United States Air and Radiation EPA420-P-99-031
Environmental Protection November 1999
Agency M6.IM.003
vvEPA Estimating Benefits Of
Inspection/Maintenance
Programs For Evaporative
Control Systems
> Printed on Recycled Paper
-------
EPA420-P-99-031
November 1999
of for
M6JM.003
Assessment and Modeling Division
Office of Mobile Sources
U.S. Environmental Protection Agency
NOTICE
This technical report does not necessarily represent final EPA decisions or positions.
It is intended to present technical analysis of issues using data which are currently available.
The purpose in the release of such reports is to facilitate the exchange of
technical information and to inform the public of technical developments which
may form the basis for a final EPA decision, position, or regulatory action.
-------
Estimating Benefits of Inspection/Maintenance Programs
for Evaporative Control Systems
Section 1 Introduction
This draft report outlines our proposal for 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 loss1.
The MOBILE6 proposals for handling I/M programs for exhaust emissions are not
described here, but can be found in reports M6.EVI.001 and M6.EXH.007.
MOBILE6 will calculate adjustments to the evaporative emission estimates for the
following I/M tests applied to Light Duty Gasoline Vehicles and Light Duty Gasoline Trucks:
Check of On-board Diagnostics (OBD) II indicator light.
• Check of gas cap for presence, damage or leaks2
Pressure test from fuel-pipe inlet ("fill-pipe" tests)
• Purge test
The gas cap, fill-pipe pressure test and purge 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"
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 separate treat missing cap
emissions separately in MOBILE6. We hope to have enough data to address this issue in future
models.
November, 1999 1 M6.EVI.003
-------
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.
At this writing, we hope MOBILE6 will include the capability to model evaporative
emission credits for selective I/M programs such as remote sensing and change-of-ownership
programs. This paper proposes the methodology to be used if these features are included.
However, time constraints may make it impossible to code evaporative emission credits for these
types of selective programs in the initial release of MOBILE6. The MOBILE6 user guide will
explain exactly what features are available at the time the model is released.
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.
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 preliminary results and 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.
November, 1999 2 M6.EVI.003
-------
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.
1.2 Data
The data available on emission reductions from evaporative I/M programs is limited in
quantity, as well as in the type of data available.
1.2.1 State Data
As of July 1999, there were 19 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 has an active I/M purge test. Only Arizona and Delaware
have 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 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) 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 its appendices, the Illinois and Arizona data was used to
determine several parameters for the evaporative I/M methodology. Arizona data was 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
3AZ, CA, CO, CT, DE, DC, GA, IL, IN, ME, NM, OR, RI, TX, UT, VA, VT, WA, WI
November, 1999 3 M6.EVI.003
-------
by model year.
1.2.2 Laboratory Data
In addition to data from state programs, there are data on repair effects from an ongoing
EPA test program,4 which provides 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 collection of vehicles sampled a wide range of model years (1983-
1995), fuel delivery systems, test status groups and evaporative problems. As proposed, the
structure of the model cannot use this data directly to generate evaporative I/M repair effects.
Rather, this data is used as a check of the MOBILE6 assumptions. The data and this comparison
are presented in Appendix C.
1.2.3 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. Evaporative
emissions for a vehicle of a given age are calculated as follows:
ENoIMage = Emissions test_status x NoIM Weighting test _status^age
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
4EPA contract 68-C5-006, Automotive Testing Laboratory, Work Assignment 1-8.
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.
November, 1999 4 M6.EVI.003
-------
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 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
Model Year is required to know whether a group of vehicles is subject to the
enhanced evaporative emission standard and OBD. The phase-in of these vehicles is
described in Table 1.1.
6This is not exact since, as described in Section 4.0, the vehicle model year does not
exactly match the calendar year of sale and because 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, currently 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 and user inputs for test-
status fractions may 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.
November, 1999 5 M6.EVI.003
-------
Table 1.1
Phase-in of Vehicles With
Enhanced Evaporative Controls
and OBD*
Model Year
1995
1996
1997
1998
1999
Percentage
0%
20%
40%
90%
100%
* Phase-in schedule from 40 CFR 86.096-8
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 (age = age/2).
These vehicles also have OBD systems designed to detect failures in evaporative emission
controls. In areas without an EVI program, M6.EVP.006 assumes OBD will reduce the
incidence of new pressure or purge failure in vehicles equipped with OBD by the
percentages listed in Table 1.2. (These reductions are the same as for exhaust emissions
and are currently being tested. They may be revised before MOBILE6 is finalized.)9 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 are also available in worksheet, "from
M6.EVP.006" of the Excel Workbook, M6IM003.xls. For I/M purposes, the Gross Liquid
Leaker fraction must be removed and the weighting factors must be re-normalized to 100
percent.
9If the assumptions in M6.EVP.006 change, these changes will carry through to the I/M
calculations but will not affect the basic methodology described here.
November, 1999
M6.IM.003
-------
Table 1.2
Preliminary Assumptions: 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
1.3 Tampering and Anti-Tampering Programs
In MOBILES and previous versions of MOBILE, crankcase, hotsoak 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.10 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 is available and because crankcase emissions are such a
small fraction of total emissions, we propose that the crankcase emissions and anti-tampering
program effects on these emissions will be unchanged in MOBILE6.
However, we propose removing the tampering effects for diurnal and hotsoak emissions in
MOBILE6, as well as removing 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 is 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.
10
See AP-42, Appendix H, Tables 1.2B1, 2.2B1, 3.2B1 and 4.2B1.
November, 1999
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 = ^[Emissions teststatus x IMWeightingteststatusage
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 tests that a vehicle is affected by also depend on other factors, including the fraction of
vehicles that show up for testing (compliance), the fraction that can be tested on a given test
(testability), and the fraction of failing vehicles that actually get repaired (waivers and
compliance).
This section describes how program characteristics such as participation, testability and
observed fail rates are used to calculate new basic weighting factors IM teststatuSr age.
Section 3.0 describes how information on inspection frequency and grace periods are used
to transform the basic revised weighting factors IM test xtate 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
November, 1999 8 M6.EVI.003
-------
every combination of gas cap, fill pipe, purge and OBD test11 as follows:
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, participation 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. This fraction depends on program design. For example, are vehicles brought in
only with Remote Sensing?12 or tested only at change of ownership? Are vehicles excluded
through "Clean Screening" or "High Emitter Profiles"?
Note that applicability is not used to model the effects of periodic testing (annual,
biennial, etc.) or to model the effects of age restrictions on testing. Applicability also is not used
to describe the combination of change of ownership or remote sensing programs with periodic I/M
programs, Modeling these cases is more complicated and is described in Section 3.
Applicability is 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 0 for all previous model years.
Determining the correct applicability is straightforward for most program designs. When
modeling a change of ownership program, MOBILE6 will use the 6 month inspection frequency
input for exhaust and double it to compute annual applicability for change of ownership. To
model a remote sensing program that detects vehicles with high exhaust emissions and brings
nNot all of these combinations make sense from a policy perspective, but the model is
designed to handle them all, partly because differences in testability may reduce the kinds of tests
a vehicle actually experiences.
12Remote Sensing Device programs generally do not detect evaporative emissions. For
MOBILE 6, we assume the vehicles selected for inclusion or exclusion (clean screening) with an
RSD program are a random sample of the evaporative emission strata of that age and model year.
November, 1999 9 M6.EVI.003
-------
them in for a set of exhaust and evaporative lane tests, MOBILE6 will require user inputs of
partial applicability at six month intervals.
Our proposal for MOBILE6 is designed to model any evaporative I/M program that is
likely to be used, but there are some combinations of tests and recruitment schemes that will not
be programmed in the model. For example, MOBILE6 will not be set up to model a program
where a given model year is subject to a change of ownership program for a gas cap test, and an
RSD program for a fuel-inlet pressure. The program assumes that if the program is only partially
applicable for two or more tests, the same cars are subject to all tests. Thus the user inputs are
limited to the following options for a given model year:
1. The same full (100%) or partial applicability for all tests.
2. Full or partial applicability for one test and full applicability for the others.
3. Full or partial applicability for one test and zero applicability for the others.
2.1.2 Participation Rate
The Participation Rate is the fraction of the fleet required to be tested that actually show
up for testing. Note, this characteristic is different than "Compliance" and "Fraction Repaired"
which are discussed below. MOBILE6 will be designed to accept Participation Rate values
ranging from 0 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 participation rate is difficult to measure. We know that some 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
participation rates may be in the 80 to 90 percent range. For purposes of this report, we will use a
value of 85 percent. However, we welcome comments on this proposal, and particularly welcome
any data on participation rates for active I/M programs.
2.1.3 Testability
Testability is the fraction of vehicles that show up that can not be tested. This fraction
depends on the vehicle model year and the type of test under consideration. User inputs for
testability are allowed. Defaults are 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
November, 1999 10 M6.EVI.003
-------
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
can not 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. See table below. Note,
manufacturers recommend no fill-pipe tests for OBD-II equipped vehicles.
• Purge test-There is no I/M lane data on testability rates by model year, but we
expect that purge test testability would be similar to testability with the fill-pipe
pressure test and would show the same decline with the phase in of OBD-II
equipped vehicles. Thus, the MOBILE6 defaults will be the fill-pipe pressure test
rates described above.
• OBD check -MOBILE6 will handle the phase-in of OBD-II vehicles by computing
EVI weightings for vehicles with and without OBD-II and averaging these together
using the OBD phase-in schedule listed in Table 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 propose using the assumption that 15 percent of all
OBD systems do not successfully identify failures. Thus, the proposed default
testability rate for OBD-n vehicles is 85 percent. We plan to reconsider this value
when current studies of OBD-II vehicles are complete.
For the I/M calculations it is necessary to determine what fraction of vehicles receive what
combination of tests. In particular, we must to establish the correlation between testability
on one test and testability on another. For example, if vehicles can't 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 assumes so. In addition, we assume a high correlation of testability
between the purge test and fill-pipe pressure test since both require similar equipment such
as flexible hoses. For the portion of the fleet equipped with OBD, testability on pressure
and purge tests is independent of OBD testability.13 We assume that most vehicles that
"Pressure and purge testability on these OBD-equipped vehicles is very low, and pressure
and purge testing of these vehicles is discouraged (see "IM240 & Evap Technical Guidance,"
1998). Thus, in general, pressure and purge tests will not be applicable to OBD-equipped
vehicles. However, if a pressure or purge test program is modeled for OBD-equipped vehicles,
there is no expected relationship between pressure and purge 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/purge and OBD as independent
November, 1999 11 M6.EVI.003
-------
can be tested on purge, 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.
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.
parameters.
November, 1999
12
M6.IM.003
-------
2.2 Failure Rate
Failure rate is the fraction of vehicles that are actually tested that fail the test they are given
and are thereby eligible for repair. This varies by model year and by test (and combination of
tests).14 Failure rates for traditional I/M tests (gas cap tests, fill-pipe tests and purge 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 Purge Test Acronym
Pressure Status
Test Status
Pass Pass PP
Pass Fail PF
Fail Pass FP
Fail Fail FF
""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.
November, 1999 13 M6.EVI.003
-------
Possible In-Use Test Result
Gas Cap Fill-Pipe Purge Acronym
Test Pressure Test
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
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 a number of assumptions:
1. As stated before, gross liquid leakers are not relevant for these I/M calculations, so the no
I/M test status weighting factors are normalized without GLLs.
2. We assume that any I/M lane purge test would be the same as the "lab" purge test.
Therefore, we assume that the lane purge test failure rate would 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).15 That is, we assume that all gas cap
15The 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
difference in emissions between vehicles with and without these small leaks for hot soak,
November, 1999 14 M6.EVI.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.16
Based on the above assumptions we can construct Table 4 describing which in-use test outcomes
could be matched with which test-status groups.
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.).
16Some 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.
November, 1999 15 M6.EVI.003
-------
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./
Fail 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 test-status group
weighting factors. To do this, we make a few additional assumptions.
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. For the default case, we assume the Illinois gas cap fail rate(IL) (see Appendix A)
identifies all vehicles that fail the gas cap test.
3. For the default case, we assume the Arizona fill-pipe failure data (AZFP) (see Appendix
A) describes the distribution of fill-pipe failures.
November, 1999
16
M6.IM.003
-------
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
ny ~ 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. If the user inputs for total fraction of fill-pipe
and gas cap failures does exceed the fraction for cannister pressure failures, the user inputs
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.17
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 4. Note, the empty cells of Table 4 indicate categories for which we
predict zero vehicles.
17As explained in M6.EVP.006, we propose assuming that 15 percent of all OBD systems
malfunction; that is, 15 percent of the OBD-equipped vehicles that would fail the cannister
pressure test or the purge test do not have illuminated indicator lights. This rate will be
reconsidered based on a current study of OBD vehicles. While the faulty MIL rate is actually a
measure of false passes, in the I/M computations, the same result is obtained by treating the rate
as a testability fraction.
November, 1999 17 M6.EVI.003
-------
Table 2.4: Frequency of In-Use Test Failure/Lab Test Status Combinations
Pre-Repair Frequency as a function of observed rates
Category
PP_PPP PP
PF_PPF PF
FP_PPP FP-FP/(FP+FF) x (AZFP + IL x AZOnly)
FF-PPF FF-FF/(FP+FF) x (AZFP + IL x AZOnly)
FP_PFP FP/(FP+FF) x (AZFP - IL x (1-AZOnly))
FF_PFF FF/(FP+FF) x (AZFP - IL x (1 -AZOnly))
FP_FPP FP/(FP+FF) x (IL x AZOnly)
FF_FPF FF/(FP+FF) x (IL x AZOnly)
FP_FFP FP/(FP+FF) x (IL x (1 -AZOnly))
FF_FFF 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 (default based on Illinois data)
AZFP= Fill-pipe pressure test fail rate (default based on Arizona data)
AZOnly=Fraction of gas cap failures with fill-pipe pass (default based on Arizona Data)
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 E 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 using the phase-in schedule in Table 1.)
The default rates for AZFP, IL, and AZOnly are listed in Appendix A of this document.
November, 1999
18
M6.IM.003
-------
Alternate user inputs could be provided for the AZ, IL, and AZOnly rates. Note that any user
input for AZFP and IL must be in the form of rates that increase steadily with age (that is,
monotonically increasing sequences); otherwise MOBILE6 calculations may give unexpected or
incorrect results.
2.2.2 OBD Checks
The MOBILE6 proposal for modeling the benefits of OBD systems and I/M tests that
include checks of OBD systems was proposed in detail 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/EVI case, we use the Failure Rates (Estimates of Strata Size by
Vehicle Age-From a non-I/M Area) in Appendix E 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.18 More research is needed on the marginal benefit of combining gas cap checks with
OBD checks.
2.3 Repair Benefits—Adjusting Weighting Factors
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 "relevant" categories are. Again,
because data are severely limited, it has been necessary to make a number of assumptions. In
particular:
18Also see footnote 13 on small gas cap leaks.
November, 1999 19 M6.EVI.003
-------
1. When failed with a purge test, repair moves vehicles predicted to have purge failures from
a purge fail category to a purge pass category. For example, from "pass gas cap/pass fill-
pipe/fail purge" (PF_PPF) to "pass gas cap/pass fill-pipe/pass purge" (PP_PPP).
2. 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).
3. 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 (or gas cap and purge 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 the vehicle credit for a full pressure repair if they are subject to both
kinds of tests (in which case FP_FFP is repaired to PP_PPP, and FF_FFF is repaired to
PF_PPF, or, with an additional purge test and repair, PP_PPP). 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.)
4. 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. If future research suggests
this assumption is incorrect, the benefit can be adjusted using the fractional benefit input
as discussed in Section 2.4.
5. As explained in M6.EVP.009, gross liquid leakers are not identified by gas cap checks,
purge tests, fill-pipe tests, or OBD checks. Thus, we assume that their fraction is
unchanged by an I/M program.
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, 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.
November, 1999 20 M6.EVI.003
-------
Table 2.5: Effects of Test and Repair on MOBILE6 Evaporative IM Categories
Pre-Repair
Category *
PP_PPP
FP_PPP
FP_FPP
FP_FFP
FP_PFP
PF_PPF
FF-PPF
FF_FPF
FF_PFF
Pre-Repair
Test-Status**
PP
FP
FP
FP
FP
PF
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
fill pipe, no gas cap
other
OBD or fill pipe
other
OBD or purge
other
OBD
purge
other
OBD, or purge and gas cap
gas cap, no purge
purge, no gas cap
other
OBD, or purge and fill-pipe
fill-pipe, no purge
purge, no fill-pipe
other
Post-Repair
Category*
PP_PPP
PP_PPP
FP_PPP
PP_PPP
FP_FPP
PP_PPP
FP_PFP
FP_FPP
FP_FFP
PP_PPP
FP_PFP
PP_PPP
PP_PPF
PP_PPP
FP_PPP
FF-PPF
PP_PPP
PP_PPF
FP_FPP
FF_FPF
PP_PPP
PP_PPF
FP_PFP
FF_PFF
Post-Repair
Test-Status**
PP
PP
FP
PP
FP
PP
FP
FP
FP
PP
FP
PP
PF
PP
FP
FF
PP
PF
FP
FF
PP
PF
FP
FF
November, 1999
21
M6.IM.003
-------
Table 2.5 Continued
Pre-Repair
Category *
FF_FFF
Pre-Repair
Test-Status**
FF
Tests Experienced
OBD, or purge and gas cap and fill pipe
purge and gas cap
purge and fill-pipe
gas cap and fill-pipe
gas cap only
fill-pipe only
purge only
none
Post-Repair
Category*
PP_PPP
FP_PFP
FP_FPP
PP_PPF
FF_PFF
FF_FPF
FP_FFP
FF_FFF
Post-Repair
Test-Status**
PP
FP
FP
PF
FF
FF
FP
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.
2.4 Waivers, Non-Compliance and Fraction Repaired
"Fraction Repaired" is the fraction of failing cars that get repaired. It is reduced from one
to account for waivers and for non-compliance (vehicles that "disappear" after an initial failure,
but before a retest.)
For traditional I/M tests, the default is the same value as for exhaust programs. As for
exhaust, we assume 10 percent non-compliance and that 10 percent of compliant failed vehicles
receive waivers. Users will have the option to enter their own waiver and non-compliance 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. Any user inputs for this value for exhaust I/M will also be
used for evaporative I/M.
In addition, to allow flexibility as new information becomes available, the model will also
have a user input for "fractional benefit" for each of the tests. For example, if future information
shows that gas cap repairs actually obtain only half the difference between pressure fail and
pressure pass emissions, the fractional benefit for gas cap tests could be set to 0.5. In the code this
wouldn't actually change the benefits assigned to the gas cap repairs, but it would reduce by half
November, 1999
22
M6.IM.003
-------
the fraction of vehicles assigned to the repaired category. Use of this input for official EPA
submittals would require explicit approval from EPA's Office of Mobile Sources.
The "Fraction Repaired" will be calculated using the following equation:
FractRepaired= (1-NonComp) *FractBen[(l-W)+(W* WBen)]
Where:
NonC omp =NonC ompli ance Rate (default i s 0.10)
FractBen =Fractional Benefit (default is 1)
W =Waiver rate (default is 0.10)
WBen =Repair Benefit for waived vehicles (default is 0.20)
Thus, using default values, the Fraction Repaired is 83%. For OBD tests, we propose not
calculating the Fraction Repaired based on these parameters, but instead choosing a value
consistent with the owner response rate for exhaust emissions. Our current proposal for owner
response rate to OBD malfunction lights for exhaust emissions is a 90 percent in I/M areas. We
welcome comments on this choice for evaporative I/M estimates.
2.5 Technician Training
In MOBILE6, as in MOBILES, the evaporative I/M repair benefits assume that 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, see worksheets "sample calc" and "sample calc OBD" of the
attached Excel 97 spreadsheet, m6im003.xls. The first worksheet is designed for modeling
weighting factors for vehicles without OBD or ETP (pre-1996). The second is designed for
vehicles with OBD and ETP (1999 and later). Model year 1996-1998 should be modeled as a
weighted average of the results from the two spreadsheets.
The worksheets are designed for users to input data into the yellow cells. Users can select
a scenario from those listed in worksheets "1995 s descriptions and results" or "2025 s
descriptions and results" specify the evaluation years, vehicle ages, and program parameters
(applicability, participation, non-compliance and waiver rates). Blue cells are values that are
looked up in other worksheets. The green cells are the resulting weighting factors for that specific
age and evaluation year.
November, 1999 23 M6.EVI.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, biennial and "N-iennial"
periodic I/M, as well as change of ownership I/M (COEVI), and remote sensing (RSD).
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 once, several months ago, but
those sold in October-December of 1996 will have experienced a more recent second inspection.
Because of the distribution in 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 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 the oldest and newest cars will 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, and while we do not believe it has a significant impact
November, 1999 24 M6.EVI.003
-------
on our results (see footnote 24 in Section 4), we welcome specific suggestions of 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.
The MOBILE6 proposed 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.19 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.20 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 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.
19These 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
20As 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.
November, 1999 25 M6.EVI.003
-------
Figure 3.1
Fraction
in Test
Status
Group
General Sawtooth Method for
Evaporative
JDX=2
JDX=3
JDX=4
FRN FRC
Fraction with
(no sawtooth)
Evaluation
— Year -Model
Year+1
JDX=5
Age in Years
November, 1999
26
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)21. 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. MOBILE6 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,
triennial, etc.) the "teeth" of the sawtooth are more widely spaced. The mathematical details of
this methodology are described in Appendix B.
3.2 Selective I/M programs
As an alternative or an addition to programs that inspect the entire fleet at regular intervals
(periodic testing), areas may establish programs that select a specific portion of a model year for
more frequent or less frequent inspections. These selective I/M programs include remote sensing
programs, clean screening programs, change of ownership programs and high emitter profiling.
Each of these is described in the report on exhaust I/M (M6.IM.001).
While remote sensing may be select vehicles with high exhaust emissions, selection by
model year or age are the only commonly used programs to screen for high evaporative emissions.
These types of screening programs are modeling by varying test applicability or applying a grace
period as described in Appendix B. But a remote sensing program or change of ownership
program may bring exhaust high emitters in for an extra evaporative test or may exempt the
vehicles from a regularly scheduled test. We propose that MOBILE6 will model these such
programs as if the vehicles affected are a representative sample of the evaporative test status
groups for that age and model year.
If coding time allows, MOBILE6 will use three different methodologies to compute
evaporative emissions in three different kinds of selective I/M programs.
21The after-repair I/M weighting will actually increase for some test-status groupings. The
fraction of pass pressure/pass purge should always increase, and, 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.
November, 1999 27 M6.IM.003
-------
3.2.1 Selective I/M without Periodic Testing
The first, and simplest, methodology is for programs that do not require periodic tests, but
require only selective tests. In this case, the program is treated as an annual inspection with the
applicability parameter (see Section 2.1.1) determined by the fraction of vehicles subject to the
emission test on an annual basis. For example, for a program that required testing only at change
of ownership and used the MOBILE6 default rates for change of ownership programs, MOBILE6
would use an applicability fraction of 0.16, twice the six month default probability of eight
percent as explained in M6.EVI.001. The basic sawtooth method would be used to distribute these
tests across the model year. If users are selecting a fraction of the model year some other way
(including remote sensing for exhaust emissions), the user must enter the fraction (from 0 to 1) by
age for which evaporative system tests are applicable. The default value is zero.
3.2.2 Selective Exemptions from Periodic I/M
For periodic I/M programs that selectively exempt a fraction of a model year (for example,
clean screening and high emitter profiling), a similar method is used, except the exempted fraction
(values from 0 to 1) is subtracted from the applicability otherwise used.. MOBILE6 will not
calculate the exempted fraction, this must be estimated outside the model. Users may input the
exempted fraction by age. The default value is zero.
3.2.3 Selective I/M with Periodic Testing
Areas may want to model programs such as RSD in combination with a periodic testing
program. As for the other selective programs, users may input the fraction of vehicles receiving
an extra test by age. The default value is zero. However, the actual computations for this type of
program are the most complicated because the benefits of the selective testing must be combined
with the benefits of periodic testing.
As for exhaust I/M, we assume that selective testing does not affect the reductions
achieved with periodic testing, except to decrease the deterioration between inspections.
Effectively, we assume that the fraction of the fleet inspected with RSD or COEVI is repaired to
the test status weighting otherwise expected from periodic inspection and that the fraction of the
fleet selected for additional inspection experiences no new failures between the most recent
periodic inspection and the evaluation date.22 We also assume that, on average, the entire I/M
22This assumption may overestimate the evaporative emission benefit of selective I/M
programs since it does not account for new failures that may occur in the selected vehicles in the
time between their selective inspection and the next periodic inspection. However, this
overestimate is likely to be negligible, especially in programs with annual or biennial periodic
inspections. Furthermore, the overestimate may be met by the deterrence effect of a selective
November, 1999 28 M6.EVI.003
-------
fleet will have no more than one selective inspection in the interval between the periodic
inspection and the evaluation date.
3.3 Algebraic and Numeric details
A detailed mathematical treatment of selective I/M with periodic testing is provided in
Appendix B.
The Excel workbook m6im003.xls includes two worksheets ("Pre-OBD Annual Sawtooth"
and "OBD Annual Sawtooth") that provide example calculations of the annual sawtooth case.
These may be modified by entering alternate evaluation months, test frequencies and grace
periods.
program.
November, 1999 29 M6.EVI.003
-------
Section 4 Preliminary comparisons to MOBILES
When MOBILE6 is coded, the model will compute 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 and fuel conditions and patterns of
vehicle activity. When coding is complete, it will be easy to compare different evaporative I/M
program designs and to compare results between MOBILE6 and MOBILES. In the meantime, we
have performed a preliminary analysis to provide a general indication of how MOBILE6
evaporative I/M benefits are likely to compare to benefits estimated using MOBILES.
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.
Charts 4.1 and 4.2 illustrate the differences between MOBILES and MOBILE6 as a
function of age. Table 4.1 lists the age-weighted average of the failing weighting factors for
MOBILES and MOBILE6, and the following charts illustrate the fraction of failing vehicles as a
function of age for MOBILES and our proposal for MOBILE6. The MOBILES "with I/M" rates
were computed for a test-only program of annual pressure and purge tests with 96 percent
compliance. The MOBILE6 rates were computed for a 1995 calendar year program of annual fill-
pipe, gas cap and purge tests with 85% participation, 10 percent non-compliance, 10 percent
waivers, 20 percent benefit for waived vehicles and a one year grace period.
23"Credit for Gas Cap Check plus Purge Test," memo from Phil Lorang to Regional Air
Directors, December 1994.
November, 1999 30 M6.EVI.003
-------
Table 4.1 Age Weighted Test Status Fractions (1995 fleet)
Without I/M With I/M
MOBILES MOBILE6 MOBILES MOBILE6
Fail Purge Tests (PF+FF) 0.096 0.100 0.018 0.060
Fail Pressure Tests (FP+FF) 0.131 0.141 0.027 0.095
While these tables and charts do not compare evaporative emissions between the two models, they
do illustrate several important differences between the models.
1. The"no I/M" 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.
2. 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 our proposed MOBILE 6 approach, failures will accumulate in a subset of
vehicles despite the existence of an I/M program. We believe this is a more realistic
result.
3. 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. MOBILE6 assumes that repaired vehicles fail at the same rate as non-repaired
vehicles of the same age. In the MOBILES case, this leads to periodic dips in the "with
I/M" curve. In the proposed MOBILE6 case, our assumptions lead to a significant
decrease 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
November, 1999 31 M6.EVI.003
-------
Chart 4.1 Pressure Failures, MS vs M6,1995 Calendar Year
0.700
0.600
0.500
0.400
0.300
0.200
0.100
0.000
0.500
0.450
0.000
10
15
Age
20
25
Chart 4.2, Purge Failures, MS vs M6, 1995 Calendar Year
30
M6 no IM
m5 no IM
m6 w IM
m5 w IM
M5 delta
M6 delta
25
30
November, 1999
32
M6.IM.003
-------
the fleet at these ages is small, we believe that the MOBILE6 approach is an acceptable
simplification;24 however, we welcome comments on how to better model repeat failures.
4. As the charts indicate, these results suggest that for an annual purge, 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.
24To test the assertion that the MOBILE6 repeat failure assumption has a negligible
impact, we took the MOBILE6 "with I/M" fraction of failures illustrated in Chart 4.1 (Pressure
Failures, M5 vs M6, 1995 Calendar Year), and set the fraction of failures as a constant for ages
greater than 20. This increased the total fraction of remaining pressure failures by only a tenth of
a percent.
November, 1999 33 M6.EVI.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, draft 10/1/97, EPA420-P-99-026.
M6.EVP.002 "Modeling Hourly Diurnal and Interrupted Diurnal Emissions Based on Real-time
Data," Larry Landman, draft 5/21/98, EPA420-P-98-011.
M6.EVP.003 "Evaluating Multiple Day Diurnal Evaporative Emissions Using RTD Tests, "Phil
Enns, draft3/9/99, EPA420-P-99-003.
M6.EVP.004 "Update of Hot Soak Emissions Analysis/Terry Newell, draft 3/8/99,
EPA420-P-99-005.
M6.EVP.005 "Modeling Diurnal and Resting Loss Emissions from Vehicles Certified to
Enhanced Evaporative Standards, "Larry Landman, draft 5/17/99,
EPA420-P98-012.
M6.EVP.006 "Estimating Weighting Factors for Evaporative Emissions in MOBILE6," Larry
Landman, draft 7/15/99, EPA420-P-99-023.
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, draft 7/15/99, EPA420-P-99-024.
M6.EVP.009 "Gross Liquid Leakers," Larry Landman, draft 7/15/99, EPA420-P-99-025.
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," Trade Jackson, draft 5/3/99, EPA420-P-99-011.
M6.IM.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.
AP-42 "Compilation of Air Pollutant Emission Factors, Volume II: Mobile Sources"
November, 1999 34 M6.EVI.003
-------
pending 5th edition, updated 1998. Available at http://www.epa.gov/oms/ap42.htm
"EVI240 & 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.)
November, 1999 35 M6.EVI.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 3.4 of M6.IMP.003. 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 default rates. This appendix explains how default
values were determined for the values AZFP, IL and AZGCOnly.
Gas Cap Tests ("IL" gas cap fail rate)
For gas cap tests, the default 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-l.
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 leveling-off in fail rates that we see in the oldest vehicles.
7. Because the data from the latest model years (ages 1 and 2) includes vehicles built to the
enhanced evaporative emission standard and equipped with OBD, the regression was run
again without these 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.
November, 1999 36 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
O
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
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
1.27
0.982
* Calculated from Illinois data. "Age" equals calendar year minus model year.
November, 1999
37
M6.IM.003
-------
(/)
*J
•5
L.
£1
3
C
350000
300000 --
250000
200000
150000 --
100000
50000
Figure A-1
Illinois Gas Cap Fail Rates-Possible Fits
50
T-T-T-T-T-CNICNICNICNI
0)
o
o CM
Number Gas Cap Tests
-Actual IL Rate (%
-SPSS rates logistic (no constant)
• logistic (capped at 30)
-logistic, capped at 30, skip age 1
and 2
age
November, 1999
38
M6.IM.003
-------
Thus, to represent the fail rates of 1995 and earlier model years, MOBILE6 will use the fail rates
computed with the capped logistic regression skipping ages 1 and 2. The associated logistic
equation is:
FxlOO-
xifl
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 1.2.3. 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
25The OBD detection rate is the product of the OBD failure rate and the owner response
rate. When MOBILE6 is coded, these values may be user inputs. In any case, they should be
consistent with the rates used for exhaust.
November, 1999 39 M6.EVI.003
-------
FxlOO =
- + b,
u U
x (1-0.765)
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 0.8487 in the
equation accounts for the difference in fail rates at age 3.
F*100 =
X(l-0.085)-0.8487
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 1.27 in the equation is the difference in fail rates at age 6 due to the
owner response rate under warranty.
FxlOO =
-
-1.27
This leads to the fail rate percentages listed in Table A-2..
November, 1999
40
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.
November, 1999
41
M6.IM.003
-------
Table A-3
Phase-in of Vehicles With
Enhanced Evaporative Controls
and OBD*
Model Year
1995
1996
1997
1998
1999
Percentage
0%
20%
40%
90%
100%
* Phase-in schedule from 40 CFR 86.096-8
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 LDTls 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.
4. The Arizona fail rate was computed as the number of failures divided by the number of
November, 1999
42
M6.IM.003
-------
vehicles with successful tests for each age.
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 more failures at ages -1 and 0 than at age 1, we
will set the values for ages 0 and -1 to be the same as the value at age 1.
November, 1999 43 M6.EVI.003
-------
Table A-4
Arizona Pressure Fail Rates, MY 1981-1995, Possible Fits
Age
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Number of
Vehicles
Tested
0
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
0.013
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
November, 1999
44
M6.IM.003
-------
0.16
0.14
Figure A-2
AZ Pressure Fail Rates, Possible Fits
70000
- 60000
] Pressure tested (count)
- AZ pressure fail rate
pressure failures, quadratic fit
-pressure failures, cubic fit
pressure failures, logistic fit (no bound)
0.02
-101234567
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
non-I/M area are expected to decrease the occurrence of failures by 76.5 percent in years 1-3 and
November, 1999
45
M6.IM.003
-------
8.5% in years 3-6. Thus we will use the following equations to predict fail rates26:
For light-duty vehicles and trucks, model years 1999-and-later, age 2-3:
F =
.0115-0.0012 Nj- +0.0005 NJ-
V 2 / 12;
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 =
0.0115-0.0002 - +0.0005 -
V 2 I \ 2 >
xO.915
-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 =
0.0115-0.0002+0.0005
-0.0084
26,
OBD assumptions are subject to change. These equations will be updated as necessary.
November, 1999
46
M6.IM.003
-------
Table A-5
Fill Pipe Pressure Test Failure rates
for MOBILE6*
Age
-1
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
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.
November, 1999
47
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 are 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.27
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.
27Per email from Jeff Reeves, Gordon Darby, Inc., 11/11/98.
November, 1999 48 M6.EVI.003
-------
Table A-6 Gas Cap Only Failures as a Fraction of Gas Cap Failures
Age
-1
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
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
0.98(
0.98!
0.982
0.97S
0.97!
0.96S
0.963
0.95!
0.94'
0.93'
0.92'
0.9K
0.903
0.89(
0.87!
0.86(
0.843
0.82(
0.80*
0.78*
0.76*
0.74(
0.72^
0.701
0.67(
0.651
0.62^
0.984C
-0.001S
-0.000!
0 950^
November, 1999
49
M6.IM.003
-------
23 Mar 99 SPSS for MS WINDOWS Release 6.1
Method.. QUADRATI
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
Residuals 13
.04584293
.00205838
.02292147
.00015834
F= 144.76411 SignifF= .0000
Variables in the Equation
Variable B SEE Beta T Sig T
AGE -.001948 .003236 -.164156 -.602 .5575
AGE**2 -.000499 .000166 -.817201 -2.997 .0103
(Constant) .984878 .013561 72.626 .0000
23 Mar 99 SPSS for MS WINDOWS Release 6.1
Page 10
November, 1999
50
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 the document M6.IM.003.
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, (January = 1, July=7)28, we compute FRC,
the fraction of the year since the model year changed on October 1 as follows:
DIFF = MEVAL -10
FRC (D IFF) > 0) = DIFF/i2
FRC (D IFF) < 0 = 1 + DIFF/U
FRN = 1 - 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.
2. MOBILE tracks model year through a model year index, JDX.
JDX=Calendar Year of Evaluation -Vehicle Model Year +1
28MOBILE 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.
November, 1999 51 M6.EVI.003
-------
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 = Q,FRC < 0.25) =.0.5 x FRC
AvAge(JDX = I,FRC > 0.25) =.0.5 x FRC
AvAge(jDX = \,FRC <0.25)= FRC x[(jDX - l) + (0.5 x FRC)]
+ FRN x [(JDX - 1) - (0.5 x FRN)]
AvAge(JDX > 1) = FRC x [(JDX - l) + (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 following29:
NoIMWeighting(JDX = Q,FRC < 0.25) = NoIM[Q.5 x FRC]
NoIMWeighting(JDX = l,FRC > 0.25) = NoIM[Q.5 x FRC]
NoIMWeighting(jDX =l,FRC < 0.25) = FRC x [NoIM(JDX -I + 0.5 x FRC)]
+FRN x [NOIM (JDX - 1 - 0.5 x FRN)]
NoIMWeighting(JDX >l) = FRCx [NoIM(JDX - 1 + 0.5 x F#C)]
+FRN x [NoIM (JDX -1 - 0.5 x FRN)]
Where:
NOIMWeighting is the weighted average of the two segments of the model year,
NoIM is NoIM from Section 2.2
29This 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.
November, 1999 52 M6.EVI.003
-------
5. MOBILE6 will not actually calculate the noIM or IM 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 M6IM002.xls
Annual I/M program starting with Agel
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 (FEN). 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, fl/f 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.30
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 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.
30This 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 seems to
reach 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 methodology.
November, 1999 53 M6.EVI.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
O
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.
3.
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. If there have been no prior inspections, we calculate the average time
since the vehicle was first sold. For an annual inspection program, Delta! is calculated as
follows for each of the two segments:
DeltaTl = l-FRN/2
DeltaT2 = FRC/2
For example, for a January evaluation date, DeltaTl=0.625, DeltaT2=0.l25; for July
DeltaTl=0.875, DeltaT2=0375.
More generally:
DeltaTl = l-FRN/2 + (JDX-2-AIM1)
DeltaT2 = FRC/2 + (JDX-1-AIM2)
The deterioration D(JDX, segment) is the additional growth in the failing test status group
November, 1999
54
M6.IM.003
-------
during the time DeltaT 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 + DeltaT I)- NoIM (AIM I)
D(JDX,2) = NoIM (AIM 2 + DeltaT 2 - NoIM (AIM 2)
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[0.5 xFRC]
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 -l-0.5xFM)]
IMWeighting(JDX > l) = FRC x [lM(AIM2)+ D(JDX ,2)]
+ FRNx[lM(AIM\) + D(JDX,\)]
Where:
EVIWeighting is the weighted average of the two segments of the model year.
EVI(AIM) is the test status weighting for age at EVI (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).
AIM1 is the age at the most recent I/M test for vehicles in segment 1 (vehicles
sold between the date of evaluation and Sept. 30).
AIM2 is the age at the most recent I/M test for vehicles in segment 2 (vehicles
sold between Octoberl and the date of evaluation).
JDX is a model year index. JDX = Calendar year- Model Year +1.
D(JDX,1) is the deterioration, the increase in failing vehicles in the time since the
most recent I/M test for Segment 1.
November, 1999 55 M6.EVI.003
-------
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 >MaxAge, AIM 1 =MaxAge
If AEVI2>MaxAge, AEVI2=MaxAge
N-enniel Inspections
This section describes how MOBILE6 will compute evaporative test status weightings for
vehicles subject to biennial, triennial, and other inspections scheduled at 12*N month periods.
The methodology is very similar to that described in Appendix D of M6.IM.001 for exhaust I/M
programs. We expect MOBILE will calculate benefits only for annual and biennial programs, but
this analysis would apply to greater values of N as well.
In N-ennial IM 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.
November, 1999 56 M6.EVI.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 EVI
inspection. For the N-ennial case, AIM1 and AIM2 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
AEVI2=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.
November, 1999
57
M6.EVI.003
-------
Table B-2
JDX
1
2
o
J
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
To model a program where vehicles are tested biennially, with 50 percent tested at odd
ages (the "1,3,5" case) and 50 percent tested at even ages (the "2,4,6" case)., MOBILE6 will
compute fail rates for both cases and average them together. This is the same approach used for
exhaust emissions.
Program Start and End
An evaporative I/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 that benefits trail off as new
failures arise and are not repaired. As for exhaust I/M, we assume the benefits decline by a third
each year for three years.
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.
Selective I/M with Periodic Testing
November, 1999
58
M6.IM.003
-------
The method for modeling selective I/M in the absence of periodic testing is
straightforward and is not described here.
In periodic IM programs all vehicles are tested every N years following an initial grace
period, GPRD. In selective IM programs, vehicles are only tested if they meet certain criteria
such as a recent change of ownership (COIM) or detection as an exhaust high emitter using a
remote sensing device (RSD). Selective IM testing is often done in areas where a periodic
program is also in operation. One important difference between selective and periodic programs is
that a vehicle can be tested at any time during the year rather than just on the anniversary of its
purchase. However, for modeling purposes it will be assumed that selective EVI tests only affect
vehicles of integer and half-integer ages.
The modeling of evaporative emissions with change of ownership and RSD is very similar
to the modeling for exhaust emissions described in M6.EVI.001, except, instead of estimating an
I/M credit, we estimate a new weighting factor. For this reason, we must change the way that the
benefits of the selective program are aggregated.
1. Let PSTL(JDX,AGE) and PSTE(JDX,AGE) denote the respective probabilities that a
Segment 1 and Segment 2 vehicle with model year index JDX and age AGE
(=0.5,1.0,1.5....) will have an EVI test as a result of identification in the previous six
months by either a COEVI or RSD program. These quantities can be defined more
precisely as
PSTL(JDX,AGE) =FSTL(JDX,AGE) *STR(JDX,AGE)
PSTE(JDX,AGE) =FSTE(JDX,AGE) *STR(JDX,AGE)
where FSTL and FSTE are the Segment 1 and Segment 2 fractions of all the JDX model
year vehicles eligible by virtue of having reached the age AGE for a selective EVI test and
STR(JDX,AGE) =RSD(JDX,AGE)+COIM (JDX,AGE)
is the normalized probability that an eligible vehicle will be selected for a test as a result of
change of ownership or detection by a remote sensing device. For example, if
PSTE(JDX=3,AGE=1.5)=0.01
then 1% of all the JDX=3 vehicles will be assigned a revised test status percent
appropriate for IM tests at Age = 1.5 as a result of the fact they were all purchased new in
the same Segment 2 of JDX=3 and received an extra I/M test at the same age of 1.5 years.
2. The calculation of STR, the normalized probability that an eligible vehicle will actually be
November, 1999 59 M6.EVI.003
-------
selected for a selective I/M test differs for remote sensing and change of ownership. It is
discussed in detail below. For COIM, we use the same value as for exhaust. The value is
independent of age. The default value is eight percent every 6 months. For RSD, users
must enter the 6 month probability that an eligible vehicle will be brought in for an
additional evaporative test. This may vary by age (JDX).
3. To calculated PSTL and PSTE, we must calculated FSTL and FSTE, the fractions of the
model year that are eligible for a selective IM test in which 6 month age bin. Selective
IM tests benefit vehicle emissions if they take place in the interval between the most
recent periodic I/M inspection and the evaluation date. For Segment 2 vehicles:
If AIM +0.5 < AGE < JDX-1 and AIM < JDX-1
FSTE(JDX,AGE) = FRC
If FRC>0.5 then a fraction
FSTE(JDX,AGE) = FRC-0.5
of the JDX model year vehicles will also be eligible for one additional test at age
JDX-0.5.
The arguments pertaining to the Segment 1 vehicles are similar.
For AIM+0.5 0.5 then only a fraction
FSTL(JDX,AGE)=0.5
of the JDX model year vehicles will be eligible.
In a biennial, triennial or other testing program with testing frequency greater than one
year, a vehicle has multiple opportunities for a selective I/M test in the interval between
periodic tests. The probability CPSIME(JDX,AIM) that a vehicle purchased new in the
FRC segment of the JDX model year receiving a selective JJVI test between its previous
periodic EVI test date at age AIM and the emissions evaluation date at the end of the FRC
November, 1999 60 M6.EVI.003
-------
segment of the JDX=1 model year is equal to the sum over PSTE(JDX,AGE) for all the
values of AGE satisfying equations described for Segment 2, above. This is given by
"•* B
CPSIME(JDX ,AIM) = ^[PSTE(JDX,AGE = AIM +0.5*M)]
M=l
where ME=2*(JDX-AIM-0.5) is the maximum number of possible selective IM test dates.
The arguments for vehicles in the FRN model year segment are similar with all the
possible values of the AGE variable described for Segment 1 above. This leads to the
probability
CPSIML(JDX,AIM) = ^[PSTL(JDX,AGE = AIM +0.5* M)]
with ML=2*(JDX-AIM-1.5).
It is a further requirement that vehicles cannot receive the benefits from more than one EVI
test. Consequently, the values of CPSEVIE and CPSIML cannot exceed FRC and FRN
respectively.
The arguments in this section up to here have been quite formal. It is instructive to
consider an example. Consider the case of a biennial IM program with selective IM testing
and a grace period of zero years. Let us set JDX=4 and select a vehicle that was bought
new in the FRN model year segment (that is, between the evaluation date and Sept. 30).
Table B-2 indicates that this vehicle will have received its previous periodic EVI test at age
AIM=1 year. The probability that this vehicle will be tested as a result of the selective EVI
program is
CPSIML(JDX=4,AM=1) = PSTL(JDX=4, AGE=1.5) + PSTL(JDX=4, AGE=2.0)
+ PSTL(JDX=4, AGE=2.5)
By contrast, the vehicles belonging to the FRC segment of the JDX=4 model year segment
are all old enough to have received their second periodic IM test at age AEVI=3. In this
case
CPSIME(JDX=4,AM=3) = PSTE(JDX=4,AGE=3.5)
where this expression contains only one term because a relatively short period elapses
between the date when the vehicles receive their periodic EVI test and the emissions
November, 1999 61 M6.EVI.003
-------
evaluation date.
5. Each PSTE(JDX,AGE) term in gives the probability that a vehicle bought new in the late
segment of the JDX model year will receive a selective EVI test at age AGE. For
simplicity, we will assume that the selective IM tests bring those vehicles tested and
repaired to the "IM" fail rate at the evaluation date. In effect, we remove the deterioration
that would occur between periodic inspections for the fraction of the fleet that was subject
to a selective I/M test.
IMWeighting(JDX = Q,FRC < 0.25) = NoIM[Q5 XFRC]
IMWeighting(JDX = l,FRC > 0.25) = {(l-CPSIME)X NoIM[Q5xFRC]} + [CPSIME X NoIM(Q)]
IMWeighting(jDX = l,FRC < 0.25) = FRC X [lM(AIM2) + D(JDX ,2)]-[CPSIME XD(JDX,2)]
+FRN X [(1 - CPSIML) X NoIM(JDX - 1 - 0.5 X FRN) + CPSIML X AWM(O)]
IMWeighting(JDX > l) = FRC X [lM(AIM2) + D(AIM2,FRC)]-[CPSIME XD(JDX,2)]
+FRN X [iM(AIMl) + D(AIMl,FRN)]- [CPSIML X D(JDX,1)]
November, 1999 62 M6.EVI.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. This is part of an ongoing program and additional data will be available in the coming year.
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. When the MOBILE6 work on
running loss and hot soak emissions is further along, similar comparisons could be made for the
effects of evaporative I/M on those emission modes.
31A model-year split at 1985 was considered to account for the effect of cumulative
improvements in evaporative emission control during the 1980s.
November, 1999 63 M6.EVI.003
-------
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:
1. First, we have limited time, and don't 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 welcomes comments on whether the proposed MOBILE6 evaporative EVI 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.
November, 1999 64 M6.EVI.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.
>ya
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 Sye.
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
OO
88
90
91
92
88
86+
All
All
All
2V
2V
Lane
Pregg
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
0.52
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.
November, 1999
65
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.
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
oo
OO
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
CARB Pass
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
arams percent arams percent arams 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.
November, 1999 66
M6.IM.003
------- |