Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 118
5
Estimating Inspection and Maintenance
Emissions Reductions Using the
MOBILE Mode!
The U.S. Environmental Protection Agency (EPA) mobile-source emissions
factor (MOBILE) series of computer models historically have been used by
state and local air-quality planning agencies to estimate emissions benefits of
inspection and maintenance (~/M) programs. In this chapter, the regulatory
context of EPA's MOBILE program is discussed. Comparisons of I/Mpro-
gram evaluation data with MOBILE predictions are then provided, followed
by a detailed explanation of proposedprocedures for estimating I/M program
effects in MOBILES, EPA's latest version of the model. The chapter also
includes abrief~iscussion of Californians model for estimating on-roadmobile-
source emissions and I/M program effects, the EMFAC model.
USE OF MOBILE IN REGULATORY APPLICATIONS
The Clean Air Act and its amendments require that areas that have not
met the National Ambient Air Quality Standards (NAAQS) develop state
implementation plans (SIPs) that describe how they will attain compliance.
The ~ 990 Clean Air Act Amendments (CAAA90) prescribe minimal control
measures and attainment dates, depending on the severity of the NAAQS
exceedance. Among other things, these SIPs must contain three main items:
118
OCR for page 119
Estimating I/M Emissions Reductions Using the MOBILE Model ~19
(~) a detailed and comprehensive current-year emissions inventory; (2) a
detailed and comprehensive future-year (for the prescribed attainment year)
emissions inventory forecast using federal, state, and local emissions-control
programs; and (3) an analysis offuture-year air quality showing attainment of
the NAAQS by photochemical modeling. ~ To ensure that emissions reductions
are occurnng, SIPs must also specify emissions targets for every third year
toward the attainment year, and so-called rate-of-progress inventories must
then be submitted to EPA.
A second legislative requirement in the CAAA90, known as conformity,
prohibits transportation projects if they impede progress toward meeting emis-
sions targets and attaining the NAAQS. Forthe projects to proceed, metropol-
itan planning organizations (MPOs) must evaluate the emissions effects of
transportation plane, projects, end programs, andpass a conformity demonstra-
tion with the U. S . Department of Transportation. Conformity is demonstrated
if mobile-source emissions that are forecasted to result from transportation
plans, programs, end projects do not exceed mobile-source emissions budgets
established in the SIP . Conformity lapses if it cannot be demonstrated that the
SIP mobile-source emissions budget will not be exceeded, or if 3 years have
passed since the last conformity demonstration. During a conformity lapse,
projects that are already under construction can proceed, but new projects
requiring federal funding or approval cannot be advanced until the conformity
lapse has been remedied.
For both of these applications, states and regions outside California use
EPA's MOBILE emissions factor mode} for estimating emissions and emis-
sions reductions from mobile-source control programs such as I/M (California
has its own emissions factor model, EMFAC).2 EPA introduced the first
version ofthe model, MOBILE I, in ~ 978. Since then, there have been a series
of mode} revisions with changes to modeling assumptions, methods, and the
ways changes in the vehicle fleet are accounted for (e.g., with adoption of new
emissions standards and other federal control programs). Many ofthe mode!
revisions have incorporated data from testing programs that were designed to
Carbon monoxide SIPs can use rollback modeling, which assumes that reductions
in emissions produce a directly proportional reduction in pollutant concentrations
(above background levels), to demonstrate future-year attainment.
2Although states are not mandated by any law or regulation to use MOBILE, SIPs
developed with some other mobile-source emissions model would not be accepted by
EPA (except for California, which must use the EMFAC model for their SIPs).
OCR for page 120
120 Evaluating Vehicle Emissions I/M Programs
assess characteristics of vehicle emissions that previously had been ill charac-
terized or were underestimated. For the past several years, EPA has been
working on the most significant mode] revision in its history. The new model,
MOBlLE6, is expected to be released in 200 ~ for use in regulatory applica-
tions. The recent National Research Council (2000) report and the Holmes and
Russell (200~ review of MOBILE describe the history of the model's revi-
sions and provide more details about the uses and implications of MOBILE as
a regulatory emissions modeling tool.
In the SIP process, MOBILE is used to estimate what are referred to as
SIP credits. States use the model to estimate the emissions reduction in a
future year with implementation of an I/M program (or changes to an existing
I/M program). These SIP credits based on MOB TEE are only an estimate of
the real emissions reductions. Actual emissions reductions from an I/M pro-
gram can be measured only with real data from vehicles that have and have
not been through the program. SIP credits are very important to states, be-
cause if they do not accumulate enough credits to demonstrate future-year
attainment, they can be penalized economically by withdrawal of federal trans-
portation funds and limitations on new construction requiring environmental
permits. On the other hand, if states claim too much credit for I/M and the
emissions reductions are not fully realized, then progress toward attaining clean
air standards is hindered.
It should be noted that MOBILE estimates emissions factors in grams per
mile by vehicle class (e.g., passenger cars, light-duty trucks, and heavy-duty
diesels). To estimate or-road mobile-source emissions, these emission factors
are then multiplied by estimates of vehicle miles traveled (VMT) by vehicle
class. In most urban areas, VMT estimates are derived from transportation
demand models. This chapter addresses issues in the MOBILE estimates of
I/M program effects. There are just as many issues and problems in the esti-
mation of VMT, but coverage of these issues is outside the scope of this re-
port.3
MODEL PREDICTIONS COMPARED WITH
PROGRAM EVALUATION DATA
As discussed above, one of the more important uses of MOBILE is for
states to generate SIP credits for an I/M program to be implemented in a
Wee EPA (1992c) for guidance on development of VMT forecasts.
OCR for page 121
Estimating I/MEmissions Reductions Using the MOBILE Model 121
future year. In the ~ 992 enhanced I/M regulatory impact analysis, EPA esti-
mated that enhanced I/M would reduce light-duty vehicle (LDV) exhaust
hydrocarbon (HC) emissions by 28%, carbon monoxide (CO) emissions by
3 ~ °/O, and nitrogen oxide (NOX) emissions by 9°/O by the year 2000 from a non-
I/M fleet (EPA ~ 992b). This prediction was made with version 4. ~ of the
model. EPA's predicted emissions reductions for enhanced I/M using
MOBlLE5, released shortly afterward, were likewise overly generous. Table
5-] shows MOBlLE5b predicted reductions in emissions estimates from the
non-~/M case for light-duty gasoline vehicles (LDGV, passenger cars) for
calendar years ~ 995 and 2000 under various I/M scenarios. The table shows
the expected increased emissions reductions with more advanced testing, with
the largest reductions occurring for the biennial IM240 with technician training.
Predicted emissions reductions for calendar year 2000 are larger than for
calendar year ~ 995, primarily because the base emissions (in the non-~/M
case) are smaller in future years with fleet turnover.
There have been only a few comparisons of emissions reductions esti-
mated from program data or remote-sensing measurements to MOBILES
predictions. These comparisons are shown in Table 5-2 for several I/M pro-
grams across the counky; evaluations for most ofthe I/M programs listed in
the table were discussed in Chapter 3. Ofthe studies referenced in Table 5-2,
the analyses ofthe Arizona IM240 program are arguably the most detailed and
rigorous; these analyses show slight overpredictionsbyMOBlLE5 of CO and
HC emissions reductions and significant overprediction (by a factor of 2) of
NOx reductions. Analyses of ColoradolM240 date also show signif~cantover-
prediction of IM240 effects. Such overpredictions ofthe effectiveness of I/M
programs hinder progress toward achieving air-quality goals, as states are
granted too much SIP credit for planned I/M programs and therefore do not
enact additional needed controls.
MOBlLE6 was not available to the committee during most ofthe commit-
tee's work. However, the draft MOBlLE6 model, just released, shows deteri-
orationrates signif~cantlylower than thoseinMOBlI E5. Figure 5-l compares
VOC and NOX emission rates in MOBILES and in draft MOBlLE6 with and
without the effects ofthe Tier 2 and 2007 heavy-duty rulemakings. The figure
shows that emission rates in draft MOBlLE6 are significantly higher in past
and current years and significantly lower in future years (after about 2005~.
If the emissions deterioration rates are closer to reality in MOBlLE6 than in
MOBlLE5, this could be a major contributing factor to the MOBlI~E5 overesti-
mation of I/M effects. EPA has been criticized in the past for overly pessimis-
tic assumptions on deterioration rates for 198 ~ and later vehicles (see, e.g.,
OCR for page 122
122 Evaluating Vehicle Emissions I/M Programs
TABLE 5-l MOBlLE5b Predicted Exhaust Emissions Reductions for
LDGVs in ~ 995 and 2000 for Various I/M Programsa
Year 1995
Year 2000
CO HC NOx CO HC NOx
(%) (%) (%) (%) (%) (%)
Idle, annual 17.9 17.8 0.8 18.8 19.1 1.2
Idle, biennial 14.9 14.5 0.8 16.8 16.9 1.3
Idle/2500, annual
Idle/2500, biennial
Loaded idle, annual
Loaded idle, biennial
IM240 (1.2/20/3), biennial,
without technician training
IM240 (1.2/20/3), biennial,
with technician training
Acceleration simulation
mode 2525/5015 (25,50,1),
biennial
25.5
21.7
23.7
20.2
32.8
39.6
33.6 30.7
22.3
18.6
22.2
0.6
0.6
0.6
18.5 0.6
31.0 13.1
29.0 24.8
26.0
26.0
23.3
36.8
36.3 19.5 45.4
22.1
24.6
21.9
32.4
39.2
1.0
1.1
1.0
1.1
24.8
19.6 41.9 39.4 24.4
aFleet average grams-per-mile emission factors with I/M relative to non-I/M.
Note: All MOBILESb runs used default fleet mix and registration distributions, 19.6
mph average speed, 75°F temperature, 8.7 pounds per square inch RVP, no RFG or
oxygenate, and default operating fractions. All I/M programs were assumed to start
in 1992, 20% stringency, 0% waiver rates, 100% compliance, test only, centralized.
Sierra Research ~ 994a). With these Tower emissions rates for future years in
MOBlI-E6, as shown in Figure 5-l, the I/M credits are likely to be lower in
MOBILES than in MOBILES. Early indications are that MOBlLE6 will indeed
reduce the emissions-reduction benefits from I/M compared with MOBlLE5
(Clean Air Report ~ 999~. There are serious policy implications if MOBlLE6
SIP credits for I/M programs are significantly lower than MOBlLE5.
One indication of I/M effectiveness in MOBlLE6 compared with evalua-
tion of benefits using program data can be gleaned from the most recent audit
of the Colorado I/M program. In this audit, EPA's Serious Area CO Mode!
was used to estimate the benefits of the state's I/M program. The Serious
Area CO Mode! is a version of MOBlLE5 that has some of the key features
of MOBILES forCO emissions, includingiower deterioration rates. The 1999
OCR for page 123
Estimating I/MEmissior~s Reductions Using the MOBILE Model 123
TABLE 5-2 Estimated Emissions Reductions Attributable to I/M As a
Percent of MOBlLE5 Predictionsa
HC
(%)
NOx CO
(%) (%)
Phoenix, AZ: Centralized IM240
Random sample of 1995 program data (EPA
1997a)
All 1996-1997 program data with fast-
pass/fast-fail converted to estimated full
IM240 (Wenzel l999b)
Random sample of 1996-1997 vehicles given
full IM240 (Wenzel l999b)
Colorado: Centralized biennial IM240
All 1997 program data (ENVIRON 1998)
Idle
IM240
Remote sensing in 1989 in I/M and non-I/M
areas (Zhang et al. 1996b)
Atlanta, GA: Decentralized idle (BAR97)
Comparison of remote-sensing measurements
in I/M vs. non-I/M areas (Corley and
Rodgers 2000)
Minneapolis, MN: Centralized annual idle
Comparison of ambient CO concentrations
(Scherrer and Kittelson 1994) with
MOBILESb fleet reductions (O'Connor et al.
85 46 100
83 43 90
89 46 83
86-103
76-93
76-84 3-6 105-121
21
Cars, 209
Trucks, 72
14
aBoth program and MOBILES estimates are fleet average grams per mile emissions.
Colorado audit estimated an 8% reduction in CO emissions for the IM240
program, compared with ~ 7°/O for the Serious Area CO Mode! (Air Improve-
ment Resource 19994.
MOBILE AM INPUTS
To obtain emissions factors from MOBILES, including credits for an I/M
program, the user provides three types of input (~) program descriptive inputs,
OCR for page 124
124 Evaluating Vehicle Emissions I/M Programs
All Mighty Vehicles
Viable ~ ~ ~C) Edit
(Ir~udingEvapor~ive and Exhaust Enissions)
6-
5
4- _
'\\
-
~ 3
1 FF;
MOEl1 Fly
~~ MOBI1 F`;wffl
rer2&mrule
.
., . \
it_
z
. x
~~;a
1
o-
1990 1996 2~100 2005 21)10 2D15 ID ~ 2Do
Yea
FIGURE 5-1 VOC and NOX emission factors in MOBILES and draft MOBILE6. The line
rule-makings. Source: Beardsley2001.
(2) program effectiveness input, and (3) fleet characterization inputs. The
model estimates emissions credits for the effects of up to five I/M programs
specified by the user. For example, if an area has two-speed idle testing for
older vehicles and IM240 testing for newer vehicles, then the user provides
program specifications for each of these two types of I/M programs.
The following I/M program descriptive inputs must be provided to
MOBlLE6:
Program start year (calendar year when program begins).
OCR for page 125
Estimating I/MEmissions Reductions Using the MOBILE Model 125
45
4
&5-
3-
All Hgh~yVehid"
~d" ~ nitrogen (~) him
, _ M - LE6 ~
~ ' ~ ! Fit;
i.\, I
An.` I
25
Q
In
2
~5
1 -a ~LE6~ffl
ma..
Tier 2 & [D rule
Fib 1
all\
An\
\ \
.. ..
>.
=~ ~ I.
1990 1995 2000 ~ ;Ino 2tn5 202D ~ 2D30
Ca er - Yea"
marked "MOBILE6" is draft MOBILE6 without the effects ofthe Tier 2 and 2007 heavy-duty
· First (earliest) and last (latest) model years of vehicles subjectto the
requirements of the program.
· Model years exempted from the program.
· Test type (idle, 2500/idle, acceleration simulation mode (ASM), IM240,
and on-board diagnostics (ODD)).
· Program type (inspection only, inspection and repair (computerized or
manual)).
· Frequency of inspection (annual, biennial, change of ownership).
· Vehicle classes covered (LDGV, light-duty gasoline truck (I DGT)
weight classes, heavy-duty gasoline vehicle (HDGV) weight classes).
OCR for page 126
126 Evaluating Vehicle Emissions I/M Programs
· Cutpoints for HC, CO, and NOX for IM240 testing.
· Remote-sensing parameters.
The user is required to provide three MOBILE inputs related to program ef-
fectiveness: compliance rate, waiver rate, and stringency level. This section
provides brief discussions of these parameters, including what they are and the
common methods used by state agencies to determine their appropriate values.
Compliance Rates
Compliance rate, typically the most important ofthese parameters in terms
of emissions reductions, is defined as the level of compliance with the inspec-
tion program. However, compliance is a difficult concept, and it is not clear
that EPA and the states have used a consistent definition or measurement of
it in the past. Figure 5-2 shows a conceptual classification of vehicles in an
area that has anI/M program. There are fourtypes of noncompliantvehicles:
~ ~ ~ those that are not registered, (2) those that avoid the program by registering
outside the area,4 (3) those that are registered but never take the test, and (4)
those that take the test and fait but never complete the test cycle with a pass-
ing test.s EPA guidance for MOBILE 5b (EPA ~ 997b) stated that the compli-
ance rate specified should include all registered vehicles that successfully
complete an [/M cycle, including both passing and waived vehicles. This
definition includes only one ofthe four types of noncomplying vehicles shown
in Figure 5-2, and in the past, states have tended to estimate the compliance
rate as the proportion of registered vehicles that actually take an I/M test and
thus underestimate the true noncompliance rate and overstate the I/M benefits.
In MOBlLE5, there is no default value for the compliance rate; it must be
specified in the input file. However, EPA asks states to provide documenta-
tion if the compliance rate is over 96%. As a result, this 96% value has be-
4An analysis of Dayton, Ohio, area registration statistics showed that when the
enhanced I/M program was implemented, registrations decreased by 10% in the coun-
ties in the I/M program and increased by a similar amount in the surrounding non-I/M
counties (McClintock l999b).
5Two other categories of vehicles could be considered in noncompliance: those
that receive inadequate or ineffective repairs; and those that pass the inspection be-
cause of emissions variability, so they are never repaired. However, because both of
these end up passing the test, they are not included in Figure 5-2.
OCR for page 127
Estimating I/MEmissions Reductions Using the MOBILE Model 12 7
All vehicles
driving in the area
Registered
Take initial test
Pass
~ '1
rail
Repair and pass
Get waiver
(possibly some | | Scrap
repair)
FIGURE 5-2 Classification of vehicles in an area subject to I/M. Shaded boxes indi-
cate noncomplying vehicles. Source: Adapted from Hamngton et al. 1998.
come a de facto default, because states could claim up to 96% compliance
without any documentation.
Similarly, MOBlLE6 defines noncomplying vehicles as "vehicles which
show up for the initial test, but drop out of the process prior to a successful
passing result or a waiver" (EPA 1 999e). However, EPA now recognizes a
second type of noncomplying vehicle~ne that does not show up for its initial
test—and says that the input compliance rate should take these vehicles into
account (EPA ~ 999e). For MOBTLE6, there is no default noncompliance rate;
the rate must be specified in the input file. However, it is unclear what sort of
documentation will be required by states in their SIPs to justify the rate that is
used.
Although improved over MOBILES, the MOBlLE6 definition of compli-
OCR for page 128
128 Evaluating Vehicle Emissions I/M Programs
ance rate remains an underestimate ofthe true noncompliance rate, as it does
not include unregistered vehicles and vehicles that avoid the program by regis-
tering out ofthe area. States will likely continue to estimate the compliance
rate based on the number of registered vehicles, as there is no incentive to
measure the proportion of vehicles that are unregistered or that are registered
outside the region. Additionally, the registered fleet might not always accu-
rately account for scrapped, moved, or change of ownership vehicles that are
no longer in operation in the region. License-plate reading as part of remote-
sensing measurement programs will help allow estimates of both of these
components ofthe fleet being driven in the region. Once the total vehicle fleet
is known, it is relatively easy to determine the complying vehicles from pro-
gram data the number of vehicles that get tested and are either repaired or
waived are complying. Some ofthese might have been incorrectly or fraudu-
lently passed, and remote sensing could also help identify them.
Waiver Rate
Waiver rate refers to the fraction of vehicles that fail their initial tests but
were never fully repairedbecause the repair cost limit (or some other criterion)
has been met; these vehicles have complied with the program requirements but
are still failing vehicles. This parameter is discussed in Chapter 3. in the
model, separate waiver rates are used for pre-1981 and post-1980 LDVs.
These rates must necessarily come from I/M program records. EPA recom-
mends that, for historical inventory development, program-specific data be
used to derive the waiver rates. For future inventories, the historical rates may
be used.
Stringency Rate
Stringency rate (or failure rate) is the expected failure rate for pre-1981
model-year vehicles.6 Stringency rate is defined as the test failure rate ex-
6Stringency rate is input only because the older model years use a methodology
(from the late 1970s) that calculates benefits based on failure rates rather than test
procedures and cutpoints. For newer vehicles, MOBILE6 uses identification rates
based on the proportion of total emissions from failing vehicles (not number of failing
vehicles) identified.
OCR for page 135
Estimating I/MEmissior~s Reductions Using the MOBILE Model 135
~ 999a). Although only a small number of vehicles (22~7 in the remote-sensing
data set were matched to the IM240 program records and there are issues
that complicate the analysis (e.g., comparing emissions reductions from IM240
tests with remote-sensing measurements) these results indicate that the
MOBlLE6 repair-effectiveness rates might be overstated. Chapter 7 contains
a discussion of emissions-repair studies.
The waiver vehicles are those that still fail the emissions test after a set
minimum amount was spent on only partially successful repairs. Although not
fillly repaired, they are assumed on average to have some repairs. The pro-
posed default for MOB TLE6 is that the waived vehicle emissions rate is 20%
less than the failed vehicle emissions rate. This default is an assumption, as
there was no available analysis of emissions from waived vehicles from oper-
ating IM240 programs at the time EPA prepared the draft MOBILES I/M
credits (EPA ~ 999e). Figure 5-5 shows an example ofthe rates estimated for
repaired and waived vehicles, for the same ~ 990- ~ 993 PET LDGV example as
in Figure 5-4. Note that the change in emissions as a function of mileage for
the repaired vehicles is not the deterioration rate for the repaired vehicles;
rather, these emissions rates are used to determine the I/M credit at a given
age. Note also for this example that the EPA methodology results in the re-
paired vehicle emissions multiple ofthe normal emissions rate increasing from
0 to about ~ 00,000 miles and then decreasing until the repaired and normal-
emitter rates are the same at about 170,000 miles. After 170,000 miles, the
calculation actually results in repaired vehicle rates being less than normal-
emitter rates, but they are set to the normal-emitter rates.
T4entif~cation Rate
The high-emitter identification rate (IDR) is the proportion of emissions
from high emitters in the fleet that are correctlyidentified. If the cutpoints are
set so that all high emitters are properly identified (i.e., fait the test), then the
IDR is ~ 00%. The TDR depends on the test method used (IM240, ASM, idle
testing) and the test cutpoints (but not model-year group or vehicle technol-
ogy). The lower the cutpoints, the higher the IDR; however, lower cutpoints
also increase the chances that normal emitters will fail the test. To estimate
7In general, a large sample of vehicles should be measured repeatably using re-
mote sensing to help establish emissions trends and repair effectiveness.
OCR for page 136
136 Evaluating Vehicle Emissions I/M Programs
IDRs for MOBILES, EPA used a database of 9 ~ 0 model-year ~ 98 ~ and later
cars and trucks that had both an IM240 test and an FTP test from EPA emis-
sions factor testing in Ann Arbor, Michigan, and Hammond, Indiana, and also
Arizona data on randomly recruite~vehicles. These identification rates, used
to calculate the average emissions ofthe fleet after a cycle of I/M testing and
repair (described below), are estimated from a regression analysis ofthe Toga-
rithms of the test cutpoints (cut) as follows (EPA 1999e):
HO IDR = ~ .145 ~ - 0. ~ 365 x In(HCcut) - 0.1069 x In(COcut)
CO IDR = I . ~ ~80 - 0.1073 x In(HCcut) - 0.1298 x In(COcut) (5-2)
NOX IDR= 0.5453 + 0.7568 x NOcut - 0.3687 x NOcut2 + 0.0406 x NOcut3
Another method that states commonly use to increase identification of
failing vehicles is to require a passing inspection for change of vehicle owner-
ship. This method can increase the fraction ofthe fleet that is inspected and
increases the likelihood of failure identification and repair. In the draft
MOBlLE6 documentation, change of ownership is assumed to be a fixed
fraction of the fleet based on an analysis of Wisconsin data, but there have
been comments suggesting that this be a user input to reflect the actual change
of ownership rates in an individual area.
Noncomplying Vehicles
As described above, the compliance rate input to MOBlLE6 is assumed
to represent vehicles that fail the initial test and do not complete the testing
process, obtaining either a passing test or a waiver, and also those vehicles that
do not show up for the required I/M testing. Although some of the no-show
vehicles could be normal emitters, the draft MOBlLE6 documentation indi-
cates that EPA considers all noncomplying vehicles as high-emitting vehicles
that are unaffected by the I/M program, and the input compliance rate should
be set with this understanding. The high emitters, therefore, consist ofthree
types of vehicles: (~) the identified high emitters that are repaired (but with
emissions rates higher than normal emitters), (2) the identified high emitters
that are partially repaired and receive waivers, and (3) noncomplying high
emitters.
OCR for page 137
EstimatingI/MEmissions Reductions Using the MOBILE Model 137
This treatment of noncomplying vehicles is different from the MOBILES
assumptions in severalways. InMOBlLE5, the noncompliance rate is defined
as a share ofthe fleet as a whole; noncomplying vehicles are assumed to have
higher emissions than normal vehicles. In MOBlLE6, noncompliance is part
of the high-emitter fraction only, and although MOB TLE5 assumes that the
failure rate ofthe noncomplying vehicles is higher than that ofthe complying
vehicles, it did not assume that all of them are high emitters as is the case for
MOBILE6.
Average Emissions after I/M
Average emissions after I/M are defined for each vehicle class, model-
year group, and technology type from a combination of normal emitters, re-
pairedvehicles, waived vehicles, end high emitters not repaired (either because
I/M failed to identify them or because they are noncomplying). Table 5-3
shows the five subsets of vehicles that contribute to the average and the
weighting factor for each subset. Once the average emissions rate after I/M
has been calculated for each vehicle class/model-year group/technology-type
combination, then sales weights are used to calculate the fleet average emis-
sions after I/M across all LDVs in the fleet.
A significant problem with the draft EPA methodology is that the IDR has
been defined es the fraction of emissions from the identified high emitters, yet
this same IDR is used as FID in Table 5-3, which is supposed to represent the
fraction of high-emitting vehicles identified. Because the distnbution of emis-
sions from high-emitting vehicles is so skewed, the infraction of emissions from
high emitters identified is substantially greater than the fraction of high-emitting
vehicles identified. Using the estimated IDR for Fain the estimation of aver-
age emissions after I/M results in an overestimate for the I/M credit (i.e.,
estimated average emissions after I/M are too low).
Application of the I/M Credit
The I/M credit in MOBlI E6 is the difference in estimated emissions
before and after I/M. Emissions before I/M are the basic emissions rates
described previously, and average emissions after I/M are calculated as the
weighted average across subsets of vehicles as shown in Table 5-3.
OCR for page 138
138 Evaluating Vehicle Emissions I/M Programs
TABLE 5-3 Calculation of Average Emissions after I/M
Vehicle Subset
Weighting Factor
Emissions Rate
Normal emitters, no change in emissions after
I/M
High emitters not identified by I/M, no
. . .
c range in emissions
Noncomplying high emitters, no change in
. .
emlsslons
High emitters identified and given cost
waivers, some repair below high-emitter level
High emitters identified and successfully
repaired
1 - - H
FH * (1 FID)
FH X FID X FNC
FH X FID X FW
FH X FID X FR
.—N
EH
EH * 0.80
ER
Note: The average for each vehicle-class, model-year, technology group is the
weighted average emissions rate across five subsets of vehicles.
FH = fraction of high emitters before I/M
FID = fraction of high emitters identified
by I/M
FNC = fraction of identified high emitters
. , .
In noncompliance
FW = fraction of identified high emitters
. .
given a waiver
FR = fraction of identified high emitters
fixed
FNc+Fw+FR= 1
EN = emissions rate for normal emitters
EH = emissions rate for high emitters
EW = emissions rate for waiver vehicles
ER = emissions rate for repaired
vehicles
ER 2 EN by constraint
MOBILES models the effect of I/M as a reduction in emissions at the time of
inspection; this is referred to as the I/M credit. Emissions are modeled to
increase between inspections at the same deterioration rate as vehicles not
subject to an I/M program. This results in the so-called "sawtooth" pattern
(also the basis for I/M credits in MOBILES) shown schematically for a biennial
program in Figure 5-6.
For an annual program, the I/M credit is calculated and applied once per
year, and there is half the time for vehicle deterioration before the next test
cycle. MOBILES, the current regulatory model, has a very small increase in
the emissions reduction in I/M benefits for an annual enhanced ]/M program
instead of a biennial program—only a 2-6% increase in emissions reductions,
OCR for page 139
Estimating I/MEmissions Reductions Using the MOBILE Model 139
c,
-
co
o
._
u,
u'
.e
_
Before repair
After repair
_
,._
-
_
~: ~spechon effect |
~ ,
1
~ I
Vehicle Age
FIGURE 5-6 Schematic of I/M credit algorithm (sawtooth) for a cohort of vehicles in
a biennial program.
depending on the pollutant. If the I/M credit for an annual instead of biennial
program in MOBILES is equally small, then states have no incentive to test
vehicles more frequently and repair high emitters quickly. Although no analy-
ses have been published that indicate significantly greater emissions reductions
for annual programs, if repair durability is less than 2 years, annual programs
are likely to reduce emissions more than only a few percent from biennial
programs. When MOBlLE6 is released, the additional credit modeled for
annual programs should be compared with real-worId data.
This sawtooth pattern for modeling I/M (the same approach used in
MOBlLE5) suffers from a number of problems and is inappropriate for a
number of reasons, including the following:
· Vehicles with end withoutI/M are assumed to deteriorate et the same
rate, but it is very likely that the repaired fleet will deteriorate at a rate differ-
ent from that of the fleet that has not been repaired. Some repairs will be
effective and lasting and others will be ineffective and cause emissions to
increase beck to the unrepaired level. A comparison of Arizona IM240 data
with remote-sensing data shows that repair effectiveness diminishes over time
(Wenzel 1999b).
OCR for page 140
140 Evaluating Vehicle Emissions I/M Programs
· There is no explicit allowance in the model for repaired vehicles to
revert beck to high-emitter status. Analysis of Arizona IM240 data shows that
42% of cars that initially failed their IM240 and then resumed and passed,
failed again in their initial test in the next cycle 2 years later (Wenze} ~ 999b).
· The I/M credit algorithm does not allow for vehicles being scrapped
or sold outside the area (and still used in the I/M area) rather than repaired.
MOBILES includes vehicle scrappage, but it is not modeled as a function of
high-emitter status.8
· There is no estimate of the effect of vehicles being repaired just be-
fore I/M testing so that they will pass the test the first time. Some of these
repairs will not be done or not be long-lasting, and the vehicle will revert back
to high-emitter status just after the test.
OBD Effects
MOBlLE6 includes emissions reductions for vehicles equipped with
OBDIT systems ( 1996 model years and later). OBDTI is discussed further in
Chapters 2 and 4. Estimates for these emissions reductions depend on three
parameters, which have assumed levels because in-use data are not yet avail-
able (EPA 1999i):
· The ability ofthe OBD system to identify high emitters is assumedto
be a fixed fraction of high emitters at 85°/O. The remaining 1 5°/O of vehicles
that are high emitters but are not detected by the OBD system are assumed
to remain as high emitters.
· The response rate is the fraction of owners who will respond to a
malfunction indicator light (MTL) and have the vehicle repaired. MOBILES
assumes that owners are much more likely to respond to a MIL in an OBD-
based I/M area, where repairs are required. In OBD-based I/M areas,
MOBlLE6 assumes that the response rate is 90°/O over the lifetime of the
vehicle. Without such an I/M program encouraging repair, the response rate
is assumed to be 90°/O up to 36,000 miles (the standard full vehicle warranty
period), ~ 0°/O from 36,000 to 80,000 miles (the age limit for federally mandated
emissions-control system warranty), and zero after 80,000 miles.
MOBILES includes vehicle scrappage for vehicles destroyed in an accident or
retired from the fleet. Emissions credits for scrappage programs are estimated outside
the model.
OCR for page 141
Estimating I/MEmissio7?s Reductions Using the MOBILE Model 141
The emissions level after a repair in response to a MTE is assumed to
be ~ .5 times the appropriate 50,000-mile emissions standard; this is the thresh-
old level for illuminating the MIL. Because of the Tow emissions for new
vehicles and the low response rate at higher mileage, as currently modeled, the
emissions reduction associated with OBD is low in the absence of an I/M
program.
The MOBlLE6 approach does not take into account the ability of the
system to identify a failed component and take corrective action to minimize
the effect ofthe emissions. For example, when an oxygen sensor fails, some
OBD systems can revert to a known open-Ioop calibration that has good, but
not optimal, emissions. Thus, some OBD identified failures might have little
emissions increase, even if the owner ignores the MIL.
Figure 5-7 (EPA 1999f) shows MOBlLE6 projected nonmethane HC
basic emissions rates for light-duty Tier ~ vehicles with OBD systems from
EPA's draft MoBlLE6 documentation (EPA ~ 999i). The figure shows that
MOBlLE6 will generate a small emissions reduction for OBD systems in
areas without OBD-based I/M and much larger emissions reductions in areas
with OBD-based I/M. Such emissions reductions might tee overly optimistic
because they might tee based on optimistic assumptions about owner response
to the MIL in the I/M areas and pessimistic assumptions about response in
non-~/M areas.
Antitampering Programs
Antitampering benefits in MOBlLE6 are intended to be as similar to
MOBlLE5 as possible with the same fractional reduction in high emitters
associated with antitampering programs for vehicles before the ~ 996 mode]
year. After ~ 996, OBD is assumed to catch all tampered vehicles. This de-
scription ofthe approach, obtained from conversations with EPA staff, was not
available to the committee in written form.
Evaporative Emissions and I/M
Evaporative emissions are modeled with three distinct groupings: normal
(functioning), purge-failure, end pressure-faiTure vehicles. Purge failure refers
to failure ofthe system that allows regeneration ofthe carbon canisters used
OCR for page 142
142 Evaluating Vehicle Emissions I/M Programs
~8
0.7
Me
~5
0.4
Em
0.3
~ i ~ 2 A $ 5. Q
.
7.5
Mileage x 11]-
10.0 12.5
No OBO J No IM — —OBO / No IM - - - OBO I IM
]~.0 17,5 20.0
FIGURE 5-7 MOBILE6 non-methane HC basic emissions rates for light-duty Tier 1
vehicles with OBD systems. Source: EPA l999f.
to capture evaporative emissions through vapor purge into the combustion
system. A purge failure can be a result of a failed valve or disconnected hose
that leads to the intake manifold on the engine. A pressure failure refers to the
loss of integrity ofthe system end caninclude a missing, split, or disconnected
hose; a missing or failed gas cap, or a leak in the tank.
Purge and pressure failures have been measured in I/M programs through
the use of flow rate and pressure tests on the vehicle during the exhaust emis-
sions test. The test procedures can introduce their own problems, as not all
vehicles have accessible components. Perhaps the greatest problem, however,
is that the purge test is very invasive, with many hoses and components actu-
ally being damaged when these tests were first tried. Although MOB ILE6 will
mode! emissions reductions associated with pressure and purge tests, few
areas are actually perfo~ing these tests. The future use of purge and pres-
sure checks appears doubtful. A gas-cap check and a targeted physical in-
spection offer the most likely benefit of I/M programs on evaporative emis-
s~ons.
OCR for page 143
Estimating I/MEmissions Reductions Using the MOBILE Model 143
CALIFORNIA'S EMFAC MODEL FOR ESTIMATING I/1\]
EMISSIONS REDUCTIONS
Historically, California has had more restrictive air-quality and automobile
emissions standards than the rest of the United States. The Clean Air Act
allows California to regulate automobiles in the state and use its own computer
models to predict emissions inventories. The California Air Resources Board
(CARB) has developed its own emissions inventory model, called EMFAC.
Similar to MOBILE, EMFAC was developed well over a decade ago and has
continually been improved over the years. The current version of EMFAC is
EMFAC2000 (available at ht~p://arbis.arb.ca.gov/msei/msei.htm).
EMFAC has a number of differences with MOBILE, as outlined in NRC
(2000~. For estimating emissions reductions from California's Smog Check
I/M program, EMFAC uses a similar modeling methodology as MOBILE, with
a sawtooth representation of inspection and repair over the life of a vehicle
(section 4 in the online EMFAC2000 documentation at ht~p://arbis.arb.ca.gov/
msei/doctabletest/doctable_test.htmI). EMFAC has four categories for high
emitters, compared with three in MOBlLE5 and one in MOBlLE6. The per-
centage of each technology group inthese high-emitter groups (or regimes, in
EMFAC terminology) is determined from vehicle surveillance program data
and I/M recapture fleet data. These percentages are then multiplied by reg~me-
specific identification rates (i.e., the percentage of vehicles that will fait a given
I/M program), as described in section ~ in the online documentation (http://
arbis.arb.ca.gov/msei/doctabletest/doctable_test.htmI) . The identification rates
in EMFAC2000 are based on failing fractions of vehicles in the various Smog
Check programs.
For estimating the repair effectiveness, EMFAC uses a two-step process.
During CARB's surveillance programs, high-emitting vehicles were exten-
sivelly repaired to determine the maximum gain achievable through a "perfect"
repair. These perfect repair values are then modified by "correction eff~cien-
cies," which are a function ofthe I/M program being simulated.9 These effi-
ciencies vary on the basis of such factors as I/M repair cost limits and esti-
mates oftechnician training. After the identification and repair percentages
have been determined, the percentages of vehicles in the different high-emitter
9Although it is questionable whether these "correction efficiencies" have been
measured in a realistic repair setting.
OCR for page 144
144 Evaluating Vehicle Emissions I/M Programs
regimes are modified. Last, the same standard deterioration rate is applied
across all emissions regimes. Total reductions across entire fleets are then
calculated by appropriately weighting the vehicle technology groups by VMT
for a specific mode! year.
Like MOBILE, EMFAC has overpredicted emissions reductions from the
state's I/M programs, thus generating SIP credits that were too large and
contributing to problems in meeting air-quality standards. With EMFAC7G
(the version prior to EMFAC2000), substantial emissions reductions were
modeled for California's 1984 biennial two-speed idle VM program—12%
reduction in HC, ~ ~ % reduction in CO, and 5°/O reduction in NOX. With a draft
version of EMFAC2000 and additional analyses of VM evaluation program
data from tests of over ~ ,000 vehicles, CARB now suggests a ~ 5°/O reduction
for HC exhaust, 9°/O for CO, and 7°/O for NOX attributed to the 1 984 program
(CARB 2000b). Early independent evaluations of this program showed no
emissions benefit (Lawson 1993; Lawson et al. 1995, 1996a).
As discussed in Chapter 3, CARB (2000c) recently evaluated the CaTifor-
nia enhanced VM program. A draft version of EMFAC2000 predicts emis-
sions reductions in ~ 999 from the program at ~ 9°/O for HC,6% for NOX, and
~ 8°/O for CO, but CARB's analyses of roadside tests showed emissions reduc-
tions of only 14°/O for HC, 6% for NOX, and ~ 3% for CO. However, these
percentage reductions should be treated with caution due to a potential location
bias in the roadside sampling that took place.
SUMMARY
The MOBILE model will continue to be used to determine future emis-
sions-reduction credits that states will receive from implementing VM or from
modifying their current I/M programs. MOBILE is a static, not a dynamic,
mode} and is therefore a simplified representation of emissions changes from
VM. Historically, MOBILE has overestimated emissions reductions from VM
programs. It remains to be seen whether MOBILES, whichis a major revision
from MOBILES, will also overestimate VM benefits or whether it will be a
more accurate representation of VM benefits. Indications are thatMOBTLE6
will estimate lower emissions reductions from VM programs than are estimated
by MOBII~E5.
Estimates of model inputs, such as compliance rates, have a large effect
on the VM credits estimated by MOBILE. To date, states have been allowed
OCR for page 145
Estimating I/MEmissions Reductions Using the MOBILE Model 145
lo use optimistic estimates ofthese inputs, instead of justifying them with anal-
ysis of program data or other assessments. The model inputs should be set by
default to pessimistic values, thus providing an incentive to states to document
evidence for inputting more optimistic values. Such evidence should include
program evaluations in the state or reliance on program evaluations from other
states with similar I/M programs. Model-based forecasts should be closely
linked to I/M program performance measurements.
There are also a large number of assumptions internal to the model that
significantly affectMOBlLE's emissions projections and I/M credits. These
include the absolute number and average emissions of high emitters, average
emissions of initially failing end passing vehicles, average repair effects, identi-
bcation rates under different cutpoints, and OBD effects. Actual data from
state programs and special studies could be used to improve model parameters
and assumptions. In the long term, the overall I/M estimation methodology in
MOBILE should be improved based on I/M evaluation data. For example,
empirical data already show that the sawtooth modeling approach is not reaTis-
tic. Human behavior, an important factor missing in the model, should be
incorporated into future models. Further, embedded assumptions in the mode}
should be given parameters as much as possible so that users can improve
their I/M benefit forecasts with the latest available data.
In the short term, sensitivity analyses should be done to demonstrate the
effects of changes in model inputs and in assumptions built into the model. The
results should be incorporated into the MOBILES guidance documentation and
related documents.
A mode} such as MOBILE will continue to be needed for forecasting
future-year emissions end the effects of mobile-source contro]programs such
as I/M. The mode! should not, however, be used to evaluate actual perfor-
mance. Instead, program evaluation studies should be done to estimate current
program effects, and results from actual I/M performance should be used to
calibrate the MOBILE estimates. As stated elsewhere in this report, guidance
from EPA is needed to accomplish these goals.
Representative terms from entire chapter:
emissions rates
