United States        Air and Radiation       EPA420-R-02-014
          Environmental Protection                March 2002
          Agency                      M6.IM.001
xvEPA    MOBILE6 Inspection /
          Maintenance Benefits
          Methodology for
          1981 through 1995 Model
          Year Light Vehicles
                                $5b Printed on Recycled
                                Paper

-------
                                                                 EPA420-R-02-014
                                                                       March 2002
                                      /
                                 for
                                      Light

                                M6.IM.001
                                Edward L. Glover
                                Dave Brzezinski
                        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.

-------
                          EXECUTIVE SUMMARY

              List of Issues, Key Points, Assumptions and User Inputs
                        Regarding MOBILE6 I/M Credits
       The methodology described in this document (M6.IM.001) covers 1981-95 model
year cars and light-duty trucks. No significant FTP based data were available for the 1994
and 1995 model year vehicles, but these were included for I/M purposes with the earlier
model years because of their general lack of On-Board Diagnostic (OBD) systems.  The
document also discusses I/M credits for running and start emissions. I/M credits are based
on a simple distribution model in which every vehicle in the fleet is either a high emitter
(FTP emission greater than 2  times HC or NOx standards or 3 times  CO standards) or a
normal emitter.  The emission levels of the high and normal emitters are based on FTP data
collected independently by EPA,  AAMA and API as part of the organizations' in-use
vehicle emission assessment programs. The frequency and distribution of high and normal
emitters in the fleet is based on a large database of EVI240 data collected in Dayton, Ohio
in 1996 and 1997. The basic  emission levels used in the model are a function of vehicle
mileage, vehicle technology, and model year.

       The basic assumption behind I/M is that a fraction of the high  emitters in the fleet
are identified and repaired down to lower emission levels during the I/M process. This
process reduces the average emission level of the fleet by reducing the fraction of High
emitters in the fleet.  The I/M benefit is the difference in fleet emission levels between the
No I/M baseline emission level and the  after I/M fleet average.

       MOBILE6 will allow various I/M scenarios to be modeled. Some of these are new
to the MOBILE model series.  The others have been changed or revamped in a significant
manner. MOBILE6 will allow for some new features.

New Features:

1.      Internal operation - Except for the constant ASM / EVI240 ratio file, used to model
       an ASM based I/M program, and the TECH12 credit file used to model pre-1981
       model year vehicles, there are No external I/M credit files to attach to the main
       program for 1981 and later model year vehicles.

2.      I/M credits given for the EVI240 test, the ASM tests, the Idle tests and OBD testing.

3.      Custom user supplied cutpoints for EVI240  can  now be entered directly in the
       program.  For example, the combination (1.5 g/mi HC, 55 g/mi CO, and 3.2 g/mi

-------
      NOX) can be entered for an IM240 scenario. Custom ASM test cutpoints cannot
      be entered directly into MOBILE6.

4.     Ability to model up to seven different exhaust and evaporative I/M programs
      simultaneously.

5.     Ability to model the exemption of the first "n" model years / ages in an I/M
      program.  The "n" can be up to the first 25 model years / ages.

6.     User input and default values for non-compliance with testing requirements, and
      cost waivers on failures can be specified.

7.     I/M credits given for cost waivered vehicles.
Development of Important Parameters

1.      The I/M methodology and associated parameters presented in this document are
       heavily based on four other EPA documents. These are "Determination of Running
       Emissions  as a Function of Mileage  for 1981-93 Model Year LDV  and LDT
       Vehicles" - M6.EXH.001, and "Determination of Start Emissions as a Function of
       Mileage and Soak Time  for  1981-93  Model  Year Light Duty Vehicles."  -
       M6.STE.003.  Also, the OBD and OBD I/M assumptions are discussed in the EPA
       MOBILE6.0 documents M6.EXH.007 and M6.EXH.009.  The '007' document
       covers the  Hydrocarbon (HC) and Nitrogen Dioxide (NOX) pollutants, and the
       '009' document covers the Carbon Monoxide  (CO) pollutant.   The  reader is
       encouraged to obtain these documents from the EPA Web site and review them.
       The paper M6. STE.003 contains the start emission parameters (average normal and
       high start emission level), and the associated statistics.

2.      Grouping Parameters - Most of the grouping of the data was done by model year
       and technology groups.  Ported fuel injection (PFI) technology was split from
       throttle body injection (TBI) and carbureted technology.  Model year groups were
       chosen based on engineering judgement regarding technology changes, or were
       grouped based on similar certification emission standards.
3.      Basic emission rate and I/M analyses were done for both cars and light trucks
       separately. The same analysis approach was used for each vehicle type; however,
       different model year grouping were selected for cars and trucks because of the

-------
       different certification standards which were in effect. Also, MOBILE6 contains the
       MOBILES I/M estimates for Heavy-Duty Gasoline Vehicles. These were NOT
       updated in MOBILE6.
4.      Basic Emission Rates - FTP emission factor data comes from significant EPA and
       industry testing (3,000+ FTP tests).  It was corrected for recruitment bias (see
       M6.EXH.001) based on IM240 testing from Dayton, Ohio (211,000 IM240 tests).
5.      Average emissions of Normals and Highs for start and running emissions - EPA
       / AAMA FTP data sample was used.
6.      Identification Rate of High emitters - These are based on a sizeable database (900
       vehicles) which received both the FTP and EVI240 tests at an EPA contractor
       facility.
7.      After I/M Repair Effects for running emissions - These are based on thousands of
       EVI240 tests from Arizona on vehicles which were repaired to pass I/M.
       After I/M Repair Effect for start emissions - These are based on FTP data collected
       by EPA.
9.      Sawtooth Methodology  - The Sawtooth algorithm has been removed from the
       MOBILE6 model exhaust I/M calculations for the 1981 and later model years. This
       change is a new and important feature of MOBILE6 since the previous Draft
       version of this document (EPA420-P-99-007) was released. The ' Sawtooth' was
       originally developed as part of the MOBILE2 model and was used in the MOBILES
       model. It was a methodology that attempted to account for fleet deterioration
       between successive I/M programs, and the standard practice of the auto industry to
       introduce its new model  of vehicle in October of the previous calendar year.

       The Sawtooth algorithm was dropped from the exhaust model for a number of
       reasons. The primary reason is that it cannot accurately be programmed into the
       MOBILE6 model. This is because the sawtooth algorithm requires knowledge of
       emission  levels  and  high emitter rates from one  subsequent and two previous
       calendar years. Unfortunately, the structure of the MOBILE6 model is such that

-------
only the current calendar year is available to the program in a given run.  Changes
to this structure to incorporate a multiple calendar run algorithm would require a
complete re-design and re-write of the MOBILE6 code. Without a complete re-
design, incorporation of a more accurate version of the 'sawtooth' would have an
extremely adverse effect on the execution time of MOBILE6.

Also,  one of the important assumptions underlying the sawtooth methodology was
the assumption that the emission deterioration of a fleet that did not have I/M was
the same as the emission deterioration of a fleet between I/M inspections. On the
surface, this sounds like a reasonable assumption given that no data to prove or
disprove it currently exists. However, another line of reasoning suggests that once
a vehicle has failed, it potentially could have a higher propensity to fail again even
if it is brought back to specifications. This higher propensity is likely due to
conditions beyond an I/M program's ability to control, such as poor manufacturer
design, build, etc. ('a lemon'), or poor  general maintenance or careless operation
by the owner. To get an accurate picture of vehicle deterioration in an I/M program,
a detailed multi-year study is required which tracks individual failures and passing
vehicles, and determines the proper level of re-failure and its emission effects.  In
the absence of such a study, EPA now believes that the assumption of  equal
deterioration rates between the fleet 'on average'  and the  previously repaired
vehicles is not likely valid, and has chosen to remove  it from the MOBILE6 model.

A proper study of the long term behavior of vehicles in I/M has never been  done.
However,  some limited work in this  area has been done by Tom  Wenzel at
Lawrence Berkeley Labs using Arizona I/M and remote sensing  data.  This new
work now suggests that I/M failures re-occur at a higher rate than the general fleet.
For example, the work suggests that the re-failure rate is in the range of 30 to 40
percent; whereas the overall fleet failure rate is in the range of 15 to 20 percent.
(Wenzel, Tom.  "Evaluation of Arizona's Enhanced I/M Program", presentation at
the 9th CRC On-Road Vehicle Emissions Workshop, April 21,1999).  These higher
rates  suggest that repaired failures are not as stable as assumed, or that  many
failures are not repaired as completely in first place as assumed.

The  sensitivity of the  emission  result to the sawtooth algorithm was  also
investigated.   It was discovered in  the course of testing  that the sawtooth
methodology has only a very  marginal effect on the size of the I/M benefits or the
after repair emission levels. For example, it reduces them slightly, typically only
one or two percent, to account for deterioration between  calendar years.   The
change in  I/M benefits from the Sawtooth is so marginal because the slope of the
emission deterioration between calendar years is small. It is this 'slope' between
inspections that the Sawtooth is attempting to model.

-------
      Despite the theoretical and practical problems  associated with the Sawtooth
      algorithm, it was retained in the MOBILE6 model in three minor areas. First, it is
      still present in the I/M Evaporative calculations.  Second, it is still present in the
      exhaust I/M calculations for pre-1981 model year vehicles.  Finally,  it is still
      present to an extremely limited degree in the Biennial I/M correction factors for the
      1981 and later model years  (see Point #10 below).  Although, retention of the
      Sawtooth in these areas  of the  model  produces  some  inconsistency,  it was
      maintained in the model primarily because its removal would require considerable
      additional engineering analysis and re-programming. Also, in the case of the pre-
      1981 model years and the Biennial I/M correction factors, its effect will likely be
      non- existent for most current calendar year runs of the MOBILE6 model, or have
      an otherwise negligible effect.
10.     One of the reasons to use the Sawtooth methodology in MOBILE6 was to account
       for the effects of a Biennial I/M program. In the absence of the Sawtooth, a new
       methodology was used that ratioed the Biennial I/M reductions from MOBILES
       with the Annual I/M reductions from MOBILE6. The resulting factor was applied
       to the annual I/M benefits in MOBILE6 to produce the reduced Biennial I/M
       reductions.

11.     Waiver Repair Levels -  In MOBILE6, cost waivered I/M failures will get some
       repair benefit. A value of a 20 percent reduction has been chosen. This value may
       updated in the future, if real data provides another value.

12.     High Emitter Non-Compliance Rate - The definition of this parameter has been
       changed. In the draft version of this document, Non Compliance was defined as the
       fraction of the fleet which either do not show up for the I/M test in the first place
       (non participants), and the fraction of the failures which show up for the  test, fail
       the test, but never show up again with either a successful repair or a waiver. In this
       case, the non participating vehicles were assumed to have the same emissions as the
       fleet average, and the fraction of the failures that did not show up were assumed to
       be High emitters. In the final version, vehicles in non-compliance will only include
       those vehicles which do not show up for  the test, and it is assumed that they have
       the fleet average emission level.  Vehicles that do not show up for the retest may
       also be considered non-compliant, and be assumed equivalent to those that do not
       show up for the initial test.  MOBILE6 does not contain a default value for this
       parameter, but requires the user to specify one.  The valid range is from 0 to 50
       percent.

-------
13.    High Emitter Waiver Rate - This is now a required user input.  It is the percent of
      I/M F AILURES that received a cost or hardship waiver to the full requirements of
      an I/M program.  The basis for this rate is NOT the percent of the total fleet or the
      percent of the tested fleet.
14.    MOBILE6 will assume that the ASM tests will have the same relative performance
      to the EVI240 that they did in MOBILES. This is necessary because no new ASM
      I/M test data matched with FTP data are available since MOBILES was released.
      New Idle and 2500RPM/Idle test data are available and new performance estimates
      have been computed, and will be installed in the MOBILE6 model. The ASM and
      Idle I/M test performance in comparison to the EVI240 will be computed in the
      MOBILE6 model by adjusting the I/M test identification rate (IDR) factors.
15.    The ASM tests assume the same after I/M repair emission levels as the EVI240 tests.
      Only the IDR rates are different. The Idle test after repair rates are the same as the
      MOBILES Idle test repair rates, and these are generally higher (less effective) than
      the corresponding ASM and EVI240 repair rates.
16.    The MOBILE6 model will not have the capability of modeling a remote sensing test
      based program or a change of ownership I/M program. This omission is the result
      of insufficient time and resources to create this feature in the model.   Code was
      developed to model RSD and change of ownership I/M.  However, it proved to be
      unreliable and was removed from the MOBILE6 program development at the end
      of the process. If future versions of MOBILE6 are developed, they may contain the
      capability to model non-periodic inspection programs.
17.    The MOBILE6 model WILL HAVE the capability to model 1996 and later model
      years using an exhaust I/M program.  However, MOBILE6 will NOT have the
      capability of modeling an OBD type I/M program on pre-1996 model years. This
      change is a  new and important feature of MOBILE6 since the previous Draft
      version of this document (EPA420-P-99-007) was released.

-------
1.0    INTRODUCTION
       This document describes EPA's new methodology for estimating exhaust emission
Inspection / Maintenance (I/M) credits.  This includes the methodology for various tests
such as the IM240, the Idle test, the 2500 RPM/Idle test, and the ASM test. It includes the
methodology used for all cars and light trucks for model years 1981 through 1995. The I/M
credits for the pre-1981 model years are not being revised for MOBILE6.  The I/M credits
for post-1995 model years with OBD systems, and the evaporative emission I/M test credits
will be discussed in a separate documents "Determination of Emissions,  OBD, and I/M
Effects for Tierl, TLEV, LEV, and ULEV Vehicles" - EPA documents M6.EXH.007,
M6.EXH.009, and "Inspection / Maintenance  Credits for Evaporative Control System
Tests" - EPA document M6.EVI.003.

       MOBILE6 will handle I/M credits differently than previous MOBILE models. One
major difference is the discontinuation of the TECHS model.  The TECHS model was a
complex external FORTRAN program which calculated and exported the exact I/M credit
values. These credit values were then built into the MOBILES block data  code or read as
an external file. The new credit methodology will instead be built into the MOBILE6 code,
and will operate automatically every time an I/M program is called by  the MOBILE6
program.   This change will give the MOBILE6 user the ability to vary the effect of
cutpoints and other program parameters through changes to the MOBILE6 input file.  No
longer will it be necessary to develop special I/M credits using  the TECHS model, and
attach them to the MOBILE program.

       The new I/M credit methodology will also be updated to reflect the new basic
emission  rates (see "Determination of Running Emissions as a Function  of Mileage for
1981-1993 Model Year Light-Duty Vehicles-Report Number M6.EXH.001"). In addition
to being lower in magnitude, the new emission rates separate start and running emissions.
MOBILE6 will account for these emissions separately, and produce separate start and
running I/M credits.

       This document is structured into six primary  sections, and an Appendix section.
Section 2 briefly describes the databases  used in the analysis and development of the
credits. Section 3 describes the methodology for development of the running exhaust I/M
credits based on the EVI240 test. Section 4 describes the periodic I/M credit calculation is
mostly mathematical terms. Section 5 describes the methodology for development of the
start exhaust I/M credits.  Section 6 describes the methodology for the development of
credits for the other types of I/M tests (Idle, 2500/Idle, and ASM).  Section 7 presents user
and peer review technical comments and EPA's response to the comments.  The document
also contains an Appendix section which  is listed A through D. Appendix A contains

-------
sample data plots, Appendix B contains sample calculations, and Appendices C and D
contain statistical diagnostics for many of the parameters used in this model.
2.0    DATA

       Four databases were utilized to develop the IM240 based credits. The first database
was a large emission factor database which contained over 5,000 initial FTP tests on 1981
through 1993 model year cars.  It was used in the I/M credit analysis to determine the
average emissions of the "Normal" emitting vehicles and the "High" emitting vehicles.
This is the same database which was used to generate the basic emission rates prior to the
application of the High Emitter Correction  Factor.  It is described in greater detail in
"Determination of Running Emissions as aFunction of Mileage for 1981-1993 Model Year
Light-Duty Vehicles" -  report number M6.EXH.001.

       The second database was a smaller I/M database. It was used to determine the high
emitter identification rates for the EVI240 test.  It contained 910,  1981  and later cars and
trucks which had both an EVI240 test and a running LA4 test (derived from the FTP test).
It contained data from EPA emission factor testing in Ann Arbor, Indiana and Arizona in
which vehicles were randomly recruited and tested on both the FTP test and the EVI240 test.

       This second vehicle emission database contains many of the same FTP / lane EVI240
test pairs that were used for the MOBILES I/M credits.  In an attempt to update the
MOBILE6 credits with newer model year data, additional vehicle data with FTP / lab
EVI240 test pairs were added where FTP / lane EVI240 were not available. Use of a lab
EVI240 versus a lane EVI240 for I/M credit purposes introduces some additional uncertainty
in the analysis since a lab EVI240 test is less similar to an actual state conducted EVI240 I/M
test than a lane EVI240. However, inclusion of the FTP / lab test data, enabled the analysis
to include some post 1991 model year vehicles and additional  light  trucks rather than
extrapolate these points. Thus, it was concluded that these benefits outweighed the slight
increase in uncertainty caused by using lab EVI240 data.

       The third database was the Arizona EVI240 database obtained  from official state
testing. It contained several thousand before-and after-repair EVI240 tests, and was used to
determine the repair effects for the running LA4 EVI240 credits.  It contains data from a
special test program that the State of Arizona conducts on a continuous basis to evaluate
the performance of their I/M program. In this program, vehicles are randomly selected to
receive the full EVI240 test both initially, and  if they fail, after all subsequent repair cycles
until they pass. EPA document - EPA 420-R-97-001 "Analysis of the Arizona EVI240 Test
Program and  Comparison with the TECHS Model" provides some detail regarding this
testing.

-------
       The fourth database of about 970 EPA tested vehicles contained both IM240 and
FTP data before and after repair. It was used to calculate the effects of repair on start
emissions.
3.0    I/M ALGORITHM FOR RUNNING EMISSIONS

3.1    Definition of Categories

       The basic purpose of I/M is to identify and repair high emitting vehicles with broken
emission control systems.   These types  of vehicles  are  termed "High" emitters,  and
typically have average emission levels which are considerably higher than the overall mean
emission levels.  The remainder of the fleet is considered to be the  "Normal" emitters.
These are low and average emitting vehicles, and their emission control systems are
generally functioning properly.  The overall  fleet emission factor is  assumed to be a
weighted average of the high and normal emitters. For comparison, the use of two emitter
classes differs from the methodology used in the previous TECHS and MOBILES models.
In those models, there were four emitter classifications (Normal, High, Very High, and
Super).
       The MOBILE6 model will generate specific I/M credits based on pollutant, model
year group, and technology type. Credits for the three pollutants HC, CO, and NOX will
be produced.  Also, credits for the 1981 through 1993 model years will be stratified into
seven separate groups.  These are: 1988-93 (PFI), 1988-93 (TBI),  1983-87 (FI), 1986+
(CARS), 1983-85 (CARS), 1981-82 (FI), and 1981-82 (CARS).  PFI means ported fuel
injection, TBI means throttle body fuel injection, (FI) means all closed-loop fuel injected,
and (CARB) means closed-loop carbureted and all open-loop vehicles combined together.

3.2    General I/M Algorithm

       Figure  1 is a general graphical view of the I/M algorithm for running emissions.
Specific algorithms for each of the model year / technology / pollutant groups will be
programed into the MOBILE6 model. Four lines are shown in Figure 1 which show the
basic emission rate, the normal emitter emission rate, the high emitter emission level, and
the after repair emission levels of the high emitters which were identified and repaired. The
basic emission rate is  shown as Line A. This line represents the average emissions of the
fleet without an  I/M  test.  It includes both the normal vehicles and the  high emitting
vehicles.
                                      10

-------
       Line B in Figure 1 represents the average emissions of the normal vehicles.  These
are the vehicles which are very unlikely to fail any EVI240 test cutpoint in the range used
by I/M programs, and should not require any significant emission related repair if they did
fail.  The line is shown as a linear function of mileage to reflect the gradual deterioration
that normal vehicles experience due to general wear.   In the data analysis these vehicles
were defined as normal emitters for a specific pollutant if their FTP HC emissions were less
than twice the applicable new car certification standard, or their FTP CO emissions were
less than three times the applicable new car certification standard, or their FTP NOX
emissions were less than twice times the new car certification standard. In MOBILE6, it
is assumed that these vehicles never fail I/M; no repair adjustment are made to them.

       Line C in Figure 1 represents the average emissions of the high vehicles. These are
the vehicles which likely have "broken" emission control systems, and that should fail the
EVI240 test cutpoint, and receive repair. In the data analysis these vehicles were defined as
high emitters for a specific pollutant if their FTP HC emissions or FTP CO emissions
exceeded twice or three times the applicable new car certification standard, respectively,
or their FTP NOX emissions were two times the new car certification standard. Because
high NOX emissions often occur with low HC and/or low CO  emissions, and sometimes
even HC can be high and CO normal, the three categories were kept separate.  Thus, a
vehicle could be a high HC emitter, but a normal CO and NOX emitter.

       The selection of twice or thrice FTP certification standards for the boundary level
between normals and highs is an engineering choice based  on the literature on I/M and
repair. Other reasonable boundary levels could also have been chosen. No formal analysis
was done to prove that these levels were optimum. One of the reasons they were chosen
is because they were used in MOBILES, and have generally  been shown in the past to be
a good dividing point between high emitting broken vehicles and lower emitting vehicles
which are not broken. Simple statistical analysis done on the data indicate that the two
means are statistically different.

       Line D represents the average emissions of the portion of high emitting vehicles that
are identified and repaired because of the I/M process. This line is calculated as a function
of vehicle age, and is a percentage (e.g., 150%) of Line B.  The portion of the fleet which
is identified by I/M will be repaired to a lower level on average.  However,  this level  is
generally not as low on average as the average of the normal vehicles. The justification for
this assumption was an analysis of Arizona EVI240 before and after repair data collected
during 1995 and 1996. (See EPA report EPA-420-R-97-001 "Analysis of the Arizona
EVI240 Test Program and Comparison with the TECHS Model" for a description of this
dataset).
                                       11

-------
3.3    Calculation of Basic Running LA4 Emission Rates

      Line A in Figure  1 represents the basic non-I/M emission rate for a given
combination of vehicle type / pollutant / model year group / technology group.  The units
represented in Figure 1 are running LA4 emissions in grams / mile. The calculation
methodology and databases used to determine these emission rates are fully documented
in the report "Determination of Running Emissions as a Function of Mileage for 1981-1993
Model Year Light-Duty Vehicles," report M6.EXH.001. The reader is  encouraged to
review this document for more  details.   Selected emission  rates were taken from
M6.EXH.001 and used in this current report as examples.
                                FIGURE 1
                 GENERAL I/M CREDITS SCHEMATIC
          LA4
       EMISSIONS
                                  MILEAGE
                                   12

-------
3.4    Calculation of Running LA4 Emission Rates for Normal Emitters

       Line B in Figure 1 represents the average emission rates for Normal emitters. These
are the low emitting vehicles in the fleet which should not fail an I/M program. Line B was
calculated by least squares regression of the emissions of the normal emitters versus
mileage in the FTP dataset. Sample sizes were satisfactory in all cases. The regression was
done for each pollutant / model year / technology group. The regression coefficients for
cars are shown in Table la and light trucks in Table Ib. The column labeled ZML contains
the zero mile coefficients, and the column DET contains  the deterioration coefficients
(slope) from the regressions (units are grams per mile per IK miles).  A sample scatterplot
of the car data and the regression line is shown in Figure A-l through A-3 in Appendix A.
Table la
Regression Coefficients for RUNNING LA4 Emissions from Normal Emitter Cars
MY
Group

1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group

PFI
TBI
FI
Carb
Carb
FI
Carb
HC Coefficients
ZML
0.0214
0.0042
0.0942
0.0774
0.1266
0.0970
0.1539
DET
0.001385
0.001701
0.001439
0.000812
0.001214
0.002250
0.001271
CO Coefficients
ZML
0.4588
0.0000
1.4448
0.5666
0.7276
1.5762
1.3932
DET
0.02293
0.01990
0.01959
0.01371
0.01691
0.02150
0.01389
NOX Coefficients
ZML
0.2006
0.2253
0.4798
0.4960
0.5555
0.4597
0.5834
DET
0.00376
0.00381
0.00188
0.00170
0.00273
0.00633
0.00233
                                       13

-------
Table Ib
Regression Coefficients for RUNNING LA4 Emissions from
Normal Emitter Light Trucks
MY
Group

1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group

PFI
TBI
FI
Carb
Carb
HC Coefficients
ZML
0.02989
0.04664
0.13384
0.26835
0.49182
DET
0.002376
0.002998
0.003280
0.002701
0.006485
CO Coefficients
ZML
0.4927
0.7663
1.6222
1.3553
7.4202
DET
0.02678
0.03442
0.04311
0.06660
0.03293
NOX Coefficients
ZML
0.3024
0.3150
0.3150
1.2872
1.6159
DET
0.003904
0.003171
0.003171
0.00010
0.000025
3.5    Calculation of Running LA4 Emission Rates for High Emitters

       Line C in Figure 1 represents the average emission rates for High emitters. These
are the vehicles in the fleet which likely have problems with their emission control systems,
and have emission levels which are considerably higher than the vehicles which do not have
problems. In the analysis they were defined as those vehicles exceeding either twice FTP
standards for HC or three times FTP standards for CO or twice NOX standards. The line
used in MOBILE6 is a flat horizontal line (constant emission level) because the emissions
of a high emitter were not found through regression analysis to be  a strong function of
mileage. One possible reason for the poor correlation is an insufficient sample size of high
emitters over a large mileage range.  This  sample size makes the regression determined
mileage coefficients statistically  unreliable.   The other  possible reason  is that the
relationship does not exist, and that high emitter emission levels are fairly constant values
(at high rates).

       Various analyzes of failing cars in EPA test programs support the use of a flat
emission rate for high emitters.  Typically, what was found during the test programs on the
newer closed loop vehicles is that if something goes seriously wrong with the emission
control  system it is likely to be catastrophic, and immediately leads to high emissions.
Furthermore, the problems are likely to be fairly discrete in their occurrence (i.e., not
mechanical wear in the carburetor that creates large numbers of high emitters over time, or
built-in obsolescence at a particular mileage).
                                        14

-------
       The weaknesses of this simplified approach are:  (1) that a certain percentage
(extremely small) of the brand new vehicles will be modeled as being high emitters.  This
result occurs because at zero miles, the regression developed estimate of normal emitter's
emission level is below the FTP and Ohio data developed  estimate of the corresponding
mean fleet emission level;  and (2) massive quantities of state IM240 data on failing
vehicles suggest that the average EVI240 emission level of a failure is a function of age
rather than a flat line. However, these data are unpreconditioned EVI240 results rather than
fully preconditioned FTP results.  They are strongly influenced by the pass/fail cutpoint
which is a function of model year, and may or may not completely represent a high emitter
as defined in this document. Nevertheless, because of the importance of this assumption,
future generations of the MOBILE6 model may use a non  constant average high emitter
level if the data warrants .

       Table 2a shows the average emissions of the high emitters (cars only) for the 21
pollutant / model year / tech groups.  Trucks are shown in Table 2b. Because of the small
sample size of high emitters in most groups, some model year / technology groups  were
combined into another model year group and across technology groups, and an overall
mean was computed for the combined group.   This combination was particularly true for
NOX emissions. For the cars and for each pollutant, the 1986-89 Carb and the 1983-85
Carb were combined and averaged together. Likewise the  1981-82 Carb and 1981-82 FI
Car groups were combined and the emissions  from the  high emitters were averaged
together. For the trucks, in some cases the fuel inj ected trucks were combined together and
a common mean high emitter emission level was computed for each pollutant.  This
combination had the effect of producing more consistent means across groups.  The high
emitter HC emission level for the  1988-93 MY PFI group is also a special case. Due to a
relatively small sample size of 1988-93 model year high emitters, and a very low average
high emitter HC emission level (the average high emitter HC emission level was lower than
the average emitter HC emission level at moderate mileages), the 1986 and 1987 model
year PFI vehicles were added to the sub-sample of 1988-93  model year PFI vehicles.  The
principal effects of this operation were to almost double the  number of high emitters in the
sub-sample,  increase  the average high emission HC level from 1.10 g/mi HC to 1.74
g/mile, and to reduce the fraction  of HC high emitters in the fleet from a theoretical 100
percent to a more reasonable level.

       The impact of this approach of averaging between groups and adding selected
vehicles to particular groups is that some high emitting vehicles contribute to the average
high emitter level of their own model year group, and to another model year group.  This
does not affect the average non-I/M running emission estimates because the normal and
high emitter split is not used to calculate the average non-I/M estimates. However, it does
affect the I/M emission rate and I/M benefits because it changes the portion of a particular
model year group's emission distribution between normals and highs.   This changed
                                      15

-------
emission distribution will affect the fraction of fleet emissions in MOBILE6 which are
identified and repaired by I/M.  It is difficult to predict the size of the overall  emission
impact (I/M and Non I/M) from this data combination because it simultaneously  increases
the average high emitter emission level, but decreases the fraction of high emitters in the
fleet.  This change also impacts (increases) the start emissions and the start I/M credits
because it changes the fraction of high start emitters in the fleet (fraction of start high
emitters is equal to the fraction  of running LA4 high emitters), but does not affect the
average start high emitter level.

       An analysis of the Ohio  EVI240 data was also done to try and estimate the high
emitter levels for running LA4 and start emissions. This was done because of  the small
numbers of high  emitters in the EPA and AAMA FTP (running LA4 and Start) data
samples. In this analysis, a large sample of Ohio vehicles were segregated into normal and
high emitters, and the average high emitter emission levels were determined and compared
with the FTP based estimates.  They compared favorably.  However, the analysis was
plagued with uncertainties such as how to separate the normals from the highs when FTP
data are  not available, the inability  to split PFI from TBI in the Ohio EVI240  data,  a
questionable transformation of EVI240 results into running LA4 and start emissions, and
unknown and possibly inconsistent conditions between lab testing and EVI240 lane testing.
Because of these problems the Ohio EVI240 data were not used to estimate the average high
emitter emission levels.
Table 2a
Mean RUNNING Emissions of High Emitter Cars
MY Group
1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group
PFI
TBI
FI
Carb
Carb
FI
Carb
HC Mean
1.740
3.394
2.372
1.845
1.845
2.372
2.372
CO Mean
36.106
46.527
37.933
27.653
27.653
37.933
37.933
NOX Mean
2.846
2.872
2.951
2.872
2.872
2.951
2.951
                                       16

-------
Table 2b
Mean RUNNING Emissions of High Emitter Light Trucks
MY Group
1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group
PFI
TBI
FI
Carb
Carb
HC Mean
2.120
3.241
2.446
2.012
3.710
CO Mean
33.283
33.283
43.870
39.415
80.726
NOX Mean
2.846
2.846
2.846
4.988
5.014
3.6    Calculation of After Repair Percentages and Emission Levels

       Line D in Figure 1 represents the average after repair emission level of high emitters
that are properly identified and repaired. In comparison, Line C represents those high
emitting vehicles that are not identified and repaired properly, or belong to owners who
evade the program after failing the initial test.  Line D is calculated by scaling up the
normal  emitter emission level  (Line B) using a multiplicative factor process which is a
function of age, pollutant and  cutpoint level (derived from Arizona EVI240 data).  The
normal  emitter emission level  is the basis for the after repair emission level, and is the
lowest emission level to which  high emitting vehicles can be repaired after adjustment for
age and mileage. This assumes that the I/M process on average does not turn aged vehicles
into brand new ones.

3.6.1   After I/M Repair Multiplicative Adjustment Factor

       The after I/M repair multiplicative adjustment factor is a function of vehicle age and
I/M cutpoint. It is calculated using a two step process. The first step is to calculate the
multiplicative adjustment factor for the standard set of EVI240 cutpoints which the State of
Arizona used in its EVI240 program. These are the phase-in cutpoints of 1.2 g/mi HC / 20
g/mi CO and 3.0 g/mi NOX. The second step involves computing and applying another
ratio which is a function of EVI240 cutpoint. It will allow the MOBILE6 program to assign
a different after repair emission level as a function of EVI240 cutpoint. The combined after
I/M repair multiplicative adjustment factor is multiplied by the normal emitter emission
level to calculate the after repair emission levels.
                                        17

-------
 Phase-in Cutpoints

       Equations 1 through 3 are the multiplicative adjustment factors used to calculate the
after repair emission level for HC, CO and NOX under phase-in cutpoints.  They were
calculated from a large sample of Arizona EVI240 data.  The same coefficients are used for
both cars and light trucks. The percent after repair I/M emission levels for the high emitters
which were identified by I/M and repaired were developed by: (1) Stratifying the sample
by age into 15 groups (ages 1 through 15); (2) Computing for each age group the average
emission level of the vehicles passing their initial Arizona I/M test; (3) Computing for each
age group the  after repair passing  emission values  of the Arizona I/M failures;  (4)
Computing for each age group the ratio of the emissions of the repaired high emitters over
the emissions of the initial passing vehicles; (5) Regressing the ratios versus age for each
of the three pollutants to produce Equations 1 through 3.

       Equations 1 through 3 are used to produce Line D for the phase-in cutpoints
(1.2/20/3.0) by following the two steps.
       First, Line D is calculated as a percentage of Line B using Equations 1 through 3.

       HC ratio      =     2.2400 - 0.07595 * (vehicle age)                  Eqn 1
       CO ratio      =     2.1582 - 0.07825 * (vehicle age)                  Eqn 2
       NOX ratio     =     1.6410 - 0.04348 * (vehicle age)                  Eqn 3
       In these equations, vehicle age ranges between 1 and 15 years, and the percentage
value at 15 years is used for all ages greater than 15.  A value of 1.0 used in cases where
the computed value is less than 1.0.

       Second, the percentage values calculated in Eqns 1 through 3 (i percentage in Eqn
4) are transformed into emission units by multiplying the percentage values by the emission
values in Line B (average emission of the normal emitters) using Eqn 4. The emission level
of the Normals is a function of mileage.
After repair emissions pollutant i = i percentage * Emissions of Normals         Eqn 4
                                       18

-------
Other Cutpoint Combinations

       Equations 1 through 4 are used to produce the after repair emission levels for an
IM240 program  which uses the phase-in cutpoints of 1.2/20/3  for HC,  CO, and NOX
respectively. Another adjustment factor is used to compute after repair emission levels for
other cutpoints. It is a multiplicative factor which proportionally increases or decreases the
after repair emission level computed for the  1.2/20/3 phase-in cutpoints to account for
tighter or looser cutpoints.

       The factor used to compute the after repair emission level for cutpoints other than
1.2/20/3 phase-in cutpoints is based on a limited amount of vehicle repair data collected
by EPA in past testing programs. It was utilized to overcome the limitation of repair data
collected at only one set of cutpoints in Arizona. This dataset was the same one used to
develop MOBILES repair effects and technician training I/M credits.  The repair effects
dataset which  was used consists of 273 vehicles from model years 1981 through 1992
tested by an EPA contractor in South Bend, Indiana and at the EPA lab in Ann Arbor, MI.
All of these vehicles had before and after repair EVI240 and FTP tests.  The sample of
vehicles were  repaired to various FTP emission level targets.  None of the after repair
results included a catalyst replacement.
       The principal goal of the data analysis was to determine as a function of EVI240
cutpoint, the FTP after repair emission levels of vehicles which initially failed the EVI240
tests and were repaired to pass the EVI240 test.  For MOBILES, this analysis was done for
seven different HC/CO cutpoint  combinations and for five NOX  cutpoints.   These
combinations are repeated in this document because they are the only after repair FTP data
for a variety of cutpoints which currently exists. These cutpoint combinations are shown
in Tables 2c and 2d. Also, shown in Tables 2c and 2d are the after repair emission levels
for each cutpoint combination group, and the ratio  of a given after repair emission level to
the after repair emission level at 1.20 g/mi HC / 20 g/mi CO. For NOX, the individual
cutpoint groups are ratioed to the 3.0 g/mi NOX group.

       It also needs to  be noted that the MOBILE6.0 program  cannot model EVI240
programs where the cutpoint is lower than 0.8 g/mi HC, 15 g/mi CO or 2.0 g/mi NOX.
Table 2c and 2d contain values for low cutpoints  that are not allowed to be modeled in
MOBILE6. They are shown for purposes of completeness of the document.
                                       19

-------
Table 2c
FTP After Repair HC and CO Emission Levels and Ratios
versus EVI240 HC/CO Cutpoint Combination
HC Cutpt
(g/mi)
1.2
0.8
0.6
0.6
0.6
0.4
0.4
CO Cutpt
(g/mi)
20
15
15
12
10
10
15
After
Repair
HC (g/mi)
1.26
1
0.88
0.87
0.86
0.78
0.74
After
Repair
CO (g/mi)
13.46
11.85
11.94
11.15
10.50
11.30
11.71
HC Ratio
1.00
0.79
0.70
0.69
0.68
0.62
0.59
CO Ratio
1.00
0.88
0.89
0.83
0.78
0.84
0.87

Table 2d
FTP After Repair NOX Emission Levels and Ratios
Versus NOX IM240 Cutpoint
NOX Cutpt
(g/mi)
1
1.5
2
2.5
3.0






After Repair
NOX (g/mi)
0.91
1.22
1.48
1.68
1.86






NOX Ratio
0.489
0.656
0.796
0.903
1.000






      For MOBILE6, the ratios data in Tables 2c and 2d were regressed versus HC, CO
and NOX outpoint to produce an after repair emission level ratio for any HC, CO or NOX
cutpoint (within the range allowed by MOBILE6) which the user may enter in MOBILE6
(the MOBILE6 user is no longer restricted to a set of seven cutpoint combinations).  A least
                                      20

-------
squares linear regression was used to produce the relationships for both HC/CO and NOX.
The regression coefficients are shown in Table 2e.  The equation form for the HC Ratio
and the CO Ratio are:
       Ratio  =
For NOX it is:
       Ratio  =
A * HCCut + B * COCut + C
B * NOCut + C
Eqn 3b
       Eqn3c
A linear regression was used instead of some other functional form because it produced
high r-squared values (0.99 for HC and NOX and 0.95 for CO). Also, note that the highest
IM240 cutpoint for HC and CO are 1.2 and 20 g/mi.  Repair effects at cutpoints higher than
these will be linear extrapolation.
Table 2e
Regression Coefficients for Repair Effects Ratios
Ratio
HC Ratio
CO Ratio
NOX Ratio
A
0.4990
0.0249

B
-l.Olle-04
0.0168
0.2538
C
0.398
0.620
0.2613
rA2
0.996
0.950
0.993
3.6.2   Application of the After Repair Adjustment Factors

       The ratio equations are used in MOBILE6 to compute the after repair emission
levels for cutpoints which are different from the standard 1.2 / 20 / 2.0 cutpoints used by
Arizona. This is done by multiplying Equations 1 or 2 or 3 by Equation 3b or 3c to produce
the repair effects ratio for the non standard (1.2/20/2.0) cutpoint. The final repair level is
obtained by multiplying this ratio by the appropriate normal emitter emission level line
(Line B). The normal emitter emission level is used as the final after repair emission level
if it is larger than the calculated after repair emission.

       The following example calculation of the after repair HC emission level for an
HC/CO cutpoint combination of 0.80g/mi HC and 15 g/mi CO is shown below for clarity.
                                       21

-------
Aft Repair HC = (2.24-0.07595*age) * (0.4990*0.8g/mi - 1.01e-04*15.0g/mi + 0.398) * Norm_ave

where

Norm_ave is the average emissions of the normal emitters.  It is a function of mileage and
technology/model year group. For an eight year old 1990 PFI vehicle at 100,000 miles it
is:  0.0214 + 0.001385 * 100 = 0.159 g/mi Running HC.

0.8g/mi HC is the HC cutpoint; 15.0g/mi is the CO cutpoint.

       Substituting the value of 0.159 g/mi and 8 years old  into the  After Repair HC
equation produces an after repair emission level ofO.206 g/mi running HC at a cutpoint of
0.80 g/mi  HC and 15 g/mi CO for an eight year old vehicle with 100,000 miles.  This
compares with an after repair emission level for the same age and mileage of 0.260 g/mi
running HC  at a cutpoint combination of 1.2/20 g/mi HC/CO. In this example, the after
repair emission level (0.206 g/mi HC) is above the value of the normal emitter (0.159 g/mi
HC).  However, if the calculation produced a value  which was lower, then the normal
emitter value would be used.
3.6.3   Discussion of the After Repair Adjustment Factors

       This approach attempts to utilize the large sample of before and after repair EVI240
data collected in Arizona. These data are an improvement over the MOBILES assumptions
since they are a large sample, and are representative of the actual I/M experience.   The in-
use data reflects the fact that regular commercial mechanics performed the repairs under
actual cost conditions.  Also, the repairs were targeted to passing the actual state EVI240
test. Many of these technicians also received some training and orientation to the EVI240
program provided or encouraged by the State of Arizona prior to its implementation. The
principal assumption underlying this approach is the ratio between the after repair EVI240
emission level and the emission level of the vehicles passing the state EVI240 test is the
same as the ratio of the after repair running LA4 emission level and the normal  emitter
running LA4 emission level. This is not an unreasonable assumption;  however, there are
potential differences between the unpreconditioned EVI240 and the preconditioned running
LA4 test.

       One drawback to the approach is that the Arizona data (and other states' data) were
available at only  one cutpoint level (phase-in  cutpoints).  This made it  impossible to
determine the sensitivity of repair levels to the EVI240 cutpoint.  To overcome this obstacle
the previous FTP databases used for MOBILES were used to make the after repair effects
a function of cutpoint. A drawback to the use of these FTP data is that they are a relatively
small sample, the repairs were  often performed by expert emission control  system

                                       22

-------
technicians rather than commercial technicians, cost was usually not a factor in the repairs,
and specified numerical repair targets based on the FTP test were used. Also, running LA4
were not available so the FTP data were used directly under the assumption that the ratio
between cutpoints is same for the FTP and the running LA4.
3.6.4   Technician Training Effects

       MOBILES had built-in I/M credits available for EVI240 programs which conducted
some form of technician training for people involved in I/M repairs. In MOBILE6, the after
repair emission levels discussed previously in Section 3.6 already include the effects of
technician training. This is because Arizona conducted a technician training program prior
and during implementation of their EVI240 program from which the repair effects data are
based.

       MOBILE6 will use as a default, after repair emission levels which are those 'with
technician training'.  For I/M programs which do not conduct a technician's training
program - 'w/o technician training', the after I/M repair emission levels will be increased
by the percentages shown in Table 2f.

       The percentages shown in Table 2f are based on a limited study done by EPA to
evaluate technician training in an EVI240 program. In the program, eleven experienced
technicians in Arizona were trained on the eve of the EVI240 implementation in 1995 to
repair emission failures using a training program developed by Aspire, Inc., and taught by
an expert emission control system technician/trainer under EPA contract. Each participant
received the training and three vehicles to repair following the training. Unfortunately,
budget limitations prevented a good pre-training baseline of the technicians' performance
to be established.  The study is fully documented in SAE Paper 960091.

       The emission results shown in columns 2 and 3 of Table 2f are EVI240 test results
in units of grams per mile.  The Student Tech column shows two numbers.  The first
number is the before any repair emission level. It is shown for comparison only, and to
demonstrate that the technicians made sizeable emission reductions from repairs.  The
second number is the average after repair EVI240 emission levels of the vehicles after the
students completed their work. The Master Tech  column shows the average after repair
EVI240 levels after the instructor completed any additional repairs which were needed to
bring the vehicle into complete compliance. On a few vehicles this included a new catalytic
converter.

       The % Difference column is the percent difference between the after repair student
tech and the after repair master tech emission results with the after repair master tech results
                                       23

-------
as the basis. It demonstrates the potential difference in performance between a master tech
and a trainee (journeyman) tech. It is proposed for MOBILE6 to calculate the 'w/o tech
training' after repair levels (w/o means without) by increasing the 'with tech training' values
by the % Difference values in Table 2f
Table 2f
Technician Training Emission Effects
Pollutant
HC
CO
NOX
Master Tech
IM240 (g/mi)
0.38
3.00
1.11
Student Tech
IM240 (g/mi)
2.16/0.68
26.4/8.21
3.66/1.54
% Difference
78%
174 %
39%
       Use of these limited data in MOBILE6 for technician training effects requires two
important assumptions. First, that the after repair levels developed in the previous sections
already contain the effects of technician training.  This is a reasonable assumption since
Arizona did institute a technician training program, and the after repair emission levels are
at relatively low levels.  Second, that the difference on  a percentage basis between the
master tech performance and the student tech performance is the same as the percentage
difference between the with and w/o technician training in the overall fleet.  This
assumption is a little tenuous since the performance of typical trained technician is not as
high as the master tech in this study.   This would have a tendency to produce a larger
percentage increase than in actuality.   On the other hand, the student tech results were
collected after the training rather than before the training, and do not strictly represent un-
trained technicians.  This factor would have a tendency to produce a smaller percentage
increase than in actuality.
3.7    Waiver Repair Line

       Not shown in Figure 1 is the waiver vehicle repair line.  However, this line falls
between the high emitter level and the after proper repairs line. These are failing vehicles
which received a waiver from program requirements because a minimum amount of money
was spent on unsuccessful or only partially successful repairs.  Typically, in most I/M
                                       24

-------
programs this means that between $200 and $450 was spent on the vehicle, and it still fails
the I/M test. The waiver repair line is below the high emitter line, despite the vehicle's
failing status, because even some limited or ineffective repair translates into reduced
emissions on average.

       Because no  analysis  has yet  been  conducted on data from  operating EVI240
programs to estimate the after I/M emission level of vehicles which were waived from the
requirement to pass the test, an assumed reduction percentage will have to be used, or the
individual user will have to  provide  a value.  The default value will be a 20 percent
reduction from the high emitter line for all pollutants.

3.8    Percentage of High and Normal Emitters in the Fleet

       Figure  1 shows in a general sense the overall fleet average emission level, the
average emissions of the normal emitters, and the average emissions of the high emitters.
The fleet average emission level was developed independent of the I/M credits, and the
methodology for its development is documented in EPA document M6.EXH.OO 1. In-order
to compute the I/M credits, the percentage of high emitters and normal emitters in the fleet
must also be calculated. Fortunately, this is an easy task since the average emission rate
is a weighted  average of the normal emission rate and the high emission rate.   The
weighting factors are simply back calculated to make this true at all odometers.

       The fraction of High and Normal emitters is calculated for each combination of
vehicle type /  pollutant / model  year /  technology group using the following general
equations.

Where:

Highs = fraction of High emitters at each age point
Normals = fraction of Normal emitters at each age point
LA4 is the average emission rate at each age point (determined in M6.EXH.001)
High_ave is the high emitter emission average at each age point
Norm_ave is the normal emitter emission average at each age point

       Highs + Normals = 1                                                Eqn 5

and
       LA4 = High_ave * Highs + Norm_ave * Normals                      Eqn 6

Solving for the variables Highs and Normals produces:
                                      25

-------
       Highs = (LA4  - Norm_ave) / (Highave - Norm_ave)                     Eqn 7

       Normals = 1 - Highs                                                 Eqn 8
       For the model year groups of 1981-82 and 1983-85 HC and CO emissions, it was
found that the base emission factors at higher mileage levels become higher than the
average emissions of the high emitters.  It occurs because at high mileages the basic
emission factors are data extrapolations.  However, under the structure of the model, this
is not possible, and it implies that the fleet contains more than 100 percent high emitters.
To overcome this inconsistency, it was assumed that the average base emission factors
could not continue to rise after it reaches the average of the high emitters, and that it would
be set to the average of the high  emitters.  Typically, the cross-over point is between
150,000 and 200,000 miles, and after this point is reached, it is assumed that the percentage
of highs in the fleet for this model year group / technology is 100 percent. This flattening
of the emission factor line at very high mileages is consistent with some remote sensing
studies. A physical explanation would be that while some surviving vehicles continue to
deteriorate, the worst emitters are progressively scrapped out of the fleet in the high mileage
range.
3.9    High Emitter Identification Rates

       The high emitter identification rate (IDR) represents the ability of an I/M test to
identify (fail) vehicles which are high emitters.  It is represented as the percentage of the
total sum of emissions from the high emitters in the fleet. For example, the DDR would be
100 percent if it identified all of the running LA4 emissions from the high emitters in the
fleet. For the HC and CO I/M credits, the IDR is a function of the EVI240 HC and CO
cutpoints. For NOX I/M credits, it is a function of the NOX cutpoints only. In MOBILE6,
the user will be able  to supply the exact EVI240 cutpoints which are desired, and  the
program will automatically calculate the IDR and the credits.  The EVI240 cutpoints will
need to be in the ranges: HC: 0.80 to 5.0 grams/mile; CO: 15.0 to 100.0 grams/mile; and
NOX: 2.0 to 5.0 grams/mile.

       The I/M  IDRs equations were calculated from the 910 vehicle database that
contained vehicle emission data from both running LA4 tests (FTP tests) and EVI240 tests
on lane fuel on cars and trucks. Cars and trucks will have the same DDR rates in MOBILE6
at a given cutpoint.  However, separate cutpoints will be allowed for cars and trucks and
for each model year in a given MOBILE6 run. The analysis to develop the IDRs consisted
of several steps:
                                       26

-------
       (1) The sample was split into two groups - the high HC and CO emitters, and the
high NOX emitters. There was some overlap between the groups. These two groups were
kept separate throughout the rest of the IDR analysis.  (2) The total HC, CO, and NOX
emissions from all of the High emitters in the sample was calculated. (3) A total of 75 HC
/ CO cutpoint combinations were developed. These ranged from (0.5g/mi HC / 5g/mi CO)
to (5.0g/mi HC / lOOg/mi CO).  For NOX, eight cutpoints were used that ranged from 1.0
g/mi to 5.0 g/mi. (4) The runningLA4 emissions identification rate (IDR) was determined
for each cutpoint combination.  For example, the  strict cutpoint combination  of 0.5 g/mi
HC / 5.0 g/mi CO might identify 90 percent of the total emissions of the high emitters
whereas the lenient cutpoint combination of 5.0 g/mi HC / 100 g/mi CO might identify only
10 percent of the total emissions.  (5) The identification rate (DDR) were calculated for 75
HC/CO cutpoint combinations, and these points were least squared regressed versus the
natural logarithms of the HC and CO cutpoint.  Natural log regressions were used because
they produced better fits, and better satisfied the inherent assumptions behind least squares
linear regression.  The logarithm form  also makes sense physically given the skewed
distribution of emissions. For example, a change  of the HC cutpoint from 1.0 to 1.5 g/mi
has a larger effect on IDR than a change from 4.0 to 4.5 g/mi. The regression coefficients
are  shown in Equations 9 and 10. (6) The NOX emission identification rate (IDR) were
also calculated for eight cutpoints and fitted to a cubic equation.  The cubic form was
chosen because it provides  a very good fit, and does not create anomalous results such as
an IDR decrease as the cutpoint gets more stringent (See Appendix D). Simpler, linear fits
for  both the HC/CO cutpoint and the NOX cutpoint IDR, and a fit including all three
pollutants simultaneously  were also investigated.  These  were rejected due to poor
statistical correlation, and anomalous results for the case of all three pollutants.

       In MOBILE6, the IDRs for all 1981 and later cars and light trucks are represented
by Equations 9 through 11.  Where In(HCcut), In(COcut), and ln(NOcut) are the cutpoints
transformed into natural logarithm space.
       HC IDR= 1.1451 - 0.1365*ln(HCcut) - 0.1069*ln(COcut)               Eqn 9

       COIDR= 1.1880 - 0.1073*ln(HCcut) - 0.1298*ln(COcut)               Eqn 10


The NOX IDR equation is a cubic form:

       NOXIDR = 0.5453 + 0.7568*NOcut - 0.3687*NOcut2 + 0.0406*NOcut3  Eqn 11


The statistics for both the logarithmic fit and the cubic fit are shown in Appendix D.


                                       27

-------
3.10   I/MNon-Compliance Rates

       One potential problem in I/M is that of non-compliant vehicles. By definition, the
compliance rate is the percentage of vehicles in the fleet that complete the I/M program and
receive either a  certificate of compliance or a waiver.  The Non-Compliance rate  is
therefore, the percentage of vehicles in the fleet that do NOT complete the I/M program
with either a certificate or a waiver.

       A non-compliant vehicle may occur in one of two mechanisms. In the first method
vehicles simply do not show up for their initial test (owners ignore I/M or go out of their
way to avoid it).  If these vehicles are normal emitting vehicles (passing the I/M test) they
have no effect on the result; however, if they are high emitters then they should have the
same effect as the initial failures which never pass or get waived. Unfortunately, because
they do  not show up  for I/M it is  impossible to determine these statistics.  As an
approximation, the model assumes that  a non-compliant vehicle emits at the level of the
average vehicle in the fleet (i.e., mixture of failures and passes).

       In the second method, vehicles show up for the initial test, fail the initial test, but
never return for a successful  retest or a  waiver.  Clearly, these vehicles are failures, and
getting them and other failures repaired is the goal of I/M.  Failure to repair such vehicles
should seemingly impose a larger credit loss than a simple random participation loss that
is imposed for non-compliance mechanism one. Nevertheless, the one mitigating factor in
this case is the fact  that the outcome  of such vehicles is unknown.  For  example,  some
research done by Colorado and Arizona to identify and track such vehicles, suggests that
many are sold outside of the I/M program area or are scrapped. If such is the case, then the
excess emissions created by these vehicles has been eliminated by the I/M program. Thus,
as an approximation,  the MOBILE6 model assumes that  a  non-compliant vehicle  of
mechanism two emits at the level of the average vehicle in the fleet (i.e., mixture of failures
and passes).
                                       28

-------
4.0    I/M Credit Calculation
4.1    General Considerations of the I/M Algorithm

       In this section all of the individual parameters discussed in previous sections such
as High and Normal emitter rates and emission levels, waiver and non-compliance rates,
and I/M identification rates are  shown in mathematical form, and utilized together to
calculate the I/M benefits. This section supercedes Section 4.0 in previous Draft versions
of this  document  that discussed the 'Sawtooth" methodology.    The  'Sawtooth"
methodology has been replaced in favor of this new simpler methodology. The Executive
summary of this document contains a brief rationale for this decision.

       Throughout the calculations, the MOBILE6  program does not use "continuous"
regression lines of emissions versus mileage (No I/M and I/M) or the fleet fraction of High
emitters versus mileage. Instead,  all of the calculations are done at discrete points on these
lines. Each point on the line represents a particular  vehicle age that ranges from  1 to 26
years and a corresponding mileage that is associated with each age.
4.2    Mathematical Description of the I/M Algorithm

       The MOBILE6 model generates separate I/M credits for each combination of
vehicle type / pollutant / model year group / technology class / EPA certification standard
type for all 1981 and later model years.  The I/M credits (percent reduction) for each
combination are generated by computing the percent difference between the basic emission
rate line with No I/M (No I/M EF) and the average emission line with the effects of I/M
included (With I/M EF). Mathematically, this is shown in Equation 12a and Equation 12b.
I/M Benefit   =     (No I/M EF - With I/M EF)                           Eqn 12a

%I/M Credit  =     I/M Benefit / No I/M EF                             Eqn 12b

       The 'I/M Benefit' in units of grams per mile (or grams for start emissions) is
calculated using Equation 13a and 13b (Equation 13b is a simplified version of Equation
13a). Equation 13a shows that theoretically the EVI Benefit is the sum of the repair benefits
of the high emitting vehicles and the normal emitting vehicles.

I/M Benefit   =     (HighEF - Repair_Net) * High  +
                          (NormEF - NormRepair) * (1.0 - High)          Eqn 13 a
                                      29

-------
Where, 'HighEF' is the high emitter emission level, 'Repair_Net' is the net after repair
emission level for  the high emitters,  'NormEF'  is the normal emitter emission level,
'NormRepair' is the after repair emission level of the normal emitters, and 'High' is the
fraction of high emitters in the fleet prior to I/M.

       In MOBILE6, it  is assumed that overall  an I/M program has no effect on the
emission level  of  normal  emitters.   Thus,  the  terms NormEF  and NormRepair are
equivalent and cancel each out.  This  allows the  simplified form of Equation 13a to be
written as Equation 13b, and allows the I/M benefit to be stated as the difference between
the High emitter emission level (HighEF) and a "Composite" Repaired High emitter
emission level (Repair_Net). The term BienADJ is also added to equation 13b to correct
for program inspection frequency. An annual frequency program has a value of 1.0 for this
term.  The values for a biennial program are discussed in Section 4.4.

I/M Benefit   =     (HighEF - Repair_Net) * High * BienADJ             Eqn 13b
       The term 'Repair_Nef is the weighted composite after repair emission level based
on four possible outcomes in an I/M scenario.   Equation 14 shows the mathematical
equation used to calculate 'Repair_Nef.  The four possible outcomes are described as
follows, and are shown in Equation 14 as the 'RepairX' variables.

Repair_Net   =     Repair 1 + Repair2 + Repair3 + Repair4                 Eqn 14
1.      High emitters NOT identified by the I/M process and remain in the fleet (Equation
       15a).
2.      High emitters in general non-compliance of the I/M test requirements (i.e., they do
       not show up for the initial test (Equation 15b).
3.      High emitters  properly  identified by the I/M process,  but are not  repaired
       sufficiently to pass the test (Waivered).  However,  they are  assumed to receive
       some effective repair (Equation 15c).  In MOBILE6 this is assumed to be a 20
       percent reduction from the High emitter level. (WAVRDC =1.0- 0.20 = 0.80). See
       Section 3.7 for more discussion.
4.      High emitters properly identified by the I/M process, and are effectively repaired.
       These vehicles are responsible for the majority of the I/M benefits (Equation 15d).
Repairl      =     HighEF * %NTIDD                                 Eqn 15a

Repair2      =     HighEF * %NCOMP                                Eqn 15b


                                      30

-------
Repairs      =     HighEF * WAVRDC * %WVRS                      Eqn 15c

Repair4      =     Repaired * %Repaired                              Eqn 15d
       The values for the variable 'Repaired' emissions in Equations 15a through 15d are
presented and discussed in detail in Section 3.6 of this document.

       The variables '%XXXX' in Equation 15a through 15d are the weighting factors for
each of the Repair outcomes. They are mathematically shown (as fractions not percentages)
and described in Equations 16a through 16d.

%NTIDD is the weighting factor used to account for the High emitting vehicles which are
not identified by the I/M process. Mathematically, it is  shown in Equation 16a. The terms
'IDR' and 'NonCom' are the identification rate for the I/M test (described in Section 3.9),
and the non compliance rate (described in Section 3.10).

%NTIDD     =      (1.0 -IDR)* (1.0  -NonCom)                        Eqn 16a
%NCOMP is the weighting factor used to account for the vehicles which do not show up
for the I/M test and for those vehicles which disappear from the I/M process immediately
after the first failing test. Mathematically, it is shown in Equation 16b.

%NCOMP   =     NonCom                                           Eqn 16b
%WVRS is the weighting factor used to account for the vehicles which fail the initial I/M,
get some repair, but do not pass the final test. Mathematically, it is shown in Equation 16c.
The term 'Waiver' is defined as the waiver rate of the program (See Section 3.7).

%WVRS     =      IDR * Waiver * (1.0 - NonCom)                      Eqn  16c
%Repairedis the weighting factor used to account for the vehicles which fail the initial I/M
test,  and are effectively repaired to pass the final test.  Mathematically, it is shown in
Equation 16d.

%Repaired   =     IDR * (1.0  - Waiver) * (1.0 - NonCom)              Eqn 16d
                                      31

-------
4.3    Effect of Exemptions on I/M Credits

       I/M exemptions are a provision granted to some vehicles which would ordinarily
be subject to an I/M inspection that excuses them from all of the testing and repair
requirements of I/M. In practice, this means that the motorist does not have to bring the
vehicle in for an I/M test; however, it may require the motorist to have received a roadside
remote sensing device (RSD) "clean screening" test(s), or to have paid a fee in-lieu of the
test.

4.4    Biennial I/M Credits

       One of the benefits of the previous 'sawtooth' I/M methodology was its explicit
ability to account for vehicle deterioration between inspection cycles.    This explicit
deterioration function made it possible to account for biennial or even longer inspection
cycles by varying the deterioration function over time.  In the new methodology the
' sawtooth' has been replaced with a multiplicative correction factor. This factor is simply
the ratio  of the biennial and annual  credits from the MOBILES model. It was created by
averaging by model year the  MOBILES biennial and annual I/M credits from the EVI240
test with  phase-in cutpoints to create a single set of multiplicative  correction  factors that
is a function of age and pollutant. These are shown in Table 3a - "Annual to Biennial I/M
Correction Factors".

       The biennial credits are  applied in the MOBILE6 model by first calculating the
respective annual credits (See Section 4.2), and then applying the biennial correction factors
in  Table 3a.  This was shown in Section 4.2 in Equation  13. The values of 0.0000 for the
biennial  test  correction factors in Table  3a reflect the MOBILES  (and MOBILE6)
assumption that vehicles less than one year in age are exempt from program requirements.
 The reader should  also note that the biennial adjustments gradually rise with age, and
become almost equivalent for vehicle ages 15 and greater.
                                       32

-------
               Table 3 a
Annual to Biennial I/M Correction Factors
Age
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
HC
0.0000
0.4966
0.5877
0.6900
0.7400
0.7773
0.8000
0.8356
0.8740
0.8914
0.9200
0.9393
0.9468
0.9532
0.9595
0.9648
0.9689
0.9729
0.9755
0.9776
0.9794
0.9810
0.9828
0.9844
0.9852
CO
0.0000
0.4976
0.5991
0.7100
0.7600
0.8000
0.8300
0.8640
0.8943
0.9083
0.9300
0.9469
0.9530
0.9589
0.9632
0.9673
0.9709
0.9744
0.9769
0.9788
0.9813
0.9829
0.9836
0.9849
0.9864
NOX
0.0000
0.5167
0.6136
0.7000
0.7500
0.7804
0.8100
0.8372
0.8730
0.8966
0.9134
0.9246
0.9353
0.9439
0.9515
0.9568
0.9615
0.9670
0.9720
0.9741
0.9757
0.9781
0.9793
0.9815
0.9826
                  33

-------
5.0    I/M ALGORITHM FOR START EMISSIONS
5.1    General I/M Algorithm

       The MOBILE6 model will also compute I/M credit reductions for start emissions
in addition to the running LA4  emissions.  The start I/M credits will be small in magnitude
since the typical I/M test (i.e., EVI240, idle, etc) does not intentionally involve testing a
vehicle during start or warm-up. The I/M credits for start emissions will reflect this fact
by assuming that vehicles with high start emissions are identified in conjunction with a
running emission failure.

       The generalized structure of the start I/M credit algorithm is the same structure as
used for the running LA4 emission credits (See Figure 1).  However, the Y-axis represents
start emissions in units of grams per start and the X-axis represents mileage. Line A shows
the basic start emission factor line before an I/M reduction. Line B shows the average start
emissions of the normal emitting vehicles. Line C shows the average start emissions of the
high emitting vehicles.
5.2    I/M Start Emission Rates

       The basic emission rates  for  start  emissions  (Line  A of Figure  1) and the
methodology used to develop them can be found in the EPA document "Determination of
Start Emissions as aFunction of Mileage and Soak Time for 1981-1993 Model Year Light-
Duty Vehicles" - Report Number M6.STE.003.

       Table 4 contains the start emission regression coefficients for the normal emitting
vehicles for all seven technology and model year groups.  Table 5 contains the average start
emissions (grams per start) from the high emitting vehicles (high emitters are defined
based on twice or thrice FTP standards - see Section 3.2). Table 6 shows the average after
repair level of the high emitting vehicles in units of grams per start. The values shown in
Table 6 are based on after repair emission testing. In these cases high emitting vehicles
(high FTP emissions or EVI240 failures) were tested, repaired and retested.
                                       34

-------
Table 4a
Regression Coefficients for START Emissions from Normal Emitter CARS

MY
Group

1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group

PFI
TBI
FI
Carb
Carb
FI
Carb
HC Coefficients
ZML
1.9987
1.9019
2.3589
1.4934
1.5892
2.3543
2.1213
DET
0.006830
0.002679
0.001388
0.018238
0.009408
0.008533
0.013610
CO Coefficients
ZML
18.972
19.233
19.949
24.698
24.442
20.038
28.637
DET
0.00703
0.00000
0.00000
0.10947
0.10577
0.22673
0.22673
NOX Coefficients
ZML
1.444
2.300
1.461
1.405
0.748
1.530
1.601
DET
0.00220
0.00000
0.00141
0.00000
0.00524
0.00059
0.00000

MY Group
1988-93
1988-93
1983-87
1986-89
Table 4b
Mean START Emissions of Hish Emitter CARS

Tech
Group
PFI
TBI
FI
Carb
HC Mean
4.829
3.293
5.313
10.520
CO Mean
38.06
27.16
65.31
92.82

NOX Mean
Same as
Normals
Same as
Normals
Same as
Normals
Same as
Normals
35

-------
1983-85
1981-82
1981-82
Carb
FI
Carb
10.520
5.313
10.520
92.82
92.82
92.82
Same as
Normals
Same as
Normals
Same as
Normals

MY
Group

1988-93
1988-93
1981-87
1984-93
1981-83
Table 5a
Regression Coefficients for START Emissions from

Tech
Group

PFI
TBI
FI
Carb
Carb
Normal Emitter Light Trucks

HC Coefficients
ZML
2.873
4.073
2.599
3.916
6.817
DET
0.00000
0.01309
0.00964
0.00854
0.00154
CO Coefficients
ZML
32.178
42.456
23.497
78.286
98.432
DET
0.0168
0.1411
0.0613
0.2564
0.3240


NOX Coefficients
ZML
1.597
4.294
1.384
0.143
1.082
DET
0.00000
0.00324
0.00000
0.00436
0.00000
36

-------

MY Group
1988-93
1988-93
1981-87
1984-93
1981-83
Table 5b
Mean START Emissions of Hish Emitter Trucks

Tech
Group
PFI
TBI
FI
Carb
Carb
HC Mean
5.212
5.212
5.826
9.406
17.865
CO Mean
83.862
83.862
60.319
162.115
179.549

NOX Mean
Same as
Normals
Same as
Normals
Same as
Normals
Same as
Normals
Same as
Normals
37

-------
Table 6
START Emission Regression Coefficients for High Emitters After Repair
Cars and Trucks
MY
Group

1990-93
1990-93
1986-89
1986-89
1983-85
1983-85
1981-82
1981-82
Tech
Group

PFI
TBI
FI
Carb
FI
Carb
FI
Carb
HC Coefficients
(g/start)
ZML
2.60
2.60
3.11
3.11
2.70
2.70
2.70
2.70
DET
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
CO Coefficients
(g/start)
ZML
18.90
18.90
30.05
30.05
28.33
28.33
28.33
28.33
DET
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.0000
NOX Coefficients
(g/start)
ZML
1.48
1.48
1.49
1.49
1.84
1.84
1.84
1.84
DET
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
5.:
Fraction of High and Normal Emitters in the Fleet
       The basic start emission factor is computed from a weighted average of the highs
and normals. The fraction of high emitters (fraction of normal emitters = 1 - fraction of high
emitters) in the fleet is the weighting factor. The fraction of high start emitters is assumed
to be the same fraction as the one used for the running emissions calculations.  Tables 3a
and 3b and Appendix A in EPA document M6.STE.003 "Determination of Start Emissions
as a Function of Mileage and Soak Time for 1981-1993 Model Year Light-duty Vehicles"
show and explain the fraction of HC and CO high emitters in the fleet at selected mileages
/ ages for each pollutant. The fraction of NOX high emitters is not shown because for NOX
the Normals and Highs are assumed to have the same emission rate (no start NOX highs
are assumed to exist).
5.4    I/M Start Identification Rates

       The algorithm for start emissions is based on test data that indicates that a portion
of the vehicles with high running emissions that are identified by the I/M process will also
                                       38

-------
have high start emissions, and that these will be identified and corrected in conjunction
with the repairs to pass the I/M test. Also, because significant NOX emissions usually form
only after the vehicle is warm, it was assumed that an I/M program could only reduce HC
and CO start emissions.

       A mathematical function that relates HC / CO cutpoint with the start emissions
identification rate (DDR) was developed from the 910 vehicle sample used to develop the
running emissions IDR.  The same methodology was used to develop the Start emission
DDR as was used to develop the running emission IDR (See Section 3.9 for a more detailed
explanation). This function also has the same range of HC and CO cutpoints (HC ranges
from 0.80 g/mi to 5.0 g/mi and CO ranges from 15.0 g/mi to 100 g/mi) used in the running
emission analysis. It predicts the percentage of start emissions from high emitters which
are identified at a specific HC/CO cutpoint level. This is the percentage of the emissions
from high emitters at Line C in Figure 1 that are reduced down to average fleet emission
levels (Line A in Figure 1).  The statistical results are shown in Appendix D.  The functions
are:

       StartHC IDR = 0.9814 - 0.1590*ln(HCCUT) - 0.1409*ln(COCUT)      Eqn32

       StartCO IDR= 1.1460 - 0.1593*ln(HCCUT) - 0.1707*ln(COCUT)      Eqn33

5.5    Average Start Emissions After I/M

       The equation used to calculate the average start emissions after I/M is very similar
in form to Equation 12a used to calculate the average running emissions after I/M. Several
of the parameters are the same such as the fraction  of high emitters in the fleet, the waiver
rate, the waiver repair percentage, and the non-compliance rate. The principal differences
are the different IDR rates (the start IDRs are calculated in Equations 32  and 33), and the
different  after repair emission levels. Equation 34 is used to calculate the After I/M start
emissions (S_EIM).  S_IDR is the start emission IDR from Equations 32 and 33,  and
S_RLEV is the after successful repair emission level (in units of grams per start) given in
Table 6. The after repair start emission levels in grams per start (S_RLEV) shown in Table
6 are used to model I/M start emissions instead of the running emission algorithm discussed
in Section 3.6.
                                       39

-------
6.0    I/M Credits for Non-IM240 Tests
       The previous sections discussed the general algorithm and methodology used to
develop the I/M credits for MOBILE6. The EVI240 test was used as the basis for the credits
because of the large  amount of EVI240 data which are available to develop the IDR
estimates and the after repair levels. I/M credits for other tests are also needed such as the
Idle  test, the 2500 RPM / Idle test,  and  the ASM  tests.  The algorithm used  to
mathematically implement these test types in MOBILE6 is analogous to the  EVI240
algorithm. The difference between the various I/M test types in MOBILE6 will be based
on the differences in the IDRs for each test.
6.1     Other I/M Tests

       The MOBILE6 model will also compute I/M credits for tests other than the EVI240
test.  The test options which will be built into the model are (1) Idle test, (2) 2500 RPM
/ Idle test and the Loaded / Idle test, (3) ASM tests, and (4) On-board Diagnostic (OBD)
I/M tests.  The OBD I/M test parameters and algorithm are discussed in EPA papers
M6.EXH.007 and M6.EXH.009.

       The default I/M tests in addition to the EVI240 test which MOBILE6 will able to
model are:

1.      Annual Two-Mode ASM 2525/5015 with Phase-in Outpoints
2.      Annual Two-Mode ASM 2525/5015 with Final Outpoints
3.      Annual Single-Mode ASM 5015 with Phase-in Outpoints
4.      Annual Single-Mode ASM 5015 with Final Outpoints
5.      Annual Single-Mode ASM 2525 with Phase-in Outpoints
6.      Annual Single-Mode ASM 2525 with Final Outpoints
7.      Annual Idle Test
8.      Annual 2500 RPM / Idle Test
9.      Annual Loaded / Idle Test
10.     Biennial Two-Mode ASM 2525/5015 with Phase-in Outpoints
11.     Biennial Two-Mode ASM 2525/5015 with Final Outpoints
12.     Biennial Single-Mode ASM 5015 with Phase-in Outpoints
13.     Biennial Single-Mode ASM 5015 with Final Outpoints
14.     Biennial Single-Mode ASM 2525 with Phase-in Outpoints
15.     Biennial Single-Mode ASM 2525 with Final Outpoints
16.     Biennial Idle Test
17.     Biennial 2500 RPM / Idle Test
                                      40

-------
18.    Biennial Loaded / Idle Test
19.    OBDI/M
6.2    ASM Tests

       Unfortunately, new paired ASM and FTP test data are not available on any ASM
I/M tests in-order to compute new and specific IDR rates or repair effectiveness rates. As
a result, the relative size of the I/M credits of these tests versus the EVI240 will remain the
same between MOBILES and MOBILE6. This was accomplished by first computing the
ratio of the MOBILES I/M credit value for an alternative ASM test over the MOBILES I/M
credit value for the EVI240 at final cutpoints of 0.8 HC / 15 CO / 2.0 NOX. When done for
each combination of model year, age and pollutant, this produces a large array of ratios (25
ages x  18 model year x 3 pollutants). Separate arrays of ASM/EVI240 credit ratios were
calculated for three ASM tests (ASM5015, ASM2525 and ASM Two Mode), and for both
Phase-in and Final ASM  cutpoint combinations.  A large array containing all six test
type/cutpoint combinations was then assembled for use in MOBILE6.  Rather than store
all those ratios in the MOBILE6 program, the ratio data are read into the program from a
separate data file if MOBILE6 is asked to calculate the effects of ASM I/M.  The ratios are
used in MOBILE6 to calculate ASM IDR rates. This is done by multiplying the appropriate
ASM ratio by the EVI240 ratio.

       The advantage of this approach is that it enables the ASM I/M test procedure credits
to be easily assimilated into the MOBILE6 I/M approach.  It also preserves a  similar
relative effectiveness of ASM versus EVI240 as was present in the MOBILES model. This
is reasonable since no new ASM data are available in conjunction with FTP data to update
the ASM credits.  One drawback of this  approach is that it does not update the effect of
different after repair levels, and assumes  that the ASM after repair levels are the same as
those for the EVI240. This means that the  after repair levels for the 0.8/15/2.0 HC, CO and
NOX EVI240 cutpoints will be used for the final  ASM cutpoint after  repair  levels.
Similarly, the 1.2/20/3.0 HC, CO and NOX EVI240 cutpoints will be used for the phase-in
ASM cutpoint after repair levels. Also,  it assumes that the ratio between the ASM and
EVI240 credits in MOBILES based on FTP emissions can be equally applied for both
running and start ASM credits in MOBILE6.
6.3    Idle and 2500RPM/Idle Tests

       The I/M credits for the Idle and 2500RPM/Idle tests were not developed like the
ASM credits by ratioing the MOBILES Idle test results with the MOBILES EVI240 results
and applying the ratio to the MOBILE6 EVI240 results to get the MOBILE6 Idle test credits.
                                      41

-------
Instead, the Idle and Idle/2500 RPM test credits were developed from a new analysis of the
available paired Idle / 2500RPM/Idle and FTP data sources collected by EPA from 1981
through 1998. The Loaded / Idle I/M test credits were developed in a completely analogous
fashion to the ASM I/M test credits by ratioing the MOBILES credits.  No new data were
available on the Loaded / Idle test.

6.3.1   Available Data

       Two primary EPA datasets were available. The first dataset is called the "4MID"
dataset.  The abbreviation "4MID" stands for "Four Mode Idle dataset".   It  contains
virtually all of EPA's paired Idle and FTP data collected at EPA's various labs from 1981
through 1998.  The four mode test is a special EPA Idle I/M test procedure developed for
research work that simulates in-use Idle tests. The first mode  is an unpreconditioned idle,
the second mode is a 2500 RPM segment used to precondition the third Idle mode, and
used to pass or fail vehicles for the 2500RPM/Idle test.  The third mode is a preconditioned
Idle, and the fourth mode is an idle in drive mode.  Only the 2500 RPM mode and the third
mode (pre-conditioned Idle) were used to develop the credits. Only the HC emissions from
the 2500 RPR mode were used in the development of the 2500RPM/Idle  credits.  The
analogous CO 2500 RPM mode readings were  not used because of their tendency to
produce false failures due to evaporative canister purge during the 2500 RPM mode.  The
preconditioned Idle test was used in both the Idle  test and the 2500RPM/Idle test credits.
The unpreconditioned Idle mode and the Idle in Drive modes were not used for the I/M
credit development.

       Test results from the Restart /Idle test used to test some early 1980's Ford vehicles
were not used in this analysis due to their inconsistent availability in the dataset. The effect
of this is thought to be very negligible. However,  since the basis of the IDR consists only
of High emitting vehicles, use of the Four mode test instead  of the Restart / Idle test for
Ford vehicles could potentially overstate the Idle test credits slightly if the higher readings
from the Four Mode test identify more high emitters that the Restart /  Idle test would
identify.

       The second primary dataset was the "IMLane" dataset. It consisted of I/M lane Idle
and 2500RPM/Idle test results from EPA's pilot I/M lane test program conducted in both
Hammond, IN and Phoenix, AR by ATL. These data  were paired with vehicle FTP data
collected at ATL's laboratory. The test procedure consisted of a 2500RPM mode, and a
subsequent preconditioned Idle mode.  The unpreconditioned Idle and the Idle in Drive
modes were not performed. The advantage of these data over the 4MID sample is that they
were collected in an actual I/M lane rather than  in the EPA laboratory like the 4MID
sample.
                                      42

-------
       For the final results, both databases were combined together to produce overall IDR
rates for the Idle test and the 2500RPM/Idle test. Despite the slight differences in the I/M
test procedures, the combination of the data makes sense for several reasons.  First, it
produces a larger sample of vehicles.  This is important because for this analysis only the
High emitters are used to compute the IDRs, and the number of High emitters can get small
in some model year groups. Also, both databases seem to complement each other in terms
of model year coverage. For example, the "4MID" sample has a large preponderance of its
data in the 1981 and 1982 model years; however, it does have some newer mid  1990's
vehicles and trucks. The ATL sample on the other hand contains only cars, and is mostly
represented by late 1980's to early 1990's cars. Tables 8a and 8b show the model year and
technology breakdown for both databases.
                                      43

-------


MY
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
Table 8a
Four Mode Idle / 2500RPM Idle and FTP Test Pairs

Cars
CARS
962
125
87
32
90
41
16
15
22






TBI
15
66
122
44
52
52
64
60
35
46
4
2
4


PFI
29
5
59
34
61
86
92
103
82
85
59
37
16
27
2

Trucks
CARS
120
45
10
48
63
17









TBI




13
23







1

PFI
4


1
6
41




2

2
1

44

-------


MY
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
Table 8b
IM Lane Idle / 2500RPM Idle and FTP Test Pairs

Idle Test
CARS
39
37
22
21
14
11
9
4
1
1



TBI
1
3
18
56
65
61
39
41
34
25
6
2

PFI
2
1
11
29
48
47
48
61
53
33
17
18
6

2500 RPM / Idle Test
CARS
39
37
22
21
14
11
9
4
1
1



TBI
1
O
18
56
63
61
39
40
34
25
5
2

PFI
2
1
10
29
47
47
48
60
53
33
17
18
6
6.3.2   Idle and 2500RPM/Idle Test IDRs

       The calculation of the IDRs for the Idle and 2500RPM/Idle tests is very similar to
the calculation done for IM240 IDRs in Section 3.9. One difference is that IDRs for a range
of cutpoints was not performed.  Instead only one set of Idle and 2500RPM/Idle cutpoints
were developed. These were at the CO/HC cutpoints of 1.2%CO and 220ppm HC.  Also,
IDRs for only HC and CO emissions for running  and  start were developed. Idle and
2500RPM/Idle IDRs for NOX emissions were not developed. Neither the Idle Test or the
2500RPM/Idle test will produce NOX benefits or NOX "Dis-benefits" for MOBILE6.  In
comparison, MOBILES contained NOX "Dis-benefits"  if an Idle or 2500RPM Idle test
were performed.
                                      45

-------



Test
Idle
2500/Idle

Test
Idle
2500/Idle
Table 9a
Idle and 2500RPM / Idle Test IDRs for Each Sample


IDRs Based on I/M Lane Sample
Hot Running LA4 HC
Carb
63.3
76.5
PFI
58.7
59.3
TBI
53.2
53.9
Cold Start HC
Carb
41.9
48.6
PFI
39.1
40.2
TBI
33.9
34.8
Hot Running LA4 CO
Carb
54.9
68.8
PFI
57.5
57.5
TBI
60.6
60.6
Cold Start CO
Carb
29.1
29.1
PFI
23.6
23.6
TBI
20.9
20.9



Test
Idle
2500/Idle

Test
Idle
2500/Idle
IDRs Based on Four Mode Sample
Hot Running LA4 HC
Carb
48.8
66.1
PFI
74.3
74.3
TBI
52.2
61.6
Cold Start HC
Carb
20.2
24.4
PFI
42.6
42.6
TBI
17.7
25.4
Hot Running LA4 CO
Carb
53.4
63.8
PFI
81.1
81.1
TBI
40.7
55.7
Cold Start CO
Carb
21.4
27.1
PFI
57.8
57.8
TBI
30.1
33.9
       Table 9a shows the Hot Running LA4 and Cold Start IDR rates for the Idle and
2500RPM/Idle tests for each of the two datasets. It is further broken down into three
technology groups.  These are Carbureted, Throttle Body Injection (TBI), and Ported Fuel
Injection (PFI).  The IDRs were not made a function of model year because of the small
sample sizes in many individual model years.  Table 9b shows the IDR results for the
combined dataset. The two datasets were combined together based on total emissions from
                                      46

-------
the high emitters rather than on the number of vehicles in the sample. The IDRs are shown
as a percentage in both tables, but will be programmed into MOBILE6 as fractions. They
represent the fraction of emissions from high emitters  which are identified by the
prospective I/M test. Separate IDRs for each pollutant and technology were developed for
Hot Running LA4 emissions and Start emissions based on Bagged FTP data. The PFI and
TBI Identification rates were subsequently combined together for analysis to create a larger
and more statistically significant sample size. Table 9b shows the results separately for PFI
and TBI, and Table 9c shows the average value used in MOBILE6. The values in Table
9b were weighted together by the overall sample size to produce the values shown in Table
9c.
Table 9b
Idle and 2500RPM / Idle Test IDRs Based on the COMBINED Sample



Test
Idle
2500/Idle

Test
Idle
2500/Idle
IDRs Based on I/M Lane Sample
Hot Running LA4 HC
Carb
54.6
70.2
PFI
63.5
63.9
TBI
52.8
56.8
Cold Start HC
Carb
25.5
30.3
PFI
40.8
41.3
TBI
29.5
32.3
Hot Running LA4 CO
Carb
54.0
65.9
PFI
63.0
62.9
TBI
53.5
58.8
Cold Start CO
Carb
23.3
27.6
PFI
37.8
37.8
TBI
25.1
26.8
Table 9c
Idle and 2500RPM / Idle Test IDRs Based on the COMBINED Sample



Test
Idle
2500/Idle
IDRs Based on I/M Lane Sample
Hot Running LA4 HC
Carb
54.6
70.2
PFI
58.3
60.5
TBI
58.3
60.5
Hot Running LA4 CO
Carb
54.0
65.9
PFI
58.4
60.9
TBI
58.4
60.9
                                      47

-------

Test
Idle
2500/Idle
Cold Start HC
Carb
25.5
30.3
PFI
35.3
36.9
TBI
35.3
36.9
Cold Start CO
Carb
23.3
27.6
PFI
31.7
32.5
TBI
31.7
32.5
6.3.3   After Repair Emission Level for Idle and Idle/2500 Tests

       The Idle Test after repair emission levels for MOBILE6 were calculated from a
dataset which was used for MOBILES development. It consisted of 36, 1981 and later
vehicles which initially failed the idle test, were repaired, and passed the final idle test at
standard cutpoints. These data were collected as part of an EPA test program conducted
to evaluate the effect of repair on idle test failures. The repairs were conducted by qualified
technicians. The vehicle sample mean FTP emission values after Idle test I/M repair were
found to be 1.89 g/mi HC and 20.0 g/mi CO. These compare with means of 1.26 g/mi HC
and 13.46 g/mi CO for the EVI240 at the 1.2/20 HC and CO cutpoint. Idle test repair effects
for NOX emissions are not computed because MOBILE6 will not give NOX benefits or
disbenefits to an idle test program.

       The ratio of the idle test after repair FTP emission level to the EVI240 after repair
FTP emission level at 1.2/20/3.0 cutpoints is computed from the data and used to generate
the after repair idle test emission level for running LA4 emissions.  A consistent ratio based
on the FTP will be used for all mileages, vehicle types, and model years.  The ratios which
are used for HC and CO  are:
       HC Ratio:    1.89 g/mi / 1.26 g/mi
       CO Ratio:    20.0 g/mi / 13.46 g/mi
1.5
1.5
       They are used in MOBILE6 to generate the idle test after repair running LA4
emission level by multiplying the ratio by the EVI240 after repair  emission  level at
1.2/20/3.0 cutpoints. The same after repair emission levels will be used for the Idle test and
the Idle/2500 RPM test.
6.4    OBD I/M Tests

       This document does not explicitly cover vehicles which are equipped with an OBD
system. However, most OBD equipped vehicles will continue to receive exhaust based I/M
                                       48

-------
tests such as the IM240 or the Idle test for much of their early lives.  Thus, the topic is
mentioned briefly in this document as an introduction. For more complete details on EPA's
modeling of OBD equipped vehicles (1996+ model years) please read EPA document
M6.EXH.007 "Determination of Emissions, OBD, and I/M Effects for Tierl, TLEV, LEV,
and ULEV Vehicles" and EPA document M6.EXH.009.

       The OBD system is an electronic diagnostic system built into most 1996 and later
and some 1994 and 1995 model year vehicles.  It is designed to (1) continuously monitor
the performance of the car's emission control system, and detect serious problem(s) which
cause the vehicle's FTP emissions to exceed 1.5 times its applicable certification standards,
(2) register a code in the vehicle's computer and turn on a dashboard warning light to notify
the owner. The system will also have the capability to be electronically accessed in an I/M
lane. The vehicle will be required to pass the OBD test (no trouble codes are present) in-
order to pass the state I/M program requirements.

       In MOBILE6 an I/M program conducting an OBD check on properly equipped OBD
vehicles will be assigned an DDR of 85 percent (fraction 0.85). This value will be given
regardless of whether an exhaust I/M test such as the EVI240 or the ASM test is performed
or not performed.  Also, the with and without technician training levels in an OBD I/M
program will be equivalent.  It is assumed that the technicians specializing in OBD
diagnosis and repair will either be fully qualified, or not involved in the industry.
6.5    Tampering Rates and Anti-Tampering Program Credits in MOBILE6.
       Vehicle Tampering and Anti-Tampering Programs (ATP) have long been associated
with Inspection / Maintenance programs (I/M). This is because for many years tampering
was often the cause of excess emissions from vehicles.  To help understand the nature and
extent of the tampering problem numerous field studies were done by EPA during the
1970s and 1980s to quantify the problem.  The results from these studies were incorporated
into the MOBILE series models.

       Unfortunately, for MOBILE6, no new studies were available that quantify the extent
of vehicle tampering in the fleet.  This is largely the result of the belief that deliberate
vehicle emission control  system tampering is no longer much of an issue. Also, it is now
felt that much of the effects of tampering are properly captured in the High Emitter rates,
High Emitter emission levels, and the High Emitter Correction Factor that are discussed
earlier in this document and in other MOBILE6 documents. (M6.EXH.002 - M6.EXH.005).
                                      49

-------
       As a background, the High Emitter Correction Factor was a multiplicative factor
that was added to the Base Emission rates originally developed for the MOBILE6 model.
It was developed because it was thought that the underlying vehicle data used to develop
the base emission factors contained a disproportionate percentage of low emitting vehicles
and consequently did not contain a high enough percentage of High emitting vehicles. It
was developed by comparing the Base Emission factor data collected in the EPA and
AAMA labs with a large sample of in-use EVI240 data collected in Dayton, Ohio.

       The tampering algorithm used in MOBILE6 is as follows:

1.      For the Pre-1981 model year vehicles there is no change from the MOBILES model
       in terms of the tampering rates or ATP effectiveness assumptions.

2.      For the 1981  through 1995 model year vehicles, there is a tampering offset that is
       built into the emission factors (i.e., high emitter correction factor).  Thus,  the
       tampering subroutines do not add any additional tampering correction factors like
       in MOBILES.  However, the same  subroutines are still used in the MOBILE6
       model to calculate the  ATP and I/M benefits in reducing the occurrence of
       tampering. These subroutines subtract a portion of the high emitter correction
       factor.
3.      For the 1996 and later model years there is assumed to be no tampering in the fleet
       This  assumption was made because  strong engineering reasons and anecdotal
       evidence  suggests that deliberate tampering of emission control devices is not
       common on today's late model vehicles. This is because the reasons for tampering
       such as the ability to misfuel, perceived improved performance and perceived cost
       savings on vehicle operation  do not  exist anymore. Also, the advent of OBD
       systems should also discourage tampering, because the immediate result of
       tampering is an OBD warning light. The effect of this assumption is that tampering
       effects will be completely removed from the MOBILE6 model by calendar year
       2021.
                                      50

-------
7.0    Response to Peer Review and Stakeholder Comments

       Section 7.0 discusses issues and comments submitted by interested parties during
the formal stakeholder review period, and by paid reviewers of this document.
1.      A key element missing from the overall methodology is the inability of the model
       to account for any possible actions that vehicle owners may take to  adjust their
       vehicles to just 'look clean' for the test.

       It is true that the so called phenomenon of 'clean for a day' is not accounted for in
       the MOBILE6 model.  Part of the difficulty with modeling this phenomenon is
       obtaining definitive data on it. The problem stems from the fact that the vehicles
       that only 'look  clean'  (adjusted only to pass the test) are identical  in terms of
       numerical test score to vehicles that pass either on their  initial test or upon a retest.
       Therefore, it is difficult to identify these vehicles from standard test programs and
       even from large scale  I/M test samples.  One possible way of determining the
       impact of these vehicles  (if they even  exist) is through some  type of very
       sophisticated remote sensing program, and subsequent and immediate confirmation
       test follow up. Multiple tests might be necessary in-order to eliminate natural test-
       to-test variability. Unfortunately, such data are not available.

       Another factor influencing the clean for a day phenomenon is advancing technology
       in vehicles and  in I/M programs.  On-board diagnostic (OBD) tests make such
       ' clean for a day'  strategies by non-complying motorists more difficult to achieve or
       less cost effective. For example, it will be much more difficult and expensive for
       a motorist to alter a vehicle's electronic OBD system to obtain a false pass reading,
       than it would be to adjust a vehicle's carburetor to obtain a temporary low emission
       reading. Also, with the advent of advanced technology and emission control system
       designs that are fully integrated into the operation of the vehicle and significantly
       affect the performance of the vehicle, one must ask the question "what are the real
       and perceived benefits of fixing a vehicle to pass a test only for a short time versus
       fixing it permanently?"

2.      One  peer  reviewer  suggested that the results from the  statistical  analysis be
       presented in tabular form in the document. This will allow for easier  review.

       EPA agrees with this suggestion.  The revised version of this document will contain
       important statistic results in tabular form.

3.      One peer reviewer asked why, given the huge size of the Arizona EVI240 database,


                                       51

-------
       the after repair analysis did not consider technology as an independent variable.

       EPA agrees that technology might be an important variable in determining after
       repair emission levels. Unfortunately, the database could not be resolved down into
       technology categories that were fine enough for use. The necessary resolution was
       by PFI, TBI and Carbureted technology. Some auto industry experts consider an
       even finer breakdown of the PFI category to be useful. The VEST decoder used to
       process the 17 digit individual vehicle VINs could not produce fuel delivery system
       resolution.  As a result, the after repair emission levels were made a function of
       model year.  Model year implicitly contains technology information since the
       progress of automotive technology has been steady for years. For example, most
       1990's cars are PFI technology and many 1980's cars are TBI.

4.      One peer reviewer expressed concern over the fact that only laboratory data that
       may not be replicated in actual repair effectiveness was used to determine the repair
       effects  at lower cutpoints.   They  suggested that such data could lead to  an
       overestimation of the benefits of lower cutpoints.  For the next version of MOBILE,
       EPA should obtain data from vehicles that undergo actual field repairs to various
       cutpoints for use in the model.

       EPA recognizes that the use of  lab repair data  may be problematic.  However,
       obtaining actual field data from programs that use substantially different cutpoints
       may be a problem, since virtually  all states use the same or similar sets of cutpoints.

5.      One peer reviewer points out that the calculation methodology (for the high emitter,
       normal emitter, average fleet emission level and the Ohio data high emitter
       correction factor) implicitly assumes that the effect from the Ohio data high emitter
       correction factor  is  only an increase in  the number  of high  emitters.   This
       assumption in turn leads to the next assumption that the average emission level of
       the high emitters are the same in  Ohio and in the EPA/AAMA samples.

       Mathematically, this observation is certainly  true.  The Ohio data high emitter
       correction factor was developed  based on the assumption that the EPA / AAMA
       samples contained an under-representative fraction of high emitters. This under-
       representation is thought to occur because motorists who tamper (commit an illegal
       act) and otherwise severely mal-maintain their vehicle are probably less likely to
       lend it to the government or the auto industry for research purposes. However, the
       EPA analysis did assume that the EPA / AAMA data base contained enough high
       emitters so as to characterize the  emission level of a high emitter,  but could not be
       reliably  used to determine the  frequency of such high emitters in the  fleet.
       Therefore, the effect of holding the average High emitter and average Normal
                                       52

-------
       emitter emission level constant while boosting the overall average emission higher
       (high emitter correction factor) leads to the mathematical result of a greater number
       of high emitters in the fleet.

       The assumption that the average high emitter emission level is constant before and
       after adding the High Emitter correction factor to the algorithm is the same as
       assuming that a high emitter in the EPA/AAMA database is equivalent to a high
       emitter in the Dayton, Ohio database. Unfortunately, this assumption could not be
       determined directly, since the Dayton data is based on the EVI240 and EVI240 fast
       pass driving schedule, and the EPA/AAMA sample is based on the FTP, running
       LA4 and Start emission factors.  However, the assumption that on average high
       emitters have generally consistent emission levels can be investigated by looking
       at various state EVI240  data on failing vehicles.

6.      One stakeholder reviewer questioned the assumption of a constant High emitter
       emission level with respect to mileage.

       This assumption is in general  sensitive to the  definition  of a High emitter.
       However, given EPA's definition,  statistical analysis of the High emitter data
       showed that the emission level of a high emitter was not a function of mileage.  The
       rate of high emitters in the fleet was a function of mileage and is  modeled in
       MOBILE6 as such.

7.      The peer and stakeholder reviewers state that the DRAFT I/M algorithm made the
       questionable assumption that the deterioration rate of failed vehicles is the same as
       that of a fleet average vehicle.

       EPA is sensitive to the criticism and widespread comments that have been received
       regarding this assumption.   Clearly, sound logical arguments can be made for
       revising it so that failed vehicles are given a higher or possibly lower probability of
       failing a subsequent test than a vehicle selected randomly from the overall fleet.
       Thus,  in the final version of MOBILE6, this assumption was rejected.

       However, in the next generation of models, EPA will likely take a rigorous look at
       the overall question of deterioration rate between I/M failures and the general fleet.
       The overall model would probably benefit from a more sophisticated approach
       regarding the role of repeat failure, non complying vehicles and waivered vehicles.
       Data sources that would prove useful for this analysis are (1) long term I/M test
       results over three complete I/M cycles on a sizeable sample that show the progress
       of both failing and passing vehicles over time, (2) good test data on the benefits of
       partial repair of waivered vehicles and the frequency of such vehicles. (3) solid data
                                       53

-------
       on the frequency, whereabouts and emission levels of non complying vehicles and
       non participating vehicles.

8.      One peer reviewer mentioned the lack of information regarding start emissions and
       the lack of statistic results presented in a tabular form.

       EPA report M6.STE.003 has been updated to include regression statistics on start
       emissions in tabular form.

9.      The peer and stakeholder reviewers expressed concerns regarding the use of least-
       squared regressions to simulate ASM / EVI240 test credits. He felt that the actual
       ratios  should  be built into  the MOBILE6 program  rather than the regression
       coefficients from the ratios.

       EPA agrees on this point.  The actual ASM  / EVI240 ratios will be used in
       MOBILE6.  They will be read into the program from an external data  file.  The
       original approach of using regression coefficients to model the ASM credits
       introduced unnecessary errors into the algorithms, and offered very little  reduction
       in the  code size or flexibility in the programming.

10.    One peer reviewer  suggested  that the ASM / EVI240 ratio should be applied in
       MOBILE6  as the  overall  I/M credit  rather  than as the relative ASM  I/M
       identification  rate (DDR).

       This is a reasonable suggestion. However, due to the design of the MOBILE6 code,
       it is impractical to implement and would require a substantial rewrite of the code.
       In addition, the  results  from  both methods (ASM/IM240 ratio as an overall
       correction factor and the ASM/EVI240 ratio as the IDR) should yield essentially the
       same results, since both the ASM tests and the IM240 test use similar after repair
       rates.

11.    The peer reviewer  commented that  Idle and Idle/2500 RPM test  credits were
       available only at one set of I/M standards (i.e., 1.2% CO and 220 ppm HC).

       This was done because these are the lowest Idle test I/M  standards which are
       covered by the 207(b) warranty provisions.  Thus, it is believed that very few states
       will want to use alternative Idle test standards.

12.    A stakeholder reviewer wondered about the impact of the new emission factors and
       I/M credit methodology on the size of the I/M performance standards and rate of
       progress issues.
                                       54

-------
      This is not a direct MOBILE6 issue, but instead falls in the area of I/M and state
      program guidance.  Subsequent to the release of MOBILE6, EPA will likely
      develop the necessary policy guidance to resolve these types of questions.

13.    Several stakeholders commented that they would like the MOBILE6 program to
      have the capability of modeling a more exact I/M start date that can be resolved
      down to the monthly level.

      Unfortunately, the program code cannot model  an I/M program start year to the
      monthly level resolution. The user is encouraged to pick the closest January 1 start
      calendar year date to the actual start date.

14.    Several users commented on the desire to better control I/M DDR rates and other
      parameters such as the fraction of High emitters in the fleet by using MOBILE6
      inputs.

      Unfortunately, the I/M IDR rate and the fraction of High emitter in the fleet cannot
      be directly changed in the model using standard inputs.

15.    Several reviewers mentioned the need for the model to be able to disable the impact
      of the 1990 Clean Air Act for Rate of Progress SIPs and other I/M program issues.

      Although, not mentioned in this document,  this feature will be allowed in
      MOBILE6.

16.    One reviewer  expressed  concern  about an  I/M credit  discount  applied to
      decentralized I/M program vis-a-vis centralized program.  This discount was a
      standard feature in MOBILES.

      The MANDATORY 50 percent discount for decentralized I/M programs that was
      built into MOBILES has been removed in MOBILE6.  It has been replaced by a
      new EFFECTIVENESS command that allows the user to set their own level of
      program effectiveness or discount.

17.    Several reviewers commented on credit issues if two ASM tests are  performed and
      wondered about the relative size of the ASM test credit and the EVI240 test credit.

      The MOBILE6 program will allow the modeling of the two mode  ASM test, and
      the test will receive more I/M credits than a single mode ASM test.  The ASM test
      credit is not a function of the ASM cutpoint because there were too many ASM
      cutpoints; however, the EVI240 test is a function of cutpoint. Thus, the answer to
                                      55

-------
       the IM240 versus ASM test comparison is ... it depends on the IM240 outpoint
       under evaluation. However, at comparable phase-in or final outpoints, the credit is
       almost the same for both tests with the EVI240 receiving slightly more credit than
       the ASM test.

18.    Several reviewers have asked if additional I/M credit will be given to a state that
       conducts both exhaust I/M testing and OBD I/M testing on the same vehicles.

       Theoretically, some small additional credit may be possible by conducting two or
       even more I/M tests on a given vehicle. However, because of a lack of data on this
       topic, and the general inability of the MOBILE6 program to model two different
       I/M program types on the same vehicle model year, no additional credit will be
       given to States that conduct both tests.

19.    One stakeholder reviewer asked about "Appropriate I/M", andtheMOBILES policy
       of given I/M credit to LEV vehicles by reducing their deterioration rate.

       The  concept of "Appropriate I/M" was  not explicitly included in MOBILE6.
       However, LEV, Tier2 and other advanced vehicle technology types will still be able
       to receive I/M credit.  After the release of MOBILE6, EPA will provide guidance
       and policy regarding the use of MOBILE6.

20.    One reviewer suggested that human behavior should be included in the MOBILE
       I/M modeling process. This behavior might include the motorist taking advantage
       of 'test to test variability' effects (i.e., continued retesting without repairs until the
       vehicle passes the test), the effect of motorists registering  outside of the I/M
       program area, and the effect of motorist's who never show up for the test in the first
       place.

       The effect of registering outside of the I/M area or never showing up for the test in
       the first place can be accounted for in MOBILE6 using the non compliance and
       participation rate inputs.

       Theoretically, test to test variability will always be an issue with an exhaust test
       with a defined cutpoint standard. Every vehicle exhaust measurement has a natural
       uncertainty associated with it and upon multiple  retesting this uncertainty could
       overlap both the "pass" level "fail". The larger the variability the more likely an
       untrue passing or failing reading could occur.  However, it is believed that most
       'true' High emitters will have  a high enough emission level  and small enough
       variability so that repeated testing is not likely to produce a false passing reading.
       On the other hand, it is  also hoped that multiple repeat testing eliminates false
                                       56

-------
       failures and lack of preconditioning failures before the repair process begins by
       giving the vehicle another opportunity to pass.

       The advent of OBD I/M testing should also help mitigate the issue of test to test
       variability. Because it is an electronic test, it produces only an obj ective pass or fail
       result.  It is also believed that the vehicle OBD systems and the OBD I/M test
       equipment and procedures will be designed properly to minimize both false passing
       tests and false failing tests. Only time and yet to be collected data will answer these
       questions.

21.    One stakeholder reviewer wondered if the without technician training emission
       levels are below the I/M cutpoints emission levels.

       This comparison is not particularly straightforward, because MOBILE6 calculates
       and reports emission in terms of FTP cycle 'unit',  and typical I/M test reports
       emissions in terms of concentration units or in the case of EVI240 gram per mile
       numbers based on a different cycle. Nevertheless, even with the increases in the
       after repair levels due to the no technician training effects, these levels are lower
       than a failing high emitter's level.

       OBD  provided  a special case for the no technician  training effects  since the
       presence of the MIL light is triggered if the FTP emission are greater than 150% of
       the certification standard. To solve this dilemma the no technician training effects
       were eliminated for an OBD I/M program under the assumption that virtually all
       technicians that repair modern vehicles equipped with OBD will have to have some
       training on the  OBD  systems, the use of the diagnostic  tools,  and general
       investigative and repair skills.  Gone are the days when a virtually uneducated
       mechanic could simply turn a few carburetor screws and replace an air filter and
       call it an I/M repair.

22.    As a result of the need for MOBILE6 Loaded / Idle I/M test credits by a couple of
       State  I/M programs, the Loaded / Idle test credits  will be inserted into the
       MOBILE6 program.  However, these credits will be identical in all respects to the
       2500/Idle I/M credits. The rational for this assumption is that there are no new data
       available to develop special Loaded / Idle test I/M credits, and in practice the loaded
       portion of the test is just a preconditioning phase rather than an additional pass / fail
       requirement. The pass / fail determination for the test is based solely on the results
       of the idle mode. This is completely analogous to the 2500 / Idle test in which only
       the idle portion of the test is used to pass or fail a vehicle.
                                        57

-------
                     APPENDIX A
     Running LA4 Emissions from 1990-93 MY PFI Normal Emitters
       Figure  A-1
       HC Emissions from  Normals
           20
       MILEAGE
             40
60
                    80
100
120
140
160
180
200
      Figure A-2
      CO Emissions from Normals
o
o
20.
18.
16.
14.
12.
10.
 8.
 6.
 4.
 2.
0    20    40    60   80   100   120   140
      MILEAGE
                       160
                                               180
                                                      200
                          58

-------
Figure A-3
NOx Emissions from Normals
   20
MILEAGE
40
60
80
100
120
140
160
180  200
                59

-------
                               APPENDIX B
                  Sample Calculations for Fleet High Emitter Rate

       This sample calculation shows the steps for determining the percentage of High
Emitters in the fleet for HC running emissions at an age of 5 for 1988-1993 PFI technology
passenger cars.

Calculating the average emission rate (A) for HC:

mileage <21,270            A = 0.05158 + 0.0013 * M

21,270 100,000           A= 0.05158 + 0.0013 * 21.27 + (100.0 - 21.27) * 0.0036 +
(M- 100.0) * 0.0036

       Where 'M' is the mileage divided by 1000.  See the document "Determination of
Running  Emissions as a Function of Mileage for 1981-1993  Model Year Light-Duty
Vehicles" for the derivation of this equation.

From Table 3, the average mileage of passenger cars 5 years old is 67,547 miles.

             A = 0.05158 + 0.0013 * 21.27 + (67.547 - 21.27) * 0.0036

             A = 0.249 g/mi HC

Calculating the average normal emitter rate (B) for HC using the coefficients from Table
la and using the mileage from Table 3:

             B = 0.0214 + 0.001385 * 67.547

             B = 0.115g/miHC

Choosing the high emitter rate (C) using the values from Table 2a:

             C= 1.740 g/mi HC

Calculating percentage of Highs using Equation 7 in Section 3.8.

             High Fraction = (A - B) /  - B)
             High Fraction = (0.249 - 0.115) / (1.740 - 0.115) = 0.0823 or 8.23

                                      60

-------
                       Appendix C
Statistical Detail: Standard Errors, P values and Standard Deviations

Table C-l
Standard Deviations of Means

RUNNING Emissions of High Emitter CARS
MY Group
1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group
PFI*
TBI
FI
Carb
Carb
FI
Carb
HC
Sample
Size
58
38
118
212
212
118
118
CO
Sample
Size
44
43
97
233
233
97
97
HC
Standard
Deviation
4.049
6.487
4.832
4.530
4.530
4.832
4.832
CO
Standard
Deviation
47.350
53.284
51.883
41.593
41.593
51.883
51.883
NOX
Standard
Deviation
1.069/11
1.012/15
0.895/44*
0.768/60*
0.768/60*
0.895/44*
0.895/44*
* Second number is the NOX high emitter sample size

Table C-2
Standard Deviations of Means

RUNNING Emissions of High Emitter TRUCKS
MY Group
1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group
PFI
TBI
FI
Carb
Carb
HC
Sampl
e Size
2
2
17
18
10
CO
Sampl
e Size
3
3
3
11
4
HC
Standard
Deviation
0.966
2.259
1.1776
1.244
1.482
CO
Standard
Deviation
18.498
18.498
3.502
25.000
21.314
NOX
Standard
Deviation
NA/1*
NA/1
NA/1
NA/1
NA/1
* Truck sample had only one high emitter in each group.
                             61

-------
Table C-3
Regression Statistics from Normal Emitting Cars - CO EMISSIONS
MY
Group

1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group

PFI
TBI
FI
Carb
Carb
FI
Carb
Sample
Size
CO
1590
431
640
93
233
107
815
S.E
Slope
CO
0.00119
0.00233
0.00212
0.00340
0.00446
0.00612
0.00231
SE
ZML
CO
0.05662
0.13169
0.13010
0.21690
0.17510
0.30090
0.09617
SigT
Slope
CO
0.0000
0.0000
0.0000
0.0001
0.0002
0.0007
0.0000
SigT
ZML
CO
0.0000
0.8301
0.0000
0.0105
0.0000
0.0000
0.0000
Table C-4
Regression Statistics from Normal Emitting Cars - HC EMISSIONS
MY
Group

1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group

PFI
TBI
FI
Carb
Carb
FI
Carb
Sample
Size
HC
1582
435
622
91
233
104
838
S.E
Slope
HC
7.066e-5
1.254e-4
1.564e-4
3.306e-4
3.825e-4
3.786e-4
1.492e-4
SE
ZML
HC
0.00335
0.00708
0.00919
0.02050
0.01490
0.01887
0.00628
SigT
Slope
HC
0.0000
0.0000
0.0000
0.0159
0.0017
0.0000
0.0000
SigT
ZML
HC
0.0000
0.5540
0.0000
0.0000
0.0000
0.0000
0.0000
62

-------
Table C-5
Regression Statistics from Normal Emitting Cars - NOX EMISSIONS
MY
Group

1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group

PFI
TBI
FI
Carb
Carb
FI
Carb
Sample
Size
NOX
1610
440
693
94
247
107
973
S.E
Slope
NOX
2.210e-4
3.608e-4
3.777e-4
0.00106
0.00107
0.00136
4.281e-4
SE
ZML
NOX
0.0106
0.0203
0.0235
0.0680
0.0442
0.0669
0.0193
SigT
Slope
NOX
0.0000
0.0000
0.0000
0.1120
0.0119
0.0000
0.0000
SigT
ZML
NOX
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
63

-------
Table C-6
Regression Statistics from Normal Emitting TRUCKS - CO
EMISSIONS
MY
Group

1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group

PFI
TBI
FI
Carb
Carb
Sample
Size
CO
329
465
90
122
163
S.E
Slope
CO
0.00421
0.00298
0.01184
0.01802
0.0274
SE
ZML
CO
0.2045
0.1274
0.8112
0.9354
1.2391
SigT
Slope
CO
0.0000
0.0000
0.0005
0.0003
0.2319
SigT
ZML
CO
0.0166
0.0000
0.0486
0.1500
0.0000
Table C-7
Regression Statistics from Normal Emitting TRUCKS - HC
EMISSIONS
MY
Group

1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group

PFI
TBI
FI
Carb
Carb
Sample
Size
HC
330
464
76
115
157
S.E
Slope
HC
3.479e-4
2.486e-4
8.651e-4
8.258e-4
0.00150
SE
ZML
HC
0.01689
0.01061
0.05490
0.0407
0.06656
SigT
Slope
HC
0.0000
0.0000
0.0003
0.0014
0.0000
SigT
ZML
HC
0.0778
0.0000
0.0172
0.0000
0.0000
64

-------
Table C-8
Regression Statistics from Normal Emitting TRUCKS - NOX
EMISSIONS
MY
Group

1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group

PFI
TBI
Fl
Garb
Garb
Sample
Size
NOX
331
466
93
132
166
S.E
Slope
NOX
9.0910-4
6.508e-4
0.00263
0.00205
0.00210
SE
ZML
NOX
0.0441
0.0279
0.1825
0.1134
0.0960
SigT
Slope
NOX
0.0000
0.0000
0.0032
0.2511
0.9910
SigT
ZML
NOX
0.0000
0.0000
0.0478
0.0000
0.0000
65

-------
                                       Appendix D
  Statistical Diagnostics for Running and  Start Emission I/M Identification
                      Rate Effectiveness (1DR) Determination
-> REGRESSION
->   /DESCRIPTIVES MEAN STDDEV CORR SIG N
->   /MISSING LISTWISE
->   /STATISTICS COEFF OUTS CI R ANOVA
->   /CRITERIA=PIN(.05) POUT(.IO)
->   /NOORIGIN
->   /DEPENDENT hcrun_id
->   /METHOD=ENTER In hccut In cocut
          ****    MULTIPLE   REGRESSION   **

Equation Number 1    Dependent Variable..   HCRUN_ID   HCRun ID

  Descriptive Statistics are printed on Page    2

Block Number  1. Method:  Enter     LN HCCUT LN COCUT
Variable(s) Entered on Step Number
   1..    LN_COCUT
   2..    LN HCCUT
Multiple R           .90947
R Square             .82713
Adjusted R Square    .82246
Standard Error       .06411
Analysis of Variance
Regression
Residual
        177.03226
      DF      Sum of Squares
       2            1.45516
      74             .30413

          Signif F =  .0000
                  Mean Square
                       .72758
                       .00411
Variable
                     Variables in the Equation
                    B
                             SE  B
                                     95% Confdnce Intrvl B
LN_HCCUT
LN_COCUT
(Constant)
  -.136503
  -.106888
  1.145095
.010483
.007869
.026063
-.157390
-.122568
1.093164
-.115615
-.091209
1.197027
Variable

LN_HCCUT
LN_COCUT
(Constant)
     T  Sig T

-13.021  .0000
-13.583  .0000
 43.936  .0000
   REGRESSION
     /DESCRIPTIVES MEAN STDDEV CORR SIG N
     /MISSING LISTWISE
     /STATISTICS COEFF OUTS CI R ANOVA
     /CRITERIA=PIN(.05) POUT(.IO)
     /NOORIGIN
     /DEPENDENT corun_id
     /METHOD=ENTER In hccut In cocut
          ****    MULTIPLE   REGRESSION   **

Equation Number 1    Dependent Variable..   CORUN_ID   CORun ID
                                              66

-------
Block Number  1.  Method:  Enter
                                      LN HCCUT LN  COCUT
Variable(s) Entered on Step Number
   1..     LN_COCUT
   2..     LN HCCUT
Multiple R            .90658
R Square              .82188
Adjusted R Square     .81707
Standard Error        .06736
Analysis of Variance
Regression
Residual
        170.72789
       DF      Sum of Squares
        2             1.54920
       74              .33574

           Signif F =  .0000
         Mean Square
              .77460
              .00454
                     - Variables in the Equation  	

                      B        SE B     95% Confdnce  Intrvl  B
LN HCCUT
LN COCUT
(Constant )
- .107306
- .129819
1.188020
.011014
.008268
.027384
- .129253
- .146293
1.133456
- .085360
- .113344
1.242584
- .477976
- .770339

Variable

LN_HCCUT
LN_COCUT
(Constant)
      T  Sig T

 -9.742  .0000
-15.702  .0000
 43.384  .0000
-> * Curve Estimation.
-> TSET NEWVAR=NONE  .
-> CURVEFIT /VARIABLES=noid  WITH nocut
->   /CONSTANT
->   /MODEL=CUBIC
->   /PRINT ANOVA
->   /PLOT FIT.

Dependent variable.. NOID

Listwise Deletion of Missing Data

Multiple R            .99902
R Square              .99805
Adjusted R Square     .99658
Standard Error        .01860
            Analysis of Variance:

              DF   Sum of Squares
Regression
Residuals
           .70707598
           .00138343
                          Method..  CUBIC
Mean Square

  .23569199
  .00034586
F =     681.46957       Signif F =   .0000

	 Variables in the Equation  	

Variable                  B        SE B       Beta          T   Sig  T
NOCUT
NOCUT**2
NOCUT**3
(Constant)
       .756842     .102036   3.175112
      -.368671     .037175  -9.352562
       .040631     .004083   5.358327
       .545291     .082060
                7.417  .0018
               -9.917  .0006
                9.951  .0006
                6.645  .0027
                                                  67

-------
-> REGRESSION
->   /MISSING LISTWISE
->   /STATISTICS COEFF OUTS CI R ANOVA
->   /CRITERIA=PIN(.05) POUT(.IO)
->   /NOORIGIN
->   /DEPENDENT hc_strt_
->   /METHOD=ENTER  In hccut In cocut
           ****   MULTIPLE   REGRESSION    ***


Listwise Deletion of Missing Data

Equation Number 1    Dependent Variable..   HC_STRT_   HC Strt  ID

Block Number  1.  Method:  Enter      LN HCCUT LN COCUT
Variable(s) Entered on Step Number
   1..     LN_COCUT
   2..     LN HCCUT
Multiple R            .85506
R Square              .73113
Adjusted R Square     .70669
Standard Error        .11633

Analysis of Variance
Regression
Residual
      DF      Sum of Squares
       2              .80951
      22              .29769

          Signif F =  .0000
Mean Square
     .40476
     .01353
Variable
                       Variables in the Equation  	

                      B        SE B     95% Confdnce  Intrvl B
LN HCCUT
LN COCUT
(Constant)
- .158962
- .140941
.981406
.028853
.024734
.084067
- .218799
- .192237
.807061
- .099126
- .089645
1.155752
- .609838
- .630732

Variable

LN_HCCUT
LN_COCUT
(Constant)
     T  Sig T

-5.509  .0000
-5.698  .0000
11.674  .0000
-> REGRESSION
->   /MISSING LISTWISE
->   /STATISTICS COEFF OUTS CI R ANOVA
->   /CRITERIA=PIN(.05) POUT(.IO)
->   /NOORIGIN
->   /DEPENDENT co_strt_
->   /METHOD=ENTER  In hccut In cocut
           ****   MULTIPLE   REGRESSION    ***


Listwise Deletion of Missing Data

Equation Number 1    Dependent Variable..   CO_STRT_   CO Strt  ID

Block Number  1.  Method:  Enter      LN HCCUT LN COCUT
Variable(s) Entered on Step Number
   1..     LN_COCUT
   2..     LN HCCUT
                                                  68

-------
Multiple R            .84999
R Square              .72249
Adjusted R Square     .69726
Standard Error        .13266

Analysis of Variance
                    DF      Sum of Squares      Mean  Square
Regression            2             1.00799            .50399
Residual            22               .38718            .01760

F =      28.63762       Signif F =   .0000


	 Variables in  the Equation  	

Variable              B        SE B     95% Confdnce  Intrvl  B        Beta

LN_HCCUT       -.159301     .032905     -.227541      -.091061    -.544428
LN_COCUT       -.170728     .028208     -.229228      -.112229    -.680635
(Constant)     1.145947     .095873       .947118      1.344777
	 ln	

Variable           T  Sig T

LN_HCCUT      -4.841  .0001
LN_COCUT      -6.053  .0000
(Constant)    11.953  .0000
                                                  69

-------