0 CDA
EPA420-R-01-056
  November 2001
 Determination of NOx and HC  Basic  Emission Rates,
   OBD and I/M Effects for Tier  1  and  later LDVs and
                               LDTs

                  Final  Report M6.EXH.007
                            John W. Koupal
                            Edward L. Glover
                     Assessment and Standards 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      Introduction

       This report presents the updated MOBILE6 methodology for NOx and exhaust HC basic
emission rates (BERs) for Tier 1 and later light-duty vehicles (LDVs) and light-duty trucks
(LDTs), and the effects of On-Board Diagnostic (OBD) systems and Inspection/Maintenance
(I/M) on these vehicles.   This report supercedes the draft version of the report (EPA-420-P-99-
009) published in March 1999, and reflects several updates based on stakeholder comments,1
new data and updated methodologies.  These updates are summarized as follows:

       1) NOx emission rates have been revised based on the analysis of a substantially larger
       database of vehicle and trucks certified to a 0.4 gram/mile 50,000 mile NOx certification
       standard.
       2) The upward adjustment of NOx high-emitter frequencies to account for test program
       recruitment bias  has been reduced based on reanalysis of Tier 0 and Tier 1 high emitter
       data.
       3) The upward adjustment to NOx high-emitter frequencies to account for the presence of
       an I/M program in the test sample has been eliminated.
       4) HC emission rates have been reduced in response to the correction of an error in the
       high-emitter frequencies presented in the draft report.
       5) The emission  level after repair of an OBD-detected malfunction has been revised to
       assume repair to the normal emitter average emission level  at a given milage, capped at
       1.5 times the 50,000 mile certification standard.
       6) The response/repair rate for vehicles flagged as malfunctioning by OBD systems in
       areas with OBD/IM programs has been increased from 90 to 99 percent, to allow for
       treatment of cost waivers and program noncompliance in a  manner consistent with
       exhaust-based I/M programs..

The updated NOx emission rates,  encompassing issues (1) through (3), have a lower intercept
(zero-mile level) but higher deterioration rate than those presented  in the draft report. Total
emissions over the life of an average vehicle  are reduced slightly with the updated rates. The HC
emission rates are more  significantly lower than originally proposed, and the change in OBD
repair level results in a slight increase in the benefit of OBD repair.2

       In additional to stakeholder review, the draft final version of this report (published in
December 1999) underwent paid peer review by an external expert. The results of this peer
review are contained in Appendix E. This report has been updated to reflect key
       Comments were received on the draft report from the State of Colorado and Applied Analysis, Inc. via the
MOBILE6 comment process; General Motors and the Alliance of Automobile Manufacturers (AAM) also
submitted substantive comments via the comment process for the Tier 2 /Sulfur proposed rulemaking. Detailed
responses to the GM and AAM comments are contained in the Tier 2 Summary and Analysis of Comments. All of
the comments are available in the Tier 2 Docket.

       2A comparison between the draft and updated emission rates as well as ARB's EMFAC7G emission rates
is contained in "Comparison of EPA, ARB and AIR Emission Rates", Memorandum from John Koupal to the Tier
2 Docket (A-97-10)

-------
recommendations from this review.
2      Overview

       The methodology discussed in this paper will be applied to generate BERs across all
vehicle classes (LDV and LDT1 through 4) for all Tier 1 and later standards, including the
TLEV, LEV, and ULEV standards under the California LEV I and Federal NLEV program,
Federal Tier 2 standards, and LEV n standards recently adopted by California. For brevity,
however, the results presented here focus on the Tier 1, LEV and ULEV standards for each
vehicle class. Per peer review comments,  emission rates for the primary Tier 2 standards have
been added in Appendix B.

       Sufficient in-use data from LDVs or LDTs complying with Tier 1 or later standards on
which to base these emission rates are not available. Thus, the methodology used in the
development of Tier 1 and later BERs is based on the differences in certification standards across
standard level and vehicle class. For NOx, Tier 1 and  later BERs were based on a sample of
1,122 LDVs and 62 LDTls certified to a 50,000 mile  0.4 gram/mile NOx standard (the Federal
Tier 1 standard). For HC, Tier 1 and later emission rates were based on BERs developed for
1988 through 1993 Ported Fuel Injection (PFI) LDVs.  Using certification standard as the base
for Tier 1 and later BERs has two implicit ramifications. First, the BERs reflect the fuel which a
vehicle is certified on the meet the standard: Indolene for Tier 1 standards, and California Phase
IIRFG for LEV and later standards. Second, the HC BERs are  expressed as non-methane
hydrocarbon (NMHC) for Tier 1 LDV/LDTs, and non-methane organic gas (NMOG) for LEV
and later LDVs/LDTs.

       On-Board Diagnostics systems were required on all LDVs and LDTs sold outside
California beginning in 1996.  Tier 1 vehicles began entering the fleet in 1994, and for two years
(1994 and 1995) were not equipped with OBD.  For MOBILE6, it will be assumed that all 1996
and later LDVs and LDTs are equipped with OBD systems, which are designed to detect
emission system malfunctions resulting in emissions at or above 1.5 times the applicable
emission standard.3 If this criteria is met, a light on the vehicle's dashboard (the malfunction
indicator light, or MIL) is illuminated to alert the driver that an emissions system repair is
required. Thus, the rate of emission deterioration for Tier 1 and later vehicles must take into
account the impact OBD systems will have overall in-use emissions, including a) the
effectiveness of these systems in detecting emission malfunctions, b) the owner response rate to
illuminated MILs, and c) the effectiveness of repair in  addressing the detected problem.

       Beginning in 2001, all Inspection/Maintenance programs will require an OBD  system
       3The "1.5 times the standard" criteria was initially required by ARB, while EPA adopted a different
malfunction threshold approach.  However, manufacturers were allowed to meet the federal program through
compliance with ARB's requirement, and most chose this option.  EPA's requirement has recently been amended
to harmonize with ARB by requiring the "1.5 times the standard" criteria for vehicle sold federally. For MOBILE6,
it will be assumed that all vehicle equipped with OBD since 1996 comply with the "1.5 times"  malfunction criteria.

-------
check for OBD-equipped vehicles.  In I/M areas, this will greatly increase the rate at which
illuminated MILs are addressed, hence further improving the average rate of in-use deterioration
for Tier 1 and later vehicles.

       Most Tier 1 and later vehicles will be equipped with an OBD system and, if in an I/M
area, subject to OBD-based I/M rather than traditional exhaust I/M. However, some Tier 1
vehicles will not be equipped with OBD (model years 1994 and 1995).  To model emissions
under these scenarios, a methodology for generating basic emission rates was developed for the
following cases:

       No OBD/No EVI (Base) applies to pre-OBD Tier 1 vehicles (1994 and 1995 model
       years).  It is also used as a basis for the computation of BERs under the OBD-only and
       OBD/EVI cases.

       OBD-Only (OBD/ No EVT) applies to 1996 and later OBD-equipped vehicles where an
       I/M program is not present.

       OBD/IM applies to 1996 and later OBD-equipped vehicles where an I/M program which
       conducts OBD checks is present. An exhaust test may or may not be performed; it is not
       differentiated from an I/M program with both OBD checks and exhaust testing, since
       additional I/M reductions are not given for exhaust testing if OBD checks are performed.

       Exhaust I/M represents a situation in which only an exhaust test is conducted in an I/M
       program (an  EVI240, ASM, or idle test).  This will apply to 1994 and 1995 model year
       Tier 1 vehicles in all calender years.  The details on the derivation of emission benefits
       from these programs are contained in a separate document, M6.EVI.001.

This report gives an overview of the basic approach for generating BERs, then describes the
specific details of BER development for NOx and exhaust HC for each case.  It is important to
reinforce that the analyses performed for Tier 1 and later emission rates as well  as OBD benefits
are based largely on engineering judgement.  Aside from Tier 1 light-duty vehicles, adequate in-
use data upon which to base emissions over the life  of vehicle and trucks certified to recently
promulgated standards are not available. With regard to the benefits of OBD, the lack of
adequate data to assess performance of and owner response to the OBD system  over the life of a
vehicle necessitates that technical judgement be employed.

3      Basic Emission Rate Derivation Concept

       The basic concept underlying the generation of Tier 1  and later BERs is similar to the
approach used to develop the I/M credits for 1981 through 1993 vehicles.4  For the No OBD/No
EVI case, this concept segregates in-use vehicles into "normal" and "high" emitters.  High
       Glover, E., and Brzezinksi, D., "MOBILE6 Inspection/Maintenance Benefits Methodology for 1981
through 1993 Model Year Light Vehicles", Draft MOBILE6 Report No. M6.IM.001, March 1999. Hereafter
referred to as "Tier 0 I/M Report"

-------
emitters are vehicles that have emission control systems which are malfunctioning in some way,
and are producing average emission levels which are considerably higher than the overall mean
emission levels.  The threshold for defining a high emitter for NOx and HC is 2.0 times the
intermediate life (50,000 mile) certification emission standard.  The remainder of the fleet are
considered "normal" emitters, defined as vehicles with emissions below 2.0 times the
intermediate life certification standard.  It is important to note that both pollutants are considered
independently when determining whether a vehicle is a high emitter. Thus, a vehicle could be a
high NOx emitter, but a normal HC emitter.

       Although the segregation of vehicles into the "normal" and "high" categories (and their
thresholds) is a somewhat arbitrary modeling method, the concept that average in-use emissions
are driven by a group of vehicles emitting well above the applicable standard is supported by data
from many years of EPA vehicle test and repair programs. This phenomena is contributed to by
the "go/no go" nature of emission control technology components such as EGR valves and air
pumps, and the high sensitivity of emissions to degradations in performance of other critical
components such as the catalyst, oxygen sensor(s) and fuel injectors.  Two important
assumptions in the development of BERs for Tier 1  and later vehicles are a) the rate at which
vehicles malfunction and become high emitters is independent of the certification standard level,
and b) the average emission levels for high emitters becomes higher relative to the standard as
the certification standard becomes lower.

       Under this methodology, average in-use emissions are computed as a weighted average of
high and normal emitters.  Figure 1 is a general graphical  view of the concept with the y-axis
representing emissions in grams per mile, (grams for start emissions) and the x-axis representing
mileage.5 The three lines presented in Figure 1  reflect a) the average or basic emission rate, b)
the normal emitter emission rate, and c) the high emitter emission level.
       5MOBILE6 uses vehicle mileage as a surrogate for vehicle age. Age and mileage are used interchangeably
throughout this document.

-------
                                         Figure 1
               GENERAL EMISSION FACTOR SCHEMATIC
        EMISSIONS
                                          MILEAGE
       The basic emission rate is shown as Line A. This line represents the average emissions of
the fleet as a function of both normal emitters and high emitters.

       Line B represents the average emissions of normal-emitting vehicles. These are the
vehicles which have emission control systems which are generally performing as designed. The
line is shown as a linear function of mileage to reflect the gradual deterioration that normal
vehicles experience, primarily due to catalyst degradation over the life of the vehicle. Normal
emitter emissions are generally expressed by a least squares linear regression of emissions
versus mileage.

       Line C represents the average emissions of high-emitting vehicles.  This line is a flat
horizontal line because emissions from these vehicles do not appear to be a strong function of
mileage, based on previous analysis of Tier 0 data6 and born out by analyses of Tier 1 NOx data
presented in the following section.  The underlying phenomena expressed here is that emission
control malfunction will drive high emissions regardless of vehicle mileage; as discussed in
subsequent sections, what changes as the vehicle ages is the probability of malfunction, rather
than the emission levels resulting from a malfunction.

       Line A represents the weighted average of lines B and C, based on appropriate weighting
factors for normal and high emitters. On a fleet-wide basis, this weighting factor represents the
fraction of high emitters in the fleet, as a function of vehicle age; on a per-vehicle basis, this
weighting factor can be considered to be the probability the vehicle will be a high emitter at a
given age. This weighting factor can be derived at any given vehicle age A (represented by
vehicle mileage) by transforming Equations 1 and 2 into Equations 3 and 4 below.
        Tier 0 I/M report

-------
       Highs + Normals = 1                                               Eqn 1

and

       AVE = High_ave * Highs + Norm_ave * Normals                      Eqn 2

Solving for the variables Highs and Normals produces:

       Highs = (AVE - Norm_ave) / (High_ave - Norm_ave)                  Eqn 3

       Normals = 1  - Highs                                               Eqn 4

Where:

Highs = fraction of High emitters, age =A
Normals = fraction of Normal emitters, age = A
AVE is the average emission rate, age = A
High_ave is the high emitter emission average (independent of age)
Norm_ave is the normal emitter emission average, age = A


4      NOx BERs and Emitter Fractions: No IM/No OBD Case

4.1    TierlLDVs

       The No EVI / No OBD case was developed first because it did not require accounting for
high emitters which underwent repair due to OBD MIL-on; hence, the methodology closely
followed the basic emission rate derivation concept outlined in the previous section. Tier 1
LDVs served as the basis for developing BERs across vehicle class (LDT1 through 4) and
standard level (LEV, ULEV).  Thus, the derivation of these BERs is the first step in the
generation of all BERs for all  OBD/EVI cases, vehicles classes and standards.  For this analysis,
Tier 1 and later BERs were first generated in FTP space and subsequently split into running and
start components, as  discussed in  Section 7; this deviates from the approach used 1981 through
1993 vehicles, for which start and running BERs were developed independently. The start and
running BERs will be used by MOBILE6 as the basis for estimating emissions from Tier 1 and
later LDVs and LDTs.

       The database used to generate No OBD/No EVI BERs for Tier 1 LDVs has been greatly
expanded since the draft report, which relied on 186 LDVs and LDTs tested by the California Air
Resources Board (ARB) as part of Surveillance Programs 12 through 14.  The dataset used for
the updated analysis  included these vehicles plus 40 additional vehicles tested by ARB, 884
vehicles tested by the auto industry, and 74 vehicles tested by EPA.7 In total, 1,122 LDVs and 62
       715 EPA vehicles had repeat tests.

-------
LDTs were used for the analysis.  All were model year 1988 and later certified to a 50,000 mile
intermediate useful life NOx standard of 0.4 g/mi; 1,041 under California's Tier 0 standards, and
143 under the Federal Tier 1 standard.8

       The first step in assessing the updated dataset was to establish whether it was appropriate
to consider all of the vehicles together, or disaggregate the analysis by vehicle class (LDV or
LDT).  This issue was raised by Applied Analysis, GM and AAM who all suggested that LDVs
and LDTs behave differently in-use, and should be modeled separately. We assessed this issue
by performing a multiple regression analysis on NOx emissions for the entire LDV/LDT1 normal
emitter sample using a model with vehicle class (LDV or LDT) as a factorial and mileage and the
cross product of mileage and class as continuous variables.  The purpose of this test was to
determine whether there is a statistically significant difference between either the intercept or
deterioration  rate for LDVs and LDTs certified to the same standard. The results are shown in the
table below:
Parameter
Intercept
Mileage
Vehicle Class (Factorial)
Vehicle Class * Mileage
P-Value
O.0005*
O.0005*
0.254
0.447
Significant?
Yes
Yes
No
No
* P-Value equals zero to three significant digits

       The lack of significance for vehicle class and class/mileage cross product indicates that
there is no statistically significant difference in the intercept or deterioration rate for in-use LDVs
and LDTs certified to the 0.4 g/mi standard. We therefore concluded that it was appropriate to
aggregate the LDV and LDT data points at the 0.4 g/mi standard level.

       The next step was to establish whether it was appropriate to disaggregate the analysis by
certification standard class (Tier 0 and Tier 1).  This issue has been raised by AAM, who
suggested that the advent of Tier 1 standards and OBD would inherently reduce deterioration
rates from Tier 0 vehicles, even if certified to the same standard. We assessed this issue by
performing a multiple regression analysis on NOx emissions for the entire LDV/LDT1 normal
emitter sample using a model with standard class (Tier 0 or Tier 1) as a factorial and mileage and
the cross product of mileage and standard class as continuous variables.  The purpose of this test
was to determine whether there is a statistically significant difference between either the intercept
or deterioration rate for Tier 0 and Tier 1 vehicles certified to the same standard. The results are
shown in the table below:
        The raw datasets are contained in the Microsoft Excel files ARB.XLS, AUTO.XLS and EPA.XLS,
located in the Tier 2 Docket.

-------
Parameter
Intercept
Mileage
Standard Class (Factorial)
Standard Class * Mileage
P-Value
<0.0005*
<0.0005*
0.166
0.504
Significant?
Yes
Yes
No
No
* P-Value equals zero to three significant digits

       The lack of significance for standard class and class/mileage cross product indicates that
there is no statistically significant difference in the intercept or deterioration rate for Tier 0 and
Tier 1 vehicles certified to the 0.4  g/mi standard.  We therefore concluded that it was appropriate
to include the Tier 0 0.4 g/mi vehicles in the analysis of Tier 1 emission rates.
4.1.1   Normal and High Emitter Emission Level

       The average FTP normal emitter emission level was obtained by separating the normal
emitters from the high emitters in the EPA/Auto/ARB sample according to the "2.0 times the
standard" criteria (i.e. all vehicles in the sample above 0.8 g/mi were defined as high emitters).
Using this cutpoint, 31 vehicles were defined as high emitters, and the remainder (1,153) were
defined as normal emitters.9  Figure 2 presents NOx emissions for the normal emitters versus
mileage,  broken down by the three data sources.
        One EPA vehicle with repeat tests had NOx results on either side of the high emitter threshold - one at
0.21 g/mi, and one at 1.31 g/mi. This vehicle was classified as a high emitter.

-------
                        Figure 3 - 0.4 g/mi NOx Normal Emitters
     o
     'en
     c/>
     E
     LU
     X
     O
            0            50000           100000          150000
                  25000           75000          125000           175000

                               Vehicle Mileage
       The EPA and Auto data were generated on relatively new vehicles (generally under four
years old); this is reflected by the scarcity of vehicles from either dataset above 75,000 miles.
The ARB dataset, on the other hand, contains a broader range of vehicle ages and mileages, with
several vehicles exceeding 75,000 miles. Despite the differences in sample makeup, we believe
the combined dataset is appropriate for determining normal emitter emissions. Our basis for this
is the fact that although the intercept of the EPA and Auto samples are much lower relative to the
ARB sample, the deterioration rates are much higher; because of this, we expect that total
emissions over the life of an average normal emitting vehicle would be comparable between the
three samples. Given that we cannot assess which program is "more representative", and because
the results from all of the samples appear reasonably consistent, the most straightforward
approach for generating the NOx BER for normal-emitting Tier 1 LDVs and LDTs is to fit a
linear regression of FTP emissions versus mileage through the entire sample,  as represented by
the "Total Population" line in Figure 2.  The result of this regression is shown in Equation 5 (the
variable 'odom' is in units often thousand miles).
       Norm_Ave(g/mi)
                       10   _
0.153 + 0.02941  *odom
Eqn 5
       We used a similar approach for determining the NOx BER for high-emitting
       10,
        Complete results for the regression equations and averages presented in equation form are contained in
Appendix C.

-------
Tier 1 LDVs and LDTs.  Figure 3 shows NOx emissions for the high emitters versus mileage,
broken down by the three data sources.

                         Figure 4 - 0.4 g/mi NOx High Emitters
3.0 •
2.5 •
1
en
§ 2.0 -
'en
en
LU
X A r-
O 1-5 '
D.
LL
1.0 •

.5
C


V
•v


T"
^
V v V

-------
representativeness of recruitment-based11 in-use emission testing program become relevant.
Our analysis of in-use emission rates for Tier 0 LDVs and LDTs for MOBILE6 found a clear
difference between average emissions of vehicles tested in voluntary recruitment-based FTP
emission testing programs, and of vehicles tested in I/M programs (in which participation is
mandatory).12  This offset has been attributed to so-called "recruitment bias," in which owners of
higher-emitting vehicles are less likely to respond to solicitations for participation in voluntary
programs. For Tier 0 emission rates in MOBILE6, this offset was translated to a "high emitter
adjustment factor". The NOx emission rates for Tier 1 and later vehicles developed above are
based solely on data from voluntary recruitment-based FTP test programs. However, a direct
comparison between data from these test programs and I/M data on Tier 1 vehicles (as was done
for Tier 0 vehicles) cannot be performed; because of the relatively recent implementation of Tier
1 and standards and the tendency for new vehicles to be waived from I/M participation early in
their life, sufficient EVI data are not available on Tier 1 vehicles to derive a high emitter
adjustment in a manner similar to that for Tier 0 vehicles. Because the FTP data used to generate
the Tier 0 and Tier 1  NOx emission rates were collected within the same test programs, we
believe it is necessary to apply a high emitter adjustment factor to the Tier 1  FTP data sample to
account for the lack of representation of high emitters, as was done with the Tier 0 FTP data.  As
discussed below, the magnitude of this adjustment has been decreased from the level presented in
the draft report.

       Our draft methodology also included a second adjustment to "remove" the impact of
California's I/M program, which the vehicles included in the ARE sample were subjected to.
GM and AAM questioned the methodology used for generating this adjustment.  While this issue
is still of concern, we are not including such an adjustment in our final methodology. This
adjustment was in fact very minor, as it was based on MOBILES I/M credits for an idle I/M
program, which provides little benefit for NOx.  Because we do not believe that MOBILES  can
be used to appropriately correct for such a bias we have eliminated the separate "I/M"
adjustment.

       The high-emitter adjustment is relevant at this stage of the calculation because they only
affect the weighting of high and normal emitters - in other words, they were used to increase
only the number of high emitters estimated in the fleet.  As discussed in Section 3 (and shown in
Equations (1) through(4)), the high/normal weighting factor is a function of normal and high
emitter emission levels and their combined average in-use emission level. Emissions for normal
and high emitters as computed in Equations (5) and (6) were not affected by this  adjustment.
The high-emitter adjustment was applied directly to the in-use average emission levels; based on
normal and high emitter emissions computed from Equations (5) and (6), the high emitter
fractions were then back-calculated using Equations (1) through (4).  This methodology is
       n"Recruitment-based" in-use emission testing programs are defined here as programs in which vehicles are
procured for testing from the general population in exchange for money and/or other incentives.  Participants are
initially contacted through mail or phone solicitation based on registration mailing lists, and participation is strictly
voluntary.

       12Enns, et al, "Determination of Running Emissions as a Function of Mileage for 1981-1993 Model Year
Light-Duty Cars and Trucks", EPA Draft MOBILE6 Report No. M6.EXH.001, March 1999

-------
detailed in the following steps:

(1)    The unadjusted average in-use NOx emission level as a function of mileage was
       computed from the combined EPA/Auto/ARB dataset.  This FTP NOx emission average
       was obtained by linear regression of the raw data versus mileage. The resulting
       regression equation is shown below:

       NOxFTP(g/mi)     =       0.117 + 0.04617 * Odom                   Eqn 7
(2) The second step was to generate an additive high emitter correction factor to account for the
potential effects of recruitment bias on the EPA/AAM/ARB sample.  While none of the
stakeholder comments disgreed entirely with the application of this adjustment, opinions as to the
appropriate level of this adjustment were varied. Colorado and Applied Analysis contended that
this adjustment should be higher than proposed, because of concerns with the representativeness
of the Dayton I/M data on which the adjustment was based; GM and AAM simply labeled our
original methodology as arbitrary.  In light of these comments, we have reevaluated how this
NOx high emitter adjustment for Tier 1 and later  vehicles should be derived. We still consider it
appropriate to base the Tier 1 adjustment on the Tier 0 high emitter adjustment in the absence of
sufficient I/M data on Tier 1 vehicles. To assess whether the Tier 0 adjustment should be
reduced for Tier 1, we compared  average FTP emission results for Tier 0 and Tier 1 high
emitters,.   The Tier 1 high-emitter BER calculated in Equation (6) (1.294 g/mi) is 56 percent
lower than average high-emitter emissions for 1988-1993 ported fuel injection (PFI) Tier 0 LDVs
planned for MOBILE6 (2.96 g/mi).13'14  Thus, the high emitter adjustment applied in the
derivation of Tier 1  and later emission rates was calculated by reducing the high emitter
adjustment for 1988-1993 PFI Tier 0 LDVs by 56 percent, versus 25 percent used in draft report.
This is reflected in Equation (8).

       HECF  =      0.00466 * Odom                                        Eqn 8

       Where:

       HECF is the high emitter correction factor.
       Odom is the mileage, in ten thousands
(3)    The corrected in-use average NOx FTP results (C_NOXFTP) were obtained by applying
       the high emitter adjustment factor from Equation 8 to the raw NOx FTP value from
       13Enns, et al, "Determination of Running Emissions as a Function of Mileage for 1981-1993 Model Year
Light-Duty Cars and Trucks", EPA Draft MOBILE6 Report No. M6.EXH.001, March 1999; Glover and Carey,
"Determination of Start Emissions as a Function of Mileage and Soak Time for 1981-1993 Model Year Light-Duty
Vehicles", EPA Draft MOBILE6 Report No. M6.STE.003, March 1999

       14Calculated values for 1988-1993 Tier 0 PFI NOx high emitters vary slightly by mileage; this represents
the value at 68,000 miles, the average in-use mileage for LDVs based on MOBILE6 travel fraction.

-------
       Equation 7, as shown in Equation 9.


       C_NOXFTP   =      (NOx FTP + HECF)                             Eqn 9
(4)    The fraction of high emitters in the fleet under NO I/M and NO OBD conditions (labeled
       HighBASE) is calculated by inserting the value of C_NOXFTP, Norm_ave, and High_ave
       (from Equations (5) and (6)) into Equation (3).  Mathematically, this is shown in
       Equation 10.

       HignBASE = (C_NOXFTP - Norm_ave) / (High_ave - Norm_ave)          Eqn 10

       The resulting high emitter fractions for the No OBD/No IM case for ages one through 25
       are shown in Appendix A, Table A-l (mileage levels as a function of age are shown in
       Appendix A, Table A-3). Because the intercept from Equation (7) is less than the
       intercept from Equation (5), Equation (10) resulted in a negative high-emitter fraction for
       vehicle ages less than two. For these years the high-emitter fraction was set to  zero, in
       effect estimating that vehicles won't become high emitters until at least the third year of
       their life.

       The average in-use FTP-based NOx emission level for  Tier 1 LDVs without OBD or I/M
can be calculated at any vehicle age using Equation (2), based  on the terms "High_ave"
(Equation 6), "Norm_ave" (Equation 5), and "Highs" (Equation 10). According to Equation (4),
"Normals" is simply 1 - Highs.
4.2    All Other Standard Levels and Vehicle Classes

       NOx BERs for No OBD / No EVI conditions are required for LDVs under post-Tier 1
standards, and all Tier 1 and later LDTs (LDT1 through 4).  At the time of this analysis, EPA
was not aware of any dataset which provided an adequate sample of in-use data for these
combinations of vehicles class and standard level. BERs for these classes were derived from the
Tier 1 LDV BERs developed above, using a set of specific assumptions about how average
emissions for normal and high emitters, and high emitter fraction, would apply across standard
level and class.

4.2.1   Normal Emitters

       It was assumed that for post-Tier 1 LDVs and Tier and later LDTs, normal emitter NOx
emissions will on average maintain the same performance relative to the applicable 50,000 mile
standard as Tier 1 LDVs.  Thus, normal emitter BERs for all post-Tier 1 LDVs and Tier 1 and
later LDTs were developed by calculating the ratio of the applicable standard level ("std") to the
Tier 1 LDV standard, and applying this ratio to the Tier 1 BER (zero-mile level and deterioration
rate), as follows:

-------
Norm_ave ZML(std) = Norm_ave ZML(tierlldv) * (Cert Std(std) / Cert Std(tierl Idv))

Norm_ave DR(std)  = Norm_ave DR(tierl Idv) * (Cert Std(std) / Cert Std(tierl Idv))
Eqn lla

Eqnllb
For example, normal-emitting LDV LEV BERs were generated by multiplying the normal-
emitting Tier 1 ZML and DR from Equation (5) by 0.5 (0.2 g/mi divided by 0.4 g/mi, the 50,000
miles standards).

       The rationale behind this approach is that basic emission levels for properly operating
vehicles should receive the full benefit of reduced standards, including lower deterioration rates
for lower standard levels.  This approach presumes that normal emitters for all standards and
vehicle classes will on average achieve the same compliance margin ("headroom") with the
50,000 mile certification standard as the normal emitters observed in the EPA/Auto/ARB  0.4
NOx data.

       With regards to trucks, this approach acknowledges that LDT emission performance
relative to the standard is expected to be similar to LDVs because of increased similarities in a)
emission control technology, b) manufacturer design practices, and c) driving and usage patterns.
In general, sufficient in-use data do not exist to empirically establish emission rates for  Tier 1
and later LDTs; some level of judgement about how these vehicles will perform in-use is
therefore required.  Stakeholder comments varied on how LDTs should be treated relative to
LDVs.  Applied Analysis suggesting that LDTs have higher emissions than LDVs; however,
their analysis was based on a comparison of I/M data on LDVs and LDTs which doesn't appear
to consider higher certification standards for trucks. Our emission rates  do project higher
emissions for LDTs certified to higher standards, but not at the same certification standard.  GM
and AAM suggested that our emission rates for LDTs were too high, based on certification data
and in-use Tier 0 data; their analysis  has several flaws, however, as discussed in Section 27.4.(J)
of the Tier 2 Response to Comments.

       To assess this issue, we performed two analyses: first comparing Tier  1 LDVs and LDTs
certified to the same standard, then LDVs and heavier LDTs certified to  their respective Tier 1
standards. A univariate regression analysis was performed on NOx emissions for the entire
LDV/LDT1  normal  emitter sample using a model with vehicle class (LDV or LDT) as a factorial
and mileage and the cross product of mileage and class as continuous variables.  The purpose of
this test was to determine whether there is a statistically significant difference between either the
intercept or deterioration rate for LDVs and LDTs certified to the same standard. The results are
shown in the table below:
Parameter
Intercept
Mileage
Vehicle Class (Factorial)
Vehicle Class * Mileage
P-Value
O.0005*
O.0005*
0.254
0.447
Significant?
Yes
Yes
No
No

-------
* P-Value equals zero to three significant digits

       The lack of significance for vehicle class and class/mileage cross product indicates that
there is no statistically significant difference in the intercept or deterioration rate for in-use LDVs
and LDTs certified to the same Tier 1 standard.
       To assess this issue for heavier trucks, an analysis of in-use emissions from heavier LDTs
was performed on a sample of 50 discrete tests of 38 late-model Tier 1 LDT2s and LDT3s tested
by EPA (37) and ARB (1).  The majority of these vehicles were under 3 years old when tested
and had an average mileage below 30,000 miles (the maximum mileage was 93,000). The
emission levels of this sample reflected its newness; all tests complied with the 50,000 mile NOx
standard of 0.7 grams per mile.  Overall, we do not consider this sample adequate for generating
in-use emission rates directly; but, it is useful for comparing emission performance relative to the
standard and early in a vehicle's life with a sample of comparable Tier 1 LDVs. To determine
whether the emission performance of the LDTs relative to the 0.7 g/mi standard was different
from the performance of comparable Tier 1 LDVs relative to the 0.4 g/mi standard, we
performed a multiple regression analysis on a sample consisting of the LDT2/3s and a subset of
Tier 1 LDVs which complied with the 0.4 gram/mile standard.  The dependent variable for this
analysis was "headroom", calculated by dividing each emission test by the standard (to normalize
across the two standard levels).  Vehicle class (LDV or LOT) was a factorial and mileage and
the cross product of mileage and class were continuous variables. The results are shown below:
Parameter
Intercept
Mileage
Vehicle Class (Factorial)
Vehicle Class * Mileage
P-Value
O.0005*
O.0005*
0.813
0.912
Significant?
Yes
Yes
No
No
* P-Value equals zero to three significant digits

       The lack of significance for vehicle class and class/mileage cross product indicates that
there is no statistically significant difference in the intercept or deterioration rate relative to the
standard for a sample of comparable Tier 1 LDVs and LDT2/3s, despite a higher emission
standard for the LDT2/3s.

       Based on these analyses, we believe it is appropriate to treat LDVs and LDTs the same
with regard to generating Tier 1 and later emission rates.

4.2.2   High Emitters

       High Emitter BERs are meant to estimate emissions from vehicles that significantly
exceed their certification standards due to malfunctioning emission control systems. A key
assumption in the development of high-emitter BERs is that, as emission  standards are lowered

-------
(or "raised" for truck BERs), emission levels for high emitters will not be changed in proportion
to the standard change. This approach is based on our judgement that post-Tier 1 vehicles are
increasingly reliant on catalyst conversion efficiency to maintain compliance with the standard,
so that tailpipe emissions are increasingly sensitive to degradation of catalyst conversion
efficiency. Comments from Applied Analysis supported this reasoning, in fact contending that
increased dependence on catalyst conversion may mean that high emitter emissions should not be
reduced with decreases in certification standard. Conversely, GM and AAM supported a
reduction in high emitter emissions fully proportional to the decrease in certification standards.

       To assess this issue, we analyzed available engine-out emission data for Tier 1 vehicles
tested as part of the Supplemental Federal Test Procedure (SFTP) review project, and a combined
sample of LEV/ULEVs tested as part of the auto industry's sulfur test program.  This analysis
shows that LEV/ULEV high emitters would actually have better catalyst conversion efficiency
than Tier 1 high emitters, although their emission levels would be higher relative to the 50,000
mile standard than Tier 1 high emitters.  This is illustrated in the following table:
NOx
Tier 1
LEV/ULEV
HC
Tier 1
LEV/ULEV
Engine-Out FTP
NOx Emissions
2.33 g/mi15
2.59 g/mi16
Engine-Out NMHC
Emissions
1.90 g/mi4
1.50 g/mi5
Projected catalyst
efficiency at 5 OK
standard
83%
92%
Projected catalyst
efficiency at 5 OK
standard
87%
95%
Projected catalyst
efficiency at high
emitter emission rate
45%
63%
Projected catalyst
efficiency at high
emitter emission rate
12%
18%
       As shown, the reductions necessary to meet the tighter LEV standards come primarily
from improvements in the catalyst.  This means that similar drops in catalyst conversion
efficiency will more adversely affect the emissions of LEVs, and result in higher emissions
relative to the 50,000 mile certification standard than for Tier 1  vehicles. This analysis shows
that degradation in NOx catalyst efficiency between normal and high emitters would actually be
less for LEVs than for Tier 1 vehicles under our assumption that LEV high-emitter emissions
would only be reduced by !/2 of the reduction in the 50,000 mile certification standards.
       15
vehicles)
         Source: "Supplemental FTP Emissions Database", CD distributed by AAMA/AIAM, January 1997 (21
       16C
         Source: AAMA Sulfur Test Program (9 vehicles). Higher NOx engine-out results for LEVs are
considered a function of manufacturer's attempts to improve HC performance and catalyst light-off through engine
calibration strategies, such as a leaner fuel mixture at startup (as indicated by reductions in engine-out HC).

-------
       A second approach to assessing our estimates of high emitter emissions is to analyze
trends in available Tier 0 and Tier 1 data (this analysis was performed only for NOx, since
available Tier 1 data for HC is from relatively new vehicles and does not include any high
emitters).  Although the combined EPA/AAM/ARB dataset does not include any post-Tier 1
vehicles, the trend towards less-than-proportional decreases in high emitter emissions as
certification standards are lowered does bear out based on a very limited sample of catalyst-
equipped (1988 and later) LDV/LDT1 high emitters:
1
NOx Standard






1.0 (Tier 0)
0.7 (Cal Tier 0


0.4 (Tier 1)


2
Percent
reduction in
certification
standard



-
30%
(from 1.0
standard)
43%
(from 0.7
standard)
3
High
emitter
sample
size



8
O


31


4
Average
emissions
(g/mi)




2.46
1.85


1.29


5
Percent
reduction in
average high
emitter
emissions from
previous
standard level
-
25%


30%


6
Percent
proportional
(Column 5 /
Column 2)




83%


70%


       Column 2 in the table above shows the percent reduction in certification standard level,
from Tier 0 through Tier 1; Column 5 shows the percent reduction in high-emitter emissions at
these standard levels. The fact that the values in Column 5 are lower than in Column 2 means
that emission levels for high emitter are not reduced in proportion to the certification standard,
which supports the underlying assumption for our development of post-Tier 1 high-emitter
emission rates.  Column 6 shows the decrease in high emitter emissions relative to the decrease
in certification standards (comparable to our estimate of 50 percent proportional for post-Tier 1
high emitters). This value decreases as the standard level drops from 1.0 g/mi to 0.7 g/mi, and
0.7 g/mi to 0.4 g/mi; in other words, decreases in high emitter emissions become less and less
proportional to decreases in the standard, for lower standards.  We expect this trend to continue
for post-Tier 1 standards, based on our analysis of engine-out emissions for post-Tier 1 vehicle
presented above; extrapolating these results, we believe it is reasonable to assume  only a 50
percent proportional drop in high emitter emissions as standards are reduced beyond Tier 1

-------
levels.

       BERs for high emitters were thus developed for post-Tier 1 LDVs and Tier 1 and later
LDTs by taking the average of the Tier 1 LDV high emitter NOx HER (1.278 g/mi) and the BER
that would result if the ratio of 50,000 mile standards were applied to the Tier 1 BER, according
to Equation 12:

High_ave(std)  = average[High_ave(tierlldv) * (Cert Std(std) / Cert Std(tierl Idv), High_ave(tierlldv)]   Eqn 12

       The result of this average is a high emitter BER which is 50 percent proportional to the
change in standard, reflecting that malfunctioning vehicles will derive some benefit on average
from lower emission standards, but not the full benefit as afforded to normal emitters.  High
emitter emissions are therefore tied closely to the Tier 1  LDV high emitter level, and the change
in high emitter emissions is "muted" relative to changes in standard.

       Normal and high emitter BERs for LDVs and LDTs complying with Tier 1, LEV and
ULEV standards are presented in Appendix B.

4.2.3   High Emitter Fractions

       The rate at which vehicles become high emitters under the No OBD / No EVI scenario was
assumed constant for all vehicles and standard classes.  Thus, the age-based high emitter
fractions developed in Equation 10 and presented in Appendix A were applied to Tier 1 and later
BERs for all classes.  The rate of emission control technology malfunction was assumed the
same between LDVs and LDTs, given that their emission technology and usage patterns are
increasingly  similar. Reduced certification standards are also not expected to influence the rate
at which emission control technology malfunctions, because  a) manufacturer's design and
durability practices are not expected to differ between Tier 1  and later standards, and b) many
cases of emission control degradation and/or malfunction are owner-induced, and hence outside
the manufacturer's liability for in-use emission performance.  It  should be noted that the high-
emitter fractions in Appendix A are shown to vary by class, due to differences in accumulated
mileage at a given age.  At the same mileage, the high emitter fractions are the same across all
classes.

       The No OBD / NO I/M average in-use NOx emission rate for vehicle/standard = (V,S)
can be calculated at any vehicle age using Equation (2), based on a) the Tier 1 LDV "High_ave"
and "Norm_ave" terms from Equations (5) and (6) adjusted as described above based on the
(V,S) standard level, and b) the base (No OBD / No EVI) high emitter fractions from Appendix A,
Table A-1.
5      NMHC/NMOG BERs and Emitter Fractions: No IM/No OBD Case

       The development of NMHC/NMOG BERs shared many of the methodological
assumptions outlined for NOx in Section 3. As with NOx, NMHC BERs for Tier 1 and NMOG

-------
BERs for LEV and later LDVs and LDTs were developed off of "base" LDV BERs; the primary
difference between the methodologies for the two pollutants was the source of the base data. At
the time of this analysis, sufficient in-use data on vehicles complying with EPA's Tier 1 NMHC
standards (for any vehicle class) were not available. The EPA/Auto/ARE dataset used for the
NOx analysis included 143 LDVs/LDTls and 38 LDT2/3's certified to the Federal Tier 1
standard. However, these vehicles were generally less than four years old at the time of testing.
As such, these data were judged to be inadequate for assessing overall in-use emission
deterioration of Tier 1 LDVs. As discussed below, however, we did use these data to validate the
zero-mile level of the predicted Tier 1 normal emitter emission rate, and to verify the
appropriateness of basing emission rates  for heavier LDTs on LDV emission rates.

       Tier  1 and later HC BERs were based on proposed MOBILE6 BERs for model year 1988
through 1993 Tier 0 LDVs with ported fuel injection (PFI).  These BERs were developed based
on several thousand vehicles tested by auto manufacturers, EPA, and through I/M programs.
The Tier 0 emission rates were considered a good starting point for developing Tier 1 and later
BERs because emission  control technology used on later Tier 0 vehicles (e.g., 3-way catalysts
and ported fuel injection) are generally similar to those used on Tier 1 and later vehicles.  A
comprehensive treatment of Tier 0 BERs and the datasets used to derive them are contained in
other MOBILE6 reports, and thus are not presented here. For this analysis  a simplifying step was
performed to generate a  linear form of Tier 0  normal-emitter BERs, since as proposed for
MOBILE6 these are expressed as nonlinear functions.17  The resulting normal and high emitting
Tier 0 BERs (expressed  as Total Hydrocarbon, or THC) used as a basis for Tier 1 and later BERs
are shown in Equations (13)  and (14).

       Norm_Ave(g/mi)    =     0.16 + 0.0186 * odom                     Eqn  13

       High_Ave(g/mi)     =     2.076                                    Eqn  14

"odom" represents mileage in units often thousand miles.

       Using the values presented in Equations (13) and (14) as a starting point, normal and high
emitter NMHC/NMOG BERs for all Tier 1 and later LDVs and LDTs were developed using the
identical methodology as for NOx (described in Sections 3.2.1 and 3.2.2) based on the ratio of
the applicable 50,000 mile standard level to the Tier 0 level of 0.41 g/mi. Since Tier 1 standards
are expressed as NMHC and LEV and later standards are expressed as NMOG, the shift from
THC to these pollutants  is accounted for in the standard ratios.

       We assessed the  validity of the predicted Tier 1 zero-mile level (intercept) by comparing
this level with emission  from nine Tier 1 LDVs/LDTls (13 tests) within the EPA/Auto/ARB
       ^Equations 13 and 14 were derived by first combining the running and start emission rates for normal-
emitting 1988-1993 PFI LDVs according to FTP weightings. The raw running and start emission rates can be found
in MOBILE6 report M6.IM.001, "MOBILE6 Inspection/Maintenance Benefits Methodology for 1981 through 1993
Model Year Light Vehicles". At a given mileage, FTP emissions were derived from start and running emission
rates according to Equation 26. For normal emitters, the combined non-linear FTP emission rates were then
regressed by mileage to create a simple linear model, resulting in Equation 13.

-------
dataset with odometer readings below 5,000 miles. This comparison is shown in the following
table:

Predicted Tier 1
LDV/T1 Zero-Mile
Level
Actual Tier 1
LDV/T1 below 5,000
miles
Average FTP NMHC
Emissions
0.098 g/mi
0.091 g/mi
95% CI Lower
Bound
-
0.083 g/mi
95% CI Upper Bound
-
0.122 g/mi
       As shown, the difference between our predicted value and the actual emission results is
not statistically significant, since the predicted average falls within the 95 percent confidence
band of the 5,000 mile data.

       In their comments, Applied Analysis provided a summary of I/M data which showed
higher emissions for LDTs relative to LDVs over the 1994-1995  model years, and suggested that
LDTs should be modeled separately from LDVs.  Again, this analysis doesn't appear to consider
higher certification standards for trucks; their results are consistent with our approach to
generating emission rates for LDTs with higher certification standards.

       To assess the appropriateness of basing emission rates for heavier LDTs on the LDV
emission rates, we compared a sample of Tier 1 LDVs/LDTls with the sample of Tier 1 LDT2/3s
discussed under Section 4.2.1.  As mentioned, we do not consider either sample adequate for
generating in-use emission rates directly; but, it is useful for comparing emission performance
relative to the standard. To determine whether the emission performance of the LDTs relative to
the 0.32 g/mi standard was different from the performance of comparable Tier 1  LDVs relative to
the 0.25 g/mi standard, we performed a multiple regression analysis on a sample consisting of
normal-emitting LDT2/3s (all were below the standard except for one vehicle well above the
high-emitter cutoff, and one at 1.5 times the standard; we excluded  the high-emitting vehicle)
and a subset of Tier 1 LDVs below 1.5 times the 0.25 gram/mile  standard.  The dependent
variable for this analysis was "headroom", calculated by dividing each emission test by the
standard (to normalize across the two standard levels).  Vehicle class (LDV or LOT) was a
factorial and mileage and the cross product of mileage and class were continuous variables.  The
results are shown below:

-------
Parameter
Intercept
Mileage
Vehicle Class (Factorial)
Vehicle Class * Mileage
P-Value
<0.0005*
<0.0005*
0.218
0.608
Significant?
Yes
Yes
No
No
* P-Value equals zero to three significant digits

       The lack of significance for vehicle class and class/mileage cross product indicates that
there is no statistically significant difference in the intercept or deterioration rate relative to the
standard for a sample of comparable Tier 1 LDVs and LDT2/3s, despite a higher emission
standard for the LDT2/3s.

       The high emitter fractions developed for 1988 through 1993 Tier 0 PFI LDVs were used
as the No OBD/No EVI emitter fractions for all Tier 1 and later vehicles, because of the expected
similarity in vehicle malfunction rates across standard level and vehicle class as discussed in
Section 3.2.3.  Subsequent to publication of the draft report, we found an error in these rates;
The corrected rates show a lower fraction of high emitters than originally reported. These
corrected rates are shown in Appendix A, Table A-2.

       The No OBD / No EVI average in-use NMHC/NMOG emission rate for vehicle/standard
= (V,S) can be calculated at any vehicle age using Equation (2), based on a) the Tier 0
"High_ave" and "Norm_ave" terms from Equations (13) and  (14) adjusted based on the (V,S)
standard level, and b) the base (No OBD / No EVI) high emitter fractions from Appendix A, Table
A-2.
6
Effects of OBD and OBD-based I/M for NOx and HC
       Separate BERs were developed for all standard and vehicle classes to account for the
effects of OBD and OBD-based I/M programs.  The methodology used to account for these
programs were identical for NOx and HC, based on reducing the fraction of high emitters in the
fleet from the No OBD/No EVI case. Thus, emission levels for normal and high emitters were not
changed under these programs,  only the fraction of highs in the fleet. This methodology
introduces a new category of vehicle into the fleet: "Repaired" emitters.  These vehicles are high
emitters that are flagged by an OBD system and undergo successful repair. For the OBD-only
and OBD/EVI cases, these vehicles are treated distinctly from normal and high emitters (although
our revision of OBD after-repair levels means that normal emitters and repaired emitters have the
same emission level for most of a vehicle's life).

6.1    OBD Effectiveness
       OBD effectiveness is defined by three parameters: a) the probability the OBD system will
detect a failure (MEL-on Rate), b) the probability an owner will respond to a MEL-on (Response

-------
Rate), and c) the average after-repair emission level for responding vehicles (Repair Level). In
general sufficient in-use data are not available to determine in-use patterns for these parameters,
although limited data has recently been published which allows some assessment of our estimates
for MIL-on rate and post-repair emissions.18  Our estimates for the projected likelihood of
malfunction detection, owner response and repair level are presented in Sections 5.1.1 through
5.1.3 for both the OBD-only and OBD/EVI cases.

6.1.1   MIL-on Rate

       For all vehicle classes and standards, we are maintaining our original proposal that OBD
systems will detect (i.e., set an  appropriate code and illuminate the MIL) 85  percent of high
emitters. Because high emitters are defined independently for HC and NOx, this response rate is
assumed to apply equally to both pollutants. The remaining 15 percent of fleet will not be
identified, and thus will remain in the fleet as high emitters. No deterioration in the ability of the
OBD system to correctly identify high emitters is assumed. Because this parameter is solely
dependent on the vehicle's OBD system, it is the same for I/M and non-I/M  areas.

       Stakeholder comments  were mixed on this issue; Colorado supported our proposal, while
AAM contended that the MIL-on rate should be increased to as high as 100 percent.  Sufficient
data are not available to empirically estimate the effectiveness of and response to OBD systems
in the field over the life of a vehicle, and the stakeholder comments did not provide any
quantitative data on which to base a change.  AAM contends that our assumptions should reflect
how OBD systems are required to perform in-use; using this logic, it could also be argued that
our estimates of in-use  emission rates should assume that vehicles only comply with the FTP
emission standard, since this is the requirement manufacturers are held to in-use.  However, in-
use emission test programs repeatedly verify that vehicles do exceed their FTP emission
standard, particularly beyond the useful life mileage point on which the standards are based.
Accordingly, it is reasonable to assume that not all OBD systems will perform as intended by the
manufacturer, particularly beyond the mileage level for which they are held liable for its
performance.

       Two potential limitations of current OBD systems must be taken into account when
predicting the real-world performance of these systems.  First, current OBD systems are required
to identify problems with individual components of the emission control system which can cause
a vehicle to exceed its emission standards by a factor of 1.5. It is possible in-use,  however, for a
combination of minor problems to cause a similar level of emission increase. Current OBD
systems will most likely not detect such a situation, as long as any individual problem is minor.
Second, catalyst  performance monitoring is still limited by available technology.  The HC
conversion efficiency of a catalyst is usually inferred from the oxygen storage capacity of the
catalyst; the assumption being that if a catalyst experiences sufficiently high temperatures to
significantly reduce its HC conversion efficiency, the same high temperatures will have
significantly reduced its oxygen storage capability.  This is supported by lab-based correlations,
       1 8
        Gardetto. E., and Trimble, T., "Evaluation of On Board Diagnostics for Use In Detecting Malfunctioning
and High Emitting Vehicles", EPA Report EPA420-R-00-013, August 2000

-------
but may not be as accurate a predictor in-use. Adequacy of the CO and NOx conversion
efficiencies of the catalyst are not even indirectly measured, but are assumed to be the same as
the HC conversion efficiency.

       EPA's recently published results for the assessment of OBD performance on high-
emitting vehicles echoes some of these themes.  This study reports that of 21 vehicles over 2.0
times the applicable certification standard, the OBD system caught 19 (a 90 percent success rate).
Of 31 vehicles over  1.5 times the applicable certification standard, the OBD system caught 27
(an 87 percent success rate).   The study confirms the concern about the catalyst monitor
discussed above; some of the "missed" vehicles were cases in which the CO and/or NOx
emissions had high emissions, but were not flagged because the HC emissions were not above
the OBD threshold.  It is also important to note that, as OBD technology is relatively new, these
vehicles were all under four years old when tested. As a result, these vehicles are less likely to
exhibit the synergistic effects of multiple minor problems discussed above. Our estimate of OBD
failure detection rate must take into account the average performance of the OBD system under
all conditions, over the entire life of a vehicle.  Given the uncertainty in how these systems will
perform in-use (particularly at higher mileages, where the highest concentration of emission
malfunction will occur), and the technical considerations detailed above, our estimate that on
average 85 percent of emission failures will be detected by the OBD system is not unreasonable.
It goes without saying, however, that this issue will need to be revisited as OBD information
becomes available on older vehicles.

6.1.2   Response/Repair Rate

       In order to obtain emission reductions from a vehicle equipped with an OBD system, not
only must the system correctly identify the vehicle, but the motorist must also respond to the MIL
and take corrective action in a timely manner (response rate) and the vehicle must be fixed
correctly (repair rate). MOBILE6 uses separate estimates for these rates depending on whether
an OBD-based I/M program is being modeled, with different connotations depending on which
case in invoked. For OBD I/M programs, response rate is handled in MOBILE6 through the
estimates of noncompliance and cost waivers applied to the I/M program inputs.  Hence, for the
OBD I/M case the value discussed here reflects only the rate of repair for MIL-on vehicles which
show up to the I/M station and do not qualify for a cost waiver. This repair rate for OBD/EVI
areas is assumed to be 99 percent over the entire life of the vehicle; it was not made 100 percent
simply to reflect the slight possibility that a vehicle with OBD failure is not detected as such in
an I/M lane. The original proposal was for a combined response/repair rate of 90 percent, which
accounted for the combination of cost waivers and noncompliance. This change has been made
to  allow cost waivers and noncompliance to be handled consistently with exhaust-based I/M
benefit calculations discussed in M6.EVI.001; to not make this change would results in double-
counting of cost waiver and noncompliance effects.

       For non-DVI areas, this value reflects a combined response/repair rate for vehicles in
which the MIL is illuminated. This repair/response rate is assumed to be a function of vehicle
warranty.  It is assumed that an owner is much more likely to  respond to a MIL-on when repairs
will be paid for by the manufacturer. Three mileage bins were therefore developed: 1) 0 through

-------
36,000, the standard bumper-to-bumper warranty period; 2) 36,001 to 80,000, for which federal
law mandates that catalysts and electronic control modules (ECMs) remain under warranty; and
3) above 80,000, for which no warranty is in effect (extended warranties are not accounted for in
this methodology).

       Under 36,000 miles, it is assumed that 90 percent of MIL-on vehicles will be repaired.
This is based on the judgment that for new vehicles still under warranty, owners will have little
hesitation in addressing a MIL-on. The 10 percent loss accounts for a small percentage of
owners who will not respond to a MIL-on even with the warranty incentive.

       Between 36,000 miles and 80,000 miles, it is assumed that  10 percent of MIL-on vehicles
will be repaired. This response rate is greatly reduced from the pre-36,000 mile level to account
for the discontinuation of warranty coverage on several emission-related components (e.g.
secondary air, EGR, oxygen sensors, fuel injectors), and reduced willingness of owners to make
emission-related repairs on an aging vehicle in the absence of an I/M program.

       Above 80,000 miles, it is assumed that no MIL-on vehicles will be repaired.  This
assumption reflects the end of warranties, the lower economic value of the vehicle, and the
(further) reduced willingness of owners to make emission-related repairs in the absence of an I/M
program.

       Stakeholder comment on these estimates was mixed.  Colorado supported these estimates,
citing results from their I/M program which show that many newer vehicles are failing their I/M
program due to MIL lights, which "seems to point to general disregard to MIL indications".
AAM and GM contend that response rates during the warranty period should be higher (up to
100 percent), and that post-warranty response rates in non-I/M areas should decrease more
gradually, mirroring the gradual depreciating value of the vehicle. GM also contends that our
response/repair rate in I/M areas assumes that owners will receive repair almost immediately
upon MIL illumination. The issue of owner response, particularly in non-I/M areas, is at this
time strictly a matter of judgement. Recommendations made by AAM and GM reflect their own
judgement, with no data provided to substantiate their claims.  With regard to the duration prior
to repair, MOBILE6 is structured to estimate emissions based on a "snapshot" of the fleet once
per year, meaning that in essence the distribution between "normal", "high" and "repaired"
vehicles is assessed once per year. Implicit within this structure is the assumption that vehicles
have up to one year (or six months on average) to become a high emitter and receive repair.  This
presumes an annual I/M program; MOBILE6 will allow the flexibility for other program
intervals.

6.1.3   Repair Level

       We have revised our estimates regarding the level to which vehicles will be repaired
through response to an OBD system failure.  Our proposal estimated that vehicles on average
would be repaired to  1.5 times the 50,000 mile standard, where it would remain constant (not
deteriorate) for the remainder of its life. This emission level is the maximum allowed before the
OBD light should come on. Colorado supported this approach, while AAM and GM commented

-------
that it is overly conservative, and recommended that the after-repair level be 1.0 times the FTP
standard, without deterioration.

       Based on our analysis of after-repair data from EVI240 programs, we believe that after-
repair emission levels will be lower than our original estimate, but that emissions from these
vehicles will deteriorate in a manner more consistent with normal emitters.  Our revised
approach to modeling after-repair emissions is to therefore assume that upon repairs, high
emitters are returned to the normal emitter emission level at the appropriate mileage point.  On
average this approach actually presumes lower after-repair emissions than for dynamometer-
based exhaust I/M programs; we consider this to be appropriate given that OBD systems will
likely improve diagnosis of emission malfunctions, a conclusion reinforced by EPA's recently
published OBD study.

       The after-repair emission levels are capped at 1.5 times the standard, the threshold for
OBD failure detection.  This occurs at approximately 150,000  miles for NOx, and nearly 240,000
miles for HC (the levels at which the normal emitter equations presented in Equations (5) and
(13) equal 1.5 times the applicable standard - 0.6 and 0.615 g/mi, respectively).  After these
mileage points a third emitter category is required - "repaired" emitters. Repaired emitters are
assumed to have constant emissions at the after-repaired emission level,  although a subset of
these vehicles "migrate" back to the high emitter category.  The emission level after an OBD-
induced repair above these mileage thresholds is assumed to be 1.5 times the applicable 50,000
mile certification standard.  The repaired emitter "BER caps"  are presented across standard and
vehicle class in Appendix B.

6.2    High Emitter Fractions for OBD and  OBD-based I/M

       Equations 15 through 17 were used to calculate the high emitter growth rate under the
OBD and OBD-based I/M scenarios (HighOBD).  Overall, the high emitter fraction in a given year
is a function of a) the number of high emitters in the previous year, b) the base high emitter
"growth rate" in the absence of OBD or I/M, and c) the OBD effectiveness assumptions outlined
in  Section 5.1. The subscript 'i' is the vehicle age.  High(O) is assumed to be zero. MOBILE6
will assign a value of 'odom' for each age 'i'.

       Nonhigh(i)           =      1.0 - HighBASE(i)                                Eqn 15

       Delta_High(i)        =      HighBASE(i)-HighBASE(i-l)                       Eqn 16

       Growth_High(i)     =      Delta_High/Nonhigh(i-l)                       Eqn 17
HighOBD(i)  = HighOBD(i-l) + [(l-OBD)*MIL*Growth_High(i)*(l-HighOBD(i-l))] +
             [(l-MIL)*Growth_High(i)*(l-HighOBD(i-l)]                     Eqn 18

Where:

-------
HighOBD(0)   =      0.0
MIL         =      0.85
Nonhigh     =      the fraction of normal and repaired vehicles
Growth_High =      the growth rate of high emitters (or, the rate at which "nonhighs" migrate
                    into the high emitter category)

'OBD' is the OBD response rate; 0.90 for OBD-based I/M, and 0.90/0.10/0.0 for mileage bins (0
- 36K), (36K - 80K), and (80K+).

An elaboration on Equations 15 through 18 is as follows: for a given vehicle age, the fraction of
high emitters is a) the number of highs from the year before, plus b) the number of MIL-on highs
added in that year due to OBD nonresponse (a function of "nonresponse" rate, MIL-on rate, and
the high emitter growth rate applied to the available pool of normal and repaired vehicles),  plus
c) the number of highs added  in that year that the OBD system did not detect (a function of MIL-
"off" rate and the high emitter growth rate applied to the available pool of "non-highs"). The
high emitter growth rate for a given year is the absolute increase in high emitters under the  No
OBD / No EVI case from the previous year divided by the fraction on nonhighs - i.e., the available
pool of vehicles which can become high emitters.

       Once the high emitter  fraction is calculated for the OBD or OBD/EVI cases, the fraction of
repaired emitters can be calculated as the difference between the fraction of high emitters that
would occur without OBD or  I/M  (HighBASE, from Equation (10)) and the fraction of high
emitters with OBD and/or I/M from Equation 18.  In equation form,

Repaired(i)   =      HighBASE (i) - HighOBD (i)                                Eqn 19

The rate of normal emitters remains constant between the No  OBD / No I/M,  OBD-only and
OBD/EVI case; only the number of high emitters decrease, directly replaced by repaired emitters.
The emitter fractions for normal, high and repaired emitters for the OBD Only and OBD/EVI
cases are shown in Appendix  A (Tables A-l and A-2) for NOx and HC, by vehicle class.

6.3    BER Calculation for OBD and OBD-based I/M

       Calculation of average in-use FTP-based NOx and NMHC emission rates at a given
vehicle age for the OBD-Only and OBD/EVI cases are similar  to the methodology for No OBD /
No I/M vehicles (Equation (2)); the primary differences are a) use of HighOBD rather than
HighBASE emitter fractions, and b)  addition of a term to account for repaired emitters. As
mentioned, the normal and high emissions rates are unchanged from the No OBD / No I/M case.
This computation is as follows:

AVE  =     HighOBD * High_ave + Normal *Norm_ave + Repaired * Rep_ave        Eqn 20
Reflecting the change in our after-repair assumptions, Rep_ave is equal to Norm_ave below
150,000 miles for NOx and 240,000 miles for HC; above these thresholds, Rep_ave is equal to

-------
1.5 times the applicable 50,000 mile certification standard.

7      NOx and HC BERs for Exhaust-Only I/M

       Since an OBD check is currently an unproven concept in an I/M program, some I/M
credit scenario must be developed for those areas that use traditional exhaust I/M test procedures.
This scenario will be likely used frequently until calendar year 2001.  By that time, it is assumed
that I/M test procedures utilizing OBD checks on vehicles equipped with OBD will be in place.

7.1    No OBD with Exhaust I/M

       The "No OBD / Exhaust I/M" emission levels for this scenario are calculated using the
methodology described in draft MOBILE6 document M6.EVI.001 ("MOBILE6 Inspection /
Maintenance Benefits Methodology for 1981 through 1993 Model Year Light Vehicles"). This
methodology utilizes I/M exhaust test identification rates and after repair effectiveness levels
based on data collected from the Arizona I/M program.  The "No I/M" and the "With Exhaust
I/M" emission rates are used to calculate the I/M emission level and I/M credits for situations
where exhaust-only I/M tests are being performed on Tier 1 vehicles without OBD. The only
vehicles which will fall in this category are the 1994 and 1995  model years certified to Tierl
standards.  In this case, the structure of the I/M credits is identical to the Tier 0  I/M credits with
the exception that the Norm_ave, High_ave, and fraction of highs in the fleet (HighBASE) are
different from analogous Tier 0 parameters.  Equation 21 defines this case mathematically.

 AVE = Norm_ave*(l-HighBASE) + High_ave*HighBASE*(l-IDR) + HighBASE*IDR*W*High_ave*RW +
              Norm_ave*R*HighBASE*IDR*FIX + High_ave*HighBASE*IDR*NC                 Eqn21

IDR is the identification rate of high emitters using an exhaust emission test.
R is the after repair emission level of vehicles repaired to pass an exhaust I/M test.
Fix is the fraction of vehicles which are repaired.
NC is the fraction of vehicles which are in non-compliance following their I/M test
W is the fraction of vehicles which receive a cost or other type of waiver.
RW is the after repair level of the vehicles which get waived.  It is shown as a fraction of the high emitter level.

(see report M6.IM.001 for a full explanation of these terms).

7.2    OBD and Exhaust I/M

       In this scenario, the vehicles in  the fleet are OBD compliant, but the state continues to
conduct an exhaust I/M test; this is most likely scenario  prior to calendar year 2001. For this
scenario, the same I/M equations and assumptions used to model the 1981-93 Tier 0 vehicles are
used.  The primary difference is the fraction of high emitters is reduced somewhat due to the
effects of OBD  program  (i.e., HighOBD is substituted for a higher rate of high emitters used in
TierO.

AVE   =      Norm_ave*(l-HighBASE) + Rep_ave*(Repaired) + High_ave*HighoBD*(l-IDR) +
              HighoBD*IDR*W*High_ave*RW +
              Norm_ave*R*HighoBD*IDR*FIX + High_ave*HighoBD*IDR*NC           Eqn22

-------
8      Derivation of Running and Start BERs for NOx and HC

       MOBILE6 will not use FTP-based BERs, but rather separate BERs for start and running
operation which are recombined according to estimates of in-use activity data. This requires that
the Tier 1 and later BERs developed on an FTP-basis, factors are required to derive running and
start BERs from FTP-based BERs.  For HC, this report contains significant modification to the
approach published previously. Our original estimates were based on an independent analysis of
a small number of Tier 1 and LEV vehicles. Subsequent validation showed that the resulting start
and running emission rates for Tier 1 and later vehicles were not internally consistent with start
and running emission rates for pre-Tier 1 vehicles; hence, we have revised our factors for
deriving start and running BERs to be consistent the start/running split for Tier 0 vehicles. For
NOx, the running adjustment factors were derived from 1988-1993 PFI Tier 0 LDV emission
rates at 100,000 miles.  A single adjustment was chosen for NOx since the ratio of running
emissions to FTP emissions was relatively stable over mileage.  For HC, however, the ratio of
running emissions to FTP emissions varied significantly over mileage (Appendix D).  As a result
we developed a running correction factor (RCF) from the 1988-1993 PFI Tier 0 LDV emission
rates as a function of mileage, as shown in Equation 24. Equations 23 and 25 will be used to
generate running BERs for Tier 1 and later standards for all vehicle classes:

       Running NOx BER (g/mi)          =     0.9 * FTP NOx BER              Eqn 23
       Running NMHC/NMOG RCF       =     (6E-05x3 - 0.0032x2 + 0.0656x + 0.2536)  Eqn 24
                                              x=mileage

       Running NMHC/NMOG BER (g/mi)=     RCF* FTP NMHC BER           Eqn 25

Start BERs (in grams per start) are related to FTP and Running BERs as shown in Equation
(26):19

FTP BER = (Running BER*7.5 +Start BER*0.43+Start BER*0.57*HS) / 7.5           Eqn 26

Where:
Running BERs =     the results of Equations (23) and (25)
7.5           =     total miles of the LA4
0.43/0.57     =      Bag I/Bag 3 weighting across total FTP
HS          =      the ratio of Bag 3 emissions to Bag 1 emissions, based on 1988-1993 Tier
                    0 LDV PFI BERs proposed for MOBILE6 (0.16 for HC, 0.204 for NOx)

Using this equation and the running and FTP BERs developed above, start factors were derived
for NOx according to Equation (27):

       Start NOx BERs (grams)            =     1.37* FTP BER                  Eqn 27
       19Glover, E., and Carey, P., "Determination of Start Emissions as a Function of Mileage and Soak Time for
1981-1993 Model Year Light-Duty Vehicles", MOBILE6 report number M6.STE.003

-------
For HC, since the running BERs are a function of mileage, a start correction factor (SCF) were
also derived as a function of mileage as shown in Equation (28). Using this equation and the
running and FTP BERs developed above, start factors were derived for HC according to
Equation (29):

       Start NMHC/NMOG SCF          =     (-0.0008x3 + 0.0474x2 - 0.9518x + 10.752) Eqn 28
                                              x=mileage

       Start NMHC/NMOG BER (grams)  =     SCF * FTP BER                  Eqn 29

These factors were applied equally to each emitter class: normal, high and repaired. The
resulting BERs for Tier 1, LEV and ULEV across all classes are shown in Appendix B.

-------
                     APPENDIX A:
a) Emitter Fractions: No OBD/No IM, OBD Only, OBD/IM Cases
          b) Vehicle Mileage As A Function of Age

-------
Table A-l: NOx Emitter Fractions
Age
(Years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
LDV
All
Normal
1.000
1.000
0.975
0.946
0.916
0.885
0.853
0.821
0.788
0.754
0.720
0.685
0.649
0.613
0.576
0.539
0.501
0.463
0.424
0.386
0.346
0.307
0.268
0.228
0.188
0.148
Base
High
0.000
0.000
0.025
0.054
0.084
0.115
0.147
0.179
0.212
0.246
0.280
0.315
0.351
0.387
0.424
0.461
0.499
0.537
0.576
0.614
0.654
0.693
0.732
0.772
0.812
0.852
OBD Only
High
0.000
0.000
0.006
0.033
0.061
0.090
0.120
0.154
0.188
0.222
0.258
0.294
0.331
0.368
0.406
0.444
0.483
0.523
0.562
0.602
0.643
0.683
0.724
0.765
0.806
0.847
Repair
0.000
0.000
0.019
0.021
0.023
0.025
0.027
0.026
0.025
0.024
0.023
0.021
0.020
0.019
0.018
0.017
0.016
0.015
0.013
0.012
0.011
0.010
0.008
0.007
0.006
0.005
OBD/IM
High
0.000
0.000
0.004
0.009
0.014
0.019
0.025
0.030
0.037
0.043
0.050
0.057
0.065
0.073
0.082
0.092
0.102
0.113
0.124
0.137
0.151
0.166
0.183
0.202
0.224
0.251
Repair
0.000
0.000
0.021
0.046
0.071
0.096
0.122
0.149
0.176
0.203
0.230
0.258
0.286
0.314
0.342
0.370
0.397
0.425
0.451
0.477
0.503
0.527
0.549
0.570
0.588
0601
LDT1/2
All
Normal
1.000
0.995
0.956
0.916
0.874
0.831
0.786
0.740
0.693
0.645
0.596
0.547
0.498
0.448
0.399
0.351
0.304
0.258
0.214
0.172
0.132
0.094
0.059
0.026
0.000
0000
Base
High
0.000
0.005
0.044
0.084
0.126
0.169
0.214
0.260
0.307
0.355
0.404
0.453
0.502
0.552
0.601
0.649
0.696
0.742
0.786
0.828
0.868
0.906
0.941
0.974
1.000
1 000
OBD Only
High
0.000
0.001
0.037
0.074
0.112
0.156
0.202
0.248
0.296
0.345
0.394
0.444
0.495
0.545
0.594
0.643
0.691
0.738
0.782
0.825
0.866
0.904
0.940
0.973
1.000
1 000
Repair
0.000
0.004
0.007
0.010
0.013
0.013
0.012
0.011
0.011
0.010
0.009
0.008
0.008
0.007
0.006
0.005
0.005
0.004
0.003
0.003
0.002
0.001
0.001
0.000
0.000
0000
OBD/IM
High
0.000
0.001
0.007
0.014
0.021
0.028
0.037
0.046
0.055
0.066
0.077
0.089
0.102
0.116
0.131
0.148
0.166
0.186
0.208
0.233
0.261
0.294
0.336
0.394
0.490
0490
Repair
0.000
0.004
0.037
0.070
0.105
0.141
0.177
0.214
0.252
0.289
0.327
0.364
0.400
0.436
0.469
0.501
0.530
0.556
0.578
0.595
0.607
0.611
0.605
0.579
0.510
0510
LDT3/4
All
Normal
1.000
0.991
0.949
0.905
0.859
0.812
0.763
0.713
0.661
0.607
0.552
0.496
0.439
0.381
0.321
0.261
0.200
0.139
0.077
0.015
0.000
0.000
0.000
0.000
0.000
0000
Base
High
0.000
0.009
0.051
0.095
0.141
0.188
0.237
0.287
0.339
0.393
0.448
0.504
0.561
0.619
0.679
0.739
0.800
0.861
0.923
0.985
1.000
1.000
1.000
1.000
1.000
1 000
OBD Only
High
0.000
0.002
0.041
0.082
0.124
0.172
0.222
0.274
0.327
0.381
0.437
0.494
0.552
0.612
0.672
0.734
0.796
0.858
0.921
0.984
1.000
1.000
1.000
1.000
1.000
1 000
Repair
0.000
0.007
0.010
0.013
0.017
0.016
0.015
0.014
0.013
0.012
0.011
0.010
0.009
0.007
0.006
0.005
0.004
0.003
0.002
0.000
0.000
0.000
0.000
0.000
0.000
0000
OBD/IM
High
0.000
0.001
0.008
0.015
0.023
0.032
0.041
0.051
0.062
0.074
0.087
0.102
0.118
0.137
0.158
0.183
0.213
0.252
0.304
0.392
0.489
0.489
0.489
0.489
0.489
04R9
Repair
0.000
0.008
0.043
0.080
0.117
0.156
0.196
0.236
0.277
0.319
0.360
0.402
0.443
0.482
0.520
0.555
0.586
0.609
0.618
0.592
0.511
0.511
0.511
0.511
0.511
0511

-------
Table A-2: NMHC/NMOG Emitter Fractions
Age
(Years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
LDV
All
Normal
0.983
0.981
0.971
0.953
0.935
0.918
0.901
0.885
0.869
0.854
0.839
0.825
0.811
0.798
0.785
0.773
0.761
0.749
0.738
0.728
0.718
0.708
0.698
0.689
0.681
0.672
Base
High
0.017
0.019
0.029
0.047
0.065
0.082
0.099
0.115
0.131
0.146
0.161
0.175
0.189
0.202
0.215
0.227
0.239
0.251
0.262
0.272
0.282
0.292
0.302
0.311
0.319
0.328
OBD Only
High
0.004
0.004
0.007
0.024
0.041
0.057
0.072
0.089
0.105
0.121
0.136
0.151
0.165
0.179
0.192
0.205
0.217
0.229
0.240
0.251
0.261
0.271
0.281
0.290
0.299
0.308
Repair
0.013
0.014
0.022
0.023
0.024
0.026
0.027
0.026
0.026
0.025
0.025
0.024
0.024
0.023
0.023
0.023
0.022
0.022
0.022
0.021
0.021
0.021
0.021
0.020
0.020
0.020
OBD/IM
High
0.003
0.003
0.005
0.008
0.011
0.013
0.016
0.019
0.022
0.025
0.027
0.030
0.032
0.035
0.037
0.040
0.042
0.044
0.047
0.049
0.051
0.053
0.055
0.057
0.059
0.061
Repair
0.015
0.016
0.024
0.040
0.055
0.069
0.083
0.096
0.109
0.122
0.134
0.145
0.157
0.167
0.178
0.188
0.197
0.206
0.215
0.223
0.232
0.239
0.247
0.254
0.261
0.267
LOT 1/2
All
Normal
0.983
0.978
0.959
0.935
0.912
0.890
0.868
0.848
0.828
0.810
0.792
0.775
0.760
0.745
0.731
0.719
0.707
0.696
0.686
0.677
0.669
0.662
0.655
0.649
0.644
0.627
Base
High
0.017
0.022
0.041
0.065
0.088
0.110
0.132
0.152
0.172
0.190
0.208
0.225
0.240
0.255
0.269
0.281
0.293
0.304
0.314
0.323
0.331
0.338
0.345
0.351
0.356
0.373
OBD Only
High
0.004
0.005
0.023
0.045
0.067
0.089
0.111
0.132
0.152
0.171
0.190
0.207
0.223
0.238
0.251
0.264
0.276
0.287
0.298
0.307
0.315
0.323
0.330
0.336
0.341
0.359
Repair
0.013
0.017
0.018
0.020
0.021
0.021
0.020
0.020
0.019
0.019
0.018
0.018
0.018
0.017
0.017
0.017
0.016
0.016
0.016
0.016
0.016
0.015
0.015
0.015
0.015
0.015
OBD/IM
High
0.003
0.003
0.007
0.010
0.014
0.018
0.022
0.026
0.029
0.033
0.036
0.039
0.042
0.045
0.048
0.051
0.053
0.055
0.057
0.059
0.061
0.063
0.064
0.066
0.067
0.071
Repair
0.015
0.019
0.034
0.054
0.074
0.092
0.110
0.127
0.143
0.158
0.172
0.185
0.198
0.210
0.221
0.231
0.240
0.248
0.256
0.263
0.270
0.275
0.281
0.285
0.289
0.302
LDT3/4
All
Normal
0.983
0.977
0.955
0.929
0.904
0.880
0.858
0.836
0.815
0.796
0.777
0.759
0.742
0.726
0.711
0.697
0.683
0.671
0.659
0.647
0.596
0.498
0.485
0.473
0.462
0.452
Base
High
0.017
0.023
0.045
0.071
0.096
0.120
0.142
0.164
0.185
0.204
0.223
0.241
0.258
0.274
0.289
0.303
0.317
0.329
0.341
0.353
0.404
0.502
0.515
0.527
0.538
0.548
OBD Only
High
0.004
0.006
0.026
0.050
0.073
0.098
0.121
0.143
0.164
0.184
0.203
0.222
0.239
0.255
0.271
0.286
0.299
0.313
0.325
0.337
0.389
0.490
0.503
0.515
0.526
0.537
Repair
0.013
0.018
0.019
0.021
0.023
0.022
0.021
0.021
0.020
0.020
0.019
0.019
0.019
0.018
0.018
0.017
0.017
0.017
0.017
0.016
0.015
0.012
0.012
0.012
0.012
0.011
OBD/IM
High
0.003
0.004
0.007
0.012
0.016
0.020
0.024
0.028
0.031
0.035
0.039
0.042
0.046
0.049
0.052
0.055
0.058
0.061
0.063
0.066
0.078
0.102
0.105
0.109
0.112
0.115
Repair
0.015
0.020
0.038
0.060
0.080
0.100
0.118
0.136
0.153
0.169
0.184
0.198
0.212
0.225
0.237
0.248
0.259
0.269
0.278
0.287
0.326
0.400
0.409
0.418
0.426
0.433

-------
Table A-3: Draft MOBILE6 Cumulative Mileages (10,000 miles)
Age (Years)
1
2
3
4
5
6
7
8
9
10
11
12
13
LDV
1.491
2.908
4.256
5.537
6.755
7.912
9.013
10.059
11.054
12.000
12.899
13.753
14.566
LDT1/2
1.950
3.788
5.519
7.146
8.672
10.100
11.436
12.681
13.839
14.914
15.910
16.829
17.676
LDT3/4
2.133
4.120
5.970
7.692
9.297
10.791
12.183
13.478
14.685
15.809
16.856
17.830
18.738
Age (Years)
14
15
16
17
18
19
20
21
22
23
24
25

LDV
15.338
16.072
16.770
17.434
18.064
18.664
19.234
19.776
20.291
20.781
21.247
21.690

LDT1/2
18.453
19.165
19.815
20.406
20.941
21.425
21.861
22.252
22.602
22.914
23.191
23.438

LDT3/4
19.583
20.371
21.104
21.786
22.422
23.014
23.566
24.079
24.557
25.003
25.418
25.804


-------
                 APPENDIX B:
TIER 1, LEV/ULEVI and Tier 2 BERs By Emitter Category

-------
Table B-l: Tier 1 & LEV/ULEV
Vehicle
Class
LDV/T1
LDT2/3
LDT4
Standard
Class
Tierl
LEV/ULEV
Tierl
LEV/ULEV
Tierl
LEV/ULEV
50K
Standard
(g/mi)
0.4
0.2
0.7
0.4
1.1
0.6
Mode
FTP
Running
Start (grams)
FTP
Running
Start (grams)
FTP
Running
Start (grams)
[ NOx Basic Emission Rates
"Normal" BER
(g/mi)
ZML
0.153
0.138
0.210
0.077
0.069
0.105
0.268
0.241
0.367
DR
0.0294
0.0265
0.0403
0.0147
0.0132
0.0201
0.0517
0.0465
0.0708
"High"
BER
(g/mi)
1.29
1.16
1.77
0.97
0.87
1.33
1.78
1.60
2.44
"Repaired"
BER Cap
(g/mi)
0.600
0.540
0.822
0.300
0.270
0.411
1.050
0.945
1.439
SAME AS LDV/T1 TIER 1
FTP
Running
Start (grams)
FTP
Running
Start (grams)
0.421
0.379
0.577
0.230
0.207
0.315
0.0809
0.0728
0.1108
0.0441
0.0397
0.0604
2.43
2.19
3.33
1.62
1.46
2.219
1.650
1.485
2.261
0.900
0.810
1.233

-------
Table B-2: Tier 2 NOx Basic Emission Rates
Vehicle
Class
LDV/T1
LDT2
LDT3
LDT4
Standard
Class
Tier 2
Interim
Tier 2
Interim A
Interim B
Tier 2
Interim A
Interim B
Tier 2
50/120K
Standard
(g/mi)
0.05/0.07
0.2/0.3
0.05/0.07
0.4/0.6
0.14/0.20
0.05/0.07
0.4/0.6
0.14/0.20
0.05/0.07
Mode
FTP
Running
Start (grams)
"Normal" BER
(g/mi)
ZML
0.019
0.017
0.026
DR
0.004
0.003
0.005
"High"
BER
(g/mi)
0.73
0.65
1.00
"Repaired"
BER Cap
(g/mi)
0.075
0.068
0.103
SAME AS LDV/T1 LEV (M6.EXH.007)
SAME AS LDV/T1 TIER 2
SAME AS LDV/T1 TIER 1 (M6.EXH.007)
FTP 0.054
Running 0.048
Start (grams) 0.073
0.010 0.87 0.210
0.009 0.78 0.189
0.014 1.19 0.288
SAME AS LDV/T1 TIER 2
SAME AS LDV/T1 TIER 1 (M6.EXH.007)
SAME AS LDT3 INTERIM B
SAME AS LDV/T1 TIER 2
  50,000 mile standard levels used to derive Tier 2 BERs for all bin categories according to
                       the methodology presented in this report (g/mi):
Bin
r
2a
3a
4a
5
6
7
8
9b
10b
LDV
0
0.014
0.021
0.029
0.05
0.08
0.11
0.14
0.2
0.4
LDT1
0
0.014
0.021
0.029
0.05
0.08
0.11
0.14
0.2
0.4
LDT2
0
0.014
0.021
0.029
0.05
0.08
0.11
0.14
0.2
0.4
LDT3
0
0.014
0.021
0.029
0.05
0.08
0.11
0.14
0.2
0.4
LDT4
0
0.014
0.021
0.029
0.05
0.08
0.11
0.14
0.2
0.4
a 50,000 miles certification "standards" estimated by multiplying full useful life standard by ratio of Bin 5
intermediate life / full useful life standards (0.05 / 0.07 = 0.71)
a Interim standard bins

-------
Table B-3: Tier 1 & LEV/ULEV I NMHC/NMOG Basic Emission Rates
Vehicle
Class
LDV/T1
LDT2
LDT3
LDT4
Standard
Class
Tierl
(NMHC)
LEV
ULEV
Tierl
(NMHC)
LEV
ULEV
Tierl
(NMHC)
LEV
ULEV
Tierl
(NMHC)
LEV
ULEV
50K
Standard
(g/mi)
0.25
0.075
0.04
0.32
0.10
0.05
0.32
0.16
0.10
0.39
0.195
0.117
Mode

"Normal" BER
(g/mi)
ZML DR
"High" "Repaired"
BER BER
(g/mi) (g/mi)
FTP 0.098 0.0113 1.67 0.375
Running FTP*RCF (See Section 8)
Start (grams) FTP*SCF (See Section 8)
FTP 0.029 0.0034 1.23 0.113
Running FTP*RCF (See Section 8)
Start (grams) FTP*SCF (See Section 8)
FTP 0.016 0.0018 1.14 0.060
Running FTP*RCF (See Section 8)
Start (grams) FTP*SCF (See Section 8)
FTP 0.125 0.0145 1.85 0.480
Running FTP*RCF (See Section 8)
Start (grams) FTP*SCF (See Section 8)
FTP 0.039 0.0045 1.29 0.150
Running FTP*RCF (See Section 8)
Start (grams) FTP*SCF (See Section 8)
FTP 0.020 0.0023 1.17 0.075
Running FTP*RCF (See Section 8)
Start (grams) FTP*SCF (See Section 8)
FTP 0.125 0.0145 1.85 0.480
Running FTP*RCF (See Section 8)
Start (grams) FTP*SCF (See Section 8)
FTP
Running
Start (grams)

FTP
Running
Start (grams)
FTP
Running
Start (grams)
FTP
Running
Start (grams)
0.063 0.0073 1.44 0.240
FTP*RCF (See Section 8)
FTP*SCF (See Section 8)
SAME AS LDT2
LEV
0.152 0.0177 2.03 0.585
FTP*RCF (See Section 8)
FTP*SCF (See Section 8)
0.076 0.0088 1.53 0.293
FTP*RCF (See Section 8)
FTP*SCF (See Section 8)
0.046 0.0053 1.33 0.176
FTP*RCF (See Section 8)
FTP*SCF(See Section 81

-------
Table B-4 - Tier 2 NMOG Basic Emission Rates
Vehicle
Class
LDV/T1
LDT2
LDT3
LDT4
Standard
Class
Tier 2
Interim/Tier 2
Interim A
Interim B
Tier 2
Interim A
Interim B
Tier!
50/120K
Standard
(g/mi)
0.075/0.09
0.2/0.3
0.16/0.23
0.125/0.156
0.05/0.07
0.16/0.23
0.125/0.156
0.075/0.09
"Normal" BER
Mort, (8^)
ZML DR
BER
BER Cap
SAME AS LDV/T1 LEVI
SAME AS LDV/T1 LEV I
SAME AS LDT3
LEVI

FTP 0.049 0.057 1.35 0.188
Running FTP*RCF (See Section 8)
Start FTP*SCF (See Section 8)
SAME AS LDV/T1 LEV I
SAME AS LDT3
LEVI

SAME AS LDT3 INTERIM B
SAME AS LDV/T1 LEV I
  50,000 mile standard levels used to derive Tier 2 BERs for all bin categories according to
                       the methodology presented in this report (g/mi):
Bin
r
T
y
4a
5
6
7
8b
8C
9d
10e
LDV
0
0.007
0.04
0.051
0.075
0.075
0.075
0.1
n/a
0.075
0.125
LDT1
0
0.007
0.04
0.051
0.075
0.075
0.075
0.1
n/a
0.075
0.125
LDT2
0
0.007
0.04
0.051
0.075
0.075
0.075
0.1
n/a
0.1
0.125
LDT3
0
0.007
0.04
0.051
0.075
0.075
0.075
0.125
0.1
0.14
0.16
LDT4
0
0.007
0.04
0.051
0.075
0.075
0.075
0.125
0.1
0.14
0.195
a 50,000 miles certification "standards" estimated by multiplying full useful life standard by ratio of ULEV
intermediate life / full useful life standards under LEV program (0.04 / 0.055 = 0.73)
b temporary standards for LDT3/4
c final standards for LDT3/4
d interim standard bin with optional LDT2 standard
e interim standard bin with optional LDT4 standard

-------
            APPENDIX C:
Detailed Results for NOx Regression Model

-------
       Equations 5, 6 and 7 are the only "new" regression model results presented in this paper
for basic emission rates used directly in MOBILE6.  The detailed regression results for these
equations are presented below.  Other equations used to derive the basic emission rates presented
in this paper are based on analyses documented in other MOBILE reports.  Information
pertaining to the source of these equations are also included, but not the detailed results.

                   Equation 5 - NOx Normal Emitter Model

                                  Descriptive Statistics

NOX
MILEAGE
Mean
.2523
33923.55
Std. Deviation
.1410
22759.61
N
1167
1167
Model Summary
Model
1
R
.475
R Square
.225
Adjusted R
Square
.225
Std. Error of the
Estimate
.1241
                                        ANOVA
Model
1
Regression
Residual
Total
Sum of
Squares
5.225
17.947
23.172
df
1
1165
1166
Mean
Square
5.225
1.541E-02

F
339.173


Sig.
.000


                                      Coefficients

Model
1
(Constant)
MILEAGE
Unstandardized
Coefficients
B
.153
2.941E-06
Std.
Error
.007
.000
Standar
dized
Coeffici
ents
Beta

.475
t
23.382
18.417
Sig.
.000
.000
95% Confidence
Interval for B
Lower
Bound
.140
.000
Upper
Bound
.165
.000

-------
Equation 6: Analysis of significance for mileage term in high emitter
                                model
                           Descriptive Statistics

NOX
MILEAGE
Mean
1.2942
79987.19
Std. Deviation
.5977
33999.56
N
31
31
                              Correlations

Pearson Correlation
Sig. (1-tailed)
N
NOX
MILEAGE
NOX
MILEAGE
NOX
MILEAGE
NOX
1.000
.031

.434
31
31
MILEAGE
.031
1.000
.434

31
31
                        Variables Entered/Removed
Model
1
Variables Entered
MILEAGE
Variables Removed

Method
Enter
Model Summary
Model
1
R
.031
R Square
.001
Adjusted R
Square
-.033
Std. Error of the
Estimate
.6076

-------
                 ANOVA
Model
1
Regression
Residual
Total
Sum of
Squares
1.038E-02
10.705
10.716
df
1
29
30
Mean
Square
1.038E-02
.369

F
.028


Sig.
.868


               Coefficients

Model
1
(Constant)
MILEAGE
Unstandardized
Coefficients
B
1.250
5.472E-07
Std.
Error
.283
.000
Standar
dized
Coefficie
nts
Beta

.031
t
4.421
.168
Sig.
.000
.868
95% Confidence
Interval for B
Lower
Bound
.672
.000
Upper
Bound
1.829
.000
Equation 6: NOx High Emitter Average
           Descriptive Statistics

NOX
Valid N
(listwise)
N
31
31
Minimum
.80

Maximum
2.77

Mean
1.2942

Std. Deviation
.5977


-------
Equation 7: NOx All LDV/LDT (0.4 g/mi Standard) Model
                     Descriptive Statistics

NOX
MILEAGE
Mean
.2793
35115.51
Std. Deviation
.2360
24229.87
N
1198
1198
                        Correlations

Pearson Correlation
Sig. (1-tailed)
N
NOX
MILEAGE
NOX
MILEAGE
NOX
MILEAGE
NOX
1.000
.474

.000
1198
1198
MILEAGE
.474
1.000
.000

1198
1198
                   Variables Entered/Removed
Model
1
Variables Entered Variables Removed
MILEAGE
Method
Enter
Model Summary
Model
1
R
.474
R Square
.225
Adjusted R
Square
.224
Std. Error of the
Estimate
.2079
                          ANOVA
Model
1
Regression
Residual
Total
Sum of
Squares
14.982
51.687
66.668
df
1
1196
1197
Mean
Square
14.982
4.322E-02

F
346.669


Sig.
.000



-------
                               Coefficients



Model

1

(Constant)

MILEAGE
Unstandardized
Coefficients


B
.117

4.617E-06

Std.
Error
.011

.000
Standar
dized
Coefficie
nts
Beta



.474

t


11.072

18.619

Sig.


.000

.000
95% Confidence
Interval for B


Lower
Bound
.096

.000

Upper
Bound
.138

.000
            Information pertaining to other equations

Equation 8: The high emitter correction factor is derived from an analysis presented in
MOBILE6 report M6.EXH.001, "Determination of Running Emissions as a Function of
Mileage for 1981-1993 Model Year Light-Duty Cars and Trucks", Enns et. al.

Equations 13 and 14 were derived by first combining the running and start emission rates
for normal-emitting 1988-1993 PFI LDVs according to FTP weightings. The raw running
and start emission rates can be found in MOBILE6 report M6.IM.001, "MOBILE6
Inspection/Maintenance Benefits Methodology for 1981 through 1993 Model Year Light
Vehicles". At a given mileage, FTP emissions were derived from start and running
emission rates according to Equation 26. For normal emitters, the combined multi-linear
FTP emission rates were then regressed by mileage to create a simple linear model,
resulting in Equation 13.
Equations 24 and 28 were developed to create a simple fit to multi-linear modeled data.

-------
                     APPENDIX D:
Running and Start Correction Factors Used In Developing Running
            and Start BERs from FTP-Based BERs

-------
      HC Running Emission / FTP Emission ratio for 1988-1993 PFI LDVs
0.4
0.2
          RCF used in MOBILE = 6E-05x3 - 0.0032x2 + 0.0656x + 0.2536
                        10         15        20
                               Mileage /10
25
30
         NOx Running Emission / FTP Emission ratio for 1988-1993 PFI
                                 LDVs
                        RCF used in MOBILE = 0.9
 0.2
                         10         15         20
                                Mileage/10
 25
 30

-------
      HC Start Emissions/ FTP Emission ratio for 1988-1993 PFI LDVs
12
10
          SCF used in MOBILE = -O.OOOSx3 + 0.0474x2 - 0.9518x + 10.752
                       10         15        20
                             MNeage/10
  25        30
     NOx Start Emissions/ FTP Emission ratio for 1988-1993 PFI
                             LDVs
                 SCF used  n MOB LE = 1.37
                     10         15        20
                           Mileage /10
25        30

-------
   APPENDIX E:
Peer Review Comments

-------
           REPORT REVIEW

M6.EXH.007 — TIER 1 AND LATER HC AND NOx
 EXHAUST EMISSIONS EQUATIONS IN MOBILES
       Prepared for U. S. Environmental Protection Agency
           Under Order Number OA-0249-NATX

                 March 21, 2000
                  L. S. CARETTO
               7805 Cowper Avenue
               West Hills, CA91304

-------
                          Table of Contents
Table of Contents	i

Introduction	  1

Body of the Review	  2
      Overall clarity	  2
            Minor editorial comments  	  4
      Overall methodology 	  5
      Appropriateness of the data sets selected	  6
      Data analyses, including statistical approaches and models	  7
      Appropriateness of conclusions	  8
            Ratio approach for basic emissions  	  8
            Effects  of OBD and I/M	  12
            Other issues  	  14

Conclusions and Recommendations for Further Studies	  15
      Short-term improvements   	  15
      Long-term improvements	  15
Peer Review of EPA MOBILE6 Report M6.EXH.007               March 21, 2000       Page i

-------
                              Introduction

The U. S. Environmental Protection Agency (EPA) requested an independent peer
review of the report entitled "Determination of NOx and HC Basic Emission Rates, OBD
and I/M Effects for Tier 1 and later LDVs and LDTs." This report, numbered
M6.EXH.007, is one part of the work being done to provide an updated version of
EPA's mobile source inventory model, MOBILES. It proposes equations to compute HC
and NOx emissions for light-duty vehicles (LDV) and light-duty trucks (LOT), with and
without onboard diagnostics (OBD) or inspection maintenance (I/M) programs. These
are late-model year and future vehicles certified to a 0.4 g/mi NOx standard (Tier 1),
low-emission vehicles (LEV), and ultra-low-emission vehicles (ULEV).

EPA provided a copy of the report and overall directions for this review.  EPA also
provided copies of comments that stakeholders had  made on a  previous draft of the
M6.EXH.007 report and copies of data files used in their NOx analysis.  The directions
for this review asked for the following areas to be addressed:
   1.  report clarity
   2.  overall methodology
   3.  appropriateness of the data sets selected
   4.  the data analyses conducted, including the statistical approaches used and
      models selected
   5.  appropriateness of the conclusions with specific attention to and comments on:
      o  the ratio approach for basic emissions
      o  the effects of OBD and I/M on the basic emission rates.
   6.  recommendations for any alternate data sets or analyses.
The overall directions for review also asked for a separation of recommendations for
improvements that could be  made in the short term versus  longer-term improvements.

The body of this review is in  the next section, subdivided into the six topics listed above.
That is followed by a conclusions section that identifies the short-term and long-term
recommendations.
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21,2000        Page 1

-------
                         Body  of the Review
Overall clarity

The report assumes that the reader is familiar with mobile source emissions and the
procedures for calculating such emissions. For this reader, the report is presented in a
reasonably clear fashion.  Some editorial comments are listed below.

All data shown in the report should be rounded to match the values used in MOBILES.
For example, the NOx emission level for high-emitting Tier 1 vehicles is listed as 1.294
g/mi on page 9, and is rounded to 1.29 g/mi in Table B-1.  A check of the fractions of
high and normal NOx emitters for Tier 1 LDVs, given in Table A-1, shows agreement
when the high emitter value of 1.294 g/mi is used. When the value of 1.29 g/mi is used,
some values of this fraction are off by 0.001 from the values reported in Table A-1.
This discrepancy indicates that the values in Table B-1 have been rounded and are not
the numbers used in the actual MOBILES calculations. The numbers in the report
should be revised to match those used in MOBILES, with the same number of
significant figures.

The analysis on pages 14 and 15 compares emissions from Tier 1 light-duty trucks in
different weight classes (LDT1/2 vs. LDT3/4). This analysis could be rearranged to
introduce and reenforce the concept that the regression analysis is done in terms of
"headroom," which is defined as the ratio of the actual emissions to the emissions
standard. Although the discussion defines this at one point and concludes that there
are no  differences "relative to the standard," it takes careful reading to understand this
section. To clarify, the discussion (starting in the middle of page 14) could be reworded
to read as follows:

      To assess this issue for heavier trucks, an analysis of in-use emissions
      from heavier LDTs was performed in terms of the ratio of the emissions to
      the emission standard.  This variable, called the "headroom," allows a test
      of the hypothesis that there are no significant differences between the
      different truck classes when the emissions are adjusted by dividing the
      data for light-duty vehicles by the appropriate emission standards.

The summary of regression coefficients for basic emission rates  in Table B-1  could be
augmented by stating the formula and data used for the emission calculations.  For
example:
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21,2000       Page 2

-------
      The fleet emission rate, E, is computed from the emissions of high-
      emitting vehicles, EH, normal-emitting vehicles, EN, and the high-emitter
      fraction, f, as follows:

                               E=fEH + (1-f)EN

      The high emissions fraction for NOx is found in terms of the odometer
      reading, odo (in units of 10,000 miles), from the following equation.
                           f=  -0.036  +  0.21420C/0
                                1.141 - 0.02941 odo
 A similar paragraph, with the appropriate equation for NMHC or NMOG, could be
 included on the same page as Table B-2.

 Consider using the same vertical scale for all the charts in Appendices C and D.
 Alternatively, use the same scale for all NOx charts in Appendix C, which could be
 different from the scale used for the NMHC and NMOG charts in Appendix D. The
 similar scale would allow a ready visual comparison among charts of the differences in
 emission rates for the different technology classes. For example, the benefit of OBD
 and I/M appears much larger for LDV/LDT1 ULEV NMOG than it does for LDT2/LDT3
 Tier 1 NMHC; it is actually less (in grams per mile or total lifetime grams).  In this same
 example the relative effect of OBD and I/M is greater for LDV/LDT1 ULEV NMOG than
 it is for LDT2/LDT3 Tier  1 NMHC. If the goal is to display the relative differences
 visually, the different vertical scales could be retained. If the vertical scale is not
 changed, the format for the vertical axis should be changed on some  charts  so that
 numbers are not repeated. Either use two significant figures in the axis labels or
 change the value used for the increment.

 The discussion of the method used for the analysis of OBD (pages 24 and 25) is not
 very clear,  and it has two errors. The first error is the definition (in the fourth line  after
 equation 18) of "Nonhigh" as the fraction of normal and repaired vehicles.  Equation 15
 correctly defines Nonhigh(i) as 1.0 - HighBASE(i).  This is different from  the fraction of
 repaired plus normals, which is 1.0 -  HighOBD(i).  The second error is in equation 17.  It
 should read GrowthJHigh(i) = Delta_High(i)/Nonhigh(i-1).  These subscripts  maintain
 consistency for the entire set of equations.  A simpler alternative would combine
 equations 15 to 17 into a single equation:
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21,2000       Page 3

-------
              Growth_High(i) =
                                      - HighBASE(i--\)
 With the equation written as shown above, the fundamental assumption about the
 method used for calculating the high fraction under OBD can be illustrated by the
 following equation.


    HighOBD(i) - HighOBD(i--\)  = ^^      ^  = HighBASE(i) - HighBAS^i-^
        1 - HighOBD(i-l)              ~             1  -  HighBASE(i-l)


 This equation emphasizes that the relative growth rate of high emitters used for OBD
 calculations is assumed to be the same as the relative growth rate found without OBD.

 Combining equations 15 to 17 into one equation eliminates the "Nonhigh" term and the
 need to define it.  The current definition of "Growth_High" refers to nonhighs and would
 have to be changed. The entire discussion of the material between equations 15 and
 19 could be improved by integrating the equations with the text.

 In the discussion of running emissions and start emissions on page 28, it may be
 appropriate to remind the reader that "start" emissions are defined as cold-start
 emissions that are subsequently adjusted for soak time.  This makes it easier to
 understand equation 25. Also, the ratio 0.43/0.57 is defined as the "Bag 3/Bag 1
 weighting across total FTP." It may be clearer to define this as the ratio of cold-start
 trips to hot-start trips used to derive the bag weightings in the  FTP.

 Minor editorial comments

 Page 6, footnote 7 states that 15 of the 74 vehicles tested by EPA had duplicate tests,
 but it does not say how these duplicate tests were treated in the data set. Were 74 or
 89 data points from EPA vehicles used in the regression analysis?

 Page 9, first line under Figure 3:  "generating" should be "generated."

 Page 1 1 , first sentence: "adjustment is" should be "adjustments are."

 Page 15: the first line is the footnote from the table on the previous page.

 Page 16: the last two rows of the table contain references to footnotes 4 and 5 as data
 sources for engine-out NMHC emissions;  these references do not seem correct.
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21,2000        Page 4

-------
 Page 22, fourth line from the bottom: "will smaller" should be "will be smaller."

 Page 25, equation 18, second line: The closing "]" should be a ")]".  Presumably this
 equation has not been simplified to the form shown below in this report to better
 illustrate the two components of the growth: MILs that do not illuminate and owners
 who do not respond.

              HighOBD(i-1) + (1  - OBD*MIL)Growth_High(i)*(1-HighOBD(i))
 Overall methodology

 This report confronts a significant problem of mobile source emission models. Such
 models are required to make estimates of emissions in future years, but the data to do
 so are not available. Thus, approximations and engineering judgements must be used
 to estimate these future emissions.

 The overall approach for Tier 1 vehicles is a modification of the exhaust emission
 method used for 1981 -1993 vehicles where a substantial database was available.   For
 those model years, the original analyses were used to determine running and start
 emissions separately.  In this report, the FTP emissions are used as the basis for the
 analysis and the equations derived for FTP emissions are then modified to determine
 running and start emissions.

 The analysis continues the division of the fleet into normal and high emitters, used in
 previous analyses.  At the top of page 11, the report notes that this adjustment affects
 only the proportions of normal and high emitters and not the emissions of these two
 regimes.  As noted in a previous review,* this is a fundamental assumption  of the
 method used to compute the high-emitter fraction.  Errors in this assumption would
 lead to errors in the computed high-emitter fraction. EPA staff should continue their
 efforts to obtain data that would justify this assumption, as they proposed doing in the
 M6.IM.001 report.

 In general, the  emissions of future vehicles are computed from the ratio of the future
 emission standard to the emission standard for the existing vehicles in the database.
 This is a reasonable approach to take in the absence of data.  Details of this approach
 are discussed further below under the "Appropriateness of Conclusions" heading.
       L. S. Caretto, "Report Review. Inspection and Maintenance Analyses for 1981-1993 Light
Vehicles in MOBILE6," Review of the M6.IM.001 report for U. S. Environmental Protection Agency under
order number 9A-0738-NATX, September 20, 1999.
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21,2000       Page 5

-------
 The determination of future emissions must be based on extrapolation of existing data.
 The methods used for such extrapolations can be readily questioned, but they cannot
 be easily justified. There may be more than one reasonable choice to make in these
 extrapolations and the ones used in this report represent reasonable choices using the
 best judgement of the authors.
 Appropriateness of the data sets selected

 The report uses three vehicle data sets for modeling NOx emissions. These data sets
 are from tests by EPA, the California Air Resources Board (ARB), and the automobile
 industry.  EPA used the data for passenger cars certified to the Tier 1 NOx standard of
 0.4 g/mi.  The three data sets contained results from 1,122 passenger cars and 62
 light-duty trucks. The overall sample size is sufficient for computing the regression
 equation for normal emitters. The scatter in the data, which is shown in Figure 1, is
 typical for emissions data. Based on this figure, the ARB and EPA data sets appear to
 have more scatter than the results obtained from the automobile industry.

 Because  of a lack of sufficient data, particularly at high mileages, NMHC emission
 equations for Tier 1 vehicles were obtained by adjusting the similar equations for 1988-
 1993 vehicles with emission control technology similar to that used on Tier 1 vehicles.

 The series of reports describing the development of MOBILES have generally  not
 provided details of how data were used in the various databases. The general
 descriptions provided in the reports allow the reader to understand the approach used.
 However, an interested user may not be able to reproduce the actual results.  The
 footnote on page eight of the report, which notes that one vehicle with NOx readings of
 0.21 g/mi and 1.31 g/mi was treated as a high emitter, is an example of the kind of
 information required for another reader to check the original calculations.*

 A preliminary examination of the data for NOx high emitters, in 1988-and-later vehicles
 certified to 0.4 g/mi NOx, was not able to match the data in the report. The data
 supplied by EPA in three spreadsheets had 27 high emitters with an average value of
 1.298 g/mi. The average value is only slightly different from the value of 1.294 g/mi
 found in the report. However, the table on page 17 lists the number of high-emitting
 vehicles as 31 instead of 27.  The high  emitters found in this review were compared
 with the data in  Figure 3, which shows the emissions of individual high emitters as a
       Although this footnote explains how the data are treated, it does not say why the vehicle was
considered a high emitter. At the average value of 0.76 g/mi, this vehicle would be a normal emitter. A
comparison of the original data set and Figure 3 shows that only the 1.31 g/mi value was used for this
vehicle rather than the average.
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21,2000       Page 6

-------
 function of mileage. The two vehicles from the EPA data and the four vehicles from
 the auto industry data match the data in the figure.  However, not all the ARB data
 match the points in the figure.

 Some formal record should be kept of the actual databases used in each analysis and
 the reasons for including or eliminating any particular data points. Such information
 should be recorded, either in the M6 series of reports or separately, for any special
 treatment of data.

 Several comments on the draft report have suggested the consideration of alternative
 databases (on past emissions performance) which would lead to different extrapolation
 assumptions.  However, there are no other databases that would provide additional
 information on the actual emissions performance of the vehicles modeled here.
 Data analyses, including statistical approaches and models

 Regression analysis is the main tool used in this report. Refined techniques for more
 detailed analyses used in other MOBILES analyses were not used here, e.g., the
 multilinear regressions used for 1981-1993 model year light-duty cars and trucks in the
 M6.EXH.001  report.  Such tools were presumably not used here because the authors
 thought that they were not justified, given the approximations required to provide
 forecasts of future vehicle emissions.

 The authors use some analysis of the regression statistics to determine the
 appropriateness of model choices.  However, no statistical results such as confidence
 intervals or standard  errors are provided for the data presented here. Such statistical
 results should be reported. The uncertainty for the emission predictions for future
 vehicles that results from the assumptions made to predict those emissions will be
 greater than any statistical uncertainty. However,  estimates of this kind of "assumption
 uncertainty" could be provided. The comparison, on page 19, of Tier 1 NMHC data for
 low mileage to the model equations for Tier 1 NMHC derived from 1988-1993 vehicles
 is an example of such an estimate. A similar estimate of the uncertainty in this
 approach could be done for higher mileages by comparing the NOx emission rates
 found from data on Tier 1 vehicles with NOx emission rates for such vehicles derived
 from  1988-1993 vehicles.

 With  the data sets selected, the statistical analyses that have been done are
 appropriate.  Commenters have suggested that the emission rates should be different
 for cars and trucks and that the changes  in the normal emissions regressions are not
 proportional to the standards ratio as proposed in  the report. However, once the
 choice of a data set was made, the regression  analysis of the entire truck-plus-car data
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21,2000       Page 7

-------
 set was not able to determine that the different vehicle types had a statistically
 significant difference in regressions. This conclusion is likely due to the limited amount
 of data available on trucks certified to 0.4 g/mi.
 Appropriateness of conclusions

 There is no explicit conclusion section in three of the reports. There is an implied
 conclusion that the data analyses were effectively done to provide an appropriate
 estimate of Tier 1 and later light-duty car and truck NOx and NMOG/NMHC exhaust
 emissions for MOBILES.  Because this report deals with vehicles for which there is not
 a sufficient database to determine the emissions, the conclusions need to be more
 tentative than in other reports. Perhaps the report should contain an explicit
 acknowledgment that its results are based in part on data analysis and in larger part on
 engineering judgement.

 The report title speaks of "Tier 1 and later" vehicles, but the report only discusses Tier
 1 vehicles, LEVs and ULEVs. There is a brief mention of, but no quantitative data for,
 transitional low-emission vehicles (TLEV). The results of this report will also apply to
 Tier 2 vehicles. There should be some mention of, if not complete emissions
 equations for, Tier 2 vehicles in  this report.

 Ratio approach for basic emissions

 The ratio approach is an appropriate technique for predicting future emissions in the
 absence of data.  For normal emitters, this method could be justified more strongly by
 comparing data on previous changes in standards and seeing if the ratio method
 works. This is done in the analysis on page 19 for low-mileage, Tier 1, NMHC
 emissions.

 In addition, the "headroom" analysis  of the heaviest light-duty trucks (LDT4) NOx
 emissions on pages 14 and 15 can be considered a test of the ratio approach.  This
 analysis basically uses a regression of different data sets, with different emission
 standards. However, the regressed data is expressed in terms of the ratio of the
 emissions to  the emissions standard. The results of the analysis show that there  is  no
 difference in the regression between the LDT4 and other light-duty trucks, when the
 data are expressed as the ratio  of emissions to the emissions standard. Other data
 sets could be compared,  however the two results cited here show that the ratio
 approach appears to work for normal emitters when the emissions control technology
 is similar.
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21,2000        Page 8

-------
 One problem with this approach is that it assumes that there will be no improvements
 or degradation in the ability of the emission control system to maintain lifetime
 emissions.  Harley's remote sensing data from Phoenix has shown that later model
 year vehicles have much better CO emissions durability than early model-year
 vehicles, even though both model-year vehicles were certified to the same standard.  It
 is also possible that the future changes requiring new emission control technologies
 will need some time to become stable as compared to the vehicles considered in
 Harley's study. Thus, the emissions deterioration (slope)  could increase or decrease
 in future model years.

 Comments by AIR using the auto database noted that the ratio method was not
 applicable to cars and trucks certified to NOx standards between 0.4 and 1 .7 g/mi.
 Observations on this data set could be used to develop an alternative extrapolation
 procedure to the ratio method.  Such a procedure would have to analyze how close the
 normal emission level would come to the actual standard  for past and future vehicles.
 Since there will be technology changes with the new standards, it is not clear that an
 extrapolation procedure developed by this approach would be any better than the ratio
 method. Without any definitive data, the ratio method is the simplest approach to use
 for normal emitters.

 An additional question about the ratio approach arises in the treatment of high
 emitters.  As noted in the report, high emitters represent failed emission control
 systems.  The emissions from two average vehicles with failed control systems may
 not be in direct proportion to their certification standards.  Thus a weighting factor, w, is
 used to predict the high emissions from a future vehicle, HF, in terms of the high
 emissions from an existing vehicle with data, HD, and the  emission standards for the
 future and data vehicles,  SF, and SD, respectively. This weighting factor is applied to
 the existing high-emitter level, with and without the application of the standards ratio,
 according to the following equation:
                                                                             [1]
 In the absence of data there are two possible arguments that can bound w. The first
 argument is that the high emitters would have the same proportional emission
 reduction implied by the standard and assumed for normal emitters. This argument
 leads to a value of w = 1 . The second argument assumes that a failed vehicle would
 have the same emissions regardless of its emission control system. This leads  to a
 value of w = 0. In both this report and in the previous draft, EPA has used a value of
 0.5 for w, which is the midpoint of this range.  EPA conducted two analyses to support
 this choice of w = 0.5. These are discussed below.
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21,2000        Page 9

-------
 The first analysis examined data on engine-out emissions and calculated the
 difference in catalyst efficiency between two exhaust emission points: (1) the efficiency
 required to achieve the emission standard at 50,000 miles, (2) the efficiency for the
 average high emitter, assuming the value of w = 0.5 was used to compute the high
 emitter value for LEV and ULEV vehicles.*  This analysis shows that the decrease in
 catalyst efficiency required to produce a high emitter is consistent with the assumption
 that w = 0.5. However, this analysis requires the questionable assumption that the
 only cause of a high emitter is a change in catalyst efficiency; the engine-out emissions
 are assumed to remain the same.

 The second analysis, on page 17 of this report, compares limited data on high emitters
 to infer a value for w.  That analysis uses equation [1 ], after rearrangement to solve
 for w.  The result, after some algebra,  is shown below.
                                      "D
                             w = - —                                      [2]
 In the table on page 17,  the numerator of equation [2] is in column five, the
 denominator is in column two, and the value of w is in column six. The data in that
 table compare federal Tier 0 vehicles, with a NOx standard of 1 .0 g/mi; California
 Tier 0 vehicles, with a NOx standard of 0.7 g/mi; and Tier 1 vehicles with a NOx
 standard of 0.4 g/mi. The first comparison - NOx standards of 1 .0 and 0.7 g/mi -
 gives a value of w = 0.83. The second comparison - NOx standards of 0.7 and 0.4
 g/mi - gives a value of w = 0.70. The authors then argue that these two values of w
 and their associated emission levels suggest the following trend:
       Lower NOx standard from 1 .0 to 0.7 g/mi      w = 0.83
       Lower NOx standard from 0.7 to 0.4 g/mi      w = 0.70
       Lower NOx standard below 0.4 g/mi           w = 0.50
 This implied trend is used to support the choice of w = 0.5 for emission reductions
 below a NOx standard of 0.4  g/mi.

 No statistical evaluation of the data used in the analysis on page 17 is presented;
 such an analysis is likely to show that the difference between the two mean emission
 levels is not statistically significant.  Since there are not sufficient data in the report to
       The LEV and ULEV classes are combined in both the NOx analysis and the NMOG analysis.
The LEV and the ULEV classes have the same NOx standard, however their NMOG standard is
different. The NMOG analysis uses the high emitter value for LEVs.  There is no indication of the
different proportions of LEVs and ULEVs in the data set analyzed.
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21, 2000       Page 10

-------
 determine this, an analysis to determine the likelihood that the differences would or
 would not be significant was carried out. In this analysis, shown below, the value of
 the standard deviation required for the differences to be statistically not significant was
 computed.

 The confidence interval for the difference between the true or population means,
 denoted as ub and ua, can be expressed in terms of the sample means, denoted as xb
 and xa, by the following equation.
                   Ma   Xb   Xa ±
2
                                              \
                                                 [3]
 In this equation, a is the significance level, nb and na represent the number in each
 sample and tQ/2 nb+na_2 is the value of the t-distribution for the given significance level and
 sample sizes.  The standard deviation, s,  is the pooled estimate from both samples.  It
 is computed from the individual standard deviations, sb and sa, by the following
 equation.*
                 s =
(nb
- 1)
2
Sb +
("a -
1K2
n
                                      ~ 2
                                                                               [4]
 The value of s that would show no difference between the two population means (at
 the extreme end of the confidence interval) can be found by setting ub = ua in equation
 [3] and solving for s.  This gives the following result.
                      s =
                                  Xb ~ Xa
                                      A
                                                                               [5]
 Using the data for federal and California Tier 0 vehicles, in the table on page 17 of the
 report, gives xb =2.46, xa = 1.85, nb = 8, and na = 3.  The value of the t statistic for a
 significance level of 0.05 and nb + na - 2 = 9 is 2.201. With these data the confidence
 interval for ub - ua contains zero if the pooled estimate of the standard deviation is 0.41
      *This equation assumes that both samples have the same variance, a2. The pooled estimate of
the variance, s, is the estimate of this common variance.
Peer Review of EPA MOBILE6 Report M6.EXH.007
                          March 21,2000
                                Page 11

-------
 or greater.  Similarly, when the California Tier 0 vehicles are compared with the Tier 1
 vehicles (xb =1.29, xa = 1.85, nb = 31, and na = 3), the value of s required to have zero
 in the confidence interval for the difference between the true means is 0.42.

 It is likely that the standard deviation for high-emitter data will  be greater than the
 values of 0.41  and 0.42, computed above.*  If s is greater than 0.42, the differences in
 the  means used to imply a slope in the discussion on page 17 are not statistically
 different. The actual standard deviations should be used to compute the significance
 of the differences between the mean values compared  in this analysis.

 Based on this review, neither of the arguments used to  justify the value of w = 0.5 is
 very convincing. They should be retained, however, to  illustrate the possible kinds of
 analyses that could  be used to establish a value for w.  In the absence of any
 convincing data to the contrary, there is no apparent reason to change this value from
 the  initial choice of w = 0.5 assumed in the draft report.

 Effects of OBD and I/M

 The basic approach used to model the effects of OBD is to assume that the fraction of
 normal vehicles, as  a function of vehicle age or mileage, remains the same as with no
 OBD. In addition, the relative growth in the fraction of high emitters is assumed to be
 the  same with and without OBD.  This assumption  is illustrated by the equation below.
          HighOBD(i) - HighOBD(i-V  = HighBAS^i)  -    	
              1  -  HighOBD(i-l)             1  - HighBASE±i-l)


 The fraction on the right-hand side is calculated from data on the base case - no OBD
 and no I/M.  The fraction is then applied, in the OBD case, to the number of normal
 plus repaired vehicles. This allows both normal vehicles, which have never been
 repaired, and vehicles that have previously been repaired, to migrate into high
 emitters.

 Once the migration into the high-emitter group  is calculated, the OBD and I/M effects
 are used to calculate the fraction of those new  high emitters repaired.  The repair
       The preliminary examination of data for 1988-and-later NOx high emitters certified to 0.4 g/mi
NOx, discussed on page 6, found 27 vehicles (instead of the 31 vehicles listed in the table on page 17)
with an average NOx emission rate of 1.298 g/mi instead of 1.294 g/mi. The standard deviation was
0.590 g/mi and the coefficient of variation (COV) was 0.590/1.298 = 45.4%. If all the data in the table on
page 17 had this same COV, the pooled estimate of the standard deviation would be large enough to
make the comparisons of the means statistically insignificant.
Peer Review of EPA MOBILE6 Report M6.EXH.007                 March 21, 2000      Page 12

-------
 calculation is based on three parameters: (1) the illumination rate for the malfunction
 indicator light (MIL), (2) the response rate - the rate at which owners will have their
 vehicles repaired when their MIL is illuminated, and (3) the emission level of repaired
 vehicles. In the absence of data,  EPA has assumed the following values for these
 parameters:
    o  The MIL will illuminate in 85% of the high emitters.
    o  The motorist response to an illuminated MIL will be 90% in an area with I/M,
       regardless of the vehicle age.
    o  In an area with no I/M program, the motorist response will be 90% during the
       warranty period from 0 to 36,000 miles; it will then drop  to 10% between 36,000
       and 80,000 miles when the emission warranty is in effect. Beyond 80,000 miles
       the response rate, in a non-l/M area, is zero.
    o  The after-repair emission level is the lesser of (1) the  normal emission level at
       the given mileage or (2) 1.5 times the emission standard for 50,000 miles.

 The report notes that different stakeholder comments either  support or recommend
 changes in the values used for the MIL illumination rate or the OBD response rate.
 These values were not changed in the final report because the commenters did  not
 provide any data to support recommended changes. Qualitatively, the report makes
 three assumptions: (1) that the MIL will illuminate most of the time in a high  emitter; (2)
 that vehicle owners will respond (i.e., have their vehicle repaired) most of the time the
 MIL is illuminated only if there is an I/M program in place or  if the vehicle is under full
 warranty; (3) the response rate becomes very small for high-mileage vehicles. These
 qualitative assumptions seem correct. However, it is not possible to make any
 definitive recommendations to support or to change the values used in the report in the
 absence of data.

 The authors state that the after-repair emission level was an assumption, based on an
 analysis of data from IM240 programs.  No details or quantitative results of that
 analysis are mentioned. The assumption that the after-repair emissions are the same
 as those of normal vehicles is consistent with the results in the I/M report, M6.IM.001
 (July 22, 1999 revision). That report contains regression equations for the ratio of
 after-repair emission level to normal emission level.  For the  most stringent cut points
 used (HC = 0.4 g/mi, CO = 10 g/mi, NOx = 1 g/mi), the average value for the ratio of
 after-repair emissions to normal emissions for vehicles fifteen years old* and less is
 1.00 for HC, 1.27 for CO and 0.68 for NOx.  For the phase-in cut points  (HC = 1.2
 g/mi, CO = 15 g/mi, NOx = 3 g/mi), the same average ratios  are 1.66 for HC, 1.43 for
 CO and 1.43 for NOx.
       The value of fifteen years was used in this average because the regression equations in
MOBILE6 are not used outside this range. The after-repair-to-normal emissions ratio for fifteen-year-old
vehicles is used for all vehicle ages greater than fifteen years.
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21, 2000       Page 13

-------
 Other issues

 The report notes that an initial draft had a correction factor to account for the effects of
 I/M on the ARB data used in the report. This factor was dropped in the final report
 reviewed here. This was a good decision because of the anomalous results presented
 by this correction factor. According to the original correction factor equation, the
 California idle I/M program  increases NOx emissions for Tier I vehicles. In that
 equation, the effects of I/M on NOx emissions are highest at zero miles and are absent
 for odometer readings of 150,000 and greater.  An I/M program that uses only HC and
 CO exhaust measurements (with no visual or functional check that may catch NOx
 defects) can increase NOx emissions, but there should be no effect at zero miles. The
 correction factor should be 1 at an odometer reading of zero and then should increase
 as vehicles went through their biannual I/M tests.  A multiplicative correction factor
 (CF) regression of the form CF = 1 + a(odo) + b(odo)2 would have no correction at zero
 mileage.  Such a regression could possibly show an increase in the correction factor
 with odometer reading and a subsequent decrease back to 1, as was found in the
 original correction factor equation. The decision to drop the I/M correction factor
 equation is justified not only by the small size of the correction factor, but also because
 the effect of the correction factor did not represent the expected behavior of an I/M
 program.
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21, 2000      Page 14

-------
     Conclusions and Recommendations for Further
 Studies
 Short-term improvements

 The comments in the previous section deal mainly with editorial changes and
 recommendations for inclusion of more statistical information about existing results.
 No short-term recommendations call for changes in the MOBILES model itself.  In
 particular, the following recommendations can be done in the next revision of the
 report:
    o  Make the revisions recommended in the Overall Clarity subsection.
    o  Include statistical information (standard error, sample size, confidence limits, p-
       values, etc.) for the regressions and average values presented in the report.
    o  Include results for Tier 2 vehicles.
    o  Provide an explicit conclusion that the results rely, to a large degree, on
       engineering judgement.
 Long-term improvements

 That more data should be obtained on new-technology vehicles is so obvious that it
 does not need to be stated.  Besides the usual data on the long-term emission
 performance of the vehicles discussed here, it will also be necessary to obtain data on
 the parameters for OBD-based I/M programs. This is especially true for the MIL
 illumination rate and the motorist response rate.  Plans to gather future I/M data should
 ensure that data fields to collect information from the OBD system are included for all
 vehicles.  This needs to be done for both I/M programs and for future inventory-related
 data-gathering studies.

 The calculation of the high-emitter fraction in this report and in other parts of MOBILES
 assumes that the average emission level of high emitters in the correction data is the
 same as that in the base fleet.  On page 16 of the M6.IM.001 report EPA noted that it
 was investigating IM240 data from Wisconsin and Colorado to confirm this
 assumption. This work should be completed to determine if there is any significant
 error in the assumption used for the high-emitter fraction calculation.

 The I/M correction factor that was not included in the final analysis of the ARB data set
 suggests a potential problem for future versions of MOBILE.  The typical emissions
 analysis is based on starting with data on vehicles that have not been through an I/M
Peer Review of EPA MOBILE6 Report M6.EXH.007               March 21, 2000      Page 15

-------
 program. With the widespread introduction of OBD and I/M programs, it will be difficult
 to get a large data set that has not been subject to an I/M program.  It is not too early
 to think about how future data collection efforts and exhaust emission models will be
 done to account for this.

 EPA should collect a formal record of the data used in the construction of MOBILES,
 including a list of the choices made (and their rationale) for including or excluding
 groups of data and individual data points.
Peer Review of EPA MOBILE6 Report M6.EXH.007                March 21, 2000       Page 16

-------