
                 Ann Arbor.  MI  48105

-------
                    DISCLAIMER
       This  paper has not  been peer reviewed.
   It is  intended  to  present the current  thinking  of
the Office of Mobile  Sources with respect  to a number
of issues pertaining to the  accurate modeling of in-use
emission factors  for  highway  vehicles, and  to facilitate
 discussion of these  issues  among interested parties.

-------
Overview

     The models and methodologies developed by EPA  for use  in
estimating highway vehicle emissions have come under increased
scrutiny recently, with much attention given to the possibility
that these models and methods underestimate the actual level of
emissions from mobile sources.  This background paper has been
prepared to address questions regarding this potential
underestimation of in-use emissions and to describe current and
planned EPA activities aimed at improving the accuracy of the next
emission factor model, MOBILES, and resulting highway vehicle
emission inventory estimates.

     In this paper, some of the issues involved in quantifying in-
use emissions from highway motor vehicles are discussed.  The
difference between new vehicle emission standards and in-use
performance, EPA's programs for collection of emission data from
in-use motor vehicles, and the emission factor model are outlined.
Roadway and tunnel emission studies and ambient measurement
programs, and their implications for EPA's model and methodology
are discussed.  The paper concludes with a section on activities
EPA has underway or planned to improve highway vehicle emission
factor estimates and inventories.
Background

     The three primary pollutants from motor vehicles and engines
for which EPA has established emission standards are hydrocarbons
(HC),  carbon monoxide (CO),  and oxides of nitrogen  (NOx).  All of
these pollutants are emitted from vehicle tailpipes when the
engine is running (exhaust emissions).   Vehicles also emit HC
through evaporation of fuel from the engine and fuel system when
they are not running (diurnal or "breathing" losses, transitory
trip-end or "hot soak" losses,  continuous resting losses from
porous tubing or leaks of liquid fuel,  and weekly or so refueling
emissions), and from sources other than the tailpipe when in
operation (running loss emissions).  Some vehicles with disabled
or disconnected hoses also exhibit crankcase emissions of HC
("blowby" losses).

     One key to the accurate assessment of air quality problems
and to estimating reductions in air pollution is the development
of reliable emissions inventories,  which quantify the total amount
of a given pollutant under a specified set of conditions.
Emission inventories, usually expressed in tons of pollutant per
year,  are the product of two factors:  emission factors and
activity levels.  An emission factor expresses the amount of
pollution emitted per unit of activity (i.e., grams of carbon
monoxide emitted per vehicle mile travelled).  An activity level
represents the amount of the given activity that occurs over a
specified period of time  (i.e., vehicle miles travelled by highway
vehicles in a metropolitan area on a typical summer ozone season

-------
day).  The  sum of  the products of  the emission  factors and
activity  levels  for all  sources of a given pollutant constitute
the  emission  inventory for that pollutant.

      For  highway vehicles  (light-duty vehicles,  light-duty  trucks,
heavy-duty  trucks, and motorcycles, both gasoline and diesel),
emission  factors are most often expressed in grams of pollutant
emitted per mile driven  (grams per mile, or g/mi).  The activity
level for highway  vehicles is vehicle miles travelled (VMT).
Emission  factors and VMT can be estimated for each individual
vehicle type  and for all highway vehicles as a group; the Area and
Mobile Source (AMS) system developed by EPA's Office of Air
Quality Planning and Standards (OAQPS) constructs mobile source
inventories from class-specific emissions factors and VMT
estimates.  The scope of a highway vehicle emission inventory can
be as small as a single  link of a  given roadway  for a specific
hour, or as large  as an entire Consolidated Metropolitan
Statistical Area (CMSA)  for an entire year.  In  any event,
development of an  accurate emission inventory for highway mobile
sources requires the best possible estimates of  the emission
factors.

      Since  the late 1960s EPA has  issued and periodically
tightened emission standards applicable to highway motor vehicles.
However, the  emission standards applicable to new vehicles do not
represent the emission factors applicable to those vehicles once  .
they are actually  in use.  Emissions from vehicles vary over the
entire range  of conditions that vehicles operate under:   ambient
temperature,  traffic conditions (represented by  average speed),
operating mode (the mix of cold or hot starting  and warmed-up
vehicle operation), fuel volatility and composition, types and
condition of  emission control equipment and other vehicle or
engine components that affect emission levels, and other variables
all affect the emissions actually  produced by vehicles in everyday
use.  In addition,  the in-use vehicle fleet is composed of several
generations of vehicles and emission control technology, each of
which behaves differently in terms of in-use emission levels and
how these levels change over time  (as the vehicle ages and
accumulates higher mileage).  Thus highway vehicle emission
factors are estimated using computer models, which allow emission
factors for different vehicle types to be estimated under
conditions specified by the user of the model and combined into an
overall estimate of emission factors for the in-use vehicle fleet.

      Direct measurement of emission levels from  all in-use
vehicles under all possible conditions is clearly impossible.  To
estimate in-use emission levels,  EPA conducts surveys in the form
of vehicle test programs, collecting emission data from as many
vehicles as is practical over as wide a range of conditions
affecting emissions as possible.    Time and funding are the
constraints that determine what is practical in  terms of building
the emission  factor data bases.  The use of such data to create
the highway vehicle emission factor model is discussed.
                                -2

-------
      Direct  measurements  of  emissions  from highway  vehicles
operating on the  road have become available through the results of
several on-road emission  measurement projects and tunnel emission
studies.  The most recent and well publicized example is the Van
Nuys, CA tunnel study sponsored by the Coordinating Research
Council  (CRC) and conducted  in 1987.  The  results of this study
were  widely  perceived to  indicate that the highway  vehicle mobile
source emission factor models developed by the California Air
Resources Board (ARE) and by EPA may underestimate  hydrocarbon
(HC)  and carbon monoxide  (CO) emissions from highway vehicles.
(California  has had more  stringent motor vehicle emission controls
than  the rest of  the country since the 1970s.  Emission factor
models have  been  developed by both EPA and ARE.  The current
versions of  these models  are known as MOBILE4.1 and EMFAC7E,
respectively.)  The results of the tunnel  study and their
implications for  highway  vehicle emission  factor estimation are
also  discussed below.


The EPA Approach to Highway Vehicle Emission
Estimates Prior to  1990 and Its Known Limitations

      The emission standards applicable to  highway motor vehicles
are specified as  gram per mile (g/mi)  limits, which vehicles are
expected to  meet  throughout their statutory "useful life" (until
recently, "useful life" was defined as 5 years/50,000 miles for
light-duty automobiles).   Since emission levels deteriorate
(increase)  with increasing mileage accumulation on  the vehicle,
new vehicles  must emit at levels less than that of  the applicable
standard in  order for emissions after 50,000 miles  to be at or
below the standard (although this may not be true of the average
in-use vehicle).  An average emission rate for a given vehicle
type and pollutant can thus be stated as a linear equation,
consisting of a "zero-mile level" and one or more deterioration
rates (rate  of increase in emissions with accumulated mileage).

      The emissions produced by highway vehicles in  use depend  on a
large number  of variables.  EPA has developed a standard test
procedure over which vehicles are certified as complying with
applicable emission standards before their introduction to the
market,  the  Federal Test  Procedure (FTP).   The FTP  includes
specific values for many of the variables affecting emission
levels (temperature,  average speed,  cold/hot mix,  and so on).

      In order to  collect  data for modeling emissions from a given
vehicle type  in actual use,  EPA conducts the Emission Factor
Program (EFP) in which randomly selected,  privately owned vehicles
(light-duty vehicles and  light-duty trucks only)  are borrowed for
emission testing.   All vehicles participating in the emission
factor testing program have an FTP test performed on them.   Since
the FTP is used to certify new vehicles to applicable emission
standards,  the real world performance of the fleet  relative to the
applicable emission standard is thus estimated.  The emission

-------
factor program to date includes test results  for more than 10,000
vehicles over more than 20 years of testing,  covering model years
from current  (1992) to pre-catalyst  (before 1975), under a variety
of conditions.  This provides a substantial data base for the
estimation of average in-use emission levels.

     However, real people most often drive under conditions that
vary considerably from those of the FTP:  they drive in colder and
hotter weather, on longer and shorter trips,  at higher and lower
average speeds, using commercial fuel that often differs in
quality, chemical composition, and/or volatility from the test
fuel used in laboratory tests, and so on.  Other factors in
addition to the conditions and mode of operation of the vehicle
affect emissions, some of them beyond the control of the
individual driver:  Vehicles subject to annual emission
inspections (commonly referred to as inspection/maintenance, or
I/M,  programs) should have lower emissions on average than
comparable vehicles not subject to such inspections, tampering
with emission control equipment can cause much higher emissions at
a given age/mileage than would be seen from untampered vehicles,
and so on.

     EPA has developed a computer model that  takes into
consideration these many influences on in-use emission levels and
provides average emission factors for different types of highway
vehicles over a range of calendar years.  These average emission
factors do not represent the actual emission  levels of any one
specific vehicle, but in the aggregate provide emission factors
that can be applied to an entire in-use fleet of vehicles.

     To characterize the impacts of the many  variables noted above
on highway vehicle emission factors, EPA also tests in-use
vehicles over driving cycles having average speeds different than
19.6 mph (the average speed of the FTP), at different
temperatures,  and using fuels of different volatilities.  Roadside
surveys are also conducted to estimate the rates of tampering with
components of the emission control system in  the real world, and
tests are performed to quantify the emission  impact of each type
of tampering.

     The above discussion focused primarily on exhaust pollutants,
those emitted from the vehicle tailpipe as waste products of the
combustion process.   As noted previously, there are other sources
of HC emissions from in-use vehicles:  evaporative emissions (hot
soak and diurnal), running losses,  resting losses, and refueling
emissions.  EPA's emission factor testing programs include routine
measurements of evaporative and running loss  emissions,  and the
results of such testing are incorporated into the emission factor
model.   (Resting loss and refueling emissions are relatively minor
components of overall HC emissions and do not vary much by vehicle
age,  mileage,  or condition;  thus they typically are not measured
on all vehicles in the emission factor program but are included in
the model.)
                                -4-

-------
     Although  the emission  factor test programs have resulted in
large amounts  of in-use vehicle data for use in modeling
emissions, there are a number of known limitations to the scope
and accuracy of the data collected prior to 1990 and the resulting
modeled emission factors.  The most important of these limitations
are outlined in the following paragraphs.

     The  "fleet" of vehicles that have been tested in the emission
factor programs generally does not include a satisfactory number
of truly old vehicles.  Such vehicles remain on the road, and
based on our knowledge of emissions behavior are likely to be
important in the overall average emission factors for the in-use
fleet.  Related to the lack of an adequate number of really older
cars in the samples is the lack of sufficient vehicles (of any age
or model year) with very high accumulated (odometer)  miles.  For  .
example, there are relatively few vehicles in the data bases with
accumulated mileages in the 75,000-125,000 mile range,  and many of
the vehicles tested at such mileages drove unusually high numbers
of miles annually (as contrasted with more "average"  cars that
might take more years to reach such accumulated mileage'levels).
The question of whether the emissions performance of vehicles at
high mileages differs depending on how quickly such mileage is
accumulated has not been thoroughly explored.

     The recruitment of privately owned vehicles to participate in
the emission factor program, while as representative as possible
in the selection of recruitment targets, is subject to bias in
terms of owner responses.  Prior to 1990, recruitment was on the
basis of mailed and/or phoned invitation, and therefore favored
vehicle owners having free time during normal business hours.
Participation  in emission factor testing programs is voluntary,
and EPA has no means of forcing those owners that decline our
solicitations  into participating.  EPA suspects that owners who
cannot afford the best vehicle maintenance,  and those embarrassed
by poor vehicle maintenance, tended not to participate,  and that
this problem is worse for older vehicles.  The potential bias due
to such "self selection," as well as practical limitations on the
number of vehicles that can be tested,  also may result in the
number of "high" and "super" emitting vehicles in the in-use fleet
being underestimated.

     As noted  above, emission factor testing is not limited to
testing vehicles over the FTP.  Tests are also conducted at a
range of temperatures, and other driving cycles with average speed
ranging from 2.5 to 65 mph are also used to test subsamples of
vehicles.  Regardless of these efforts to characterize a range of
driving patterns and behaviors,  the ability of any specific
driving cycles to represent true in-use driving behavior is
uncertain.  Even acknowledging the inability to accurately
characterize true in-use driving patterns with a limited number of
specifically defined driving cycles, there are still two areas in
which the approach taken by EPA can be questioned:  None of the

-------
driving cycles used include acceleration rates over  3.4 mph/sec  (a
rate of acceleration known to be exceeded in real world
situations, but with unknown frequency), and no data on purely
steady-state driving ("cruising" at a given speed) are used  in the
emission factor modeling.

     The evaporative emission test procedure used in testing
emission factor vehicles for evaporative emissions,  like the FTP
for exhaust emissions,  is very specific:  after a defined
preconditioning routine, hot soak emissions are measured for one
hour at about 82°F, the temperature rise for the measurement of
diurnal emission  is 60° to 84°F  over a period of one hour.   These
conditions, both  in terms of the temperature and the durations
(hot soaks clearly can last longer than one hour, diurnal
temperatures rises occur through the daytime), are probably  not
very realistic on average for characterizing real world emissions.
By 1990, EPA had begun testing at higher temperatures, but the
one-hour duration and the specific preconditioning procedure were
still artificial and probably not representative.

     The.inability to test all vehicles at all possible conditions
means that EPA has to make certain analogies and extrapolations
for characterizing emissions at conditions not represented in the
data.  An example of an analogy is that if only late-model fuel-
injected light-duty vehicles (cars) had been tested  in sufficient
numbers to model the effects of very high temperature on running .
loss emissions,  these data would then have to be used to model
high temperature running loss emissions from carbureted light-duty
vehicles and from fuel-injected light-duty trucks as well.   An
extrapolation example is that very few exhaust emission tests are
performed at temperatures higher than 95°F;  the effects of higher
temperatures on emissions would then have to be based on an
extrapolation of the effects of temperature known to exist from
75° to  95°.   In  some cases,  extrapolations are "flat," in that
they are limited to certain maximum (or minimum) values observed
in the most extreme conditions actually tested, while in the real
world the emissions impacts are undoubtedly not flat (although the
curve shape,  or behavior of emissions in the extrapolated range,
may differ from that used to characterize the range  for which data
are available).

     For each of the known limitations described above, it is not
unreasonable to assume  that the bias introduced is on the low side
(i.e.,  biases introduced by these limitations are likely to  be in
the direction of underestimating emissions).   These biases would
have influenced all of  EPA's emission factor models up to and
including MOBILE4 (1989) .
  Emission factors from versions of the model prior  to MOBILE4
were even more seriously underestimated, since they did not
recognize the running and resting loss categories of evaporative
emission.

     The next section of this paper describes the results of on-
road and tunnel studies,  particularly the 1987 Van Nuys tunnel

-------
study, and the  implications for highway vehicle emission
estimates.
On-road and Tunnel Studies and Their  Implications

      In 1987,  the Coordinating Research Council  (CRC) sponsored
what has come  to be known as the  "Van Nuys tunnel study," in which
a highway tunnel in Van Nuys, CA was used to measure the total
emissions produced by all of the traffic passing through the
tunnel during  a total of 22 one-hour sampling periods.  By
measuring all  of the emissions exiting the mouth of the tunnel
(accounting for winds and total air  flow) and counting the number
and types of vehicles that passed through the tunnel during each
sample period, an estimate of gram per mile  (per vehicle)
emissions was  derived.

     The details of the design and execution of the tunnel study
are beyond the scope of the current discussion.  For additional
information on the study, the reader is referred to the.final
report "Measurement of On-Road Vehicle Emissions in the California
South Coast Air Basin, Volume I: Regulated Emissions."  [1]

     While not the original design purpose of the tunnel study,
one high-interest use of the emission factor estimates measured in
the study is "validation" of the emission factor estimates
produced by the models.  After fleet average gram per mile
emissions for each tunnel study sample period were calculated,
they were compared to emission factors modeled by the then-current
version of the California model, EMFAC7C.  These comparisons
reflected good agreement for NOx emissions, but suggested that HC
and CO emissions were being underpredicted by EMFAC7C.  Over the
range of conditions (average speed and temperature) occurring
during the tunnel study, the apparent underpredictions were by
ratios of between 1.4 and 6.9 for HC (average 3.8), and between
1.1 and 3.6 for CO (average 2.7).  The results of these
comparisons led many to question the accuracy of the EMFAC7C model
and its successors,  EMFAC7D and 7E.  Since EMFAC7C was based in
part on the EPA emission model (then MOBILES), the accuracy of the
emission factors calculated by MOBILES (and succeeding versions of
the model)  were called into question as well.

     Earlier tunnel studies and other roadside emission
measurements have also been interpreted as indicating that the
emission factor models may underpredict actual in-use emission
levels.  The Van Nuys study is focused on in this paper since it
was the most recent such study and it has received widespread
attention and publicity.

     Another source of information for use in evaluating the
apparent discrepancies between measured and modeled emission
factors is work that has been performed by ARE and EPA to compare
the ratios of different pollutants observed in ambient air to the

-------
ratios calculated  from emission inventories  (which are based  on
the modeled emission  factors).  [2,3]  These  comparisons  indicate
that the ambient ratios of HC to NOx, and of CO to NOx,  are higher
than the ratios calculated from the inventories.  Taken  together
with the reasonable agreement between measured and modeled NOx
emissions, this adds  weight to the argument  that modeled HC and CO
emissions are underpredicted.  Speciation profiles of ambient
volatile organic compound (VOC) emissions show an excess of those
species that arise from the evaporation of gasoline, suggesting
that evaporative emissions (non-exhaust emissions of all types)
are a likely contributor to the apparent underestimation of HC
emissions.  This is consistent with more recent emission data
obtained by EPA and resulting emission factor model revisions
since the tunnel study was conducted (particularly the addition of
running loss and resting loss emissions).

     The remainder of this section outlines  some of the  reasons
that may underlie the apparent discrepancies between the measured
(tunnel study) and modeled (EMFAC and MOBILE) emission factors.
These reasons are discussed in more detail in a 1991 EPA technical
memorandum. [4]  The  last section of this paper describes
activities currently underway, and additional ones being planned,
by EPA in continuing efforts to improve the accuracy of  the MOBILE
emission factor model.

     The widespread perception generated by  the CRC study final  ,
report that the emission factor models,  including EPA's MOBILE4,
underestimate in-use emission levels is somewhat stronger than
justified.  The oft-cited ratios of underprediction of HC and CO
emissions by the models are a range of 1.4 to 6.9 (average 3.8)
for HC and a range of 1.1 to 3.6 (average 2.7) for CO.  These
ratios,  as noted previously,  were based on modeling the  tunnel
study sample period conditions using the then-current version of
the California emission factor model EMFAC7E.  When the  same
comparisons were made using the EPA emission factor model MOBILE4
(1989),  both the average and the range of measured-to-modeled
ratios were reduced substantially,  although the indication that
the model underpredicts these emissions remained.  EPA's
reanalysis [4] shows that the discrepancies in measured-to-modeled
emission factors is reduced considerably (to average ratios of
measured to modeled emission factors of 1.7 for HC and 1.9 for CO)
when:

     (1) Emission sources now recognized and modeled by  both EPA
and ARB,  but not included in the version of EMFAC (EMFAC7C)  used
in the final tunnel study report,  are added to the modeled
emission factors.   The most important of these is running loss
emissions.

     (2) The traffic age mix is correctly accounted for.  The
emission factors produced by the model are weighted on the basis
of travel fractions for vehicles of different ages,  relecting the
relative proportion of vehicles of each age in the fleet as well
as the fact that older vehicles typically are driven fewer miles
                                -8

-------
 per  year  than  are newer vehicles.   In  a microscale  situation  such
 as the  tunnel  study,  emissions  from vehicles  of  different  ages
 should  be weighted  only by  their  fractional representation in the
 fleet;  differences  in annual mileage accumulation rates  are not
 applicable  to  such  situations.  The details of this adjustment are
 provided  in the EPA reanalysis.

      The  issues raised by attempts  to  use  the results  of on-road
 measurements and tunnel studies to  verify  the emission factors
 produced  by the models are  also more complex  than generally
 perceived.   The MOBILE and  EMFAC models were  designed  to estimate
 average emissions for broad areas  (such as an entire metropolitan
 urban area)  over relatively longer  time periods  (such  as a full
 day).   Caution is necessary in developing  appropriate  inputs  for
 the  models  to use in  estimating emission factors for micro-scale
 traffic situations, such  as those represented by on-road and
 tunnel  emission measurement studies,   A true  model  validation
 exercise  would have to be based on  actual  emission  measurements
 for  the broader geographic  and temporal scales the  models  were
 developed to represent, which poses  immense experimental
 difficulty.

      However, even  when the above is taken into  consideration, the
 tunnel  study does raise legitimate  concerns as to the  accuracy of
 the  models  in predicting  in-use emission levels.  The  known
 limitations of the EPA and ARE programs as they existed  in 1987, •
 discussed in the last section, add  credence to such concerns.
 Concerns  about the accuracy of future  year emission factor
 projections produced by the models  as  of 1987 (MOBILE3 and
 EMFAC7C)  are more speculative, since emissions from the  in-use
 vehicle fleets at that time were dominated by emission control
 systems no  longer in production, but are also justified.

      Given  that the EPA and ARE emission factor models appear to
 underpredict in-use HC and CO emissions from highway vehicles, the
 issues become (I)  what has EPA done  to improve the  collection of
 in-use  emission data and  its use in  modeling  emission  factors, and
 (2)  what  additional steps is EPA taking to further  improve its
 estimates.  These are discussed below.


Implemented and Planned Improvements in
Emission Factor Data Collection and the MOBILES Model

      EPA's modeling of in-use highway  motor vehicles emission
 factors is characterized by continuing efforts to better quantify
not only  the basic emissions performance of vehicles (i.e., at FTP
conditions,  how do in-use levels compare to certification
 standards),  but also the effects of  the many variables affecting
emissions.  The correction factors used to adjust basic  emission
rates for average speed,  temperature, and other variables  have
been revised repeatedly in response  to new information gathered in
the course of emission factor testing.  The doubts  raised by  the
                                -9

-------
tunnel study and other on-road emission measurements require
additional efforts to improve the accuracy of the emission factor
estimates produced by the model.

     One important recent change aimed at improving the accuracy
of EPA's emission factor estimates is the characterization of
running loss emissions and their incorporation into the model.
Running losses, which were not even considered in the MOBILES or
EMFAC7C models used in the CRC report's comparisons of measured
and modeled emission factors, constitute a considerable portion of
HC emissions from highway vehicles.  In some situations, such as
prolonged travel at high temperatures, the contribution of running
loss emissions exceeds that of exhaust emissions to the total HC
produced by highway vehicles.  EPA began running loss testing in
1988, and when running losses were added to the MOBILE4 model in
1989, the total HC emission estimates increased significantly.
The inclusion of these emissions is widely recognized as
appropriate and clearly helps to make the model's estimates more
closely reflective of real-world conditions.  Also, the category
of resting losses were recognized, quantified, and included for
the first time ever in the MOBILE4.1 model.

     While the characterization of running and resting loss
emissions and their inclusion in the model clearly represented a
major step in improving the accuracy of the model's estimates,
there remain a number of other areas of concern.  Issues that have
been raised concerning the accuracy of the model, programs that
are underway or planned for the near future in response to these
issues,  and how each of these actions can be expected to assist in
the goal of improving the accuracy of the emission factors
estimated by the model,  are briefly recounted below.


ISSUE:  Recruitment bias  in emission factor testing.

     Traditional recruitment programs are based on initial
contacts by mail and generally have not had high positive response
(participation) rates.   Are the vehicle samples developed by EPA's
emission factor testing programs really representative of the in-
use fleet?

Action toward resolution:

     EPA is now performing emission testing of vehicles at
centralized I/M program test stations in Hammond, IN and Phoenix,
AZ.  Since all vehicles must obtain these tests, the resulting
"lane" emission data has essentially no selection bias.  A
subsample of vehicles participating in the lane testing is then
recruited for additional laboratory testing.  The recruitment of
vehicles for laboratory testing has been much more successful
(higher participation rates)  at these lanes -- potential
participants recruited at the lanes have the immediate opportunity
to see their vehicles tested on a dynamometer and to see the
                                 10

-------
leaner vehicles provided for their use while EPA has their vehicle
for testing, decreasing reluctance to participate that may be
based in part on uncertainty about what will be done with their
vehicle during emission factor testing.

     The I/M lane  testing programs also allow  far more vehicles  to
be tested and added to the data base within time and resource
constraints, greatly increasing the number of vehicles represented
in the data base and thus increasing confidence in the resulting
estimates.  (In recent years, the "traditional" emission factor
program has tested between 500 and 1000 vehicles annually,
depending on the number and complexity of the tests performed on
each vehicle.  In the first two years of the Hammond program, data
has been obtained on nearly 9000 in-use vehicles, with over 100
vehicles per week being added to the data base.  The Phoenix
program is expected to test about 150 vehicles per week when fully
operational.)   The development of MOBILES emission factors based
on the results of these programs will be another significant
benefit of such programs.


ISSUE:   Are "super emitter" vehicles correctly represented in the  model?

     Within each model year, MOBILE4.1 assumes a distribution of
vehicles among four emitter categories:  normal, high,  very high,
and super emitters.  Normal emitters have hydrocarbon (HC)
emissions of no more than twice the standard, or 0.82 g/mi, and
carbon monoxide (CO) emissions of no more than three times the
standard, or 10.2 g/mi.  Super emitters are vehicles emitting at
least 10 g/mi HC or 150 g/mi CO.  The relative proportion of
vehicles in each of these emitter categories, the growth in the
fractions of "high, "very high," and "super" emitters over time,
and the emission rates of the "super" emitters are estimated on
the basis of emission factor program data,  the fraction of
vehicles modeled as super emitters and their modeled in-use
emission rates may not accurately reflect the occurrence and
behavior of these vehicles in the real world.

Action toward resolution:

     In the real world, vehicle emissions fall along a continuum
from low (normal)  levels to the extreme values characterized in
the model as "super" emitters.  The grouping of vehicles into four
emitter categories in the model, with vehicles slowly "migrating"
from the lower to the higher emitter categories with increasing
mileage, may underestimate emissions in two ways, by underesti-
mating the number of vehicles in the higher emitter categories or
by underestimating the in-use emission levels associated with
those categories.

     The idea of using remote sensing measurements to verify the
rate of occurrence and the emission levels of  "super" emitters has
generated considerable interest recently.  Remote sensing
                                -11

-------
researchers have obtained instantaneous emission  level
measurements from tens of thousands of  "real world" vehicles.
However, these measurements cannot be used to categorize cars as
"normal,"  "high," "very high," or "super" emitters since remote
sensing and laboratory FTP measurements do not display a good
correlation.  What could be done, in theory, would be to take
remote sensing measurements of all cars participating in the
emission factors testing program, including I/M lane cars, and to
compare the distribution of remote sensing readings from them with
the distribution from the large remote sensing test programs.

     To date, EPA has not obtained sufficient remote sensing
measurements on emission factor program test vehicles, including
the I/M lane samples, to make direct comparisons  of frequency or
of emission levels.  The Agency is considering whether and how to
implement  such measurements.  Logically, these remote measurements
should be made on roadways and under traffic conditions like those
used in the researchers' large-volume remote sensing studies, not
just at the exit of an emissions laboratory or I/M test lane.


ISSUE:   Representativeness of the FTP and other driving cycles.

     The Federal Test Procedure  (FTP), used to represent urban
traffic, and the other driving cycles used in emission testing
(i.e., cycles of differing average speeds used to develop speed
correction factors) used to represent other modes of driving
behavior, may not be representative of actual vehicle operation.

     The FTP was developed during the early 1970s in southern
California.  Traffic patterns and typical driving patterns have
changed significantly since then, calling into question the
utility of the FTP in characterizing typical urban driving
behavior.  Even if the FTP is still representative of current
urban driving in terms of such variables as operating mode,
average speed,  and average trip length, EPA is aware that the
maximum rate of acceleration contained in the FTP (3.4 mph/s) is
at times exceeded in real-world driving conditions and that
emission rates at higher accelerations can be much greater than
those at lower rates of acceleration.  The various speed cycles
used in the development of speed correction factors (SCFs), and
thus the SCFs themselves, may not be adequately representative of
the emissions behavior of vehicles at those speeds in real-world
conditions.

Action toward resolution:

     The issues of representativeness of the FTP  to characterize
typical urban driving,  the effects of "off-cycle" driving patterns
such as higher acceleration rates,  and the representativeness of
the various speed cycles used in emission factor  testing to actual
traffic at various average speeds are closely related.  EPA is
currently conducting programs involving both instrumented vehicles
                                -12-

-------
and chase cars to evaluate the degree to which the FTP is
representative of today's driving, as required by the 1990 CAA
Amendments.  While the results of these studies will not be
available for use in the model in the short term, in the long run
they will enable EPA to -develop new cycles for use in emission
factor testing, and may lead to revisions in the FTP itself.

     With respect to "off-cycle" emissions, limited emission
testing over cycles containing higher acceleration rates indicates
that emission rates during such hard accelerations are much
greater than are observed in the FTP'S accelerations.  EPA is
currently conducting a study to quantify both the emissions impact
of higher acceleration rates and the frequency of their
occurrence.  Results of this study will be used to improve
emission factor estimates.  EPA is continuing to test vehicles at
different speeds, and to evaluate the suitability of the SCFs used
in the model.  Since speed is one of the most critical
determinants of emission levels, improvements.in characterization
of in-use speeds and speed variability and in the methodology used
to convert speed-dependent emission data into speed correction
factors used in the model will improve the accuracy of the
emission factors.
ISSUE:  Actual freeway (limited access highway) speeds exceed 55 mph
in many areas.

     MOBILE4 and earlier versions of the EPA emission factor model
estimated emission factors only up to a maximum speed of 55 mph.
EPA guidance was to use the 55 mph emission factors for situations
where vehicle miles travelled  (VMT) were accumulated at higher
speeds.  The increase in the speed limit to 65 mph on many limited
access highways, and the frequent violation of the 55 mph limit in
areas where it applies, led to requests for emission factors at
higher speeds to be calculated by the model.

Action  toward resolution:

     MOBILE4.1 used data collected by the California Air Resources
Board from driving cycles with average speeds as high as 65 mph to
extend and revise the speed correction factors used in MOBILE4 and
previous versions of the model.  Where the earlier versions of the
model extrapolated the curve describing emissions as a function of
average speed to a maximum of 55 mph, and thus modeled a slight
decrease in emissions when speed was increased from 48 to 55 mph,
MOBILE4.1 speed correction factors apply to a maximum speed of 65
mph and model emissions as increasing with increasing average
speed from 48 to 65 mph.  Therefore emission factors and
inventories for limited access highways based on the use of
MOBILE4.1 will be higher than those based on MOBILE4.  EPA will
reconsider the speed corrections for "high" speeds (all average
speeds greater than 48 mph) using all available data in the
development of MOBILES.  EPA will also begin its own high-speed
                                 13

-------
emission test programs as soon as sound, representative  driving
cycles with higher average speeds are developed.


ISSUE:  The use of average trip speed to estimate emissions  as a
function of speed in the model.

     The speed  correction factors used  in  the  models  to  correct
emissions to average speeds other than  the FTP average speed  of
19.6 mph are derived from testing over  a series of driving
cycles, having  average speeds ranging from 2.5 to 65 mph.  Each
of these driving cycles represents a trip, in  that.each  cycle
begins and ends at idle and includes a  mix of  accelerations,
decelerations,  and driving at different speeds, such that the
total distance  of the cycle divided by  elapsed time gives the
average speed of the cycle.  Many users of the MOBILE model
apply the emission factors to individual or aggregated highway
links  (for example,  a given length of an arterial roadway, or
all arterial roads taken together), where emission factors based
on "cruising" at the given, speed would  be more appropriate than
the trip-based  average speed emission factors  produced by the
model.

Action  toward  resolution:

     The emission factors as a function of average speed
produced by MOBILE4.1 are the best that can be developed at
present, given  the driving cycles used  to develop the speed
correction factors and the available data.  When the results  of
the driving characterization studies now being performed become
available,  EPA will consider ways in which the emission  factors
can be refined to represent trips and highway  links more
accurately.  One possibility would be to generate driving cycles
of different average speeds for different roadway types,
reflecting differences in traffic congestion and other
conditions, then to develop correction  factors to the emission
factors that would take both average speed and roadway type into
account.


ISSUE:  Accuracy of the characterization of in-use emission deterioration
rates in the model.

     If vehicles overall exhibit significantly greater emission
deterioration (increases)  with increasing age  and/or accumulated
mileage than is modeled on the basis of currently available data
(based on FTP testing at different mileages),   then the basic
emission rate equations understate emissions for all but
relatively new vehicles.   There are two aspects to the issue  of
the  modeling of emission deterioration rates:  the problem of
extrapolating emissions as a function of mileage to very high
mileages (very few vehicles with extreme odometer mileages have
been tested in the traditional emission factor testing programs),
                                 14

-------
and separation of the emission effects of vehicle acre and vehicle
mileage.

Action toward resolution:

     The Hammond and Phoenix  I/M  lane testing programs mentioned
above include vehicles up to  fifteen years old, with real in-use
accumulated mileages.  Thus these programs are providing EPA with
data on vehicles that are both older and having higher accumulated
mileages than have generally  been included in the emission factor
testing programs.  The results of these test programs will be used
to increase the confidence in deterioration rates used in the
models, resulting in more realistic basic emission rate equations
and hence modeled emission factors.  EPA will also be utilizing
some data collected by the California ARE, which has generally
tested more older cars than EPA.


ISSUE:  Evaporative emissions,  particularly diurnal emissions,  may not
be the same under test conditions as in real-world experience.

     The diurnal portion of the evaporative emission test includes
a temperature increase, in the vehicle fuel tank only, of 60° to
84°F over one hour.   This temperature increase would typically
occur over nine to twelve hours outdoors and would affect the
entire vehicle, not just the  fuel tank; hence the resulting
emissions may not be the same.  EPA in the past has only measured
hot soak emissions for one hour after the end of operation,  while
such emissions do not just stop occurring after one hour.

Action toward  resolution:

     As required by the 1990  Amendments to the Clean Air Act, EPA
is revising the evaporative emissions test procedure used for new
vehicle certification to more accurately reflect real-world
conditions.  This will result in better control of evaporative
emissions under a wider range of in-use conditions not reflected
in the current evap test procedure.  To more accurately measure
evaporative emissions from in-use vehicles certified under the
current procedures,  EPA has very recently implemented changes in
the evaporative test procedures used for emission factor testing
to include more realistic conditions, including "real-time"
diurnal tests and measurements of hot soak emissions in the second
and third hours following engine operation, and functional purge
and pressure testing (see next item) .  Indications are that higher
emissions will be observed.   Results from such testing will be
used to revise existing evaporative emission factors to better
represent actual in-use conditions.
                                 15

-------
ISSUE:  Many vehicles, particularly those more than five years old, have
functional problems with  evaporative emission control systems.

     Recent  data  have shown.that  many vehicles,  particularly those
greater than five years  of  age, demonstrate  functional  problems
with their evaporative  emission control  systems.   Such  problems
fall into two categories,  "purge"  and "pressure"  failures.
Vehicles with either or  both of these functional  problems will
exhibit much higher levels  of evaporative emissions.


Action  toward resolution:

     As part of the Hammond and Phoenix  programs  already noted,
EPA is conducting functional pressure and purge tests of vehicle
evaporative  control systems.  This provides  data  on both the rate
of failure of these tests observed in a  large  random sample of
vehicles and on the emissions impact  of  vehicles  exhibiting these
problems.  Incorporation of this  information in the MOBILE model
will increase and improve'the accuracy of the  estimates-of
evaporative  emissions.   As  noted  earlier, examination of ambient
and modeled  emission ratios seems  to  suggest that evaporative
emissions are not adequately accounted for;  this  work will assist
in better quantifying the relative and absolute contribution of
evaporative  emissions to overall  HC emissions  from motor vehicles.


ISSUE:  Commercial and test  fuels differ in  properties (volatility, sulfur
level) that have an impact on vehicle emission  levels.

     The fuel used in certification and  standard  emission factor
testing (excluding tests designed to  characterize the impact of
fuel parameters,  such as volatility)  is  blended to a specified
formula.  Commercial fuels differ  from this  test  fuel and in
different areas of the country and at different times,  and fuel
parameters other  than volatility  affect  emissions.

Action  toward resolution:

     MOBILE4 incorporated fuel volatility as measured by Reid
vapor pressure  (RVP)  as  a user input,  and adjusted emission
factors to account for volatility effects.   As with other
correction factors (e.g., average speed, temperature),  these
corrections will  be updated as more data are available  for
analysis.   EPA will also adjust upward the base  (9.0 psi RVP)
emission factors  to reflect the effects  of other  fuel parameters,
such as sulfur content, based on a large body  of  recent testing
performed as part of the Auto/Oil Air Quality  Improvement Research
Program.
                                • 16

-------
ISSUE:  Accuracy of vehicle miles travelled (VMT) estimates and traffic
modeling; potential use of operating hours  rather than VMT as the
activity level for highway mobile sources.

     While not  related  to  emission  factor modeling per se,  this
issue is  important to the  development of accurate highway mobile
source emission inventories.  The accuracy of VMT estimates has a
directly proportional effect on the accuracy  of mobile  source
emission  inventories  (i.e., a 20 percent underestimate  of VMT
results in emissions being underestimated by  20 percent).
Interest has been expressed in the idea of using vehicle operating
hours, rather than VMT, as the activity level  for highway mobile
sources on the grounds  that operating hours might be more readily
and objectively measured.

Action toward  resolution:

     The  development of accurate VMT estimates is a local/State
responsibility.  EPA is working with the U. S. Department of
Transportation  (and other  parties) to develop and demonstrate
better methods  for estimating VMT and modeling traffic  flow.  EPA
will require States to  tie their estimates of VMT to actual
observations of traffic volumes in a more comprehensive way than
has been required in the past, which is intended to improve the
accuracy of these estimates.  The use of traffic counts (such as
in the Highway Performance Modeling System) as part of  this
process is an important component of efforts  to improve the
accuracy of this aspect of the inventory development process.

     EPA's Office of Research and Development (ORD) is
investigating the potential for using vehicle operating hours as
the activity level for highway mobile sources, as a substitute for
VMT.  This is considered a long-range research effort,  as EPA does
not believe that States or others are in any  better position to
accurately measure operating hours than to estimate VMT at this
time.  The suggestion has been made that emissions as a function
of time of operation would be less sensitive  to average speed and
driving cycle; if proven true, this would reduce the importance of
uncertainties in these areas.  The Office of  Mobile Sources does
not believe that this will prove to be the case, but will continue
to examine new information as it becomes available.

     EPA  is also involved  in other efforts to improve  emission
factors and VMT estimates  that do not fit neatly into  the
problem/action format of the preceding paragraphs.  These efforts
include our active participation in the Southern Oxidant Study,
which will include another tunnel study  (the  Fort McHenry tunnel
in Baltimore,  MD).  EPA is also working to improve and  enhance our
cooperation and coordination with efforts being undertaken by the
California Air Resources Board to improve emission factor and
activity level estimates for use in developing accurate emission
inventories.
                                -17-

-------
     The  list above, while not  exhaustive,  indicates the most
important of the efforts underway and planned to improve the
estimates produced by the emission factor model.


Conclusion

     Over the last 15 years, EPA has spent much time and effort to
develop a model that can provide accurate emission factor
estimates for highway vehicles  over a broad range of conditions.
While confident that each new version of the model (MOBILE1, 1978;
MOBILE2, 1981; MOBILE3,  1984; MOBILE4, 1989; MOBILE4.1, 1991; and
MOBILES, to be released later in 1992) is a significant
improvement over the preceding  version, EPA acknowledges that much
work remains to be done to improve the accuracy of the emission
factors produced by the model.  As described above, EPA is active
on a number of projects that will enable improvements in the
emission factor estimates for highway vehicles.

     The MOBILE models are capable of estimating emission factors
over a wide range of conditions specified by the model user, and
many of the built-in assumptions of the model can be altered by
the modeler.  Careful consideration of the situation for which
emission factors are required and how that situation can be
reflected in the emission factor modeling is essential to the
creation of the most accurate possible emission factors from the
model.
                                 18

-------
                           References
1.    Ingalls, Melvin N.,  et al.,  "Measurement of On-Road Vehicle
     Emission Factors in the California South Coast Air Basin,
     Volume I: Regulated Emissions," Final Report,  June 1989.

2.    Fujita, Eric M.,  et al.,  "Comparison of Emission and Ambient
     Concentration Ratios of CO,  NOx,  and NMOG in California's
     South Coast Air Basin,"  January 16,  1992 (submitted for
     publication to the Journal of the Air and Waste Management
     Association).

3.    Baugues, Keith, "Further Comparisons of Ambient and Emission
     Inventory NMOC/NOx Ratios,"  for presentation at the 85th
     Annual Meeting and Exhibition of the Air and Waste Management
     Association in June 1992.

4.    "On-road Emissions Measurements in the South Coast Air Basin
     ("Tunnel Study")," EPA Technical Memorandum from Terry P.
     Newell to Charles L. Gray,  Jr., August 22,  1991.
                                -19-

-------