United States Air and Radiation EPA420-P-99-016
Environmental Protection March 1999
Agency M6.EXH.008
Overview of
Methodology for
Tier 0 In-Use
Deterioration and
Key Issues for Comment
DRAFT
> Printed on Recycled Paper
-------
EPA420-P-99-016
March 1999
of for 0
for
M6.EXH.008
DRAFT
Penny Carey
Phil Lorang
Assessment and Modeling Division
Office of Mobile Sources
U.S. Environmental Protection Agency
NOTICE
This technical, report does not necessarily represent final EPA decisions or positions.
It is intended to present technical analysis of issues using data which are currently available.
The purpose in the release of such reports is to facilitate the exchange of
technical information and to inform the public of technical developments which
may form the basis for a final EPA decision, position, or regulatory action.
-------
EPA420-P-99-016
- Draft -
Overview of Methodology for Tier 0 In-Use Deterioration and Key
Issues for Comment
Report Number M6.EXH.008
April 29,1999
Penny Carey
Phil Lorang
U.S. EPA Assessment and Modeling Division
NOTICE
This technical report does not necessarily represent final EPA decisions or positions. It is intended to present technical
analysis of issues using data which are currently available. The purpose in the release of such reports is to facilitate the
exchange of technical information and to inform the public of technical developments which may form the basis for a
final EPA decision, position, or regulatory action.
M6.EXH.008 DRAFT 4/29/99
-------
Introduction
There are four draft papers that document the proposed findings on the in-use exhaust
deterioration of 1981-1993 model year (Tier 0) cars and trucks:
- M6.STE.002, "The Determination of Hot Running Emissions from FTP Bag Emissions"
- M6.STE.003, "Determination of Start Emissions as a Function of Mileage and Soak Time
for 1981-1993 Model Year Light-Duty Vehicles"
- M6.EXH.001, "Determination of Running Emissions as a Function of Mileage for 1981-
1993 Model Year Light-Duty Cars and Trucks"
- M6.EXH.002, "Analysis of Emissions Deterioration Using Ohio and Wisconsin IM240
Data"
This brief document provides an overview of the basic methodology being proposed for
comment (with references to the appropriate papers), describes the uncertainties/key issues for
comment, and compares the results to estimates obtained with MOBILES.
As background, EPA is proposing that MOBILE6 will separate vehicle exhaust emissions
into start emissions and running emissions. The papers described here provide estimates of start
emissions as a function of mileage and running emissions as a function of mileage. These are
basic emission rates; adjustments for off-cycle driving and other effects are described in other
documents.
Peer review comments on these draft papers were solicited and received from three
reviewers. A summary of the peer review comments is also available. Comparisons of the
proposed findings to other available data sets will also be made available to assist public and
stakeholder reviewers.
Methodology for Running Emissions
For running emissions, the following steps were followed:
1) Available FTP data were combined (described in M6.EXH.001 and M6.STE.003).
2) The FTP data were split into start and running portions using equations documented in
M6.STE.002.
3) Estimates of average FTP-based running emissions (under standardized laboratory
conditions) as a function of mileage were developed (described in M6.EXH.001).
4) Since these estimates of average running emissions are based on FTP tests of vehicles
obtained from public vehicle recruitment programs, there is some concern that the low
vehicle recruitment acceptance rates (typically less than 25%) in these programs may
introduce recruitment bias. In addition, because of the dates of testing, the combined sample
of vehicles had a restricted range of accumulated mileages. To address this, IM240 data from
Dayton, Ohio were examined. The Dayton data are described in detail in M6.EXH.002.
M6.EXH.008 -2- DRAFT 4/29/99
-------
5) In order to convert the Dayton IM240 data to running emissions, it was necessary to convert
the fast-pass results to full EVI240s, using equations documented in M6.EXH.002. The
predicted full IM240s (under I/M lane conditions) were then converted to estimated running
emissions (under standardized laboratory conditions) using equations documented in
M6.EXH.002.
6) The IM240-based running estimates were compared to the FTP-based estimates. Based on
this comparison, adjustment factors were developed and applied to the FTP-based estimates
(described in M6.EXH.001). In most cases, the adjustment factors are positive, i.e., they
increase the estimate of in-use emissions, particularly, emissions due to deterioration.
7) Using the final adjusted estimates of fleet average emissions at each mileage, the fractions of
normal and high emitters as a function of mileage/age were back-calculated. The normal and
high emitter fractions are used to estimate the start emission rates, and are provided in
M6.STE.003 (see start methodology below). This is done to provide consistency between the
start and running emission estimates.
Methodology for Start Emissions
For start emissions, the following steps were followed (all described in M6.STE.003):
1) Available FTP data were combined. These are the same FTP data sets used for running
emissions.
2) The FTP data were split into start and running portions using equations documented in
M6.STE.002. Again, this is similar to what was done for running emissions.
3) The start sample was separated into normal and high emitters.
4) The normal and high emitter samples were recombined using the mileage-dependent
fractions of normal and high emitters for running emissions (described in step 7 above).
Estimates of basic start emissions as a function of mileage were then developed.
5) A relationship between start emissions and soak time was also developed.
Uncertainties/Assumptions
There are a number of uncertainties/assumptions associated with this analysis. These are
key issues and are provided below, along with EPA's current view of each. By way of this
document, EPA is soliciting comments and/or information pertaining to these issues.
Issue 1: The model year groupings and segregation of fuel system types were based primarily on
engineering judgment.
EPA's Current View: This reflects a belief, shared by the automobile industry, that vehicles in
different groups are of sufficiently different design that the emission performance of one group
cannot be attributed to any other.
M6.EXH.008 -3- DRAFT 4/29/99
-------
Issue 2: The representativeness of the FTP data sample. The vehicles in the FTP data set were
generally 1-5 years old at the time of testing, making age-related analyses difficult. Some of the
high mileage data, particularly for the newer vehicles, come from vehicles specifically recruited
to meet a mileage target, and thus may not be representative of all vehicles of their model year
and age. Sample sizes are limited. Many of the vehicles were not recruited and tested under EPA
supervision. Also uncertain is whether the mail-based public recruitment used in FTP studies
introduces significant bias.
EPA's Current View: FTP tests are believed to provide the highest quality exhaust emissions
data because vehicles are preconditioned in a consistent manner and the test is designed to
simulate real world driving. The mail-based public recruitment may introduce bias, but the
magnitude and even direction cannot be determined reliably without comparison to another data
source. Any bias would be less strong at lower mileages, since all significant repairs required at
these low mileages would be covered under warranty.
Issue 3: The FTP data are not considered to represent the fleet, and need to be adjusted based on
IM240 data.
EPA's Current View: The disparity between emissions from the FTP and Dayton IM240 data
sets is quite evident, even for comparable mileages. At lower mileages, the adjustment was less
(trending towards no adjustment at zero miles) and at higher mileages the adjustment grew
linearly. This appears consistent with the theory that the FTP data may be subject to a
recruitment bias and, therefore, underrepresent the number of high emitting vehicles relative to
the real fleet (as represented by the observations in Dayton).
Issue 4: The separation of FTP measurements into start and running emissions are based on
results from 76 cars of limited model year range, which are assumed to be representative of the
national fleet.
EPA's Current View: There was no better alternative approach, because data to support another
approach did not exist.
Issue 5: The representativeness of the EVI240 data sample. The IM240 data were collected over a
period of just one year, representing a snapshot in time for the fleet, and making age-related
analyses difficult. Fuel, vehicle preconditioning, and ambient temperature are uncontrolled. The
Dayton sample is assumed to be sufficiently representative of the national fleet. The odometer
readings were judged to be unreliable, making it necessary to substitute regional-specific annual
average odometer readings by vehicle age for the reported odometers.
EPA's Current View: The Dayton sample was selected because it was the only area performing
IM240 tests, while not having previously been subject to any emissions inspection program.
Since Dayton is a biennial program, this sample contains about one-half the cars in the Dayton
I/M area. The sample size is thus quite large, relative to the FTP sample. Also, since the program
is mandatory, the data set should not be subject to significant recruitment bias. Since the
M6.EXH.008 -4- DRAFT 4/29/99
-------
comparison to the FTP data was done within each model year group by comparing mean
emissions corresponding to mean odometer values, the regional odometer estimates were thought
to adequately represent the mean odometer values for the Dayton sample.
Issue 6: The translations of the Dayton fast pass IM240 measurements to full EVI240, and from
full IM240 (under I/M lane conditions) to running emissions (under standardized lab conditions),
are based on data obtained in other state and EPA programs from several years earlier.
EPA's Current View: The translation from fast pass measurements to full IM240 required
second-by-second measurements from a program that performed random full IM240s. Wisconsin
IM240 data were chosen over data from the other two IM240 states with second-by-second data
because of the geographic, demographic, and meteorological similarities between Ohio and
Wisconsin. Alternative statistical models developed by others using Colorado and Arizona
IM240 data give similar results.
The translation from full EVI240 under I/M lane conditions to running LA4 emissions under
standard laboratory conditions required data from vehicles that were tested on both the FTP and
lane EVI240 cycles. The "Indiana" EPA data set used for this analysis was the sole data set that
met this requirement.
The "Indiana" data set allows the adjustment from lane EVI240 to lab running LA4 emissions to
be separated into two parts, one for the driving cycle difference and one for all other factors
which would include the fuel, temperature, and prior operation for the two testing situations. We
observed that these non-cycle factors did have a substantial effect on measured emissions.
Vehicles were considerably cleaner under the laboratory conditions than under the I/M lane
conditions. From a variety of indications, we believe that much of this difference is due in large
part to the unusual vehicle operation just before many IM240 tests, namely the long period of
waiting. This waiting period can accentuate the influence of fuel volatility and/or cause catalyst
cool down, which can affect the measured emissions of all cars both passing and failing. Most
cars receive their IM240 test with no preconditioning to reverse these prior-operation effects. If
we had used the Ohio EVI240 data without adjusting for the differences, the estimate of in-use
running emissions would have been higher, but not appropriately so. The long waiting-in-line
periods at IM240 lanes are not representative of normal driving. Of course, lab conditions are
also not representative of typical driving. To make the final estimates reflect normal driving,
MOBILE6 will apply correction factors for fuel and temperature to the lab running LA4
estimates.
The issue of prior operation and lack of preconditioning on all cars is a serious complication in
interpreting IM240 data. EPA invites comments on how it can be best addressed. EPA would
also welcome a large data set from IM240 testing in which all cars are preconditioned once on
the EVI240 before the actual EVI240 measurement.
M6.EXH.008 -5- DRAFT 4/29/99
-------
Issue 7: The manufacturer mix is not necessarily representative of the national mix, particularly
for 1988 and newer model years.
EPA's Current View: This pertains to the FTP data obtained by the domestic manufacturers. The
vehicles were roughly 2-3 years of age at the time of testing. It is the largest single source of FTP
data. The Ohio IM240 data has a better mix of imported and domestic models.
Issue 8: Certain data sets (in particular, the Colorado IM240 data) were not used for the analysis.
EPA's Current View: EPA obtained the Colorado IM240 data from the Colorado Department of
Public Health and the Environment and used it to validate the MOBILE6 predictions obtained
using the Dayton IM240 data. Specifically, the Colorado data were used in place of the Dayton
data to calculate the running emissions estimates. The results indicate that the running emissions
are somewhat higher when Colorado data are used. When combined with start emission estimates
to obtain composite FTP values, the effect of using Colorado data in place of the Dayton data is
considerably diminished. The MOBILE6 emission deterioration rates are still much smaller than
those predicted by MOBILES, regardless of which state's IM240 data are used. This work and
comparisons to other available data sets are described in M6.EXH.010, which is also being made
available for public and stakeholder review. The Colorado data set is subject to issues of prior
operation and lack of 100% preconditioning, discussed above in item 6.
Issue 9: The effect of road grade has not been considered.
EPA's Current View: Road grade may have a significant effect on emissions and there may be
an interaction between in-use deterioration and emissions under grade, but data and time were
not available to address this issue for MOBILE6. It will be considered for later analysis.
Issue 10: The MOBILE model has historically predicted lower emissions than those estimated
from tunnel and ambient studies. In particular, MOBILE predicts lower NMHC/NOx ratios than
those observed in the tunnel and ambient studies. The lower in-use exhaust deterioration
estimates in MOBILE6 would appear to increase, rather than decrease, this discrepancy.
EPA's Current View: This is a serious and complex issue which EPA has considered, and about
which there should be continuing discussion and research. As long as there are apparent
discrepancies between models like MOBILE and tunnel/ambient field studies, air quality
planners and decision makers should keep the possible underestimation of emissions in mind as
they decide on control programs.
It must be recognized that tunnel and ambient studies are not foolproof, and by their nature are
not direct and irrefutable tests of the accuracy of a model like MOBILE. Beyond that, there are
some specific reasons why EPA has not proposed to apply any adjustment to the emission rates
based on comparisons to tunnel and ambient studies. Tunnel and ambient studies can indicate
whether there may be a mismatch in estimates of fleet emissions, but they have not identified
how the mismatch may depend on the model year of the vehicles or on exhaust versus
M6.EXH.008 -6- DRAFT 4/29/99
-------
evaporative emissions. It is plausible that there are different errors in MOBILES emission
estimates for different model years, and for evaporative versus exhaust emissions.
The studies in question were generally done in the early to mid 1990's. Some were done in
California, which is known for the high average age of its fleet. In that time frame, a very large
fraction of modeled fleet emissions is estimated to come from cars from the mid-1980's and
earlier. In addition, a large portion of NMHC emissions were due to evaporative emissions.
Therefore, errors in the estimates for these emission categories could explain much of the
apparent discrepancies between models and tunnel/ambient studies. In particular, the failure to
recognize (until MOBILE6) that these early model years suffered from high emissions during
typical driving is a partial explanation that is unrelated to in-use deterioration of later model year
cars.
For MOBILE6, we are proposing downward revisions of the in-use deterioration rates primarily
for cars sold after the mid-1980's. For many of the available tunnel/ambient studies, the time
frame of the study was such that some of these model years were still young enough that even
MOBILES predicted them to have low in-use deterioration, i.e., they were still before their
"kink." Therefore, EPA does not presently consider the discrepancies with field studies to be a
decisive indictment of the revisions being proposed here for comment, but continues to welcome
views and data analyses from others.
Comparison of MOBILES and Proposed MOBILE6 Emission Factors
In order to compare the MOBILE6 estimates with MOBILES, it is necessary to combine
the start and running estimates to calculate composite FTP emissions. In general, the MOBILE6
estimates of FTP exhaust emissions as a function of mileage for Tier 0 vehicles fall substantially
below those predicted by MOBILES. To illustrate, Figures 1, 2, and 3 provide a comparison of
the MOBILES and proposed MOBILE6 HC emission factors for 1992, 1987, and 1981 model
year cars, respectively.
M6.EXH.008 -7- DRAFT 4/29/99
-------
Figure 1
FTP COMPARISON OF MOBILES and PROPOSED MOBILES HC EMISSION FACTORS
1992 MODEL YEAR CARS
HC (g/mi)
2
1-
0
50 100
MILES (X1000)
150
1992 Model Year (Mostly Fuel-Injected) Vehicles
As part of its overall upgrade to MOBILE6, QMS has been studying in-use deterioration. Real world information
became available from states who implemented mass emissions I/M programs. That information coupled with
traditional EPA FTP data and information received from other sources such as AAMA, API, and state I/M programs
was used to calculate the in-use deterioration data for MOBILE6.
As shown in the above graph, rather than beginning to deteriorate more quickly at the "kink" of 50,000 miles as
previously projected, newer model vehicles are not only cleaner when new, they stay cleaner longer. In 1990 and
later models, port fuel injected systems began to dominate the industry, also contributing to cleaner vehicles.
For brevity, not all model years between 1981 and 1992 are shown in these figures. The MOBILES estimates for
those model years were all very similar. The draft MOBILE6 estimates show a progressive improvement across
time, reflecting the transition to fuel-injection and the manufacturers increasing success at compliance in actual use.
M6.EXH.008
DRAFT 4/29/99
-------
Figure 2
FTP COMPARISON OF MOBILES and PROPOSED MOBILES HC EMISSION FACTORS
1987 MODEL YEAR CARS
HC (g/mi)
2
1-
3
50
100
150
MILES (X1000)
1987 Model Year
Using MOBILES, projections indicated that after 50,000 miles, MY87 vehicles would deteriorate or
become progressively dirtier over the next 100,000 miles. By 150,000 miles, their projected contribution to
the emission inventory accounted for approximately 3.2 g/mi HC.
MOBILE6 predicts that after 50,000 miles the MY87 vehicles still become progressively dirtier, but the
deterioration is not as dramatic as originally thought. Due to emerging emission control technology and
early-model fuel injection systems, the exhaust emitted by light duty cars and light duty trucks during
actual use is less than what was predicted in MOBILES.
M6.EXH.008
-9-
DRAFT 4/29/99
-------
Figure 3
FTP COMPARISON OF MOBILES and PROPOSED MOBILES HC EMISSION FACTORS
1981 MODEL YEAR CARS
HC (g/mi)
2
50
^ I
100
3
2
150
MILES (X1000)
1981 Model Year (Carbureted^) Vehicles
As this graph illustrates, MOBILES projections indicate that the rate of in-use deterioration for 1981
vehicles rose most dramatically at 50,000 miles. MOBILE6 also predicts significant vehicle deterioration
but at a much steadier rate over the life of the vehicle.
M6.EXH.008
-10-
DRAFT 4/29/99
------- |