United States        Air and Radiation       EPA420-R-02-002
          Environmental Protection                 January 2002
          Agency                       M6.EXH.001
&EPA    Determination of
          Running Emissions as a
          Function of Mileage for
          1981-1993 Model Year
          Light-Duty Cars and
          Trucks
                                 $5b Printed on Recycled
                                 Paper

-------
                                                             EPA420-R-02-002
                                                                 January 2002
                    of Running                 as a Function of
                         for
                 Light-Duty               Trucks

                             WI6.EXH.001
                                 Phil Enns
                                 Ed Glover
                                Penny Carey
                       Assessment and Modeling Division
                     Office of Transportation and Air Quality
                     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 that 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.

-------
1.0    INTRODUCTION

       The MOBILE6 emissions inventory model will allocate vehicle exhaust emissions
between engine start (start emissions) and travel (running emissions).  This split allows the
separate characterization of start and running emissions for correction factors such as fuel
effects and ambient temperature. It also enables a more precise weighting of these two aspects
of exhaust emissions for particular situations such as morning commute, parking lot and freeway
driving. This document describes the methodology used to calculate the in-use deterioration of
running emissions and presents estimates for model year 1981-1993 light-duty cars and trucks
proposed for use in MOBILE6. The deterioration of start emissions is addressed in a separate
document.1

       Section 2 describes the Federal Test Procedure (FTP) data sources and the model year
and technology groups used.   Section 3 presents the methodology for calculating running
emissions from FTP bag data.  It contains a basic overview of the FTP, defines all of the
applicable emission terms, and provides the calculations for determining the base unit of engine
running emissions.  Section 4 describes models and results for the in-use  deterioration of
running emissions as a function of mileage. Section 5 reports on high emitter correction factors
which are applied to the deterioration estimates. Section 6 displays the final results in tabular
form.
2.0    FTP DATA SOURCES USED

       FTP datasets were used to determine in-use deterioration. The FTP-based emission
estimates were then adjusted by applying high emitter correction factors derived using Ohio
EVI240 data. This section describes the FTP data sources used. Three FTP data sources were
used: (1) the test results from the EPA laboratory in Ann Arbor, Michigan; (2) data received
from the American Automobile Manufacturers Association (AAMA) based on testing conducted
in Michigan and Arizona; and (3) American Petroleum Institute (API) data collected in Arizona.
Model years range from 1981 through 1993, and vehicles include both cars and trucks.  Table
1 gives a  breakdown  for the light duty vehicle sample  by vehicle type, model year,  and
technology for the three datasets combined.

       Most of the 1990 and later model year vehicle data were supplied by AAMA, while most
of the pre-1990  data came from EPA laboratory testing. The API sample is a relatively small
sample (99 cars and trucks). Its chief appeal is that the vehicles have generally higher mileage
readings (all over 100,000 miles) than the  rest of the sample. There is a general trend from
carbureted and open loop technologies in early model years to fuel injection in more recent
Clover, E. and P. Carey, "Determination of Start Emissions as a Function of Mileage and
Soak Time for 1981-1993 Model Year Light-Duty Vehicles," Report No. M6.STE.003,
October, 1998.

                                        -2-

-------
years. Port fuel injected vehicles dominate in 1990 and later model years. Although not
explicitly shown in the table, new catalyst technology was phased slowly into the fleet starting
in the mid 1980's.

       For this analysis,  cars and trucks were each classified  into the following model
year/technology groups:

                         MY Group / Technology Type - Cars

                           1988-93 Port Fuel Injection (PFI)
                        1988-93 Throttle Body Injection (TBI)
                         1983-87 Fuel Injection (PFI plus TBI)
                     1986-93 Closed Loop Carbureted/Open Loop
                            1983-85 CL Carb/Open Loop
                              1981-82 FI (PFI plus TBI)
                            1981-82 CL Carb/Open Loop

                        MY Group / Technology Type - Trucks

                           1988-93 Port Fuel Injection (PFI)
                        1988-93 Throttle Body Injection (TBI)
                              1981-87 FI (PFI plus TBI)
                     1984-93 Closed Loop Carbureted/Open Loop
                            1981-83 CL Carb/Open Loop
       These groupings were selected on the basis of changes in emission standards or the
development/refinement of new fuel metering or catalyst technologies. Because of the relatively
large amount of 1988-93 fuel inj ected data, this category was split into PFI technology and TBI
technology for both cars and trucks. This produces separate deterioration functions based on
these fuel delivery technologies and allows the modeling of the future penetration of PFI
technology into the in-use fleet.
              3.0DETERMINATION OF RUNNING LA4 EMISSIONS

                  3.1 Overview of the Federal Test Procedure (FTP)

       The Federal Test Procedure (FTP) is a test cycle which is used to certify new vehicles
to emission performance standards.2 The FTP consists of a cold start segment (Bag 1), a hot
stabilized segment (Bag 2), and a hot start segment (Bag 3). Initially, the vehicle is stored for
   240 CFRPart 86, Subpart B, Section 86.144

                                        -3-

-------
a minimum of 12 hours before testing to simulate a 12 hour overnight soak period.  It is then
driven over the cold start segment, which lasts 505 seconds over a length of 3.59 miles, and the
emissions collected as Bag 1. Bag 2 emissions are then immediately collected from the hot
stabilized segment, which lasts 867 seconds over a length of 3.91 miles.  After a 10 minute soak,
the 505 seconds of the start segment is repeated and the emissions are  collected as Bag 3.

       The FTP composite emission rate is a weighted combination of the three measured bags
designed to represent two trips. The first trip is a cold start after a 12 hour soak, and the other
is a hot start after a 10 minute soak. Each trip is a "LA4" cycle, which  is a combination of the
505 cycle (either Bag 1  or Bag 3) and the Bag 2 cycle.  In a typical FTP test, the Bag 2 is only
measured once and the results are used for both trips. Since the 505 cycle is 3.59 miles long and
the Bag 2 cycle is 3.91  miles long, each LA4 trip is 7.5 miles long. Based on findings about
driving activity from the original FTP study, the cold start trip is weighted 43% and the hot start
trip weighted 57%. Hence the fraction of vehicle miles traveled (VMT) in Bag 1  (containing
the cold start) is:

FTP Bag 1 VMT Weighting = 43%*(3.59 miles / 7.5 miles) = 0.206

Similarly, since 57% of trips involve a hot start, the VMT weighting for Bag 3 (containing the
hot start) is:

FTP Bag 3 VMT Weighting = 57%*(3.59 miles / 7.5 miles) = 0.273

The remaining VMT represents stabilized driving (Bag 2). Since it is used for both the cold
start and  hot start trips, its VMT weighting is computed from  both:

FTP Bag 2 VMT Weighting = (43% + 57%)*(3.91 miles / 7.5 miles) = 0.521

Thus, the standard VMT weighting of the bags reported in grams per  mile (g/mi) for the full
FTP is:

FTP = (Bag 1*0.206) + (Bag 2*0.521) + (Bag 3*0.273)

where the fractions represent the proportion of vehicle miles traveled within the three modes
during the FTP trip in grams  per mile.
3.2    Overview of the Hot Running 505 and Its Use

       The FTP testing method outlined above does not allow the precise separation of start and
running emissions, since Bags 1 and 3 contain both start and running emissions.  Bag 2 of the
FTP does not contain an engine start;  however, the driving cycle used in the second bag is
significantly different from the cycle used for Bags 1 and 3. Thus, to estimate the amount of
                                         -4-

-------
    FTP emissions that can be allocated to engine start and running emissions, the concept of the
    Hot Running 505 (HR505) must be introduced.

           The HR505 refers to emissions measured from a driving test performed on the 505-
    second cycle of FTP Bags 1 and 3 without an engine start.3 Appending the FIR505 cycle to a
    standard three-bag FTP produces values that can be used to estimate the portions of Bags 1 and
    3 attributable to start emissions following a 12 hour soak and start emissions following a 10
    minute soak, respectively.

           Since the HR505 has not historically been included in FTP test programs, a method of
    estimating the FIR505 from FTP bag data was developed using data from a special test program.
    Briefly, HR505 emissions were measured in a sample of 77 cars and trucks tested under EPA
    contract.  The results from this sample were used to develop a correlation between the FIR505
    and FTP bag data.  This correlation was then used to estimate HR505 results for the larger FTP
    dataset used in this analysis.
    3.3    Basic Running LA4 Emission Rate

           The LA4 refers to a cycle comprised of the 505-second driving cycle used for Bags 1 and
    3 of the FTP and the 867-second cycle of Bag 2. Running LA4  emissions are defined as
    emissions from this 1372-second cycle with no engine start. For the MOBILE6 separation of
    start and running emissions, the running LA4 represents the running portion. For a given three-
    bag FTP, running LA4 emissions can be estimated using a VMT-weighted combination of the
    FIR505 and the Bag 2 emissions (stabilized operation). This estimate contains all of the driving
    behavior in the LA4 cycle, without engine starts. Mathematically, it is given by:

              Running LA4
               Emissions     = (HR505*(0.206+0.273)) + (Bag 2 * 0.521)
              (grams/mile)

    where 0.206, 0.273, and 0.521 are the VMT weightings for Bags 1, 3, and 2,  respectively.

    Like the FTP, running LA4 emissions are measured in units of grams per mile. This estimate
    is proposed for use in MOBILE6 as the basic exhaust emission rate from which all other running
    exhaust emission estimates are derived.

           Using the methods described in this  section, all emissions measured using the FTP and
    reported by bag can be allocated to start or running emissions before analysis. Average running
    LA4 and FTP emission rate estimates for each model year are shown in Table 2 for the light-
       3Brzezinski, D. and P. Enns, "The Determination of Hot Running Emissions from FTP
Bag Emissions", Report No. M6.STE.002, December, 1997.

                                            -5-

-------
duty cars and trucks in the EPA-industry sample used in this study.

4.0    FTP-BASED MODELS OF RUNNING LA4 DETERIORATION
       WITH MILEAGE

       This section describes the methodology EPA used to estimate the deterioration of
running emissions.  Deterioration  of running emissions as a function of mileage was
examined using a number of linear and nonlinear models. The goal was to develop a
description of deterioration that is consistent with both the available test data and with
engineering judgment of past and likely future technologies.

       In particular, for the model year/technology sub-fleets identified above, adequate
data are often absent in some part of the useful lifetime mileage range. Such data gaps raised
concerns when trying to fit a single functional form to a given data set, as it usually was
found that no simple description of deterioration adequately describes the full range. For
example, a fitted least  squares regression often tends to overestimate emissions at low
mileage.

       A number of linear and nonlinear models of deterioration were examined. The
chosen models represent a balance of simplicity and engineering judgment. They take the
general form of expressing  emissions as a piecewise linear function of mileage. At low
mileage, emissions are assumed to equal the mean level estimated from those vehicles in the
dataset with less than 20,000 miles of accumulated driving. This level applies for mileages
ranging from zero up to the mean mileage for those vehicles. This approach was thought to
give the best prediction, since the vehicles tested at low mileage should not be subject to any
recruitment bias influence.  The  20,000 mile cutoff is somewhat arbitrary, and was
developed in coordination with the FACA In-Use Deterioration Workgroup.

       At higher mileage, emissions are modeled to deteriorate linearly. While nonlinear
models were investigated, they did not provide significant improvement over simpler linear
forms. Two linear functions are used  in the final models: (1) a least squares regression
using all the data that is constrained to pass through the low mileage sample means; and (2)
a least squares regression using all  the data, with no constraints. The unconstrained linear
model was chosen as the best representation of the data at higher mileages. The constrained
line was chosen to provide a transition, when needed, between the low mileage mean and
the high mileage unconstrained regression. The connection between these two lines is made
based on their relative positions for a given technology/model  year class.  With a few
exceptions, the following steps describe the calculation of this piecewise linear function.

1.     For each of the model year/technology groups listed in Section 2.0, the mean CO,
       HC, and NOX emissions, and the mean mileage were computed for all vehicles with
       an odometer reading of less than 20,000 miles.  This value is used to model the
       group's low mileage emission level from zero miles up to the mileage determined
                                       -6-

-------
       in step 2. An exception occurs when the mean emissions for the entire sample is less
       than the mean of the low mileage sub-sample, a case that is discussed below.

2.      For each group, an (unconstrained) regression line was estimated for emissions
       versus mileage.

          a.     If this line has positive slope and its intercept is less than the low mileage
                mean emissions from (1) above, it defines estimated emissions beginning
                at the mileage where it intersects the low mileage mean.

          b.     If the (unconstrained) regression has positive slope but the intercept is
                greater than the low mileage mean, the constrained line defines emissions
                from the mean of the low mileage sub-sample to the mileage at which it
                intersects the unconstrained line. Beyond that mileage, emissions are
                estimated by the unconstrained line. Thus, the constrained line links the
                other lines for an intermediate range of mileages.

          c.     If the unconstrained regression has negative slope or the mean of the full
                sample is less than low mileage mean, emissions for all mileages are set
                equal to the mean emissions for the full  sample.  This assures that
                negative deterioration cannot occur.

       While these rules do not  encompass all  possible  scenarios,  they  do cover  all
situations  arising with the FTP data on which this analysis is based. The majority of cases
are covered by option 2(a), giving a simple two-piece function. The three-piece function of
2(b) applies to several  situations, usually  with only a small slope change from the
constrained to unconstrained line. Finally, the simple horizontal deterioration line of option
2(c) is needed for the CO fits of the  1988-93 TBI cars and the NOx fits of the 1981-83
carbureted trucks. The underlying numerical estimates are listed in Table 3.

       For the FTP data set, these rules appear to produce reasonable emission proj ections
in  most cases. The two cases in which the full sample mean is less than the low mileage
mean are caused by a few low mileage outliers.
5.0    HIGH EMITTER CORRECTION FACTORS

       Since the estimates of running emissions deterioration are based on FTP tests
obtained from public vehicle recruitment programs, there is some concern that low vehicle
recruitment acceptance rates (typically less than 25%) in these programs may introduce
recruitment bias. Whether such bias results in overestimation or underestimation of the true
emissions deterioration is a matter of debate. This section addresses this issue, describes the
methodology for adjusting emission factor estimates to account for bias,  and presents the
                                       -7-

-------
results.

       Most of the 1990 and later model year vehicle data for this analysis were supplied
by the domestic automobile manufacturers (the AAMA dataset). The manufacturers have
expressed the opinion in FACA meetings and MOBILE6 workshops that owners of vehicles
experiencing problems would be more likely to respond to the manufacturers' recruitment
efforts, especially considering that repairs were included as an incentive to participate. The
AAMA dataset is also composed of vehicles tested when they were roughly 2-3 years of age,
when gross emitters should be few in number and any recruitment bias influence should be
minimal.

       Most of the pre-1990 data were collected by EPA; the average age of the vehicles
was roughly 3-5 years at the time of testing. In this case, tampered vehicles or vehicles with
problems should be greater in number, but owners may be more reluctant to participate in
a program run by a regulatory agency, resulting in an underestimation of high emitters. The
California Air Resources Board (CARB) has tested this hypothesis by comparing estimates
from its CALEVIFAC emissions inventory model, which are based on surveillance programs
similar to those run by EPA, with emissions obtained from a California Pilot Project fleet
with a high (60%) vehicle capture rate. In general, the comparison showed that the modeled
estimates tend to underestimate emissions in  older model year vehicles  and slightly
overestimate the emissions of newer vehicles. CARB developed high emitter adjustment
factors (HECFs) for use in its EMFAC model to account for these discrepancies.

       EPA developed high emitter correction factors  using EVI240 data collected in
Dayton, Ohio during 1996-97. Like other inspection and maintenance (I/M) data, these form
a large sample of vehicles within their geographical region, and are considerably less subj ect
to  sources of bias  found in  non-mandatory programs. The data and their translation to
running LA4 estimates are described in more detail in a separate document4.

       Because of problems with the Ohio data odometer readings, the data were condensed
to  their mean running  LA4  values by age,  which then were associated with  the
corresponding region-specific mileage accumulations obtained from 1995 Nationwide
Personal Transportation Survey (NPTS) data. After smoothing these values in the manner
required for use in MOBILE6, these points were graphed with the emission rates fitted from
the FTP  data  as described  in Section 4. For  each pollutant and within each model
year/technology group, the  difference  between the Ohio mean and FTP-based fit was
computed. These values were regressed through the origin against mileage. (The line was
forced through the origin so that at zero  miles the difference is zero.) Finally, the fitted
differences were added to the fitted FTP-based values to obtain corrected values.
4Enns, P., E. Glover, P. Carey and M. Sklar, "Analysis of Emissions Deterioration
Using Ohio and Wisconsin EVI240 Data," Report Number M6.EXH.002, October, 1998.

-------
       In a few model year/technology groups, the Ohio adjustment is negative and, when
applied to the deterioration line, causes negative deterioration. For these cases, deterioration
is held equal to zero up to the mileage at which the adjusted emissions exceed the low
mileage constant level.
       Figures 1 and 2 illustrate how the adjusted and original values compare for each
model year/technology group as a function of mileage for the car sample. Ninety-five
percent confidence bands for the unadjusted lines are drawn to help judge the impact of the
corrections. If the adjusted values fall inside these bands, it suggests that the Ohio EVI240
data agrees fairly closely with the FTP data, i.e., bias is not a large problem. Otherwise, the
recruitment bias is more serious. The graphs show varying levels of disagreement between
the two  data sources.  In these graphs, the mileage interval for a given set of lines
corresponds to the average mileages assigned in the NPTS survey to the model years for that
group of vehicles. For example, the 1990 to 1993 cars range in mileage from about 45,000
to 70,000. Thus, the graphs show line fits for those vehicles in that interval.

       Figures 3 to 9 present emission estimates for each model year/technology group as
a function of mileage both with and without the high emitter correction factors for cars and
trucks. For MOBILE6, deterioration estimates with the high emitter corrections will be used.
6.0    RESULTS

       Results for each vehicle type/model year/technology group are presented in Tables
3 and 4. Included are the  slopes and intercepts of the constrained and unconstrained
regression lines, low mileage emissions and mileage intervals for each line segment. Table
3, described in Section 4, gives the unadjusted slopes. Applying the adjustment factors
effectively changes the line segment slopes. The high emitter correction factors and the
corresponding adjusted slopes are displayed in Table 4.

       Shown below is a sample calculation of running emissions. It illustrates how the
model coefficients given in  Table 4 are used.

Example:     Calculate HC running emissions for a 1985 model year Fl-equipped car with:
             a) 15,000 miles, b) 75,000 miles, and c) 125,000 miles.

From Table 4:
a)     At 15,000 miles, mileage
-------
b)     At 75,000 miles, first cornersecond corner, therefore:

       Running (g/mile) = ZML + (First Slope * First Corner) +
                         (Second Slope) * (Second Corner - First Corner) +
                         (Third Slope) * (Mileage - Second Corner)
                       = 0.1479 + (0.0000*18.89) + (0.0078)*(81.38-18.89) +
                         (0.0059)*(125-81.38)
                       = 0.8927 g/mile
                                      -10-

-------
     7.0 RESPONSE TO COMMENTS FROM PEER AND STAKEHOLDER
                                  REVIEW
(1) "MOBILE6 Determination of Running Emissions as a Function of Mileage for 1981
- 1993 Model Year Light-Duty Cars and Trucks", and (2) "Analysis of Emissions
Deterioration Using Ohio and Wisconsin IM240 Data"
Comment #37 (AAMA) June 5.1997

       Comment:  "EPA  did not describe, however,  how  they  intend  to  estimate
       deterioration for current and future technology vehicles, particularly Tier I and LEV-
       type vehicles with extended durability (100K) and onboard diagnostic controls."

       EPA's Response: EPA 's estimates of future technology vehicle emission factors
       can be found in documents M6.EXH. 00 7 and M6.EXH. 009.  The first document
       discusses the HC andNOx emission factors and the second paper discusses the CO
       emission factors.
       Comment: "AAMA differs significantly with the EPA on its approach to estimate
       in-use exhaust emissions and deterioration. EPA should not base in-use emission
       rates used in MOBILE6 on I/M240 results, for the following reasons:

       1.     EPA has not demonstrated sufficient correlation between the I/M240 and
             FTP.

       2.     EPA has little control over vehicle test fuel, preconditioning, or temperature
             at which I/M240 tests are conducted. The methods  used by  EPA in
             MOBILES to correct I/M240 results for temperature and fuel effects are
             questionable.

       3.     The use of the I/M240 results along with dubious correlation equations and
             correction factors for fuels and temperature very likely result in EPA arriving
             at emission rates for current and future  vehicles that are significantly
             different than the California Air Resources Board's (ARB) emission rates for
             exactly the same vehicles. If this is the case, then EPA  must explain why
             these emission rates are so different than ARB's emission rates. If EPA is
             convinced that the I/M240 must be used, then EPA should convince ARB to
             estimate emission rates  in the  same way, so that  some consistency in
             emission rates on identical vehicles is achieved."
                                     -11-

-------
EPA's Response:

1.     Emission results from the IM240 exhaust emission test procedure were not
       used directly to estimate the emissions of vehicles using the FTP.  The
       IM240 was usedprimarily to assure that recruitment bias in the FTP testing
       did not affect the overall emission deterioration estimate. Adjustments were
       made to the IM240 results in order to approximate FTP emission levels,
       however, these  emission levels are never used to estimate FTP emission
       levels for MOB1LE6 directly.

2.     The adjustments to IM240 results for preconditioning, fuel and temperature
       effects are applied only as an attempt to reduce, not eliminate, the effect of
       these parameters.  EPA feels that the application of these adjustments
       improved the credibility of the overall analysis.

3.     EPA does feel  that the emission estimates for similar future vehicles in
       California and in the federal fleet should be comparable. However, since
       the specific rules related to new vehicle  certification federally and in
       California are not identical, there is room for disagreement on the emission
       impact of the new rules. EPA and California are sharing data and methods
       so that EPA is confident that, once all factors have been considered by both
       groups,  that the differences between EPA 's and  California's estimates of
       emissions from future vehicles will be  negligible.
Comment:: "EPA should not base FTP emission rates on fast-pass I/M240 data, for
the same reasons as above.  AAMA believes there will be  a weak correlation
between fast-pass data and full I/M 240 data, and a weak correlation between full
I/M240 data and the FTP. FTP values developed from fast-pass data and these two
weak correlations will be subject to a high degree of error."

EPA's Response: EPA admits that the relationship between the fast pass IM240
test and the FTP test is not well characterized in the work and may possess a low
level of correlation.  This would be particularly true for high emitting vehicles or
vehicles which possess intermittent emission problems. Fortunately, in the case of
high emitting vehicles (Ohio failures),  no fast pass results were used since all
failures automatically received a full IM240 test.
Comment: "The automakers submitted extensive in-use FTP data to the EPA a year
ago which show that emission rates of 1990-1994 vehicles are significantly less than
MOBILES estimates, furthermore the data do not show any accelerated deterioration
after 50,000 miles as EPA now assumes. The emission rates developed from this
                                -12-

-------
       data were  similar to ARB's emission rates for the  same vehicles.  Data were
       submitted on over 2,000 cars and 900 light duty trucks. This data was not mentioned
       by EPA in the Workshop. This data significantly adds to the EPA data on the same
       model years, and should give EPA a much larger database to analyze for MOBILE6.
       AAMA urges that the MOBILE6 emission rates should be based on all available
       FTP testing of vehicles by various sources, including EPA. ARE, and the Industry."

       EPA 's Response:  The EPA acknowledges the receipt of these data and their use in
       characterizing in-use emission behavior.  The FTP test data provide by these
       various sources became the backbone of the MOBILE6 emission factors.  It was
       used almost exclusively to characterize the normal and high emitter emission levels.
       It was also the basis for the average emission calculation presented in  reports
       M6.EXH.001 andM6.STE.002.
       Comment: "100K and Useful Life Effects: For purposes of certification, under
       Section 206 (Clean Air Act), emission standards were established for useful life
       extended to 10 years or 100.000 miles. EPA needs to explicitly incorporate the effect
       of this added requirement in the MOBILE model as this was not done in
       MOBILES."

       EPA's Response:  The MOBILE6 1981-1993 model year emission factors were
       empirically developed from the FTP and Ohio datasets, and no assumptions were
       added to account for extended useful life. However,  the 1994 and later vehicle
       modeling was less empirical in nature, and contains factors that account for the
       extension of the useful life from 50,000 miles to 100,000 miles.
Comment #56 (J.F. KOWALCZYK. State of Oregon) Dec. 19.1997

       Comment:: "It is recommended that EPA go through the precise calculations to
       make the adjustment to mobile 6 based on Ohio I/M data and 1995 NPTS data as
       outlined above. Additionally, EPA should use the 1995 NPTS mileage
       factors in the final mobile 6 model which may necessitate speeding up
       the Acurex work. The 1995 NPTS data is clearly more up-to-date and most
       likely more accurate than the 1990 data. Regional specific NPTS mileage
       data should be used as there appears to be significant differences
       between regions.

       I also recommend that CARB's revised EMFAC model, which according to
       Mark Carlock should be available in January, be analyzed against
       mobile 6 and that full harmonization between the two models be sought
       or, at the very least, a believable explanation of the difference be
                                     -13-

-------
       provided."

       EPA's Response: The final analysis byAcurex usedinMOBILE6 included the 1995
       Nationwide Personal Transportation Survey (NPTS) database. As suggested, the
       regional NPTS results were used for estimating mileage accumulations in the
       analysis of the Ohio I/M data.

       We  have  attempted to coordinate with California as much as possible when
       constructing MOB1LE6, including sharing of data and analysis.  We have met
       repeatedly with California Air Resource staff during the development ofMOBlLE6
       to share ideas.  Some aspects of the California fleet, however, are unique to
       California and will not be identical in the federal (non-California) fleet.
Comment #58 (Data Analysis team, In-Use Deterioration and Modeling Workgroups)
January 14.1998

       Comment:   EPA should consider a broad range of mass emissions databases,
       including those reviewed by the Team and those which EPA has said it is still
       seeking to obtain and/or analyze (e.g., more complete Ohio EVI240 data, California
       Pilot I/M Program data). EPA should report to the Team at a later date its proposal
       for the role of each data base in revising MOBILE,  groupings of model  year,
       technology, model year/technology, emitter categories, etc.  The team wishes to
       review and discuss EPA's draft revision of the in-use deterioration estimates in the
       MOBILE model.

       EPA Response: EPA agrees that the data summarized in the workgroup report are
       enough to warrant a serious reassessment of the MOBlLESa emissions. EPA has
       considered the data during that reassessment. EPA has presented interim analyses
       to the Workgroup for review, including  the model year/technology groupings
       chosen, and the role of each data base in revising MOBILE. The documentation for
       the proposed in-use deterioration estimates will be provided to the Workgroup for
       review.

       Comment:   EPA  should acknowledge the overestimation  that results when
       predicting future evaporative emissions based on current experience and existing
       technology.

       EPA Response: EPA agrees that cars meeting the new evaporative test procedures
       and standards should be modeled as being lower emitting in-use than previous cars.
       This was in fact the case in MOBILES, but the size of the difference is worthy of
       reconsideration. The reconsideration ought to start with an understanding of the
       causes of high evaporative emissions  in the  older  cars, and then  apply an
                                      -14-

-------
       understanding of how changes in design and materials would affect those causes.

       Comment: EPA should take an active role in obtaining high quality data from state
       IM240 programs.

       EPA Response: EPA  has obtained state IM240 data from Colorado, Arizona,
       Wisconsin, and Ohio.  Effort in 1998  is limited due to staff time and resource
       constraints; however, we remain interested in encouraging and perhaps financially
       supporting better preconditioning, more full IM240  tests,  and more careful
       recording of vehicle information, on a sizable sample of vehicles.

       Comment: An Auto/Oil hot soak pilot study has been conducted. In addition, CRC
       has conducted a real time diurnal study that will measure  24 hour diurnal emissions
       from 151 vehicles. EPA is also conducting a diurnal emission study. The results of
       these analyses should be reviewed when  available to provide insight into evaporative
       emissions deterioration.

       EPA Response: EPA plans to use these data along with the data collected under
       EPA sponsorship.
Comment #83 (Rick Barrett. CDPHE) July 28.1999

       Comment: The 60-day comment period is not sufficient to complete the review of
       MOBILE 6 documents.

       EPA Response: We agree that due to the simultaneous posting of large amounts of
       MOB1LE6 documents that a 60-day comment period may not be sufficient time to
       complete a thorough review of the MOBILE6 documents. As such, we have accepted
       comments for several months following the original 60-day deadlines.
       Comment:  "there is likely still an under prediction in both the frequency and
       projected emissions levels of high emitters in the proposed Mobile 6 estimates."

       EPA Response:

       EPA does not believe that the emission factors and high emitter frequency contained
       inMOB!LE6 seriously under or over predict (within a reasonable statistical range)
       the emission results that have been collected from vehicles or will likely be collected
      from typical vehicle emission testing. This conclusion is backed up by considerable
       laboratory and in-use IM240 testings, and by recent repeat testing of vehicles
       (multiple tests over a long period of time).
                                      -15-

-------
      However,  it doesn't necessarily  mean  that  the  emission  factors  based on
      dynamometer tests exactly match either individual vehicle behavior or even fleet
      behavior in the  'real world'. Vehicle emission testing has confirmed that individual
      vehicle emission behavior is highly variable. Thus, any individual vehicle or small
      sample of vehicles may have results which are considerably  different that these
      averages. Also, the 'real world' may contain factors that greatly affect emissions
      which have not been accounted for in the MOBILE6 modeling due to testing and
      experimentation resource limits.

      Comment:   The EPA has several documents related  to the Tier 0  and Tier 1
      vehicles, as well as I/M effects, which show the high emitting vehicles  have a
      constant average emission rate with respect to age or mileage. CDPHE staff believes
      this assumption does not make sense. As a result of an audit of Colorado's I/M
      program conducted last year by Environ, vehicles identified as failures, and then
      repaired, were  deteriorating at a faster rate than the remainder of the fleet.  In
      addition an analysis of Colorado's data conducted by Peter McClintock has shown
      that high emitters average emissions do increase with age. CDPHE staff believes
      EPA should reevaluate these assumptions based on I/M data where the emissions
      performance of the same vehicles can be tracked through multiple inspection cycles.

      EPA Response: The assumption of zero deterioration of High emitters was made
      because absolutely no correlation could be found in the data  (FTP and a limited
      sample of Ohio I/M data) that indicated that it is a function of mileage or age. Thus,
      the assumption of a constant emission level with respect to mileage or age is the
      most prudent assumption to make.  The engineering rationale for this assumption
      is that the more serious emission related problems are not necessarily more likely
      to occur at high mileage than  at low mileage.  However,  the large observed
      variance of the high emitter level does suggest that a variety  of emission related
      problems occur, and that they can occur at any mileage (few problem occur under
      20,000 miles, though). The rising High emitter fraction with mileage also implies
      that the frequency of High emitter appearance increases with mileage.
       Comment:  The CDPHE staff feels that EPA has made reasonable assumptions
       regarding the effectiveness of OBD systems and motorist response to OBD MIL
       indications.

       EPA Response: Thanks..
Comment  #85  (Joel  Schwartz,  California  Inspection &  Maintenance Review
Committee) August 3.1999
                                      -16-

-------
Comment:  Data biases should be examined and corrected: Some cars never get
tested.  Some motorists prepare for the test by setting up their car to pass the test
without making substantive repairs. And some motorists make substantive repairs
to their cars in preparation for the test. Either (1) validate the EVI240 test-lane data
against data collected in random roadside pullovers or by remote sensing, or (2) use
roadside or RSD data to generate the emission factors for the model. Base RSD
credits on more representative RSD data and encourage the use of the CRC best
practices for collection of high-quality RSD data.

EPA Response:

(Also see responses #1 and #20 of section 7 ofM6.IM.001)

The remote sensing device (RSD) has the potential to collect massive amounts ofin-
use data that could potentially be used for the development of emission factor
models or the confirmation of emission factor models.  Unfortunately, at the time of
data collection and analysis, high quality RSD data in large quantities were not
available for  use.   Also, the quality and applicability of use of RSD data is
frequently dependent on the actions of individual operators and the specific test
locations.  Improper siting of the RSD units  can lead to results which are not
representative of overall vehicle operation (100% ramp siting for example), or do
not measure a random sample of the vehicle population.

Comment:  Base model assumptions on real-world data from vehicle emissions and
human behavior studies, and not on incorrect pre-conceptions about the way people,
vehicles, or programs ought to  behave.  EPA should provide justification for
"Engineer Assumptions". I/M "Saw-Tooth" does not represent real I/M programs.
Human behavior should explicitly included in  the model. Real-world data should
drive evaporative emissions calculation. RSD benefits are contingent on incorrect
assumptions about effectiveness of scheduled I/M programs

EPA Response:

(Also see Response  #20 of section 7 ofM6.IM.001)

The RSD benefits have been removed from the MOBILE6 model.
Comment:  MOBILE should not be used for regulatory purposes until it has been
appropriately validated. Include multiple internal validation checks so users can
ensure that MOBILE's output jibes with real-world measurements, both overall and
in detail. Make real-world measurements, rather than MOBILE predictions, the final
arbiter of the emissions reduction credit attributed to an I/M program.
                                -17-

-------
      EPA Response: The process for using a model such as the MOBILE model series
      for regulatory purposes  rather than conducting a multitude of elaborate  and
      expensive studies of 'real world' emissions has been established through Federal
      regulation over several decades. Until this process has been changed, the need for
      MOBILE6 will continue.
Comment #87 (Alison Pollack, Till Stoeckenius and Cuong Tran, Environ) August 18,
1999

       Comment: Comments from Peter McClintock of Applied Analysis (7/4/1999) and
       Tom Wenzel of Lawrence Berkeley National Laboratory.

       EPA Response: Are these comments answered or addressed? Comments from Peter
       McClintock of Applied Analysis (7/4/1999 conversion of IM240 to running LA4
       and 6/25/1999 memo to Phil Lorang regarding change vehicle registration patterns)
       Comment:   Fast  Pass  time is  incorrectly included in the regression during
       conversion of fast pass to full EVI240 using Wisconsin data. Fast pass time (variable
       name FSEC in EPA regressions) is assumed to be linearly related to the log of fast
       pass emissions.

       EPA Response: It would seem logical that the fast pass time would be a significant
       variable in  the correlation between fast pass IM240 and the Full IM240.  It's
       exclusion would clearly be a mistake. Only Phil Enns can definitively answer this
       question.  However, it is my understanding that he double checked  all of his
       regression equations, and found and corrected any errors that may have  been
       present.
       Comment: Use of the logarithmic transformation is inappropriate and reduces the
       effects of the high emitters, leading to underprediction of emissions.

       EPA Response: Because of the highly skewed nature of vehicle emission data, a
       logarithmic transformation was necessary  in-order to perform standard least-
       squares regression analysis.  In order to overcome any possible under-prediction
       of emissions, an additive and positive logarithmic transformation factor was added
       to the regression equation.  It has the effect of increasing the average emissions and
       restoring the effect of high emitters.   It is a  valid and fairly commonly used
       statistical transformation.
                                      -18-

-------
       Comment: We recommend that EPA review the proposed approach for estimating
       MOBILE6 deterioration rates, taking into account the comments received, and revise
       the data bases and modeling methods accordingly. In addition, we recommend that
       EPA perform sensitivity and uncertainty analyses to guide further data collection
       efforts. In addition, as stated above,  we recommend that EPA make all data sets
       readily publicly available, and also that full regression statistics be provided for any
       regression models using in the development of these and other MOBILE6 emission
       factors.

       EPA Response:  Throughout this process EPA has reviewed all comments and
       revised our methodology when appropriate. Allnon confidential or non proprietary
       data sets will be made public for subsequent analysis, and most regression models
       will also be available as part of the report.
Comment #89 (Robert Slott. Consultant) August 23.1999

       Comment:  "... large data sets of random full EVI240 measurements exist in a
       number of states. These data sets should be used to get a first approximation of
       deterioration  rates  in  each  of these  states." "Analysis  of remote  sensing
       measurements, corrected for vehicle specific power, should be developed to give
       more realistic  on-road deterioration."

       EPA Response: Data from IM240 testing done in Ohio and Wisconsin was used in
       MOB1LE6 to  adjust the deterioration of the national average fleet estimates in
       MOBILE6.  Time and resources were not available to do a more extensive review
       of the available 1M240 testing results from other States.  There was not a sufficient
       consensus on how to interpret remote sensing measurements in order to use them
      for estimating on-road deterioration for MOB1LE6.  The appropriate use of remote
       sensing will be considered as apart of future EPA model development.
                                      -19-

-------
        Rgum 1: RUNNING LM: FTP-BASED MOBILES and OHIO IM240 ADJUSTMENTS
                             1961-03 FUEL INJECTION CARS
HC (g/ml)
 3.0

 «•

 2.0

 1.6-

 1.0-

 as

 0.0-
CO (a/ml)
                                                        120
                                                                                   in
                                            (XIOOO)
 ao

 10

 10

  6

  O-
                                           ; ociooo)
NOX (0/mO
 1.6
 ao
                                            «clOOO)

                  88—98 MOBLEB    HHH 83—87 MOBLEO
                  88-88 ADJUSTED   A** 83-87 ADJUSTED
                  LOWER 90* CL     —• LOWER 00% CL
                  UPPER 95% CL     ~~ * UPPER 98% CL
                                                        120
*** 81-82 MOBILES
*** 81-82 ADJUSTED
— • LOWER 90% CL
— • UPPER 98% CL
                                        -20-

-------
        Figure 2: RUNNING LM: FTP-BASED MOBILES and OHIO IM240 ADJUSTMENTS
                              1981-88 CARBURETED CARS
HC (g/mi)
  100
CO (g/fnl)
                             120
                                                        •MO
                                                                                   •MO
                                       MLE8 (XIOOO)
 1O-
   10O



NOK (gtrti)

 2.0 i
 •tf-
 1.0
 OB-
 001
                                            i OcMOO)
                                     ' LOWER flMt CL
                                     1 UPPER MKCL
— • LOWER 9BKCL
— • UPPER 9BK CL
                                         -21-

-------
HC (g/ml)
        Figure 3: FTP-BASED MOBILES PROJECTIONS and OHIO IM240 ADJUSTMENTS
                          RUNNING LA4. 1868-88 PFI CARS
 0.0
                                                                            200
                                     -22-

-------
HC (g/ml)
 1.0 d
        Figure 4: FTP-BASED MOBILES PROJECTIONS and OHIO IM240 ADJUSTMENTS
                          RUNNING LA4. 1868-83 TBI CARS
 O4
 0.0
                                                                            200
                                     -23-

-------
HC (g/ml)
       Figure 6: FTP-BASED MOBILES PROJECTIONS and OHIO IM240 ADJUSTMENTS
                         RUNNING LA4, 1866-83 GARB CARS
 0.0
                                                                           200
                                     -24-

-------
HC (g/ml)
        Figure 6: FTP-BASED MOBILES PROJECTIONS and OHIO IM240 ADJUSTMENTS
                           RUNNING LA4, 1863-87 F\ CARS
NOX 
-------
HC (g/mi)
       Figure 7: FTP-BASED MOBILES PROJECTIONS and OHIO IM24O ADJUSTMENTS
                         RUNNING LA4, 1863-86 GARB CARS
 0.0
                                                                            200
                                     -26-

-------
HC (g/mi)
       Figure 8: FTP-BASED MOBILES PROJECTIONS and OHIO IM240 ADJUSTMENTS
                          RUNNING LA4, 1861-82 H CARS
 1-
                                                                           200
                                     -27-

-------
HC (g/mi)
       Figure 9: FTP-BASED MOBILES PROJECTIONS and OHIO IM240 ADJUSTMENTS
                         RUNNING LA4. 1981-82 GARB CARS
 1-
                                                                           200
                                     -28-

-------
                                                    Table 1
                  Distribution of Vehicles by Model Year and Technology for the Combined FTP Dataset

81
82
83
84
85
86
87
88
89
90
91
92
93
CARS
TECHNOLOGY
CARB
657
71
57
30
74
34

15
22




OPLP
367
71
63
5
24
7







PFI
29
8
62
35
66
92
106
113
103
250
426
347
366
TBI
15
74
127
46
56
60
76
69
38
160
91
57
29
SUB
TOTAL
1,068
224
309
116
220
193
200
197
163
410
517
404
395
TRUCKS
r
TECHNOLOGY
CARB


3
22
30
9







OPLP
124
45
8
26
33
14







PFI




6
41
4



144
92
93
TBI




13
23
6


144
141
92
90
SUB
TOTAL
124
45
11
49
82
87
10


145
285
184
183
TOTAL
1,192
269
320
165
302
280
210
197
163
555
802
588
578
M6.EXH.001                                           -29-

-------
     TOTAL      977     538  2,003     898   4,416     64     250     382    509   1,205   5,621
M6.EXH.001                                     -30-

-------
                                          Table 2
        Mean Running LA4 and FTP Emission Levels by Model Year for Light-Duty Cars and Trucks
                                 for the Combined FTP Dataset
                               CARS                                             TRUCKS
       |HC_RUN  |  HCFTP  | CO_RUN |  COFTP  | NOX_RUN |  NOFTP  | HC_RUN  | HCFTP | CO_RUN  |  COFTP
                                    NOXRUN   NOFTP
IMYR
    81     |   0.42l|   0.706|   6.489|  9.667|   0.795|   0.897|   0.759|   1.275| 10.876
                                18.1581  1.6621   1.7521
-1- -1-
82 0.588


83 0.230


84 0.533


85 0.355
-t-
0.789


0.431


0.756


0.533
-1-
5.394
16. 77^

2.760
13.22.

7.622
10.63:

5.561
-1-
8.318
1 1.74(

5.073
5 1.40.

9.968
3 1.38'

6.935
-t-
0.750
3 1.73

0.677
5 1.43

0.785
7 1.4C

0.687
-t-
0.872
2

0.806
5

0.893
5

0.770
-t-
1.163


0.865


0.419


0.923
-t-
1.732


1.361


0.802


1.281
-t-
8.987


5.759


3.597


8.999











                                           -31-

-------
        	+	+	+	+	+	+	
                                +	+	I
           0.759|   0.926|   8.738| 10.432|  0.612|  0.713
                             8.789|   1.006|   1.057|
        	+	+	+	+	+	+	
                                +	+	|
           0.456|   0.656|   7.005|  8.366|  0.698|  0.789
                             4.610|   0.53l|   0.605|
        	+	+	+	+	+	+	
                                +	+	|
         0.2121   0.4061   3.3441   4.5741   0.5641   0.6681
       86
0.5611   0.8231   6.248
       87
0.1641   0.4011   2.959
     88
     	+	+	+	+	+	+	+	+	+	+ .
                                     +	+	|
     89     I  0.1521   0.3111   2.6451  3.9111   0.5531   0.6521       .1       .1
       91
       93
90    |   0.109|   0.274|   2.087|  3.614|  0.400|  0.633|  0.163
                               . |   0.376 |       .  |
                                                                               2.245
           0.078|   0.237|   1.572|  3.145|  0.353|  0.524|  0.187|  0.800|  2.228
                             9.510|  0.486|  0.885|
92    |   0.094|   0.267|   2.599|  4.327|  0.322|  0.508|  0.152
                               .|   0.469|       .|
                                                                               2 .172
           0.061|   0.225|   0.977|  2.55l|  0.286|  0.466|  0.137|  0.420|  1.668
                             5.3631  0.4591  0.8471
M6.EXH.001
                                       -32-

-------
                                                          Table 3
                          Running Emission Deterioration Model Coefficients for HC (Unadjusted)
                                                     Light-Duty Cars
ModelYear/
Technology
88-93 PFI
88-93 TBI
83-87 FI
86-93 CARS
83-85 CARS
81-82FI
8 1-82 CARS
ZML Mean
Emissions
(gr/m)
0.0516
0.0843
0.1479
0.0815
0.1691
0.1240
0.2108
First
Slope
(gr/m/ 1000m)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
First
Corner
(1000 miles)
20.03
34.39
14.10
19.83
25.24
11.29
10.18
Second
Slope
(gr/m/ 1000m)
0.0023
0.0020
0.0079
0.0019
0.0095
0.0038
0.0110
Second
Corner
(1000 miles)
N/A
N/A
81.38
N/A
N/A
70.55
N/A
Third
Slope
(gr/m/ 1000m)
N/A
N/A
0.0060
N/A
N/A
0.0037
N/A
                                                    Light-Duty Trucks
88-93 PFI
88-93 TBI
84-93 CARS
81-87 FI
81-83 CARS
0.0932
0.0783
0.2495
0.2927
0.6587
0.0000
0.0000
0.0000
0.0000
0.0000
23.40
16.24
22.03
29.38
15.99
0.0025
0.0043
0.0136
0.0136
0.0110
N/A
55.16
N/A
N/A
N/A
N/A
0.0042
N/A
N/A
N/A
    Note: The first slope is zero, since it is assumed that the ZML emission rate is constant from zero miles to the first corner. For the
    cases with a single corner, the second slope is determined from the unconstrained regression and the corner occurs at the mileage
    where that line intersects the ZML mean emissions. For the case with two corners, the second slope was obtained using a
    regression line constrained to pass through the ZML mean emissions-mileage. The third slope is for the unconstrained regression
    line and applies at mileages above the second corner. (Unadjusted refers to estimates  obtained using the FTP dataset only.)
M6.EXH.001
-33-

-------
                                                      Table 3 (cont.)
                          Running Emission Deterioration Model Coefficients for CO (Unadjusted)
                                                     Light-Duty Cars
ModelYear/
Technology
88-93 PFI
88-93 TBI
83-87 FI
86-93 CARS
83-85 CARS
81-82 FI
81-82 CARS
ZML Mean
Emissions
(gr/m)
0.7983
2.5684
2.1416
0.6910
1.0983
1.7270
2.9361
First
Slope
(gr/m/ 1000 m)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
First
Corner
(1000 miles)
13.78
N/A
14.10
21.13
22.69
16.62
8.79
Second
Slope
(gr/m/ 1000 m)
0.0397
N/A
0.1142
0.0307
0.1739
0.0585
0.1494
Second
Corner
(1000 miles)
N/A
N/A
69.78
N/A
N/A
N/A
15.02
Third
Slope
(gr/m/ 1000 m)
N/A
N/A
0.0898
N/A
N/A
N/A
0.1459
                                                    Light-Duty Trucks
88-93 PFI
88-93 TBI
84-93 CARS
81-87FI
81-83 CARS
0.9017
1.1439
1.5384
5.2337
9.0704
0.0000
0.0000
0.0000
0.0000
0.0000
16.80
17.54
19.30
55.03
18.86
0.0357
0.0491
0.1986
0.0644
0.0635
58.68
N/A
N/A
N/A
N/A
0.0297
N/A
N/A
N/A
N/A
    Note: The first slope is zero, since it is assumed that the ZML emission rate is constant from zero miles to the first corner. For the
    cases with a single corner, the second slope is determined from the unconstrained regression and the corner occurs at the mileage
    where that line intersects the ZML mean emissions. For the case with two corners, the second slope was obtained using a
    regression line constrained to pass through the ZML mean emissions-mileage. The third slope is for the unconstrained regression
    line and applies at mileages above the second corner. (Unadjusted refers to estimates obtained using the FTP dataset only.)
M6.EXH.001
-34-

-------
                                                      Table 3 (cont.)
                         Running Emission Deterioration Model Coefficients for NOx (Unadjusted)
                                                     Light-Duty Cars
ModelYear/
Technology
88-93 PFI
88-93 TBI
83-87 FI
86-93 CARS
83-85 CARS
81-82FI
8 1-82 CARS
ZML Mean
Emissions
(gr/m)
0.2582
0.2931
0.5976
0.5522
0.5614
0.6370
0.6121
First
Slope
(gr/m/ 1000 m)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
First
Corner
(1000 miles)
18.58
21.55
34.25
26.12
12.52
16.36
8.79
Second
Slope
(gr/m/ 1000m)
0.0048
0.0047
0.0042
0.0023
0.0059
0.0129
0.0063
Second
Corner
(1000 miles)
N/A
N/A
N/A
N/A
N/A
N/A
17.00
Third
Slope
(gr/m/ 1000m)
N/A
N/A
N/A
N/A
N/A
N/A
0.0060
                                                    Light-Duty Trucks
88-93 PFI
88-93 TBI
84-93 CARS
81-87 FI
81-83 CARS
0.3782
0.3346
1.3234
0.5388
1.6660
0.0000
0.0000
0.0000
0.0000
0.0000
21.20
16.24
22.20
21.43
N/A
0.0044
0.0040
0.0040
0.0084
N/A
N/A
55.16
N/A
N/A
N/A
N/A
0.0032
N/A
N/A
N/A
    Note: The first slope is zero, since it is assumed that the ZML emission rate is constant from zero miles to the first corner. For the
    cases with a single corner, the second slope is determined from the unconstrained regression and the corner occurs at the mileage
    where that line intersects the  ZML mean emissions. For the case with two corners, the second slope was obtained using a
    regression line constrained to pass through the ZML mean emissions-mileage. The third slope is for the unconstrained regression
    line and applies at mileages above the second corner. (Unadjusted refers to estimates obtained using the FTP dataset only.)
M6.EXH.001
-35-

-------
                                                          Table 4
                                  High Emitter Adjusted Running Emission Deterioration
                                                 Model Coefficients for HC

                                                     Light-Duty Cars
ModelYear/
Technology
88-93 PFI
88-93 TBI
83-87 FI
86-93 CARS
83-85 CARS
81-82FI
8 1-82 CARS
ZML Mean
Emissions
(gr/m)
0.0516
0.0843
0.1479
0.0815
0.1691
0.1240
0.2108
First
Slope
(gr/m/ 1000 m)
0.0013
0.0013
0.0000
0.0039
0.0003
0.0094
0.0048
First
Corner
(1000 miles)
20.03
34.39
18.89
19.83
25.24
11.29
10.18
Second
Slope
(gr/m/ 1000m)
0.0036
0.0033
0.0078
0.0058
0.0098
0.0132
0.0158
Second
Corner
(1000 miles)
N/A
N/A
81.38
N/A
N/A
70.55
N/A
Third
Slope
(gr/m/ 1000m)
N/A
N/A
0.0059
N/A
N/A
0.0131
N/A
Adjustment
Additive
(gr/m/ 1000m)
0.0013
0.0013
-0.0001
0.0039
0.0003
0.0094
0.0048
                                                    Light-Duty Trucks
88-93 PFI
88-93 TBI
84-93 CARB
81-87 FI
81-83 CARB
0.0932
0.0783
0.2495
0.2927
0.6587
0.0013
0.0013
0.0000
0.0000
0.0018
23.40
16.24
36.01
40.58
15.99
0.0038
0.0056
0.0083
0.0099
0.0127
N/A
55.16
N/A
N/A
N/A
N/A
0.0055
N/A
N/A
N/A
0.0013
0.0013
-0.0053
-0.0038
0.0018
    Note: Adjusted refers to estimates obtained using the high emitter correction factors. To obtain the adjusted values, the additive
    adjustments given in this table were applied to the unadjusted slopes in Table 3. Slope values of zero were assigned in cases where
    the additive adjustments would have resulted in negative deterioration.
M6.EXH.001
-36-

-------
                                                      Table 4 (cont.)
                                  High Emitter Adjusted Running Emission Deterioration
                                                 Model Coefficients for CO

                                                     Light-Duty Cars
ModelYear/
Technology
88-93 PFI
88-93 TBI
83-87 FI
86-93 CARS
83-85 CARS
81-82FI
8 1-82 CARS
ZML Mean
Emissions
(gr/m)
0.7983
2.5684
2.1416
0.6910
1.0983
1.7270
2.9361
First
Slope
(gr/m/ 1000 m)
0.0310
0.0310
0.0000
0.0727
0.0000
0.1817
0.1414
First
Corner
(1000 miles)
13.78
N/A
19.04
21.13
25.68
16.62
8.79
Second
Slope
(gr/m/ 1000m)
0.0707
N/A
0.1091
0.1034
0.1537
0.2401
0.2908
Second
Corner
(1000 miles)
N/A
N/A
69.78
N/A
N/A
N/A
15.02
Third
Slope
(gr/m/ 1000m)
N/A
N/A
0.0846
N/A
N/A
N/A
0.2873
Adjustment
Additive
(gr/m/ 1000m)
0.0310
0.0310
-0.0051
0.0727
-0.0203
0.1817
0.1414
                                                    Light-Duty Trucks
88-93 PFI
88-93 TBI
84-93 CARB
81-87 FI
81-83 CARB
0.9017
1.1439
1.5384
5.2337
9.0704
0.0326
0.0326
0.0000
0.0545
0.1040
16.80
17.54
28.90
55.03
18.86
0.0683
0.0817
0.1327
0.1190
0.1675
58.68
N/A
N/A
N/A
N/A
0.0623
N/A
N/A
N/A
N/A
0.0326
0.0326
-0.0660
0.0545
0.1040
    Note: Adjusted refers to estimates obtained using the high emitter correction factors. To obtain the adjusted values, the
    additive adjustments given in this table were applied to the unadjusted slopes in Table 3. Slope values of zero were assigned
    in cases where the additive adjustments would have resulted in negative deterioration.
M6.EXH.001
-37-

-------
                                                      Table 4 (cont.)
                                  High Emitter Adjusted Running Emission Deterioration
                                                Model Coefficients for NOx

                                                     Light-Duty Cars
ModelYear/
Technology
88-93 PFI
88-93 TBI
83-87 FI
86-93 CARS
83-85 CARS
81-82 FI
8 1-82 CARS
ZML Mean
Emissions
(gr/m)
0.2582
0.2931
0.5976
0.5522
0.5614
0.6370
0.6121
First
Slope
(gr/m/ 1000 m)
0.0010
0.0010
0.0023
0.0021
0.0003
0.0000
0.0003
First
Corner
(1000 miles)
18.58
21.55
34.25
26.12
12.52
30.66
8.79
Second
Slope
(gr/m/ 1000m)
0.0058
0.0058
0.0064
0.0045
0.0062
0.0069
0.0066
Second
Corner
(1000 miles)
N/A
N/A
N/A
N/A
N/A
N/A
17.00
Third
Slope
(gr/m/ 1000m)
N/A
N/A
N/A
N/A
N/A
N/A
0.0063
Adjustment
Additive
(gr/m/ 1000m)
0.0010
0.0010
0.0023
0.0021
0.0003
-0.0060
0.0003
                                                    Light-Duty Trucks
88-93 PFI
88-93 TBI
84-93 CARS
81-87 FI
81-83 CARS
0.3782
0.3346
1.3234
0.5388
1.6660
0.0002
0.0002
0.0000
0.0000
0.0008
21.20
16.24
1754.24
32.21
N/A
0.0046
0.0042
0.0001
0.0056
N/A
N/A
55.16
N/A
N/A
N/A
N/A
0.0034
N/A
N/A
N/A
0.0002
0.0002
-0.0040
-0.0028
0.0008
    Note: Adjusted refers to estimates obtained using the high emitter correction factors. To obtain the adjusted values, the
    additive adjustments given in this table were applied to the unadjusted slopes in Table 3. Slope values of zero were assigned
    in cases where the additive adjustments would have resulted in negative deterioration.
M6.EXH.001
-38-

-------