United States Air and Radiation EPA420-R-01-053
Environmental Protection October 2001
Agency M6.IM.003
vyEPA Estimating Benefits of
Inspection/Maintenance
Programs for Evaporative
Control Systems
y£u Printed on Recycled
Paper
-------
EPA420-R-01-053
October 2001
Estimating Benefits of Inspection/Maintenance
Programs for Evaporative Control Systems
M6.IM.003
Megan Beardsley
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.
-------
Table of Contents
Section 1 Introduction 3
1.1 Overview 4
1.2 Data 5
1.2.1 State Data 5
1.2.2 Laboratory Data 6
1.3 Emissions in Areas without I/M 6
1.4 Tampering and Anti-Tampering Programs 9
Section 2 Basic Revised Weighting Factors 10
2.1 Fraction Tested 11
2.1.1 Applicability 11
2.1.2 Compliance Rate 11
2.1.3 Testability 12
2.2 Failure Rate 14
2.2.1 Traditional I/M tests 14
2.2.2 OBD Checks 19
2.3 Repair Benefits-Adjusting Weighting Factors 20
2.4 Waivers and Fraction Repaired 23
2.5 Technician Training 24
2.6 Sample Computations 24
Section 3 Effects of Inspect!on Frequency (Sawtooth Methodology) 25
3.1 General Sawtooth Method 25
3.2 Algebraic and Numeric details 28
Section 4 Comparisons to MOBILES 29
References 33
Appendix A Observed Fail Rates 35
Gas Cap Tests ("IL" gas cap fail rate) 35
Fill-Pipe Pressure Test Fail Rates ("AZFP" pressure rate) 41
"Gas Cap Only" Rates ("AZOnly") 47
Appendix B Mathematical Treatment of Inspection Frequency for Evaporative I/M . 50
The "No-I/M" case 50
Annual I/M program starting with Age=l 52
I/M with age exemptions 55
N-enniel Inspections 55
Program Start and End 57
Appendix C Laboratory Data on Evaporative I/M Repairs 58
Appendix D EPA Response to Comments on Draft Report 62
October 2001
M6.IM.003
-------
Estimating Benefits of Inspection/Maintenance Programs
for Evaporative Control Systems
Section 1 Introduction
This report outlines how MOBILE6 will adjust evaporative (non-exhaust) emission
estimates to account for Inspection/Maintenance (I/M) programs that include tests of automotive
evaporative control systems. Evaporative emissions include the evaporation of hydrocarbons
(HC) as diurnals, resting loss, hot soak and running loss emissions1.
A draft version of this report was released in November 1999. This final report has been
updated to include EPA's response to comments on the draft report, and to incorporate corrections
and additional detail developed as the methodology was coded into the model. Comments on the
draft and EPA's responses are listed in Appendix D.
This report focuses on the algorithms and data used to compute evaporative I/M benefits.
For details on using the model, see the MOBILE6 user guide. The MOBILE6 algorithms for
handling I/M programs for exhaust emissions are not described here, but can be found in reports
M6.IM.001 and M6.EXH.007.
MOBILE6 will calculate adjustments to the evaporative emission estimates for I/M tests
applied to Light-Duty Gasoline Vehicles (LDGV), Light-Duty Gasoline Trucks (LDGT), Heavy-
Duty Gasoline Vehicles (HDGV), and Heavy-Duty Gasoline Buses (HDGB). Estimates will be
calculated for the following tests:
• Check of On-board Diagnostics (OBD) n indicator light.
• Check of gas cap for presence, damage or leaks2
Pressure test from fuel-pipe inlet ("fill-pipe" tests)
The draft report also discussed modeling effects of a purge test. However, because an
Evaporative crankcase emissions exist for some of the oldest vehicles (non-tampered
vehicles of 1967 and earlier, tampered vehicles of later years). Evaporative I/M programs do not
effect these emissions. Anti-tampering programs (see section 1.3) do reduce these emissions.
This effect will be unchanged in MOBILE6.
Preliminary evidence suggests that vehicles with missing gas caps have higher running
loss emissions than vehicles with leaking gas caps. However, the result is based on tests of only
three vehicles (these vehicles are among those listed in Appendix C), and only a small fraction of
the fleet has missing caps, Therefore, we have decided not to treat missing cap emissions
separately in MOBILE6. We hope to have enough data to address this issue in future models.
October 2001 3 M6.EVI.003
-------
effective purge test is not available for general I/M use, benefits of a traditional I/M purge test are
not calculated in the final version of MOBILE6.
The gas cap and fill-pipe pressure tests are described in detail in the EPA report: "EVI240 & Evap
Technical Guidance," EPA report no. EPA420-R-98-010, dated August 1998. It is available on
the EPA web site http://www.epa.gov/oms/im.htm.
MOBILE 6 will calculate emissions separately for each of the types of evaporative
emissions (diurnals, resting loss, hot soak, and running loss). In this document, when "emissions"
are mentioned without further description, the need to distinguish by emissions type should be
assumed. Note that the equations given in this document do not include units because the units
for emissions (g/soak, g/hour-veh, g/mile-veh) vary with the type of emission.
Also, note that evaporative emissions vary by time of day due to variation in temperature
and activity and these parameters will affect the magnitude of emissions with and without I/M.
However, the model's calculation of I/M effects is independent of hour, temperature and activity,
so these parameters are not addressed in this report.
When the draft version of this report was released, we hoped to have the capability to
model evaporative emission credits for selective I/M programs such as remote sensing and
change-of-ownership programs. However, time constraints made it impossible to code
evaporative emission credits for these types of selective programs in the final release of
MOBILE6.
1.1 Overview
For MOBILE6, the vehicles studied to develop evaporative emission rates were recruited
based on their status on laboratory pressure and purge tests, so the emission rates used in the
model are test-status specific. Evaporative emissions are calculated as a weighted average of
emissions of the test-status sub-groups. There are four original test-status groups: pass/pass, pass
pressure/fail purge, fail pressure/pass purge and fail/fail. To calculate the benefits of evaporative
I/M, we assume that the effect of I/M is to shift vehicles from higher-emitting sub-groups to
lower-emitting groups.
Section 1.2 describes the empirical data available on evaporative I/M.
Section 1.3 summarizes the calculation of evaporative emissions in areas without I/M.
This provides a baseline for the I/M calculations.
Section 2 describes how MOBILE6 will estimate the shift of vehicles from higher-emitting
sub-groups to lower-emitting groups at the time of inspection and repair by evaluating the
effectiveness of different kinds of evaporative I/M tests and the interaction between tests.
October 2001 4 M6.EVI.003
-------
Section 3 describes how MOBILE6 will account for the fact that inspections and repairs
are distributed throughout the year and will calculate different effects for programs with different
inspection frequencies.
Section 4 provides comparisons to MOBILES in terms of the fraction of the fleet in
various test status groups.
Appendix A describes the calculation of default fail rates based on empirical data.
Appendix B provides a mathematical treatment of the calculation of inspection frequency
effects for evaporative I/M.
Appendix C compares proposed MOBILE6 repair effects to laboratory test data.
Appendix D summarizes the comments received on the draft report, and EPA's response.
1.2 Data
The data available on emission reductions from evaporative I/M programs are limited in
quantity, as well as in the type of data available.
1.2.1 State Data
As of 1998, when this analysis was begun, there were 20 states3 with active gas cap tests,
three with active fill-pipe pressure tests (Arizona, Kentucky and Delaware), and four with OBD
checks (Utah, Colorado, Wisconsin and Vermont). No state had an active I/M purge test. Only
Arizona and Delaware had both a fill-pipe pressure test and a gas cap program. In evaluating
evaporative I/M data for this report, we focused on data from Arizona, which is the state with the
longest running evaporative I/M program, and data from Illinois, which has a gas cap program
more typical of other states.
The Arizona program includes a fill-pipe pressure test where vehicles' fuel systems are
pressurized to 14 inches of water at the fillpipe and failures are defined as those losing more than
6 inches of pressure in 2 minutes or less. The Arizona also includes a gas cap test for leaks of 200
cc/min or more. Gordon Darby provided EPA with test result counts for initial tests by month and
vehicle class for 536,520 LDVs, 227,753 LDTls and 67,292 LDT2s (831,565 total) that came in
for fill-pipe pressure tests and gas cap tests between August 1997 (when the test was automated)
3AZ, CA, CO, CT, DE, DC, GA, IL, IN, ME, NM, OR, PA, RI, TX, UT, VA, VT, WA,
WI
October 2001 5 M6.EVI.003
-------
and August 1998 (when the data was requested). Model years from 1981 to 1999 were included
in the data.
The Illinois program tests gas caps for leaks of 60 cc/min or more. We were able to obtain
gas cap test result counts for 1,877,191 LDVs and LDTs of model year 1971-1997, tested from
April 1997 to April 1998.
As explained in this report and Appendix A, the Illinois and Arizona data were used to
determine several parameters for the evaporative I/M methodology. Arizona data were used to
determine the fill-pipe pressure test fail rate by age, the fill-pipe pressure test testability rate by
model year, and the percent by age of gas cap failures that also had fill-pipe pressure test failures.
The Illinois data were used to determine gas cap test fail rates by age, and gas cap testability rates
by model year.
1.2.2 Laboratory Data
In addition to data from state programs, there are data on repair effects from an EPA test
program,4 which measured pre- and post-repair diurnal, hot soak and running loss emissions for
about 20 vehicles which failed either an I/M lane fuel-inlet pressure test and/or an I/M lane gas
cap test. This study sampled a wide range of model years (1983-1995), fuel delivery systems, test
status groups and evaporative problems. MOBILE6 cannot use these data directly to generate
evaporative I/M repair effects. Rather, these data were used as a check of the MOBILE6
assumptions. The data and this comparison are presented in Appendix C.
1.3 Emissions in Areas without I/M
The details of how MOBILE6 will compute evaporative emissions are described in a series
of reports (M6.EVP.001- M6.EVP.009) which focus on the specific data and analysis used to
compute emissions for each of the evaporative emission types. This detail will not be repeated
here. However, to understand how these emissions will be adjusted to account for I/M, it is
important to understand the underlying approach used for all evaporative emissions calculations.
Evaporative emissions for a vehicle of a given age are calculated as follows:
ENoIMage = Emissions test_status x NoIMWeightingtest _status^e
4EPA contract 68-C5-006, Automotive Testing Laboratory, Work Assignment 1-8.
October 2001 6 M6.EVI.003
-------
Where Emissions vary with the test-status group,5 and NoIMWeighting is a fraction from 0
to 1 describing the fraction of a given age and model year that has a specific test status. In
MOBILE6, NoIMWeighting is calculated as a function of Age, Test Status and Model year.
Age is calculated as the calendar year of evaluation minus the model year of the
vehicle.6
Test status The vehicles studied to develop emission rates were recruited based
on their pressure and purge test status, so the emission rates are test-status specific.7 Note
that the pressure test used for this purpose is not the same as the pressure test used for I/M
tests. The emission rate stratification is based on a laboratory pressure test performed by
applying pressure from the cannister, while the I/M test applies pressure from the fuel-inlet
(or "fill pipe"). There are four test-status groups relevant for I/M calculations: pass both,
pass pressure/fail purge, fail pressure/pass purge and fail both.8
Emissions are also a function of model year, fuel delivery system, temperature, fuel
vapor pressure and other variables specific to the emission type, but independent of I/M. Resting
loss and Diurnal emissions estimates are described in M6.EVP.001, M6.EVP.002, M6.EVP.003,
M6.EVP.005 and M6.EVP.006. Hot soak emissions are described in M6.EVP.004. Running
loss emissions are described in M6.EVP.008. Crankcase emissions will be the same as in
MOBILES.
6This age value is, of course, an estimate. As described in Section 3.0, the vehicle model
year does not exactly match the calendar year of sale and sales (and I/M tests) are distributed
throughout the year. In draft M6.EVP.006, age is calculated as for an April evaluation date, as
Calendar Year- Model Year + 6 months. The effect of these differences in age calculations is
negligible.
7For some emission types, these groups may be collapsed; that is, two groups may have
the same emission rate.
8MOBILE6 also uses an additional "test status" strata that is not relevant for I/M. As
discussed in earlier papers, especially M6.EVP.009, vehicles with gross liquid leaks may
contribute a substantial portion of the evaporative emission inventory. However, at this writing
there are no I/M tests designed to detect gross liquid leakers. Thus, for I/M purposes, MOBILE6
will ignore the gross liquid leakers. While the default values in MOBILE6 include a fraction of
gross liquid leakers, the model will remove this fraction for the I/M calculations described here,
and then add it back in the same proportion once the test-status weightings for the original four
test-status groups are revised to account for I/M effects. Therefore, for the remainder of this
paper we will use "test-status" to refer to the four original pressure and purge test status groups,
and will use "Weighting" to refer to the normalized fraction of the fleet in each of the four
groups.
October 2001 7 M6.EVI.003
-------
A vehicle's Model Year determines whether a vehicle is subject to the enhanced
evaporative emission standard and OBD-II. The observed phase-in of these technologies
for light-duty vehicles is described in Table 1.1. This phase-in rate was adjusted from the
draft report to account for data from Wisconsin's I/M program that demonstrated that
vehicle manufacturers phased-in OBD equipment and compliance with the enhanced
standard more quickly than required by regulation. Although most heavy-duty vehicles are
not required to have OBD, the lightest HD vehicles (HDGV2b and HDGV3) are required
to meet the enhanced standard with the same phase-in schedule as light-duty vehicles.
Table 1.1
Observed Phase-in of Vehicles
With Enhanced Evaporative
Controls and Light Duty OBD
Model Year
1995
1996
1997
1998
1999
Percentage
0%
30%
55%
90%
100%
As explained in M6.EVP.006, vehicles subject to the enhanced evap standard will have
adjusted test-status rates. For the enhanced vehicles, the fail rates will rise at half the rate
of the previous vehicles. These vehicles also have OBD systems designed to detect
failures in evaporative emission controls. In areas without an I/M program, M6.EVP.006
states that OBD will reduce the incidence of new pressure or purge failure in vehicles
equipped with OBD by the percentages listed in Table 1.2. The resulting test-status
weighting factors as a function of age including Gross Liquid Leakers are given in
Appendix E of M6.EVP.006. They have been copied into the worksheet, "from
M6.EVP.006" of the Excel 97 Workbook, M6IM003.xls. For I/M purposes, the Gross
Liquid Leaker fraction must be temporarily removed and the weighting factors must be re-
normalized to 100 percent.
October 2001
M6.IM.003
-------
Table 1.2
Reductions in failures due to OBD in Non-I/M Areas
Age
0-3
3-6
7+
Mileage
0-36,000
36,000-80,000
80,000+
Failures of
OBD System
15%
15%
15%
Fraction of
failures detected
by OBD.
85%
85%
85%
Owner
Response Rate
90%
10%
0%
Reductions in
failures due to
OBD, No I/M
76.5%
8.5%
0%
From M6.EVP.006
1.4 Tampering and Anti-Tampering Programs
In MOBILES and previous versions of MOBILE, crankcase, hot soak and diurnal
emissions were decreased to account for anti-tampering programs.
The crankcase tampering effects in MOBILES are an additive offset that is calculated
based on the frequency of various kinds of tampering events. Based on these frequencies, the
crankcase emissions increase with mileage. For 1981 and later vehicles at 150,000 miles, average
(tampered and non-tampered) hydrocarbon emissions are estimated at 0.015 g/mi for light-duty
gas vehicles, 0.024 for light-duty gas trucks, and 0.025 for heavy-duty gas trucks.9 The frequency
of the tampering events is decreased for areas with anti-tampering programs, leading to a decline
in these values. Because no new data are available and because crankcase emissions are such a
small fraction of total emissions, in MOBILE6 the crankcase emissions and anti-tampering
program effects on these emissions will be the same as in MOBILES.
However, in MOBILE6 we have removed the tampering effects for diurnal and hot soak
emissions, as well as the hot soak and diurnal benefits for anti-tampering programs. In
MOBILES, the effects of tampering on hot soak and diurnal emissions are calculated by assuming
that vehicles with tampered evaporative control systems are a subset of the vehicles with pressure
failures. In areas with evaporative I/M pressure tests, the entire fraction of pressure failures is
decreased. In areas without evaporative I/M, but with anti-tampering programs, only the tampered
fraction is decreased. Since the tampering data are outdated and since areas currently are not
performing evaporative anti-tampering programs in the absence of fill-pipe pressure checks, there
is no need to retain this code in MOBILE6.
9See AP-42, Appendix H, Tables 1.2B1, 2.2B1, 3.2B1 and 4.2B1.
October 2001
M6.IM.003
-------
Section 2 Basic Revised Weighting Factors
This section describes the computation of basic revised weighting factors, that is,
weighting factors revised to account for the effect of I/M at the time immediately after inspection
and repair. Section 3 will take the basic weighting factors described here and describe how they
are transformed to final weighting factors that take into account the distribution of repairs
throughout the year and the effects of different inspection frequency designs.
From a computational point of view, the basic function of the I/M tests is to adjust the
fraction of vehicles that fall into the various test-status groups. Thus:
EwithIMage = ^[Emissionstest_status x IMWeighting
test _status,age
Where Emissions are the same emissions used in the No I/M case, but IMWeighting is a
revised set of weighting factors for the test status groups.
The purpose of the evaporative I/M algorithm is to determine the new weighting factors
IMWeighting for the I/M program chosen by the user. The I/M program may vary in the kinds of
tests that vehicles are subject to, as well as in the ages subject to testing and the frequency of tests.
The benefit of I/M also depends on other factors, including the fraction of vehicles that show up
for testing and are eventually passed or waived (compliance), the fraction that can be tested on a
given test (testability), and the fraction of failing vehicles that actually get repaired (waivers).
This section describes how program characteristics such as compliance, testability and
observed fail rates are used to calculate new basic weighting factors IM teststatuS: age.
Section 3.0 describes how information on inspection frequency and grace periods are used
to transform the basic revised weighting factors IM test status> age into the final revised weighting
factors IMWeighting test_statuSt age.
IMWeightingtest_status>age = f (IM test_status,age)
The basic weighting factors IM test statuS: age will be calculated based on the repairs due to
combinations of gas cap, fill pipe, and OBD tests as follows:
October 2001 10 M6.EVI.003
-------
IM = FractSubjToTest X FailRate X FractRpaired
Each of the factors that determine I/M weight is described in detail below.
2.1 Fraction Tested
The fraction tested is the fraction of vehicles of a given model year that actually are tested
in the interval specified. It is the product of applicability, compliance and testability.
2.1.1 Applicability
Applicability is the fraction (0 to 1) of the fleet required to be tested for a given test, model
year and age. MOBILE6 computes this fraction based on user input of the program description.
For example, to model a gas cap test for light duty cars and trucks of Model Year 1981 and later,
applicability for the gas cap test is set to one for these model years and to zero for all previous
model years. Note that applicability is not used to model the effects of test frequency (annual,
biennial, etc.) or to model the effects of age restrictions on testing.
2.1.2 Compliance Rate
The Compliance Rate is the fraction of the applicable fleet that show up for testing and are
eventually passed or waived10. Note, this characteristic is different than the "Fraction Repaired"
which is discussed below. MOBILE6 accepts Compliance Rate values ranging from 0.50 to 1.
This user-input rate would be the same both for exhaust and for evaporative I/M, and the same
for all model years and vehicle classes.
The compliance rate is difficult to measure. It accounts for vehicles that fail a test and do
not return for retest, but also for vehicles do not show up for initial testing, either because they
ignore the requirement or because vehicles otherwise included in the fleet are registered outside
the I/M area. Studies suggest that as many as 10-18% of the vehicles driven in a typical non-
attainment area may not be properly registered. Thus compliance rates may be in the 80 to 90
percent range. In the draft report, EPA requested comments on the default rate. None were
received. We have used a default rate of 85 percent in the final MOBILE6 model.
10In the draft report, calculations used both a "participation rate" and a "compliance rate".
For consistency with exhaust I/M calculations, these were combined in the final MOBILE6.
October 2001 11 M6.EVI.003
-------
2.1.3 Testability
Testability is the fraction of vehicles that show up that actually can be tested. This fraction
depends on the vehicle model year and the type of test under consideration. In the draft report, we
reported that user inputs for testability would be allowed. We have since determined that user
inputs are unnecessary. MOBILE6 defaults will be used as described below:
Gas cap test-Most vehicles have gas caps that can be tested. The default values for
MOBILE6 are listed in Table 2.1. They are based on Illinois tests of 1,877,191
vehicles. For model years 1984-1997, all vehicles were testable. The rate declines
for earlier model years. There is a large drop in the 1970s when evaporative
regulations for trucks were less stringent. (Note, we did not have data to develop
separate rates for LDVs and LDTs, but this might be a useful area to explore in the
future.) For 1998 and later vehicles, default testability is set to 1.
• Fill-Pipe Pressure Test-In a high-traffic I/M setting, many old and new vehicles
cannot be tested with the pressure test. The default values for MOBILE6 are listed
in Table 2.2. They come from Arizona data on 831,565 vehicles of model years
1981-1999. Model years 1981 and earlier will use the 1981 default rate. Model
years 1998 and later will use the 1998 default rate. Note, manufacturers
recommend no fill-pipe tests for OBD-II equipped vehicles.
• OBD check-MOBILE6 will handle the phase-in of OBD-n vehicles by separately
computing I/M weightings for vehicles with and without OBD-II and averaging
these together using the OBD phase-in schedule listed in Table 1.1. For those
vehicles not equipped with OBD-II, testability for the OBD check will be zero. For
vehicles with OBD-II, we assume most will be testable with an OBD check.
However, as explained in M6.EVP.006, we assume that 15 percent of all OBD
systems do not successfully identify failures. Thus, the default testability rate for
OBD-n vehicles is 85 percent.11
For the I/M calculations it is necessary to determine what fraction of vehicles receive what
combination of tests. In particular, we must establish a relationship between testability on
one test and testability on another. For example, if vehicles cannot be tested for gas cap
leaks, are they likely to be among the vehicles that can not be tested with a fill-pipe
pressure test? We assume so. For the portion of the fleet equipped with OBD, testability
nNote, the ability of OBOE systems to detect evaporative problems is expected to
improve in time as vehicle manufacturers gain experience and as stricter tolerances are required
in 2002. However, we do not have the data to model this in detail. The 85 percent is a predicted
average rate for all vehicles with OBDII systems, regardless of model year.
October 2001 12 M6.IM.003
-------
Table 2.1
Gas Cap Testability, IL data
Model
Year
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
Vehicles
2822
1582
4156
1505
4202
2623
12654
6583
23440
6545
18062
10372
45294
33484
111601
58215
159608
87439
213017
98152
226168
62406
196814
128597
361680
122
48
Fraction
Testable
0.88
0.87
0.85
0.81
0.78
0.75
0.79
0.75
0.93
0.94
0.96
0.97
0.97
0.98
1
1
1
1
1
1
1
1
1
1
1
1
1
Table 2.2
Fill-pipe Testability, Arizona Data
Model
Year
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
Vehicles
11180
13332
18440
30982
41442
50738
51687
56864
63926
59650
63379
62335
74693
81892
90748
37946
19186
3107
38
Fraction
Testable
0.61
0.63
0.68
0.71
0.73
0.77
0.79
0.78
0.77
0.76
0.79
0.80
0.79
0.76
0.64
0.44
0.35
0.12
0.16*
*From data. Due to small sample. 1998 rate
will be used for this and later years. Note,
OBD checks are expected to replace pressure
checks for OBD equipped vehicles.
October 2001
13
M6.IM.003
-------
on fill-pipe pressure tests is independent of OBD testability.12 We assume that most vehicles that
can be tested with pressure or OBD tests can also be tested on gas cap tests. The equations
proposed to handle these correlations are embedded in the worksheets "sample calc" and "sample
calc OBD" of the Excel workbook M6im003.xls.
2.2 Failure Rate
Failure rate is the fraction of vehicles that are actually tested that fail the test they are given
and are, therefore, eligible for repair. This varies by model year and by test (and combination of
tests).13 Failure rates for traditional I/M tests (gas cap tests and fill-pipe pressure tests) and for
OBD checks are treated differently.
2.2.1 Traditional I/M tests
For traditional I/M tests, the test status weighting factors used for non-I/M emission
factors are based on lab tests and do not correspond exactly to I/M tests used in I/M lanes, so we
have to construct a relationship between the MOBILE6 weighting factors and the in-use tests.
The categories that need to be mapped are listed below:
Test Status Weighting Factors
Cannister Lab Purge Acronym
Pressure Test Status
Test Status
Pass Pass PP
Pass Fail PF
Fail Pass FP
Fail Fail FF
12 Pressure tests generally will not be applicable to OBD-equipped vehicles. However, if
a pressure test program were modeled for OBD-equipped vehicles, there is no expected
relationship between pressure testability of the OBD-equipped vehicles and OBD-system
malfunction. Since OBD-equipped vehicles are considered testable unless they malfunction, we
treat testability for pressure and OBD as independent parameters.
""Failure Rate" is similar to the "ID Rate" used in discussions of exhaust I/M. However,
"ID Rate" refers to a fraction of emissions, while "Failure Rate" is used for a fraction of vehicles.
October 2001 14 M6.IM.003
-------
Possible In-Use Test Result"
Gas Cap
Test
Pass Pass Pass PPP
Pass Pass Fail PPF
Pass Fail Pass PFP
Pass Fail Fail PFF
Fail Pass Pass FPP
Fail Pass Fail FPF
Fail Fail Pass FFP
Fail Fail Fail FFF
* Note, this table and naming scheme were developed when the purge test was still considered a
viable in-use test that needed to be included in the model. We have kept the naming scheme from
the draft report, but the in-use purge test is not modeled in MOB1LE6.
The task then is to establish a relationship between the four test-status groups and the
eight possible in-use test results. To do this, we make three assumptions:
1. As stated before, gross liquid leakers are not relevant for these I/M calculations. We
assume gross liquid leakers are distributed equally among all other groups, so the "no I/M"
test status weighting factors are normalized without GLLs.
2. We assume that if an I/M lane purge test were viable, it would be the same as the "lab"
purge test. Therefore, the lane purge test failure rate is set to equal the original test-status
weighting for vehicles failing the lab purge test (That is, all vehicles in the test status
groups PF and FF would have in-use test results of either PPF, PFF, FPF or FFF).
3. We assume the gas cap and fill-pipe failures (based on I/M lane data) are a subset of the
cannister pressure test failures (based on lab data).14 That is, we assume that all gas cap
14The assumption that vehicles that fail gas cap tests are a subset of cannister pressure test
failures is a simplifying assumption proposed for MOBILE6, but is not always true. Some small
gas cap leaks are not caught by the cannister pressure test (or OBD). The emissions of vehicles
with these small gas cap leaks are currently included in the cannister pressure pass/purge pass
(PP) average emissions. To model the repair of these small leaks we would need data on the
October 2001 15 M6.IM.003
-------
failures and all fill-pipe failures are cannister pressure test failures, but some cannister
pressure test failures (FP, FF) may pass both the gas cap and fill-pipe tests.15
Based on the preceding three assumptions we can construct Table 2.3 describing which in-use test
outcomes could be matched with which test-status groups.
Table 2.3 Matching Test-status groups and in-use test outcomes
Possible In-Use Test Outcomes
Pass Gas Cap/Pass Fill-pipe/PassPurge
Pass Gas Cap/Pass Fill-pipe/Fail Purge
Pass Gas Cap/Fail Fill-pipe/Pass Purge
Pass Gas Cap/Fail Fill-pipe/Fail Purge
Fail Gas Cap/Pass Fill-pipe/Pass Purge
Fail Gas Cap/Pass Fill-pipe/Fail Purge
Fail Gas Cap/Fail Fill-pipe/Pass Purge
Fail Gas Cap/Fail Fill-pipe/Fail Purge
(Lab) Test Status Group*
Pass
Press./
Pass Purge
PP PPP
Fail Press./
Pass Purge
FP-PPP
FP_PFP
FP_FPP
FP_FFP
Pass
Press./ Fail
Purge
PF_PPF
Fail Press/
Fail Purge
FF_PPF
FF_PFF
FF_FPF
FF_FFF
* The empty cells indicate categories for which we predict no vehicles.
We can then predict the frequency of in-use test failures based on the original laboratory test-
status group weighting factors. To do this, we make a few additional assumptions.
difference in emissions between vehicles with and without these small leaks for hot soak,
running, diurnal and resting loss emissions. In the absence of such data, we will assume that
repairs to these leaks have the same effect as repairs to larger leaks, but will cap the proportion of
all gas cap and fill-pipe failures at the cannister pressure fail weighting factor. One consequence
of this approach is that the marginal benefit of adding a gas cap check to an OBD check on
vehicles is limited to a gas cap benefit for only the fraction of vehicles that we assume have
malfunctioning MIL lights. (See M6.EVP.006 for a discussion of OBD assumptions.)
15Some vehicles may fail the lab cannister pressure test but pass both the fill-pipe pressure
test and the gas cap test due tighter time constraints in the I/M lane and to other differences
between the lab and lane test procedure.
October 2001
16
M6.IM.003
-------
1. We assume the distribution of purge failures is independent of the kind of pressure failure.
That is, the fraction of pressure failures with purge failures in M6.EVP.006 equals the
fraction of fill-pipe pressure failures with purge failures, and the fraction of gas cap
failures with purge failures.
FF PFF FPF FFF
FF + FP PFF + PFP FPF + FPP FFF + FFP
2. We assume the Illinois gas cap fail rate(IL) (see Appendix A) describes the distribution of
gas cap failures.
3. We assume the Arizona fill-pipe failure rates (AZFP) (see Appendix A) describes the
distribution of fill-pipe failures.
4. We also need to determine what fraction of vehicles fail both the gas cap test and the fill-
pipe pressure test. Because we are using gas cap and fill-pipe fail rates from different
states, we need to make additional assumptions. Appendix A describes how we used the
Arizona data to calculate the proportion of gas cap failures that are both gas cap and fill-
pipe failures as opposed to "gas cap only" failures. As explained in Appendix A, we call
this last fraction "AZOnly" For the default case, we assume that "AZOnly" describes a
fraction of all gas cap failures that is the same for all state programs and is the same
regardless of purge status. In particular,
FPP FPF
AZOnlv =
FPP + FFP FPF + FFF
5. We assume the total fraction of fill-pipe and gas cap failures cannot be greater than the
fraction of (lab) cannister pressure failures. This is a simplification for modeling purposes.
If for some reason the total fraction of fill-pipe and gas cap failures does exceed the
fraction for cannister pressure failures, the values are reduced proportionally.
6. We assume that I/M lane checks of OBD indicator lights detect cannister pressure failures
and purge failures, that is, we assume they detect all vehicles in the PF, FP and FF groups,
except those vehicles with malfunctioning OBD indicator lights ,16
16We assume that 15 percent of the OBD-equipped vehicles that would fail the cannister
pressure test or the purge test do not have illuminated indicator lights. While the faulty MIL rate
is actually a measure of false passes; in the MOBILE6 I/M computations, the same result is
October 2001 17 M6.IM.003
-------
Using the assumptions described above, we can derive predictions of the frequency of in-use I/M
test failures where each of the following equations describes the fraction of vehicles assigned to
the respective cell in Table 2.3. Note, the empty cells of Table 2.3 indicate categories for which
we predict zero vehicles.
Table 2.4: Frequency of In-Use Test Failure/Lab Test Status Combinations
Pre-Repair
Category
PP_PPP
PF_PPF
FP PPP
FF-PPF
FP_PFP
FF_PFF
FP_FPP
FF_FPF
FP_FFP
FF_FFF
Frequency as a function of observed rates and age
PP
PF
FP- FP/(FP+FF) x (AZFP + IL x AZOnly)
FF- FF/(FP+FF) x (AZFP + IL x AZOnly)
FP/(FP+FF) x (AZFP - IL x (1 -AZOnly))
FF/(FP+FF) x (AZFP - IL x (1 -AZOnly))
FP/(FP+FF) x (IL x AZOnly)
FF/(FP+FF) x (IL x AZOnly)
FP/(FP+FF) x (IL x (1 -AZOnly))
FF/(FP+FF) x (IL x (1 -AZOnly))
Where:
PP= Test status weighting for pass pressure/pass purge
PF= Test status weighting for pass pressure/fail purge
FP= Test status weighting for fail pressure/pass purge
FF= Test status weighting for fail pressure/fail purge
IL= Gas cap fail rate (based on Illinois data)
AZFP= Fill-pipe pressure test fail rate (based on Arizona data)
AZOnly=Fraction of gas cap failures with fill-pipe pass (based on Arizona Data)
obtained by treating the rate as a testability fraction.
October 2001
18
M6.IM.003
-------
For a given evaluation year, MOBILE6 will compute the frequencies listed in Table 2.4 for each
model year based on the MOBILE6 default values for PP, PF, FP, FF, AZFP, IL and AZOnly.
The default test status rates (PP, PF, FP and FF) for 1995 and earlier vehicles are listed in
Appendix A of M6.EVP.006. For 1999 and later vehicles, the PP, PF, FP and FF rates are
replaced with the rates from Appendix D of M6.EVP.006, which account for the owner response
to OBD that would be predicted without I/M checks. (For 1996-1998, the rates should be a
weighted average of the OBD and pre-OBD rates using the observed phase-in schedule in Table
1.1.)
The default rates for AZFP, IL, and AZOnly are listed in Appendix A of this document. In
the draft report, we proposed that MOBILE6 include a user input for alternate rates. However, we
determined that this input was unlikely to be used, and did not program this capability into the
model.
2.2.2 OBD Checks
The MOBILE6 approach for modeling the benefits of OBD systems and I/M tests that
include checks of OBD systems is detailed for exhaust emissions in the paper M6.EXH.007. The
approach includes assumptions about OBD "false passes" and owner response rates to illuminated
MILs.
In the report M6.EVP.006, we adapted this proposal for evaporative emissions to
calculate test-status rates for OBD equipped vehicles in I/M and non I/M areas. However, the
with I/M rates proposed in M6.EVP.006 need to be adjusted to account for differences in local
I/M programs. Thus, for the No I/M case, we use the Failure Rates (Estimates of Strata Size by
Vehicle Age-From a non-I/M Area) in Appendix D of M6.EVP.006 as the Failure rates for
vehicles in model years and calendar years with an OBD check. In particular, we assume that the
OBD failure rate is the sum of the FP, PF and FF rates. As described in Section 2.1.3, we reduce
testability to account for cases where OBD does not successfully identify failures. The owner
response rate for these vehicles is described in section 2.4.
For exhaust emissions, we assume that there is no additional benefit for a traditional I/M
check when an OBD check is performed. For evaporative emissions, some additional benefit may
be gained from gas cap tests because we assume the MIL functions correctly for only a portion of
the vehicles.17 More research is needed on the marginal benefit of combining gas cap checks with
OBD checks.
17Also see footnote 14 on small gas cap leaks.
October 2001 19 M6.IM.003
-------
2.3 Repair Benefits—Adjusting Weighting Factors
The benefits of a repair depend on what test the vehicle failed and was repaired to pass.
Based on EPA program staff experience of what test combinations are used in state I/M programs,
MOBILE6 models the following test combinations:
• Gas Cap Test Only
Gas Cap Test and Fill-pipe Purge Test
Gas Cap and OBD Check
OBD Check Only
As detailed in Section 1.2 and Appendix C, there are limited data available on the emission
benefits of evaporative emission repairs. Thus, the repair benefits in MOBILE 6 will assume that
a repair moves a vehicle from one test-status category to another and the repair benefit will be the
difference between the emissions assigned to the two relevant test-status categories.
Because the in-use I/M tests are not the same as the test-status categories used for non-I/M
emissions in MOBILE6, it is necessary to determine what the appropriate categories are. Again,
because data are severely limited, it has been necessary to make a number of assumptions. In
particular:
When failed with a gas cap test, repair moves vehicles predicted to have gas-cap failures
and no fill-pipe failures from a pressure fail (FP_FPP or FF_FPF) to a pressure pass
category (PP_PFP or PF_PFF). Likewise, when failed with a fill-pipe test, repair moves
vehicles predicted to have fill-pipe failure (FP_PFP or FF_PFF) to a pressure pass
category (PP_PPP or PF_PPF).
When vehicles predicted to have both gas cap and fill-pipe failures (FP_FFP or FF_FFF)
are failed and repaired with a gas-cap only test, we assume that the gas cap problem is
fixed, and the fill-pipe problem remains. While some vehicles might see some decrease in
emissions from the gas cap repair, this is likely to be small since there is still a pressure
leak. We assign no emission benefit and the vehicle remains in the pressure fail category
(FP_PFP or FF_PFF). Similarly, if a vehicle with a gas cap problem and a fill-pipe
problem (FP_FFP or FF_FFF) receives a fill-pipe test but no gas cap test, we assume the
vehicle remains a pressure failure (FP_FPP or FF_FPF). MOBILE6 only gives such a
vehicle credit for a full pressure repair if it is subject to both kinds of tests. (In which case
FP_FFP is repaired to PP_PPP, and FF_FFF is repaired to PF_PPF.) While this approach
may slightly underestimate the benefits of an evaporative I/M program, the underestimate
is expected to be negligible because the fraction of vehicles that simultaneously fail both
the gas cap and the fill-pipe pressure tests is very small (see Appendix A for rates.)
• For evaporative emissions, we assume the vehicles failed by OBD checks will move from
a failing test-status category to the pass/pass category on repair.
October 2001 20 M6.IM.003
-------
As explained in M6.EVP.009, gross liquid leakers are not identified by gas cap checks,
fill-pipe tests, or OBD checks. Thus, we assume that their fraction is unchanged by the
I/M programs modeled in MOBILE6.
These assumptions lead to Table 2.5, which describes our model of how I/M repairs move
vehicles between the cells in Table 2.3, and thus between test-status categories. Note that because
different tests have different testability rates, different portions of the model year fleet in the same
test program will experience different combinations of tests in the same evaluation year.
MOBILE6 models the benefits separately for each combination of tests that vehicles of a given
model year could receive.
October 2001 21 M6.EVI.003
-------
Table 2.5: Effects of Test and Repair on MOBILE6 Evaporative I/M Categories
Pre-Repair
Category *
PP_PPP
FP_PPP
FP_FPP
FP_FFP
FP_PFP
PF_PPF
FF-PPF
FF_FPF
FF_PFF
FF_FFF
Pre-Repair
Test-Status**
PP
FP
FP
FP
FP
PF
FF
FF
FF
FF
Tests Experienced
any or none
OBD
other***
OBD or gas cap
other
OBD, or gas cap and fill pipe
gas cap, no fill pipe
other
OBD or fill pipe
other
OBD
other
OBD
other
OBD
gas cap
other
OBD
fill-pipe
other
OBD
gas cap and fill-pipe
gas cap only
other
Post-Repair
Category*
PP_PPP
PP_PPP
FP_PPP
PP_PPP
FP_FPP
PP_PPP
FP_PFP
FP_FFP
PP_PPP
FP_PFP
PP_PPP
PF_PPF
PP_PPP
FF-PPF
PP_PPP
PF_PPF
FF_FPF
PP_PPP
PF_PPF
FF_PFF
PP_PPP
PF_PPF
FF_PFF
FF_FFF
Post-Repair
Test-Status**
PP
PP
FP
PP
FP
PP
FP
FP
PP
FP
PP
PF
PP
FF
PP
PF
FF
PP
PF
FF
PP
PF
FF
FF
*These categories refer to the cells in Table 2.3.
**The test-status groups are defined in Section 2.2.1
*** "Other" refers to any remaining combination of evaporative I/M tests, as well as no test at all.
October 2001
22
M6.IM.003
-------
2.4 Waivers and Fraction Repaired
"Fraction Repaired" is the fraction of failing cars that get repaired. It is reduced from one
to account for waivers18. For traditional I/M tests, the default waiver rate is the same value as for
exhaust programs. As for exhaust, we assume that 5 percent of compliant failed vehicles receive
waivers.19 Users will have the option to enter their own waiver rates; these need not be the same
as for exhaust emissions.
We assume that waived vehicles undergo some repairs before being granted a waiver. We
assume these limited repairs have some fractional emission benefit, with the fraction the same as
that assumed in the exhaust I/M calculations. The default emission benefit for waived vehicles is
the same as for exhaust: 20 percent. Users will have the option to enter their own waiver benefit
rates; these need not be the same as for exhaust emissions.
The "Fraction Repaired" for fill pipe and gas cap tests will be calculated using the
following equation:
FractRepaired= (1-W)+(W* WBen)
Where:
W =Waiver rate (default is 0.05)
WBen =Repair Benefit for waived vehicles (default is 0.20)
Thus, using default values, the Fraction Repaired for gas cap and fill pipe pressure tests is
96 percent.
For OBD tests, our draft report proposed not calculating the Fraction Repaired based on
these parameters, but instead choosing a value based on the expected owner response rate for
exhaust emissions. In the final MOBILE6 model, we chose a different approach that better allows
users to adjust benefits for local program data. We assume that, in an I/M program, regardless of
vehicle warranty, the base fraction of vehicles with an illuminated OBD light that are repaired is
99 percent. This accounts for intentional tampering with the OBD light and is consistent with the
18In the draft report, "Fraction Repaired" was calculated in a slightly more complicated
manner, but it was simplified for consistency with exhaust I/M calculations.
19The draft version of this report erroneously reported a higher default rate.
October 2001 23 M6.IM.003
-------
assumptions for exhaust vehicles. The Evap OBD Fraction Repaired is then reduced to account
for waivers and waiver benefits.
FractRepaired=C* ((l-W)+(W*WBen))
Where:
W =Waiver rate (default is 0.05)
WBen =Repair Benefit for waived vehicles (default is 0.20)
C =OBD Coverage (0.99)
Thus, using default values, the Fraction Repaired for OBD vehicles is is 95.04 percent.
2.5 Technician Training
In MOBILE6, as in MOBILES, the evaporative I/M repair benefits assume that all
technicians have sufficient training to make effective evaporative emission repairs. Thus there is
no extra benefit for technician training for evaporative I/M.
2.6 Sample Computations
For sample calculations for light-duty vehicles, see worksheets "sample calc" and "sample
calc OBD" of the attached Excel 97 spreadsheet, m6im003.xls. These have been revised and
corrected since draft spreadsheets were distributed with the draft report. "Sample calc" is
designed for modeling weighting factors for vehicles without OBD or ETP (pre-1996), and is set
up to model a calendar year 1995 scenario. "Sample calc OBD" is designed for vehicles with
OBD and ETP (1999 and later) and is set up to model a calendar year 2025. Model years 1996-
1998 are modeled in MOBILE6 as a weighted average of the results from the two spreadsheets.
The worksheets are designed for users to input data into yellow cells. Users can define a
scenario using the worksheets "1995 s descriptions" or "2025 s descriptions" to specify the
evaluation years, vehicle ages, and program parameters (applicability, compliance, and waiver
rates). Blue cells are values that are looked up in other worksheets. Green cells indicate results
used elsewhere in the workbook.
October 2001 24 M6.IM.003
-------
Section 3 Effects of Inspection Frequency (Sawtooth Methodology)
As is true for exhaust I/M, the length of time between evaporative system inspections is an
important variable for determining the emission benefit of the inspections. Thus, MOBILE6 will
calculate different evaporative test status IMWeightings for annual and biennial I/M programs.
To calculate the effects of the different program designs, MOBILE uses the "sawtooth
methodology." This methodology is designed to account for two I/M facts: (1) failures occur
between inspections, and (2) because vehicles are tested throughout the year, emissions at any
point in time are actually an average of the emissions before and after testing.
The sawtooth used for evaporative system I/M is based on the sawtooth used for exhaust
emissions (see M6.EVI.001), modified to apply to the unique aspects of MOBILE6 evaporative
emissions calculations. In particular, it takes into account the fact that evaporative I/M reductions
are not calculated as a credit subtracted from no-I/M emissions, but as a re-weighting of the test-
status categories.
3.1 General Sawtooth Method
For purposes of modeling, we assume all vehicles are inspected on the first anniversary of
their purchase and periodically thereafter, always on that same date. It is also assumed that sales
occur exactly in the 12 month period from October of the calendar year previous to the model year
through September of the next calendar year. For example, in January 1999, the age distribution
of the 1997 model year vehicles will range from 1.25 years to 2.25 years. With an annual
inspection program, most of these vehicles will have been inspected only once, several months
earlier, but those sold in October-December of 1996 will have experienced a more recent second
inspection.
Because of the distribution of inspection times, the sawtooth methodology divides the
vehicle of a given model year into two groups, (1) the younger cars, those purchased between the
evaluation date (typically July 1) and September 30, and (2) the older cars, those purchased
between October 1 of the previous year and the evaluation date. As in the description of the
exhaust I/M sawtooth method, the first group is referred to as the "first segment"; the second is
the "second segment".
For exhaust emissions, we assume that the type of problems which cause I/M failures can
re-occur as often in the repaired vehicles as they do in the unrepaired fleet. We will assume that
this holds true for evaporative failures as well. Thus, it is assumed that the fleet, after repair, will
have the same rates of new failures as vehicles of the same age before repairs. For evaporative
emissions, this assumption means that, within each model year, the oldest and newest cars will
October 2001 25 M6.IM.003
-------
have low rates of new failure after repair, while middle-aged vehicles will have much higher rates
of new failures. This assumption is not completely satisfactory; however, we do not believe it has
a significant impact on our results (see footnote 24 in Section 4). In our draft report, we requested
comments on ways to modify the sawtooth methodology in order to account for fail rates that
differ for repaired and unrepaired vehicles, and for suggestions of alternative rates to be used for
the vehicles with evaporative system repairs. None were received and we have retained the
methodology described in the draft report.
The MOBILE6 approach is illustrated in Figure 3.1 which shows a generalized version of
the I/M sawtooth methodology. The top set of points represents a set of original test-status
weightings at various ages for a given test-status.20 For instance, Points A, B, and C might
illustrate the non I/M case fail pressure/fail purge test status fraction for vehicle ages of 0, 1, and 2
years.21 The lower curve represents the fraction after I/M repair (without the sawtooth). In
particular, points E and G represent the revised fail rates calculated for a given age as described in
Section 2.
The toothed line (A, B, E, F, G,....) between the no-I/M curve and the repair curve
represents the repair and subsequent deterioration of a cohort of vehicles of the exact same age.
All deterioration slopes are parallel with the no-I/M curve (i.e., segment E-F is parallel to segment
B-C). Note that the slope varies for different ages. Also note that the assumption that the lines are
parallel (that the rate of new failures is a function of age, but are not affected by I/M inspections)
does not affect the fail rate after inspection (points E and G in Figure 1), these are assumed to be
independent of inspection frequency and are the values determined in Section 2. However, the
rate of new failures does determine the slope of the I/M line between inspections and the
estimated fail rate before inspection, thus determining the value for point F and the average values
of Segments 1 and 2.
20These calculations will be made for the test-status groups that involve failure (PF, FP,
FF). The rate for passing group (PP) will be calculated by subtracting all the failing fractions
from 1.0
21 As explained in M6.EVP.005 and summarized in the introduction of this document,
there are different curves for traditional vehicles and vehicles subject to the enhanced evaporative
test procedure and vehicles equipped with OBD.
October 2001 26 M6.IM.003
-------
Figure 3.1
Fraction
in Test
Status
Group
General Sawtooth Method for
Evaporative
Fraction with I/M
(no sawtooth)
Evaluation
- Year -Model
Year+1
JDX = 5
Age in Years
October 2001
27
M6.IM.003
-------
In the illustration, the repair effect is represented by the sudden change in test status weighting at
each inspection interval (i.e, from Point F to Point G)22. However, this represents only a single
cohort of vehicles inspected on the same day. To compute the evaluation-date value of
inspections distributed throughout the year, we calculate an average. The heavy shaded portions
of the lines illustrate how the test status weighting for a model year of index 3 (JDX=evaluation
year-model year +1) is produced. The user chooses either January or July of a calendar year as the
evaluation date. This example shows a January evaluation. Segment 1 represents the vehicles sold
from January through September, which are still less than 2 years old at the January 1 evaluation
date. The vehicles sold from October through December are represented by Segment 2. These
are vehicles which are older than 2 years. The average value for each segment is calculated and
the two are weighted together by the model year fractions, FRC and FRN, that are represented by
each segment. FRC and FRN are calculated from the evaluation month (for example, for January,
FRC=0.25, FRN=0.75). This weighted average is used in MOBILE6 as the fail rate for the age in
the post-I/M case. For less frequent periodic inspections (biennial programs.) the "teeth" of the
sawtooth are more widely spaced. The mathematical details of this methodology are described in
Appendix B.
3.2 Algebraic and Numeric details
The Excel workbook m6im003.xls includes two worksheets ("pre-OBD sawtooth" and
"OBD sawtooth") that provide examples of the sawtooth calculations. These may be modified by
using the scenario worksheets ("2025 s descriptions" and "1995 s descriptions) to enter alternate
evaluation months, test frequencies and grace periods. Note that while Figure 3.1 implies that J/M
is modeled as a continuous curve, I/M effects are actually modeled only at integer ages and
benefits are interpolated between. For vehicles with age less than one, no J/M benefits are
calculated.
22The after-repair I/M weighting will actually increase for some test-status groupings. The
fraction of pass pressure/pass purge should always increase. Furthermore, for example, in an I/M
program with only a gas cap test, repairs may increase the fraction of vehicles in the pass
pressure/fail purge category as vehicles are repaired from fail pressure/fail purge status.
October 2001 28 M6.IM.003
-------
Section 4 Comparisons to MOBILES
MOBILE6 computes the emission impact of evaporative I/M programs by combining the
methodology presented in this report with updated approaches for estimating diurnal, hot soak and
running loss emissions, including emissions from gross liquid leakers and estimates of the effects
of ambient temperature, fuel composition, and patterns of vehicle activity. Thus, a comparison
of evaporative I/M benefits between MOBILES and MOBILE6 represents many more changes
than just the changes in evaporative I/M methodology described in this report. To compare
MOBILES and MOBILE6 evaporative I/M benefits for an area, it is necessary to run the two
models with the area's specific ambient conditions, fuel, vehicle fleet mix and vehicle activity.
Such comparison is beyond the scope of this report.
Instead, to help explain the technical basis for differences users will see between
MOBILE6 and MOBILES benefits, this chapter compares the algorithms used in MOBILES and
MOBILE6 and the effect of this algorithm change on the weighting factors used in the two
models.
The MOBILES methodology for estimating emissions from evaporative I/M is less
elaborate than what we have proposed for MOBILE6, but the two approaches have basic
similarities. Like MOBILE6, the MOBILES methodology is based on varying the fraction of
vehicles in the pressure and purge test status groups. In MOBILES, the weighting factors for the
no I/M case are modified to account for I/M and modified to account for I/M effectiveness and the
"sawtooth" effect of failures between inspections. MOBILES only estimates evaporative I/M
effects for pressure and purge tests. The effects of OBD tests are not computed, and gas cap
effects have to be calculated outside the model as a fraction of the pressure test effect.23
In MOBILES, test-status weightings for vehicles failing the pressure test, the purge test
and both were estimated using a zero mile intercept and slopes for below and above 50,000 miles.
These rates were based on data from Hammond, Indiana I/M lanes. MOBILE 5 then calculates a
percent reduction in the fraction of failing vehicles. This is based on a user-input compliance rate
(typically 96 percent) and a "sawtooth" effectiveness based on age that is calculated outside the
model, based on an assumed 95 percent detection rate and an assumed cyclical pattern of repeat
failures in previously repaired vehicles.
Table 4.1 lists the age-weighted average of the failing weighting factors for LDGV in
MOBILES and MOBILE6. The MOBILES "with I/M" rates were computed for a test-only
program of annual pressure tests with 96 percent compliance. The MOBILE6 rates were
23"Credit for Gas Cap Check plus Purge Test," memo from Phil Lorang to Regional Air
Directors, December 1994.
October 2001 29 M6.IM.003
-------
computed for a July 1995 evaluation of a program of annual fill-pipe and gas cap tests with 85%
compliance, 5 percent waivers, 20 percent benefit for waived vehicles and a one year grace period.
Table 4.1 Age Weighted Test Status Fractions (1995 fleet)
Without I/M With I/M
MOBILES MOBILE6 MOBILES MOBILE6
Fail Pressure Tests (FP+FF) 0.131 0.140 0.027 0.077
Chart 4.1 illustrates the difference between MOBILES and MOBILE6 in the pressure failure rate
as a function of age.
Chart 4.1 Pressure Failures, M5 vs M6, 1995 Calendar Year
0.700
0.600
0.500
0.400
0.300
0.200
0.100
0.000
M6no IM
m6w IM
m5 no IM
m5 w IM
M5 delta
M6 delta
25
30
In Chart 4.1, "M6 delta" is the difference between the MOBILE6 with I/M and MOBILE6 no I/M pressure
fraction. "M5 delta" is the corresponding difference for MOBILES.
While the table and graph do not compare evaporative emissions between the two models, they do
illustrate several important differences between the models:
October 2001
30
M6.IM.003
-------
1. MOBILES includes a purge test option. Purge tests are not modeled in MOBILE6
because, in practice, purge tests are not performed in I/M programs.
2. The"no I/M" pressure failure curves have quite different shapes in the two models. The
MOBILE6 curves show lower failure rates in early years and a steep increase in failures
after about age 10. For the oldest cars, MOBILE6 predicts much higher failure rates than
MOBILES. This is based on new data on vehicles over 20 years old collected by the
Coordinated Research Council (CRC). However, averaged across all ages, the two models
lead to very similar average fail rates for the "no I/M" case. For more information on the
"no I/M" rates, see M6.EVP.006.
3. The MOBILES model (in particular, the external model used to estimate I/M effectiveness
values for MOBILES) is based on the assumption that detection and repair of failures in
one year is independent of detection and repair in future years, thus with an I/M program,
virtually all failing vehicles will eventually be detected and repaired if they go through
enough I/M cycles. On the other hand, the MOBILE6 approach is based on testing
experience that shows some vehicles are inherently "untestable" because of their design,
and will never be repaired, no matter how many inspections they undergo. Furthermore,
the MOBILE6 methodology assumes that vehicles that are not repaired in one year due to
non-participation, non-compliance or waiver, also will not be repaired in future years.
Thus, in the MOBILE 6 modeling approach, failures accumulate in a subset of vehicles
despite the existence of an I/M program. We believe this is a more realistic algorithm.
4. The MOBILES model (in particular, the external model used to estimate I/M effectiveness
values for MOBILES) assumes a cyclical pattern of repeat failures in previously repaired
vehicles. This leads to periodic dips in the MOBILES "with I/M" curve. MOBILE6
assumes that repaired vehicles fail at the same rate as non-repaired vehicles of the same
age, leading to a noticeable downturn in the "with I/M" failures at ages where the fail rate
in the "no I/M" case has flattened out (for example, at ages greater than 20 in Chart 4.1).
Because the fraction of the fleet at these ages is small, we believe that the MOBILE6
approach is an acceptable simplification.24 In the draft report, EPA requested comments
on how to better model repeat failures. None were received. We have used our simplified
"equal fail rate" approach in the final version of MOBILE6.
24To test the assertion that the downturn in the "MOBILE6 with I/M" pressure fail rate
(illustrated in Chart 4.1, Pressure Failures, MS vs M6, 1995 Calendar Year) has a negligible
impact, we tested setting the fraction of failures to be constant for ages greater than 20. This
increased the fleetwide average fraction of remaining pressure failures by only one tenth of a
percent.
October 2001 31 M6.IM.003
-------
As the chart and table indicate, these comparisons suggest that for an annual gas cap and
fill-pipe pressure test for all ages, MOBILE6 may indicate significantly less credit for evaporative
I/M programs than MOBILE 5. However, these results look at test-status weighting factors only.
They do not include the new MOBILE6 estimates for evaporative emissions, and gross liquid
leakers. They do not include changes in RVP and temperature effects or vehicle activity. In
addition, the results discussed here are only for a specific scenario. With other program designs,
such as gas-cap only tests and programs with longer grace periods, the MOBILE6 I/M benefits
may be closer to, or even exceed those estimated in MOBILES. Users can best compare
MOBILES and MOBILE6 I/M benefits by running the two models side-by-side for the scenarios
of interest.
October 2001 32 M6.IM.003
-------
References
(Available on EPA website http://www.epa.gov/otns/tn6.httn unless otherwise noted.)
M6.EVP.001 "Evaluating Resting Loss and Diurnal Evaporative Emissions Using Real Time
Diurnal Tests," Larry Landman, April 2000, EPA420-R-01-018.
M6.EVP.002 "Modeling Hourly Diurnal and Interrupted Diurnal Emissions Based on Real-time
Data," Larry Landman, April 2001, EPA420-R-01-019.
M6.EVP.003 "Evaluating Multiple Day Diurnal Evaporative Emissions Using RTD Tests," Phil
Enns, April 2001, EPA420-R-01-020.
M6.EVP.004 "Update of Hot Soak Emissions Analysis," Terry Newell, April 2001,
EPA420-R-01-026.
M6.EVP.005 "Modeling Diurnal and Resting Loss Emissions from Vehicles Certified to
Enhanced Evaporative Standards," Larry Landman, April 2001, EPA420-R-01-
021.
M6.EVP.006 "Estimating Weighting Factors for Evaporative Emissions in MOBILE6," Larry
Landman, April 2001, EPA420-R-01-022.
M6.EVP.007 "Hot Soak Emissions as a function of Soak Time," Ed Glover, draft 6/23/98
EPA420-P98-018.
M6.EVP.008 "Running Loss Emissions," Larry Landman, April 2001, EPA420-R-01-023.
M6.EVP.009 "Gross Liquid Leakers," Larry Landman, April 2001, EPA420-R-01-024.
M6.EXH.007 "Determination of NOx and HC Basic Emission Rates, OBD and I/M Effects for
Tier 1 and Later LDVs and LDTs," John Koupal, draft 5/3/99, EPA420-P-99-009.
M6.FLT.007 "Fleet Characterization Data for MOBILE6: Development and Use of Age
Distributions, Average Annual Mileage Accumulation Rates and Projected Vehicle
Counts for Use in MOBILE6," Tracie Jackson, September 2001,
EPA420-R-01-047.
M6.JJVI.001 "MOBILE6 Inspection/Maintenance Benefits Methodology for 1981 through 1993
Model Year Light Duty Vehicles," Ed Glover and David Brzezinski, draft, March
1999, M6.IM.001, EPA420-P-99-007.
October 2001 33 M6.JM.003
-------
AP-42 "Compilation of Air Pollutant Emission Factors, Volume II: Mobile Sources"
pending 5th edition, updated 1998. Available at http://www.epa.gov/oms/ap42.htm
"IM240 & Evap Technical Guidance," EPA420-R-98-010, August 1998. Available
at http://www.epa.gov/oms/im.htm.
"Credit for Gas Cap Check plus Purge Test," memo from Phil Lorang to Regional
Air Directors, December 1994. (available from EPA on request.)
October 2001 34 M6.IM.003
-------
Appendix A Observed Fail Rates
Observed fail rates are the percent of tested vehicles that failed a given evaporative
emission test. They are used in MOBILE6 to determine the failure rates for specific tests as
described in Section 2.2 of this report. Observed fail rates are likely to vary with individual
program design, and ideally should be measured from the program being modeled. However, for
prospective modeling, it is necessary to provide rates. This appendix explains how MOBILE6
values were determined for the variables AZFP, IL and AZGCOnly.
Gas Cap Tests ("IL" gas cap fail rate)
For gas cap tests, the observed fail rates in MOBILE6 are derived from data from the
Illinois evaporative I/M program. This program tests gas caps for leaks of 60 cc per minute or
more. Data is available on 1,865,029 LDVs and LDTs of model year 1971-1997, tested from
April 1997 to April 1998.
To determine a smooth function describing how gas cap fail rates increase with age, we
used the following approach:
1. Age was calculated as test year minus model year.
2. Test results by model year and test month were grouped by age.
3. Failing vehicles were defined as vehicles that had leaking, missing, damaged or wrong gas
caps.
4. Failure rates were defined as the number of failing vehicles divided by the number of
vehicles receiving a gas cap test.
5. Regressions were run though the failure rate data. These are displayed in Table A-l and
Figure A-1.
6. The logistic regression capped at 30 percent was chosen as the best fit to represent the fail
rates. Like the un-capped logistic regression, this regression provides a close fit for the
most abundant model years. Compared to the uncapped logistic, it provides slightly higher
fail rate estimates for younger vehicles, but it provides a much better representation of the
osbserved "leveling-off' of fail rates in the oldest vehicles.
7. Because the data from the latest model years (ages 0,1 and 2) included vehicles built to the
enhanced evaporative emission standard and equipped with OBD, the regression was run
again without these vehicles. (The number of vehicles of age 0 and 1 was also orders of
magnitude lower than the number of older vehicles, thus there was much greater
uncertainty in the fail rate for these very young vehicles.)
8. The coefficients from this regression were used to develop an equation describing the fail
rate that could be used to predict the fail rate at ages 1 and 2.
October 2001 35 M6.IM.003
-------
Table A-1
AGE BASED GAS CAP TEST RESULTS
Initial Tests for April 1997 Through April 1998
First
Model
Year
1971
Totals
R2
Age
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
Number
Gas Cap
Tests
336
2782
1218
3634
1085
3856
2269
11435
5352
21995
5621
20038
11582
55440
31974
126805
53179
193384
56621
262211
59887
216051
48952
285036
85945
298243
62
36
1865029
Actual IL
Rate*
(%)
27.4
25.8
23.6
25.9
22.6
24.3
21.0
17.9
16.1
14.3
13.9
10.6
12.1
9.2
8.4
6.2
6.3
4.5
5.2
3.5
2.5
2.3
2.5
1.3
1.6
0.9
4.8
2.8
3.7
Linear
Regression
(no
constant)
24.11
23.22
22.32
21.43
20.54
19.65
18.75
17.86
16.97
16.07
15.18
14.29
13.39
12.50
11.61
10.72
9.82
8.93
8.04
7.14
6.25
5.36
4.46
3.57
2.68
1.79
0.89
0.00
0.958
Logistic
Regression
(no constant)
44.77
38.89
33.78
29.35
25.49
22.14
19.24
16.71
14.52
12.61
10.95
9.51
8.26
7.18
6.24
5.42
4.71
4.09
3.55
3.08
2.68
2.33
2.02
1.76
1.53
1.33
1.15
1.00
0.967
Logistic
Regression
(capped at
30)
26.14
25.43
24.62
23.69
22.66
21.51
20.26
18.93
17.52
16.07
14.59
13.12
11.69
10.32
9.03
7.84
6.76
5.78
4.92
4.16
3.50
2.94
2.46
2.05
1.70
1.41
1.17
0.97
0.979
Logistic,
capped at
30, skip age
0,1 and 2
26.67
25.98
25.16
24.22
23.14
21.92
20.58
19.13
17.58
15.98
14.36
12.75
11.19
9.72
8.35
7.11
6.00
5.03
4.19
3.46
2.85
2.34
1.91
1.56
1.27
0.982
* Calculated from Illinois data. "Age" equals calendar year minus model year.
October 2001
36
M6.IM.003
-------
350000
300000
Figure A-1
Illinois Gas Cap Fail Rates-Possible Fits
-------
Thus, to represent the fail rates of 1995 and earlier model years, MOBILE6 will use the fail rates
computed with the coefficients from the capped logistic regression on the data minus ages 0, 1
and 2. The associated logistic equation is:
FxlOO =
Where:
F = the fail rate
u = the upper bound, 30
b0= 1.4462
bj= 0.8051
a = age
This leads to the fail rate percentages listed in Table A-2.
For 1996-1998 model years, the I/M benefits will be calculated using both 1995 and 1999
rates. After emissions are calculated, these will be weighted together based on the enhanced test
procedure phase-in schedule given in Table A-3.
For 1999 and later years, this fail rate will be adjusted to account for the enhanced
evaporative emission standard and OBD since we would expect fewer failures in these vehicles.
As is done for test status groups (see Section 1.3 of this document and also M6.EVP.006), the
failures in these vehicles will be adjusted to double the time for failure and to reduce failures to
account for OBD. For age 0-3, we assume that 76.5 percent of new failures are detected by OBD
and repaired.25
FxlOO =
o i
x (1-0.765)
25
The OBD detection rate is the product of the OBD failure rate and the owner response
rate.
October 2001
38
M6.IM.003
-------
For 1999-and-later year vehicles of ages from 3 to 6, we assume the doubling of durability
and that 8.5 percent of new failures are detected and repaired. The 0.63 in the equation accounts
for the difference in fail rates at age 3.
F*100 =
- + b xb
u U
x(l-0.085)-0.63
For 1999-and-later year vehicles of age greater than 6, we assume no OBD repairs in the absence
of an I/M program. The 0.74 in the equation is the difference in fail rates at age 6 due to the
owner response rate under warranty.
FxlOO =
j_
u
-0.74
This leads to the fail rate percentages listed in Table A-2.
October 2001
39
M6.IM.003
-------
Table A-2
Gas Cap Fail Rates (in percent) For
MOBILE6
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
25
26
27
MY 1995
and earlier
0.68
0.83
1.03
1.27
1.56
1.91
2.34
2.85
3.46
4.19
5.03
6.00
7.11
8.36
9.72
11.20
12.76
14.36
15.99
17.59
19.13
20.59
21.93
23.14
24.22
25.17
25.98
2668
MY 1999 and
later*
0.16
0.18
0.20
0.22
0.31
0.42
0.53
0.67
0.82
0.99
1.18
1.38
1.60
1.85
2.12
2.41
2.73
3.07
3.45
3.85
4.29
4.76
5.27
5.80
6.37
6.98
7.62
829
*For 1996-1998 model years, I/M benefits will be calculated using both 1995 and 1999 rates.
Emissions will be weighted together based on the enhanced test procedure phase-in schedule
given in Table A-3.
October 2001
40
M6.IM.003
-------
Table A-3
Observed Phase-in of Light Duty
Vehicles With Enhanced
Evaporative Controls and Light
Duty OBD
Model Year
1995
1996
1997
1998
1999
Percentage
0%
30%
55%
90%
100%
Fill-Pipe Pressure Test Fail Rates ("AZFP" pressure rate)
For fill-pipe pressure tests, fail rates are based on rates measured in Arizona's I/M
program. Gordon Darby provided test result counts by month and vehicle class for 536,520
LDVs, 227,753 LOT Is and 67,292 LDT2s that received fill-pipe pressure tests and gas cap tests
between August 1997 and August 1998. Model years from 1981 to 1999 were tested in a program
that pressurizes to 14 inches of water column at the fillpipe and looks for a loss of more than 6
inches of water in 2 minutes or less.
To determine a smooth function describing how fill-pipe pressure fail rates increase with
age, we used the following approach:
1. To derive rates appropriate for vehicles built prior to the advanced evaporative test
procedure, data on model years 1996-and-later were removed from the data set, leaving
497,227 LDVs, 214,049 LDTls and 61,634 LDT2s of model year 1981-1995. For
consistency with other evaporative emissions work, we combined all data on the three
vehicle classes.
2. Age was defined as model year minus test year.
3. Fill-pipe failures were defined as vehicles that failed to pressurize or had visual failures of
missing canisters, damaged canisters or missing hoses.
October 2001
41
M6.IM.003
-------
4. The Arizona fail rate was computed as the number of failures divided by the number of
vehicles with successful tests for each age.
5. This rate was graphed versus age, and a number of statistical fits were computed with
SPSS software. The raw data and the three best fits are shown in Figure A-2 and Table A-
4. The cubic and quadratic equations offer the best fits to the data, each with an R-squared
of 0.986. In SPSS the two equations provide almost identical predicted values as shown in
Table A-4. For simplicity, the quadratic form will be used. However, because the
curvature of the quadratic form predicts slightly more failures at age 0 than at age 1, we
will set the values for ages 0 to be the same as the value at age 1.
October 2001 42 M6.IM.003
-------
Table A-4
Arizona Pressure Fail Rates, MY 1981-1995, Possible Fits
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Number of
Vehicles
Tested
0
0
22596
58576
60745
55611
50282
47890
46381
46942
43468
39917
35770
27259
18729
11310
7721
4928
Pressure Fail
Rate
0.012
0.012
0.016
0.017
0.021
0.024
0.030
0.042
0.054
0.066
0.073
0.079
0.081
0.100
0.115
0.142
R-squared
bO
bl
b2
h3
Quadratic Fit
0.011
0.011
0.011
0.012
0.015
0.018
0.022
0.027
0.034
0.041
0.049
0.058
0.068
0.079
0.092
0.105
0.119
0.134
0.986
0.0115
-0.0012
0.0005
Cubic Fit
0.009
0.009
0.010
0.012
0.015
0.018
0.023
0.028
0.034
0.041
0.049
0.058
0.068
0.079
0.091
0.104
0.119
0.135
0.986
0.0086
0.0002
0.0003
5 90F.-06
Logistic Fit
(no upper
bound)
0.008
0.009
0.011
0.013
0.016
0.019
0.022
0.027
0.032
0.038
0.045
0.053
0.064
0.076
0.090
0.107
0.127
0.151
0.982
126.536
0.8406
October 2001
43
M6.IM.003
-------
Figure A-2
AZ Pressure Fail Rates, Possible Fits
0 14 -
0 12 -
0 1 -
3
re
0 08 -
1
0 02 -
[
n -
1 F — C ==
n
^
rn
-
--'
•
-
J
-"-
i
r'
'
u
^
*
/
i
/
/
-
•
U
.
'
u
—
.
4
F
-
/
/
,
.
/
/
f
i
1 60000
- 50000
•j; | [Pressure tested (count)
ii — B-AZ pressure fail rate
& pressure failures, cubic fit
- 30000 •§
§ pressure failures, logistic fit (no bound)
- 20000
- 10000
- n
-10123456
7 8 9 10 11 12 13 14 15 16 17
Age
So, for light-duty vehicles and trucks, model years 1981-1995, the default fill-pipe pressure test
fail rate in percent will be:
F = 0.0115-0.0012a + 0.0005a2
Where
F = the fail rate in percent
a = the age in years.
The default rates are listed in Table A-5
For model years 1999 and later, these fail rates will be adjusted to account for the
enhanced evaporative test procedure (ETP) requirement and OBD.
As previously explained for gas caps and test status groups, the increased durability due to
the ETP will be expected to decrease the age effects by a factor of two. The OBD effects in a
October 2001
44
M6.IM.003
-------
non-I/M area are expected to decrease the occurrence of failures by 76.5 percent in years 1-3 and
8.5% in years 3-6. Thus we will use the following equations to predict fail rates:
For light-duty vehicles and trucks, model years 1999-and-later, age 2-3:
F =
0.0115-0.0012 y +0.0005 y
x (1-0.765)
Because the equation predicts higher rates for for ages -1,0 and 1 than for age 2, we will use the
value calculated for age 2 for these rates.
For ages greater-than-3 up to 6, the calculation includes the ETP durability effect (division of age
by two), and the non-I/M area reductions of new failures by 8.5%, but the total fail rate also is
shifted to account for the reductions in years 1-3. Thus, the fail rate includes a term (0.0074)
subtracting the difference between the fail rate at age 3 with the 90% and the 10 percent owner
response assumptions.
F =
0.0115-0.0012 - +0.0005 -
xO.9m-0.0074
For ages greater than 6, the calculation is similar, except there is no reduction in new failures due
to OBD, and the rate is adjusted by 0.0084 to account for the difference at age 6 in the fail rate
with the 10% and the zero response assumptions.
F =
0.0115-0.0012 - +0.0005 -
-0.0084
October 2001
45
M6.IM.003
-------
Table A-5
Fill Pipe Pressure Test Failure rates
for MOBILE6*
Age
-1
0
1
2
O
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Model
Years 1995
and earlier
0.011
0.011
0.011
0.011
0.012
0.015
0.018
0.022
0.028
0.034
0.041
0.050
0.059
0.069
0.080
0.093
0.106
0.120
0.136
0.152
0.169
0.188
0.207
0.227
0.248
0.271
0294
Model
Years 1999
and later**
0.003
0.003
0.003
0.003
0.003
0.003
0.003
0.004
0.005
0.006
0.008
0.010
0.012
0.014
0.016
0.019
0.022
0.025
0.029
0.033
0.037
0.041
0.046
0.050
0.055
0.061
0066
* (Calculated in M6IM003.xls, "Az pres rates", 3/26/99. Values differ from those in table A-4
due to rounding of coefficients in SPSS output.)
** Emissions for Model Years 1996-1998 will be a weighted average based on the phase-in
schedule given in Table A-3.
October 2001
46
M6.IM.003
-------
"Gas Cap Only" Rates ("AZOnly")
The first section of this appendix describes the fraction of all vehicles that have gas cap
failures. Some of these vehicles have both gas cap problems and additional problems that would
be detected with a fill-pipe pressure test. For MOBILE6 we need to determine the distribution of
these vehicles in the fleet. Specifically, for the calculations described in this report, we need to
determine the fraction of all gas cap failures that are "gas cap only" failures (rather than both gas
cap and "fill-pipe" failures). Since the only states performing both tests were Arizona and
Delaware, and Arizona data were readily available, the Arizona data were used.
1. To derive rates appropriate for vehicles built prior to the advanced evaporative test
procedure, data on model years 1996-and-later were removed from the data set, leaving
497,227 LDVs, 214,049 LDTls and 61,634 LDT2s of model year 1981-1995. For
consistency with other evaporative emissions work, we combined the three vehicle classes.
2. Age was defined as model year minus test year.
3. Total gas cap failures for a given age were defined as the sum of the vehicles failing the
gas cap test and having a missing or damaged gas cap.
4. "Failed both pressure and gas cap" numbers were provided by Arizona and summed for
each age. This number includes vehicles that failed both tests for any reason, including
visual failures such as missing gas caps and missing hoses.26
5. Because many vehicles that were tested with the gas cap test were not testable with the
fill-pipe pressure test, the "gas cap only" fraction of gas cap failures could not be
computed directly. Instead we computed a ratio of Arizona fail rates. The gas cap only
rate was defined as (gas cap failures/total gas cap tests - failed both/total pressure
tests)/(gas cap failures/total gas cap tests), ie 1- (failed both/gas cap failures)(total gas cap
tests/total pressure tests).
6. This rate was computed for each age, and regressions were run through the data
(eliminating years where data was not available). The best fit regression was a quadratic.
The actual and predicted values are listed in Table A-6 below. The SPSS statistical output
is also shown.
7. Since the "gas cap only" ratio is a ratio of fail rates, the predicted improvements due to
OBD and enhanced durability should generally cancel out. Thus, in MOBILE6, the gas
cap only ratio is used for all model years.
26Per email from Jeff Reeves, Gordon Darby, Inc., 11/11/98.
October 2001 47 M6.IM.003
-------
Table A-6 Gas Cap Only Failures as a Fraction of Gas Cap Failures
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
25
Measured
Rate
na
na
0.987
0.971
0.966
0.955
0.954
0.952
0.931
0.937
0.926
0.911
0.861
0.872
0.859
0.852
0.846
0.791
na
na
na
na
na
na
na
na
bO
bl
b2
A Hi r2
Predicted
0.985
0.982
0.979
0.975
0.969
0.963
0.955
0.947
0.937
0.927
0.916
0.903
0.890
0.875
0.860
0.843
0.826
0.808
0.788
0.768
0.746
0.724
0.701
0.676
0.651
0.624
0.9849
-0.0019
-0.0005
09504
October 2001
48
M6.IM.003
-------
23 Mar 99 SPSS for MS WINDOWS Release 6.1
Dependent variable.. RATE
Multiple R .97828
R Square .95703
Adjusted R Square .95042
Standard Error .01258
Analysis of Variance:
DF Sum of Squares Mean Square
Regression 2 .04584293 .02292147
Residuals 13 .00205838 .00015834
F= 144.76411 SignifF= .0000
Variables in the Equation
Method.. QUADRATIC
Variable
AGE
AGE**2
(Constant)
B
-.001948
-.000499
.984878
SEE
.003236
.000166
.013561
Beta
-.164156
-.817201
T
-.602
-2.997
72.626
SigT
.5575
.0103
.0000
23 Mar 99 SPSS for MS WINDOWS Release 6.1
October 2001
49
M6.IM.003
-------
Appendix B Mathematical Treatment of Inspection Frequency for
Evaporative I/M
This appendix describes a mathematical approach to calculate the effects of inspection frequency
for evaporative I/M as summarized and illustrated in Section 3.0 of this report.
Note that these calculations must be made for each of the three failing test status groups (pressure
fail, purge fail, purge and pressure fail.) For simplicity, this distinction is not detailed in the
equations below.
The "No-I/M" case
The sawtooth methodology (see Figure 3.1 in the main document) is used in the No-I/M
case to determine the test status weightings appropriate for the evaluation month. While simpler
methods could accomplish this, this case provides an opportunity to explain many details of the
sawtooth methodology before adding the additional complications of the I/M case.
1. Where MEVAL is the month of evaluation, (e.g. January = 1, July=7)27, we
compute FRC, the fraction of the year since the model year changed on October 1
as follows:
DIFF = MEVAL -10
FRC(DIFF > 0) = DIFF/\i
FRC (DIFF < 0) = 1 + DIFF/n
FRN =l-FRC
Thus, for the evaluation months of January-September, FRC>0.25, for October-December,
FRCX0.25. Because we assume sales are uniform throughout the year, FRN may be
considered the fraction of the fleet represented by Segment 1; FRC is the fraction of the
fleet represented by Segment 2 in figure 3.1.
27MOBILE calculates emissions only for January and July (FRC= 0.25 and FRC = 0.75).
However, for completeness and future coding flexibility, this report describes a method that
could be used for any month of the year.
October 2001 50 M6.IM.003
-------
2. MOBILE tracks model year through an integer index of model year, JDX.
JDX=Calendar Year of Evaluation -Vehicle Model Year +1
This means, for example, in every month of 1990, a 1990 vehicle has a JDX of 1.
3. For the specific model year and evaluation month, we compute the average age for
vehicles.
AvAge(JDX = 0,FRC < 0.25) = 0.5 xFRC
AvAge(JDX =\,FRC > 0.25) = 0.5 xFRC
AvAge(jDX = \,FRC < 0.25) = FRC x [(JDX -1) + (0.5 x FRC)]
+FRN x [(JDX - 1)- (0.5 x FRN)]
AvAge(jDX > 1) = FRC x [(JDX -1) + (0.5 x FRC)]
+FRN x [(JDX - 1)- (0.5 x FRN)]
4. Based on the previous equations for average age, we can compute the test status weighting
for each test status group in the NoIM case as the following28:
NoIMWeighting(JDX = 0,FRC < 0.25) = NoIM [0.5 x FRC]
NoIMWeighting(JDX = l,FRC > 0.25) = NoIM[0.5 x FRC]
NoIMWeighting(jDX = I,FRC < 0.25)= FRC x [NoIM (JDX - 1 + 0.5 x FRC)]
+FRN x [NoIM (JDX -1 - 0.5 x FRN)]
NoIMWeighting(JDX > l) = FRC x [NoIM (JDX - 1 + 0.5 x FRC)]
+FRN x [NoIM (JDX -1 - 0.5 x FRN)]
28This equation could be simplified for the No I/M case as was done in the description of
exhaust I/M calculations (M6.EVI.001, Appendix D), but the format given here has a parallel
structure to the format for the I/M calculations.
October 2001 51 M6.EVI.003
-------
Where:
NOIMWeighting is the weighted average of the two segments of the model year,
NoIMis NoIMfrom Section 2.2
MOBILE6 will not actually calculate the noI/M or I/M values as it makes the sawtooth
calculation, but will look up the appropriate value in a previously calculated array.
Because the array has values only for integer ages, the MOBILE code must interpolate
between these ages. This interpolation is not detailed here, but can be seen in the
"sawtooth" worksheets of Excel workbook M6IM003.xls. By assumption, the I/M test-
status weighting at age 0 is the same as the non-I/M weighting.
Annual I/M program starting with Age=l
In a periodic I/M program, the test status weighting at an evaluation date can be considered
a weighted average of the cars that recently had an inspection (the FRC fraction) and those with a
more distant (or no) inspection (FRN). For both the recent and the distant cases, we assume that
the test status weighting at the date of inspection was the I/M test-status weighting (that is, IM
from Section 2.0) for the average age at inspection (AIM).
In the time elapsed since the inspection, we assume more vehicles have entered the failing
test status categories. In particular, we assume that the additional vehicles have entered the failing
categories at the same rate as they would in the same time period without an I/M program.29
Thus, we need to figure the average age at last inspection and the average time since the
last inspection.
1. As was explained for exhaust emissions (M6.IM.001, Appendix D), we assume each
vehicle is tested on its "birthday," and AIM is the average vehicle age at the last inspection
(AIM1 is the age for Segment 1; AIM2 is the age for Segment 2.) For an annual inspection
program where vehicles are first inspected at age one year, the age at most recent
inspection is listed in Table B-l. We will define AIM=0 to mean no prior inspection.
29This is a low rate for low and high ages and a high rate in the middle years. The low
rate makes sense for low ages, but makes less sense for high ages where the noI/M rate reaches
an equilibrium we wouldn't necessarily expect with I/M. However, alternative approaches
require separate rates for vehicles that have and have not been repaired, and, therefore, would
require a more complicated algorithm that tracks these vehicles separately.
October 2001 52 M6.IM.003
-------
Table B-l Age at most recent inspection, annual I/M, first inspection at age 1 year
JDX
1
1
2
2
3
3
4
4
JDX
JDX
Segment of
Model Year
1
2
1
2
1
2
1
2
1
2
AIM
na*
0**
0**
1
1
2
2
3
JDX-2
JDX-1
* Vehicles in this category have not been sold to original owner yet.
** Vehicles in these categories are less than one year old.
2. Unlike exhaust emissions, where emissions are a function of mileage, we also need to
calculate the average time in years since the most recent inspection for both the FRN and
FRC segments. Jf there have been no prior inspections, we calculate the average time
since the vehicle was first sold. For an annual inspection program, DeltaTis calculated as
follows for each of the two segments:
DeltaTl = 1-FRNI2
DeltaT2 = FRC/2
For example, for a January evaluation date, DeltaTl=0.625, DeltaT2=0.125; for July
DeltaTl=0.875, DeltaT2=0375.
More generally:
DeltaTl = 1-FRNI2 + (JDX-2-AIM 1)
DeltaT2 = FRC 12 + (JDX-1-AIM2)
October 2001
53
M6.IM.003
-------
3. The deterioration D(JDX, segment) is the additional growth in the failing test status group
during the time Delta! since the most recent inspection. Since the change in test-status
weighting is the same for both the I/M and the NoIM case:
D(JDX,\) = NoIM (AIM 1 + DeltaTl) - NoIM (AIM 1)
D(JDX,2} = NoIM(AIM2 + DeltaT2)-NoIM(AIM2)
4. Given the previous calculations, we can compute test status weightings for an annual I/M
program as follows:
IMWeighting(JDX = Q,FRC < 0.25) = NoIM[Q.5 x FRC]
IMWeighting(JDX = l,FRC > 0.25) = NoIM[0.5 x FRC]
IMWeighting(jDX = I,FRC < 0.25) = FRC x [lM(AIM2) + D(JDX ,2)]
+FRN x [NoIM (JDX - 1 - 0.5 x FRN)]
IMWeighting(JDX > l) = FRC x [lM(AIM2) + D(JDX,2)]
+FRN x [iM(AIM 1) + D(JDX,1)]
Where:
EVIWeighting is the weighted average of the two segments of the model year.
EVI(AIM) is the test status weighting for age at I/M (AIM) after repair, waiver, and
non-compliance factors from Section 3.0, but before taking the variety of
ages into account. This value is independent of prior year I/M reductions.
We define IM(AIM=0) to be NoIM(O).
AEVI1 is the age at the most recent I/M test for vehicles in segment 1 (vehicles
sold between the evaluation month and Sept. 30).
AEVI2 is the age at the most recent I/M test for vehicles in segment 2 (vehicles
sold between Octoberl and the evaluation month).
JDX is a model year index. JDX = Calendar year- Model Year +1.
October 2001 54 M6.IM.003
-------
D(JDX,1) is the deterioration, the increase in failing vehicles in the time since the
most recent I/M test for Segment 1.
D(JDX,2) is the deterioration, the increase in failing vehicles in the time since the
most recent I/M test for Segment 2.
5. As in the No I/M case, the MOBILE model will actually look up values for integer ages
and interpolate between them. This is illustrated in the "sawtooth" worksheets of
M6EVP002.xls.
I/M with age exemptions
As is the case for exhaust I/M, emissions differ for I/M programs that exempt specific ages
from testing. The exemptions may vary depending on the test conducted. For exemptions at the
beginning of a vehicles life (grace periods), we set the age of last I/M inspection (AIM) to zero.
In particular:
If JDXGRPD+1, AIM has normally computed value
Where GRPD = the highest age that is exempt from I/M. at the beginning of the vehicle life, for
example, if a program begins testing vehicles at age 4, GRPD=3.
For age exemptions at the end of the vehicle life, we set the age of the last I/M inspection
to the age computed for the last eligible model year. That is, for a user input "MaxAge:"
If AIM 1 >Max Age, AIM 1 =Max Age
If AEVI2>MaxAge, AEVI2=MaxAge
N-enniel Inspections
This section describes how MOBILE6 will compute evaporative test status weightings for
vehicles subject to annual and inspections. The methodology is very similar to that described in
Appendix D of M6.EVI.001 for exhaust I/M programs. While MOBILE will calculate benefits only
for annual and biennial programs, this analysis would apply to other periodic inspections as well.
In N-ennial EVI programs vehicles are inspected every N years on the anniversaries of the
sale to their first owner. In this general case, all vehicles with model year index greater than
GRPD+1 should receive one inspection in the 12*N month period preceding the date when the
emissions are to be evaluated.
October 2001 55 M6.IM.003
-------
The principle that a unique value of AIM can be calculated for each model year segment
holds true for arbitrary values of N and GPRD. Thus, to compute I/M weighting factors we can
use the same equations as in step 4 of the annual case, where AEVI1 and AEVI2 denote the integer
ages in years of vehicles on the date of their previous EVI test that were purchased new in the first
and second segments of the JDXth model year.
The N-ennial case differs from the annual case in the calculation of the age at the last I/M
inspection. For the N-ennial case, AEVI1 and AEVI2 are calculated as follows:
JDXGPRD+1:
JDXGPRD+1:
Where
JDX
GPRD
N
MOD(a,b)
AIM1=0
AIM1= JDX-2-MOD((JDX-2-GPRD), N)
AIM2=0
AIM2=JDX-1-MOD((JDX-1-GPRD), N)
= calendar Year- model year +1
= oldest year exempt from I/M at the beginning of the vehicle life
= frequency of regular inspections in years
= the remainder of a divided by b
For illustration, we compute the following table of ages at most recent inspection for the first
eight model years.
October 2001
56
M6.IM.003
-------
Table B-2
JDX
1
2
3
4
5
6
7
8
Annual
GPRD=0
AIM1 AIM2
0 0
0 1
1 2
2 3
3 4
4 5
5 6
6 7
Biennial
GPRD=0
AIM1 AIM2
0 0
0 1
1 1
1 3
3 3
3 5
5 5
5 7
Biennial
GPRD=1
AIM1 AIM2
0 0
0 0
0 2
2 2
2 4
4 4
4 6
6 6
Program Start and End
An evaporative J/M program has a start and end year. We assume that there is no benefit
for the program prior to the start year. At the end year, we assume a sudden end to benefits.30
As for exhaust I/M, in the start up year itself, benefits are 0 for a January evaluation date
since no cars have been evaluated at that time. For an annual program, July benefits in the start-
up year are two-thirds what they would be in a normal year (since the Oct-Dec cars were not
tested). For a biennial program with a mix of "1,3,5" and "2,4,6" testing, the start-up effect
continues into the year after the start up year. Benefits in January of this second year are four-
fifths the benefit of a normal year. Benefits in July of the second year are six-sevenths of those of
a normal year.
30In our draft report, we proposed that benefits decline by a third each year for three years.
However, this would have been very difficult to code into the model and would have a limited
impact on results. For the final MOBILE6 we chose the much simpler "abrupt end to benefits"
approach.
October 2001
57
M6.JM.003
-------
Appendix C Laboratory Data on Evaporative I/M Repairs
As explained in Section 2.3, EPA proposes modeling the effect of an emission repair as a
change of test-status, so the emission benefit of the repair is the difference in emissions between a
vehicle in a failing test status group and a passing test status group as those emissions are defined
in the other reports on evaporative emissions. In this appendix, we compare this modeling
approach to recently collected empirical data.
EPA contracted with ATL to collect data on real time diurnal, running loss and hot soak
evaporative emissions before and after the repair of problems found with fill-pipe and gas cap I/M
tests. This repair effects data was collected under EPA Contract 68-C5-0006, Work Assignment
1-8. At the time the draft report was written, we expected to receive additional data under this
work assignment, but this was not the case.
In the study, 26 failing vehicles were recruited from Arizona I/M test lanes; two vehicles
did not return for retest after repair. Four vehicles with missing gas caps were tested only for
running loss. Table Cl on the following page gives summary data for the difference in emissions
before and after repair for the vehicles, with the difference listed both as an emission rate (g/test
or g/mi) and as a percent reduction.
The test data indicate that, while repairs are almost always beneficial, there is substantial
variation between vehicles. We examined the data in aggregate as well as grouped by the type of
test that was failed, fuel delivery system and model year (see Table C. I).31 ANOVA analysis (p <
0.05) suggests that the data can be disaggregated by fuel delivery system (for running loss
emissions), model year ( for running loss emissions in g/mi and percent), and the type of I/M test
that was failed (for running loss in g/mi and percent, and for hot soak in percent), but we did not
see any trends in the data that made sense based on an engineering understanding of the
evaporative systems.
In order to test the assumptions made in the MOBILE6 handling of evaporative I/M, we
compared repair data for real-time diurnal emissions to the emission benefits that MOBILE6 will
predict as described in M6.EVP.001. To match the test conditions, the MOBILE6 predictions
were made for a 24 hour cycle from 72°F-96°F, at 9.0 RVP. The results are given in the
following table (Table C.2). The results vary by fuel delivery system, model year and test type,
and in a number of cases. On average, the MOBILE6 predictions are fairly close to the ATL data,
although they consistently underestimate the repair effects.
At this time, we do not intend to alter the MOBILE6 proposal to account for differences
between our proposed benefits and the empirical data. This is for three reasons:
31A model-year split at 1985 was considered to account for the effect of cumulative
improvements in evaporative emission control during the 1980s.
October 2001 58 M6.IM.003
-------
1. First, we do not believe the magnitude of the possible effect warrants the time and effort
that would be needed to make a wholesale revision to the evaporative emissions modeling
approach.
2. Second, the MOBILE6 proposed rates are based on tests of hundreds of vehicles (as
described in M6.EVP.001 and the other reports on evaporative emissions) while the repair
effects data, although more directly applicable, are a much smaller dataset.
3. Third, if we were to use the repair effect data, it is not clear how to best adjust repair
effects for variation in RVP, ambient temperatures and differences between vehicle groups
in a way that is consistent with the evaporative emissions proposed in the absence of I/M.
EPA requested comments on whether the proposed MOBILE6 evaporative I/M benefits should be
adjusted to account for differences between the current MOBILE6 proposal and the empirical
data, and suggestions on how this could best be done. None were received, and we have retained
our methodology for the final version of MOBILE6.
October 2001 59 M6.IM.003
-------
Table C.I, Evaporative Emission Repair Data
Contract: 68-C5-0006 - Work Assignment 1-8
Evaporative Emission Repair Credits
Vehicle Listing
Average
Std. Dev
Average
Std. Dev
*
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Veh.
Me.
070
080
061
065
071
077
063
060
069
068
078
084
066
073
075
076
081
079
064
085
062
067
072
074
082
083
Model Fuel
00
OO
88
All
83
85
85-
87
87
88
89
89
95
86
86
91
86+
All
All
83
86
86
87
All
84
85
85-
86
00
00
88
90
91
92
88
86+
All
All
All
2V
2V
Lane
Fail
Fail
CARS Fail
PFI
PFI
FI
PFI
PFI
PFI
PFI
PFI
PFI
TBI
TBI
TBI
FI
FI
All
2V
2V
2V
2V
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Pass
Pass
Pass
Pass
CARBPass
PFI
PFI
FI
PFI
PFI
PFI
PFI
PFI
PFI
TBI
FI
FI
All
All
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
All
Lane
Gap
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Fail
Fail
Fail
Missing
Fail
Fail
Missing
Fail
Fail
Fail
Missing
Fail
Fail
Missing
Damaged
Fail
Fail
Fail
All
Delta
RTD %
Delta
HotS oak %
Delta
(g/mi)Rur %
vehicle not returned
0.08
0.08
na na
8.78
3.89
6.33
3.45
13.37
14.69
0.01
0.01
0.78
0.61
0.69
0.12
0.92
0.92
1.29
1.29
na na
5.07
6.03
5.55
0.68
1.36
1.95
0.81
0.81
na
0.91
0.95
0.93
0.03
0.88
0.88
0.88
0.88
na
1.64
1.70
1.67
0.04
1.80
1.50
0.95
0.95
0.98
0.98
0.98
0.00
0.99
0.97
vehicle not returned
19.11
-0.23
23.47
-9.52
15.25
1.31
9.68
11.25
9.01
10.09
8.20
9.94
14.75
7.19
9.98
not tested
10.64
3.82
15.16
not tested
15.16
na na
0.98
14.00
not tested
0.14
12.53
not tested
11.68
8.12
6.73
9.08
6.71
9.60
5.69
8.83
8.13
0.77
-0.07
0.91
-0.41
0.89
0.20
0.52
0.53
0.55
0.48
0.50
0.48
0.83
0.53
0.68
0.68
0.15
0.93
0.93
0.21
0.93
0.05
0.95
0.83
0.61
0.43
0.65
0.41
0.66
0.33
0.57
0.42
7.29
1.20
1.15
7.94
4.66
0.29
3.23
3.00
3.69
2.83
3.48
2.78
2.77
0.61
1.23
not tested
1.54
1.11
0.44
not tested
0.44
na na
7.29
8.15
not tested
0.03
4.90
not tested
3.20
2.71
3.27
4.00
3.40
3.18
3.01
3.34
2.81
0.97
0.67
0.78
0.94
0.93
0.48
0.82
0.17
0.84
0.16
0.84
0.15
0.79
0.43
0.62
0.61
0.18
0.24
0.24
0.87
0.98
0.08
0.98
(^2
0.53
0.39
0.61
0.39
0.61
0.32
0.74
0.26
1.80
1.47
2.90
1.83
3.70
0.33
1.92
1.01
1.87
0.89
1.78
0.90
0.48
0.65
0.56
-0.01
0.42
0.30
1.84
2.38
2.11
0.38
4.88
4.05
6.34
0.08
2.06
4.44
9.06
4.39
2.89
3.90
2.70
2.83
2.77
2.35
2.16
0.98
0.96
0.99
0.97
0.98
0.86
0.96
0.04
0.96
0.04
0.96
0.04
0.88
0.93
0.89
0.00
0.67
0.45
0.76
0.99
0.88
0.16
0.96
0.99
0.96
0.43
0.99
1.00
0.99
0.87
0.21
0.90
0.19
0.83
0.29
0.89
0.22
* This vehicle was repaired twice. Results here are for the difference in emissions between the first and second repair.
October 2001
60
M6.IM.003
-------
Table C.2, Comparison of MOBILE6 predictions and ATL data
24 hour Real Time Diurnal
MOBILE6 predictions compared to ATL data
Average
Std. Dev
Average
Std. Dev
*
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Average
Std. Dev
Veh.
No.
070
080
061
065
071
077
063
060
069
068
078
084
066
073
075
076
081
079
064
085
062
067
072
074
082
083
Model Fuel
Year Svs.
oo
OO
88
All
83
85
85-
87
87
88
89
89
95
86
86
91
86+
All
All
83
86
86
87
All
84
85
85-
86
00
00
88
90
91
92
88
86+
All
All
All
2V
2V
Lane
Press
Fail
Fail
CARD Fail
PFI
PFI
FI
PFI
PFI
PFI
PFI
PFI
PFI
TBI
TBI
TBI
FI
FI
All
2V
2V
2V
2V
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Fail
Pass
Pass
Pass
Pass
CARBPass
PFI
PFI
FI
PFI
PFI
PFI
PFI
PFI
PFI
TBI
FI
FI
All
All
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
All
Lane
Cap
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Fail
Fail
Fail
Missing
Fail
Fail
Missing
Fail
Fail
Fail
Missing
Fail
Fail
Missing
Damaged
Fail
Fail
Fail
All
Repair Data M6 Predicted ratio: ATL/M6
grams percent grams percent grams percent
vehicle not returned
0.08
0.08
na na
8.78
3.89
6.33
3.45
13.37
14.69
0.01
0.01
na
0.78
0.61
0.69
0.12
0.92
0.92
9.09
9.09
na
15.18
15.3
15.24
0.08
4.58
4.58
0.50
0.50
na
0.68
0.69
0.69
0.00
0.49
0.49
0.01
0.01
na
0.58
0.25
0.42
0.23
2.92
3.21
0.01
0.01
1.13
0.88
1.01
0.18
1.88
1.88
vehicle not returned
19.11
-0.23
23.47
-9.52
15.25
1.31
9.68
11.25
9.01
10.09
8.20
9.94
14.75
7.19
9.98
not tested
10.64
3.82
15.16
not tested
15.16
na na
0.98
14.00
not tested
0.14
12.53
not tested
11.68
8.12
6.73
9.08
6.71
9.60
5.69
8.83
8.13
0.77
-0.07
0.91
-0.41
0.89
0.20
0.52
0.53
0.55
0.48
0.50
0.48
0.83
0.53
0.68
0.68
0.15
0.93
0.93
na
0.21
0.93
0.05
0.95
0.83
0.61
0.43
0.65
0.41
0.66
0.33
0.57
0.42
4.69
4.69
4.79
4.5
4.5
4.75
4.64
0.11
6.76
4.47
6.97
4.30
8.16
8.9
8.9
8.65
0.43
15.25
15.25
na
4.5
4.64
4.72
4.75
4.64
4.70
0.10
6.42
4.33
7.16
3.61
7.06
3.90
0.50
0.50
0.51
0.48
0.48
0.51
0.49
0.01
0.53
0.08
0.53
0.08
0.45
0.49
0.49
0.48
0.03
0.69
0.69
na
0.48
0.49
0.50
0.51
0.49
0.50
0.01
0.53
0.08
0.51
0.07
0.52
0.07
4.07
-0.05
4.90
-2.12
3.39
0.28
2.08
2.42
1.74
2.25
1.59
2.20
1.81
0.81
1.12
1.25
0.51
0.99
0.99
na
0.22
3.02
0.03
2.64
2.52
1.73
1.44
1.57
1.32
1.46
1.08
1.53
1.74
1.54
-0.15
1.79
-0.85
1.86
0.39
1.04
1.09
1.04
0.96
0.94
0.96
1.87
1.08
1.37
1.44
0.40
1.36
1.36
0.44
1.89
0.11
1.89
1.68
1.23
0.86
1.23
0.77
1.30
0.65
1.10
0.84
* This vehicle was repaired twice. Results here are for the difference in emissions between the first and second repair.
October 2001 61
M6.IM.003
-------
Appendix D EPA Response to Comments on Draft Report
During the comment period on the draft report, EPA received comments from two
organizations, Pennsylvania's Department of Environmental Protection (PA), and the American
Petroleum Institute (API). Also, prior to the publication of the draft report, EPA received
comments from the California Inspection and Maintenance Review Committee (CEVIRC)
specifically directed to evaporative emissions I/M. All the comments are summarized below in
plain text with EPA responses in italic.
Pennsylvania also has a gas cap test. Please add us to your list of states with such programs. (PA)
This has been done.
Like many other states, the only test Pennsylvania is performing for evaporative emissions is the
gas cap test. While EPA has evaluated data from a state (Illinois) that also is only performing a
gas cap test, it is significant that Illinois has a centralized program, unlike many other states which
either have hybrid or decentralized programs. (PA)
In MOBILE6, the data from Illinois and Arizona was used to generate testability and
failure rates. Testability is intended to represent the characteristics of the vehicle, and is not
expected to vary with program design. And while it would be very interesting to do a comparison
of failure rates between centralized and decentralized programs, such a study was outside the
scope of the MOBILE6 analysis.
EPA needs to revise its repair assumptions applying to gas-cap-only programs and treat them
differently than full pressure and full purge. In decentralized programs (perhaps in centralized
programs, but we can't speak to that), gas cap repairs are almost always made. That is, they are
invariably included in the amount that must be spent to obtain a waiver because the expense is so
minimal. In decentralized programs, this repair may be made before a tailpipe test is even
performed (although the final result of the full test is a "fail"). Therefore, for gas-cap-only
programs, these repairs will be almost unaffected by the existence of a waiver and should not be
discounted. (PA)
As explained in the report, the default waiver rate may be adjusted with user inputs to
more accurately represent the actual rate for the program being modeled.
Again, perhaps the same methodology should not applied to gas cap only as more comprehensive
evap tests. A gas cap failure is an all-or-nothing event and the repair is replacement. Therefore,
the age of the vehicle, once the gas cap is replaced, is irrelevant. The failure data we see in our
program does have the same shape as the Illinois program, that is, is higher in earlier model years.
However, once those caps are replaced, it's the age of the cap, not the age of the car that would be
the determining factor. It seems like the deterioration should be more of a sawtooth/stairs sort of
thing than a straight line.
October 2001 62 M6.IM.003
-------
Unfortunately, data to determine the best estimate for post-repair fail rates is hard to
come by; and, even if we did have data, developing a new methodology to properly apply this data
is not trivial. We did not have the time or resources to fully explore these options in MOB1LE6;
however, we do intend to consider using separate failure rates for repaired vehicles in developing
the New Generation Model that will follow MOB1LE6.
We appreciate EPA providing the ability to model gas cap only programs. Pennsylvania may have
increasing use for this feature in the future as discussions continue on how to maintain the
one-hour standard in growing areas as well as attain the possible eight-hour standard.
I know EPA has both less experience and skepticism about decentralized programs and data from
those programs, but modelers should keep the network design in mind all the time - ask
themselves "would this behavioral assumption still be true in a test and repair network?" (PA)
As much as possible, we have tried to make MOB1LE6 include the capacity to modify
behavioral assumptions with user inputs if better, more relevant data is available.
The methodology proposed for MOBILE6 represents an improvement relative to MOBILES in
that it is based on a more realistic estimate of the fraction of the fleet that is "testable" with
respect to the evaporative emissions control canister. (MOBILES used a 95% testability estimate
while use of the Arizona I/M lane data as described in the referenced document results in a more
realistic estimate of about 67%.)
However, we have some concerns about: (1) the proposed baseline failure rates for vehicles
subjected to the evaporative system pressure test and (2) the apparent EPA assumption that the
fuel inlet pressure test will only identify a fraction of the vehicles with pressure test defects. The
proposed baseline failure rates appear to be too high for mid 1980s to mid 1990s vintage vehicles
and the assumption regarding the fuel inlet pressure test does not seem reasonable. As described
more completely in the attached memo from Sierra Research, both of these elements of EPA's
evaporative I/M proposal for MOBILE6 may have resulted from the inappropriate mixing of
different data sets. (API).
This report acknowledges that fuel-inlet pressure tests are not an exact subset ofcannister
pressure test failures. The two tests are different and may identify different vehicles. However,
given the lack of data in this area, we believe this simplifying assumption is reasonable for
estimating I/M benefits.
Comments on the baseline failure rates are addressed in the report that discusses these
rates, M6.EVP.006.
API did not mention their contractor's comment that the methodology incorrectly
accounts for gas cap durability improvements. Sierra argues that gas cap durability improved
significantly in the mid 80s but was not improved by the phase-in of enhanced evaporative
October 2001 63 M6.IM.003
-------
requirements in the late 90s. If true, this suggests thatMOBILE6 may overestimate the benefits
of gas cap checks for vehicles built in the mid 80s to the late 90s, and underestimates the benefits
of gas cap checks for vehicles in the late 90s and beyond. Further investigation of this question
would be worthwhile for a future version of the model.
In the MOBILE6 evaporative emissions module, EPA classifies vehicles based on whether they
fail the pressure or purge test. MOBILE6 assumes that gross liquid leaks occur only among
vehicles that fail one of these tests. However, EPA's paradigm for evaporative emissions is at
odds with the results of real-world studies. For example, an Auto/Oil study of hot soak emissions
from 300 vehicles found that more than half of the excess evaporative emissions come from cars
that do not fail either the pressure or purge tests. Failure of the pressure or purge tests is thus a
poor surrogate for actual evaporative emissions rates. Furthermore, no I/M program includes the
pressure or purge tests so failure of these tests does not appear to be a relevant factor in assessing
actual I/M programs.
Several studies, including recent CRC-sponsored studies, have measured actual evaporative
emissions rates of vehicles. EPA should not continue to base evaporative emissions estimates on
an errant paradigm. Instead, EPA should simply use the real world data directly (with appropriate
attention to sample validity issues of course) to determine evaporative emission rates.(CIMRC)
MOBILE6 assumes that gross liquid leaks are independent of pressure and purge test
results and uses CRC and Auto/Oil results to estimate gross liquid leak emissions (see
M6.EVP.009). Furthermore, as stated in the introductory chapter of this report, we assume that
gross liquid leakers are not corrected with l/Mprograms. Thus, our modeling methodology is
consistent with CIMRC's comments on these topics.
October 2001 64 M6.IM.003
------- |