United States Air and Radiation EPA420-R-01-022
Environmental Protection April 2001
Agency M6.EVP006
vvEPA Estimating Weighting
Factors for Evaporative
Emissions in MOBILE6
> Printed on Recycled Paper
-------
EPA420-R-01-022
April 2001
for
in
M6.EVP.006
Larry C. Landman
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
NOTICE
This technical report does not necessarily represent final EPA decisions or positions.
It is intended, to present technical analysis of issues using data 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.
-------
ABSTRACT
In parallel documents (M6.EVP.001, M6.EVP.002, and
M6.EVP.005), EPA identified the methods used in MOBILE6 for
estimating resting loss and diurnal emissions based (in part) on
the performance of the vehicles (i.e., pass or fail) on the purge
and pressure tests. EPA computes model-year and age specific
average resting loss and diurnal emissions by weighting together
the emissions of passing and failing vehicles according to their
frequency in the in-use fleet. This document describes this
approach and EPA's predictions of pass and fail rates (i.e.,
weighting factors) as functions of vehicle age.
This report was originally released (as a draft) in June
1999. This current version is the final revision of that draft.
This final revision incorporates suggestions and comments
received from stakeholders during the 60-day review period and
from peer reviewers.
-------
TABLE OF CONTENTS
Page Number
1.0 Introduction 1
2.0 Data Sources 2
3.0 Analysis 6
3.1 Strata Based on Purge and Pressure Tests ... 6
3.1.1 Vehicles Failing the Pressure Test . . 6
3.1.2 Vehicles Passing Both the Pressure
and the Purge Tests 9
3.1.3 Vehicles Failing ONLY the Purge Test . 11
3.1.4 Summary of Purge and the Pressure
Failure Rates 13
3.1.5 Vehicles Failing Both the Pressure
and the Purge Tests 13
3.1.6 Effects of Inspection/Maintenance ... 16
3.2 Modeling "Gross Liquid Leakers" 16
3.3 Combining Purge/Pressure Rates with
Gross Liquid Leaker Rates 18
4.0 Modeling Enhanced EVAP Vehicles Equipped with OBD . 21
5.0 Comparisons with MOBILES 25
5.1 Comparisons of Weighting Factors 25
5.2 Comparisons of Weighted Diurnal Emissions . . 28
5.3 Comparisons of Diurnal Emissions from Vehicles
Certified to the Enhanced Evaporative Control
Standards 34
6.0 Summary 36
7.0 References 38
11
-------
TABLE OF CONTENTS (Continued)
APPENDICES Page Number
A. Estimates of Purge/Pressure Strata Size by
Vehicle Age for 1995 and Older Model Year
Vehicles for Non-I/M Areas 39
B. Predicted Frequency of Occurrence of "Gross
Liquid Leakers" by Emission Type and Vehicle
Age (for 1995 and Older Model Year Vehicles) .... 40
C. Estimates of Purge/Pressure Strata Size by
Vehicle Age for 1999 and Newer Model Year
Vehicles for I/M Areas 41
D. Estimates of Purge/Pressure Strata Size by
Vehicle Age for 1999 and Newer Model Year
Vehicles for Non-I/M Areas 42
E. Predicted Frequency of Occurrence of "Gross
Liquid Leakers" by Emission Type and Vehicle
Age (for 1999 and Newer Model Year Vehicles) ... 43
F. Peer Review Comments from Sandeep Kishan 44
G. Comments from Stakeholders 49
111
-------
Estimating Weighting Factors
for Evaporative Emissions in MOBILE6
Report Number M6.EVP.006
Larry C. Landman
U.S. EPA Assessment and Standards Division
1.0 INTRODUCTION
In three parallel reports [1,2,3]*, the US Environmental
Protection Agency (EPA) developed methods of estimating the
resting loss and diurnal emissions from results of real-time
diurnal (RTD) tests of in-use vehicles in which the ambient
temperature cycled over a 24 degree Fahrenheit range to simulate
in real-time the daily heating and cooling that parked vehicles
experience over a 24 hour period. For many of the vehicles used
in these studies, the recruitment method was designed to recruit
a relatively large number of vehicles that had problems with
their evaporative control systems. Specifically, two tests of
the integrity of each vehicle's evaporative control system were
used to screen the candidate vehicles (a pressure test** and a
purge test). This recruitment bias did not affect the analysis
of these data as described in earlier reports; since those
analyses were performed within each purge/pressure grouping, the
selection was random within the purge/pressure and model year
groups. However, to correctly represent the entire in-use fleet
the results must be weighted. In this report, EPA develops
weightings for each stratum to estimate the emissions of the
entire in-use fleet. EPA will use these factors to weight
together the results of the RTD tests, as well as the results of
hot soak tests and running loss tests which were also derived
from measurements of a stratified sample.
For each of the parallel analyses of resting loss and
diurnal data, the sample of test vehicles was divided into four
strata. The first of these strata consisted of several vehicles
having substantial leaks of liquid gasoline (as opposed to simply
vapor leaks); these vehicles were labeled "gross liquid leakers"
(GLLs) . EPA will use the following three definitions [4] (based
* The numbers in brackets refer to the references in Section 7 (page 39).
** This pressure test was performed by disconnecting the vapor line at the
canister and then pressurizing the tank from that position with the gas
cap in its normal position. This procedure differs from the method
currently being used in Inspection and Maintenance (I/M) lanes.
-------
-2-
on the evaporative emissions test used) for such vehicles with:
resting loss emissions (i.e., the mean emissions during
the last six hours of the 24-hour RTD test) were at least
2.00 grams per hour (see also reference [1]), or
hot soak test emissions were at least 10.00 grams per one-
hour test (see also reference [5]) , or
running loss test emissions were at least 7.00 grams per
mile over a LA-4 driving cycle (or 137.2 grams per hour)
(see also reference [6]) .
These three different definitions will identify potentially
different sets of vehicles as being "gross liquid leakers" (see
Section 3.2). For the remaining three strata, we used the
results of the purge and pressure tests to match the
stratification of the recruitment process. This approach
produces the following three additional strata:
1) vehicles that pass both the purge and pressure tests,
2) vehicles that fail the pressure test (regardless of their
performance on the purge test),* and
3) vehicles that fail only the purge test.
This document reports on EPA's proposal to weight those four
strata together to obtain estimates (of running loss, hot soak,
resting loss, and diurnal emissions) for the entire in-use fleet.
2.0 DATA SOURCES
To develop the appropriate weighting factor for the stratum
of vehicles identified as "gross liquid leakers," EPA relied on
five groups of data to estimate the frequency of the occurrence
of these vehicles (see also reference [4]) :
For the "gross liquid leakers" identified by the RTD test,
EPA used a sample consisting of 151 vehicles tested by the
Coordinating Research Council (CRC) during 1996 as part of
its real-time diurnal testing program (Program E-9) [7]
combined with 119 vehicles tested by EPA. [1]
For the "gross liquid leakers" identified by the hot soak
test, EPA combined the sample of 300 vehicles tested by
Auto/Oil during 1993 as part of its real world hot soak
testing program with 197 vehicles tested by EPA. [5]
Vehicles failing both purge and pressure are discussed in Section 3.1.4.
-------
-3-
For the "gross liquid leakers" identified by the running
loss test, EPA used a sample consisting of 150 vehicles
tested by the CRC during 1997 as part of its running loss
testing program (program E-35) . [6]
The CRC also tested 50 late-model year vehicles during
1998 as part of its combined hot soak, real-time diurnal,
running loss testing program (E-41) . [8] (These results
are used in reference [4] to test the predictions of the
occurrence of "gross liquid leakers" among newer
vehicles.)
A fifth source of data consisted of the results of a
testing program run jointly by the CRC and the American
Petroleum Institute (API) [9]. This program was designed
to determine the frequency of vehicles with liquid leaks.
Actual measurements of evaporative emissions were not
performed in this program; therefore, we cannot determine
which of those vehicles identified as having liquid leaks
would have actually met any of our definitions of a "gross
liquid leaker."
To develop the appropriate weighting factors based on the
performance on the purge and pressure tests, EPA used data from
an EPA testing contractor, Automotive Testing Laboratories (ATL),
which performed purge and pressure tests on a random sample of
13,425 vehicles at its Inspection and Maintenance (I/M) lanes in
Indiana and Arizona between the years 1990 and 1995. Since the
testing protocols were changing in the early months of the
program, we omitted the first nine months of data. We then
identified the initial test of each of the test vehicles and
calculated, by vehicle age, the number of pre-1996* model year
vehicles in each of the three purge/pressure categories.
We combined the results for the I/M lane testing into a
single table (Table 1 on page 5). Omitted from all of the
columns in Table 1 are the results on approximately fifteen
percent of the vehicles for which the purge or pressure tests
were not performed. The reasons that testing was not performed
varied, and included both periodic problems with the testing
equipment as well as problems related to the vehicle (e.g.,
presence of check-valves or difficulty accessing the necessary
lines). All of the subsequent analyses were performed on the
sample of vehicles for which the purge/pressure classification
could be made. Since all of the subsequent analyses are based on
ratios from Table 1 (e.g., the number of classified pressure
failures divided by the total number of vehicles that were
Limiting the analysis to pre-1996 model year vehicles is related to the
phase-in of the enhanced evaporative control vehicles (see Section 4).
-------
-4-
varied, and included both periodic problems with the testing
equipment as well as problems related to the vehicle (e.g.,
presence of check-valves or difficulty accessing the necessary
lines). All of the subsequent analyses were performed on the
sample of vehicles for which the purge/pressure classification
could be made. Since all of the subsequent analyses are based on
ratios from Table 1 (e.g., the number of classified pressure
failures divided by the total number of vehicles that were
In examining the data in Table 1, we note that there were
relatively few vehicles more than 20 years of age. Since small
sample sizes tend to result in low statistical confidence in the
calculated ratios (i.e., the percent of vehicles at each age that
fall into each of the purge/pressure strata), those small sample
sizes are an obvious weakness of this analysis. We will address
that weakness by using the calculated variances in the ratios to
weight the analyses. (That is, the ratios from the model years
containing the most vehicles will be counted more heavily than
the ratios from the more sparsely sampled model years.)
An alternative approach (not being used) is to smooth the
data from the older vehicles by averaging the results from the 66
vehicles over the age of 20 years (all of which were from the
industry programs) to obtain a sample with:
a mean age of 23.23 years,
19 vehicles (28.8 percent) passing both the pressure test
and the purge test,
38 vehicles (57.6 percent) failing the pressure test, and
9 vehicles (13.6 percent) failing only the purge test.
That averaged failure rate on the pressure test of almost 60
percent among the vehicles over 20 years of age suggests a
substantially higher failure rate among these vehicles than was
predicted in MOBILES (i.e., under 35 percent). This difference
in estimating the failure rate (on the pressure test) of older
vehicles is due entirely to data recently obtained in the CRC
testing programs. (The EPA testing used in MOBILES had no data
on vehicles older than 17 years of age.)
-------
-5-
Table 1
Distribution of 14,061 1971 - 95 Model Year Vehicles
Vehicle
Aae*
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
Performance
Fail
Pressure
5
48
42
61
81
94
91
94
88
68
46
41
64
49
29
19
13
7
4
12
12
3
7
10
6
6
6
on Purge and
Tests
Fail Only
Purqe
9
29
33
33
42
50
76
74
46
68
44
24
23
20
5
6
3
1
0
1
3
2
0
2
2
3
0
Pressure
Passing
Both
228
1,448
1,302
1,494
1,308
1,475
1,403
1,261
888
682
369
192
152
102
62
34
17
7
2
4
7
7
5
3
2
1
1
Total
242
1,525
1,377
1,588
1,431
1,619
1,570
1,429
1,022
818
459
257
239
171
96
59
33
15
6
17
22
12
12
15
10
10
7
The quantity "Vehicle Age" is the whole number calculated by subtracting
model year from test year and then changing all negative results to zero.
-------
-6-
3.0 ANALYSIS
3.1 Modeling Strata Based on Purge and Pressure Tests
Using the data from Table 1, we calculated the rate at which
vehicles were present (by age) in the following three categories
determined by the results on the pressure test and the purge
test.
vehicles passing both the pressure test and the purge
test,
vehicles failing the pressure test, and
vehicles failing only the purge test.
These three categories are not independent. Given the results
from any two would permit the size of remaining category to be
determined. EPA chose to model the rates at which vehicles were
present in the first two of those categories. These rates by
vehicle age (in years) along with the corresponding 90 percent
confidence intervals are given in Tables 2 and 3. The confidence
intervals were calculated separately for each vehicle age rather
than having an overall calculation for the entire sample.
Calculating confidence intervals independently (as if the failure
rate at one age were not related to the failure rates of
neighboring ages) emphasizes the disparity in the sizes of some
of the samples by age, as the size of the confidence interval is
substantially controlled by the sample size.
3.1.1 Vehicles Failing the Pressure Test
Calculating (from Table 1) the rates at which vehicles
failed the pressure test (regardless of the performance on the
purge test) produces the data given in Table 2.
As previously stated, since the 90 percent confidence
intervals in Table 2 were calculated separately for each age that
was sampled, the confidence intervals are most representative of
the relative sample sizes. Rather than immediately attempting to
use a regression analysis to obtain an equation relating the rate
of vehicle's failing the pressure test to the vehicle's age, we
first examined the calculated 90 percent confidence intervals in
Table 2. (Binomial confidence intervals were used since there
were exactly two possible values, namely "PASS" and "FAIL.")
-------
-7-
Table 2
Estimating Rate of Vehicles that Fail the Pressure Test
For 14,061 1971-95 Model Year Vehicles
With 90 Percent Confidence Intervals
Vehicle
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
Sample
Size
242
1,525
1,377
1,588
1,431
1,619
1,570
1,429
1,022
818
459
257
239
171
96
59
33
15
6
17
22
12
12
15
10
10
7
Failure
Rate
2.1%
3.1%
3.1%
3.8%
5.7%
5.8%
5.8%
6.6%
8.6%
8.3%
10.0%
16.0%
26.8%
28.7%
30.2%
32.2%
39.4%
46.7%
66.7%
70.6%
54.5%
25.0%
58.3%
66.7%
60.0%
60.0%
85.7%
90 Percent
Confidence Interval
0.6% - 3.6%
2.4% - 3.9%
2.3% - 3.8%
3.0% - 4.6%
4.7% - 6.7%
4.8% - 6.8%
4.8% - 6.8%
5.5% - 7.7%
7.2% - 10.1%
6.7% - 9.9%
7.7% - 12.3%
12.2% - 19.7%
22.1% - 31.5%
23.0% - 34.3%
22.5% - 37.9%
22.2% - 42.2%
25.4% - 53.4%
25.5% - 67.9%
35.0% - 98.3%
52.4% - 88.8%
37.1% - 72.0%
4.4% - 45.6%
34.9% - 81 .7%
46.6% - 86.7%
34.5% - 85.5%
34.5% - 85.5%
64.0% - 100.0%
Examining the confidence intervals in Table 2, we found that
some of those confidence intervals are so large as to be almost
useless. (For example, for vehicles 17 years of age, knowing
that the actual failure is most likely between 25 percent and 68
percent is not helpful in predicting the true failure rate.)
However, using both the sample failure rates and the confidence
intervals, we were able to make the following four observations
that were then used to select an appropriate mathematical model:
The pressure failure rate appears to start (i.e., for new
vehicles) between two and four percent.
-------
-8-
The pressure failure rate increases gradually for the
first seven years of the vehicle's life.
The pressure failure rate then increases more rapidly for
the next ten years.
The pressure failure rate then begins to level off
(approaching 60 percent, according to the data from the
industry programs) (see third "bullet" on page 4).
This type of behavior is typical of a logistic growth function.
Prior to constructing an appropriate logistic growth
function, we first "adjusted" the values of the variable "AGE"
(in Tables 1 through 4) which are based on the test date (since
it is the integer calculated by subtracting the model year from
the test year). In the MOBILE model, the (default) age is the
integer age as of January 1 of the evaluation year. However, the
approximately 14,000 tests in Tables 1 through 4 occurred over
all 12 months of the testing years. Thus, the average (i.e.,
typical) test date was in early July (mean date of July 3 and
median date of July 10). To adjust for that six month difference
(between the default month of January and the average test date
in July), we modified that age value in the tables by adding 0.5
years (six months) to it prior to the analyses.
To account for differences in the size of the confidence
intervals (or, equivalently the sample sizes) the data were
weighted by the reciprocal of the variance. The "logistic
growth" function that best models the weighted pressure test
failure rates from Table 2 is given by the following equation:
Pressure Failure Rate = 1 + !7.733*6x^.01362*(AGE-2)] (1)
Estimates based on this equation of failure rates on the pressure
test are given in Appendix A. These estimates must be adjusted
for the "gross liquid leakers" (see Section 3.3).
In Figure 1 (on the following page), we plotted both the
curve described by equation (1) (as a solid line) as well as the
90 percent confidence intervals (as dotted lines) for the failure
rate on the pressure test (from Table 2, shifted by six months to
compensate for the July testing). (The small sample of vehicles
at least 20 years of age was combined to produce reasonably sized
confidence intervals.) That graph suggests that equation (1) is a
very good fit for the measured failure rates except at ages for
which fewer than 20 vehicles were recruited at each age. Also,
for vehicles over 20 years of age, the predicted failure rate on
the pressure test is close to 60 percent which closely
approximates the results of those 66 tests from industry programs
(see third "bullet" on page 4).
-------
-9-
0)
£
Q.
ra
Figure 1
Comparison of Predicted Rates to Confidence Intervals of Measured Rates
For Vehicles Failing the Pressure Test
(by vehicle age)
90%
60%
£ 30%
10 20
Vehicle Age (years)
30
3.1.2 Vehicles Passing Both the Pressure and the Purge Tests
As described in the preceding section, we first calculated
from Table 1 the rates at which vehicles passed both the pressure
and the purge tests, yielding the results given in Table 3.
As in the case of the failure rate on the pressure test, we
were able to make the following four observations from Table 3:
The rate at which vehicles passed both the pressure test
and the purge test starts (i.e., for new vehicles) between
92 and 96 percent.
The rate at which vehicles passed both the pressure test
and the purge test decreases gradually for the first seven
years of the vehicle's life.
The rate at which vehicles passed both the pressure test
and the purge test then decreases more rapidly for the
next ten years.
The rate at which vehicles passed both the pressure test
and the purge test then begins to level off (approaching
20 to 40 percent, according to the data from the industry
programs) (see second "bullet" on page 4).
-------
-10-
Table 3
Estimating Rate of Vehicles that Pass Both the Pressure and Purge Tests
For 14,061 1971-95 Model Year Vehicles
With 90 Percent Confidence Intervals
Vehicle
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
Sample
Size
242
1,525
1,377
1,588
1,431
1,619
1,570
1,429
1,022
818
459
257
239
171
96
59
33
15
6
17
22
12
12
15
10
10
7
Failure
Rate
94.2%
95.0%
94.6%
94.1%
91 .4%
91.1%
89.4%
88.2%
86.9%
83.4%
80.4%
74.7%
63.6%
59.6%
64.6%
57.6%
51 .5%
46.7%
33.3%
23.5%
31 .8%
58.3%
41 .7%
20.0%
20.0%
10.0%
14.3%
90 Percent
Confidence Interval
91 .7% - 96.7%
94.0% - 95.9%
93.5% - 95.6%
93.1% - 95.1%
90.2% - 92.6%
89.9% - 92.3%
88.1% - 90.6%
86.8% - 89.6%
85.2% - 88.6%
81.2% - 85.5%
77.3% - 83.4%
70.2% - 79.2%
58.5% - 68.7%
53.5% - 65.8%
56.6% - 72.6%
47.0% - 68.2%
37.2% - 65.8%
25.5% - 67.9%
1 .7% - 65.0%
6.6% - 40.5%
15.5% - 48.2%
34.9% - 81 .7%
18.3% - 65.1%
3.0% - 37.0%
0.0% - 40.8%
0.0% - 25.6%
0.0% - 36.0%
As before, the "logistic growth" function appeared to be the
best choice for modeling the rates at which vehicles passed both
the pressure and the purge tests. After adjusting for age, the
resulting equation is given below as equation (2):
0.7200
Rate of Passing Both = 1 - 1 + 13.4o*expi-0.0145*(AGE*2)]
(2)
Estimates based on equation (2) of the rates of vehicles
passing both the purge and pressure tests are given in Appendix
A. These estimates must be adjusted for the "gross liquid
leakers" (see Section 3.3).
-------
-11-
In Figure 2 (below), we plotted both the curve described by
equation (2) (as a solid line) as well as the 90 percent
confidence intervals (as dotted lines) for the rate of vehicles
passing both the purge and pressure tests on the pressure test
(from Table 3, shifted by six months to compensate for the July
testing). (The small sample of vehicles at least 20 years of age
was again combined to produce reasonably sized confidence
intervals.) That graph suggests that equation (2) is a very good
fit for the measured rates for ages at which at least 20 vehicles
were sampled. Also, for vehicles over 20 years of age, the
predicted rate of vehicles passing both the purge and pressure
tests is close to 29 percent which closely approximates the
results of those 66 tests from industry programs (see second
"bullet" on page 4).
Figure 2
Comparison of Predicted Rates to Confidence Intervals of Measured Rates
For Vehicles Passing Both the Purge and Pressure Tests
(by vehicle age)
100%
J3
-------
-12-
vehicles older than 16 years of age. (This effect suggests that
some of the vehicles that only failed the purge test in a prior
year would begin to also fail the pressure test, thus migrating
into the pressure failure stratum.)
This effect is illustrated in Figure 3 . If we combine this
estimate of the incidence of vehicles' failing only the purge
test with the estimate (from Section 3.1.4) of failing both the
purge and pressure tests, we obtain a predicted failure rate on
the purge test (regardless of any pressure test result) . This
purge test failure rate does not exhibit that quirk of a decrease
in failure rate with increasing age.
The predicted size (in percent) of the stratum of vehicles
that failed only the purge test is given in Appendix A. These
estimates must be adjusted for the "gross liquid leakers" (see
Section 3.3).
In Figure 3, for vehicles failing only the purge test, we
plotted both the 90 percent confidence intervals (calculated from
Table 1 and shifted by six months to compensate for July testing)
and the curve described by subtracting from one hundred percent
the total of equation (1) plus equation (2).
Figure 3
Comparison of Predicted Rates to Confidence Intervals of Measured Rates
For Vehicles Failing ONLY the Purge Test
(by vehicle age)
_ 30%
.2
0)
S> 20%
O
c
o
10%
"- 0%
10 20
Vehicle Age (years)
30
The preceding graph indicates that the combination of equations
(1) and (2) is a very good fit for the measured rates for ages at
which at least 100 vehicles were sampled (i.e., through age 13).
-------
-13-
Also, for vehicles over 20 years of age, the predicted rate of
vehicles failing only the purge test is close to 12 percent which
closely approximates the results of those 66 tests from industry
programs (see fourth "bullet" on page 4).
3.1.4 Summary of Purge and the Pressure Failure Rates
Combining the predicted rates from Figures 1, 2, and 3 (or
Appendix A) into a single area graph produces the following graph
(Figure 4) .
Figure 4
Predicted Distribution of Pressure and Purge Failures
(by vehicle age)
100%
75%
50%
25%
0%
Pass Both
Purge and Pressure
10 15
Vehicle Age (years)
20
25
3.1.5 Vehicles Failing Both the Purge and the Pressure Tests
When EPA analyzed the RTD data, it was determined that there
were insufficient test results to distinguish between the diurnal
emissions of vehicles that failed both the purge and pressure
tests from those that failed only the pressure test. Therefore,
those vehicles were combined into the single stratum of vehicles
that failed the pressure test (regardless of their performance on
the purge test). Since the purpose of this study is to develop
factors to weight together the estimates of the individual
stratum to predict the in-use fleet emissions, it was not
necessary to model frequency of the stratum of vehicles that
failed both the purge and pressure tests.
-------
-14-
As a service to possible future researchers and modelers who
may require an estimate of the number of vehicles failing both
the purge and pressure tests (and based on the same sample that
produced equations (1) and (2)) , EPA performed the following
analysis.
The approach was similar to the one used in Sections 3.1.1
and 3.1.2 in which a table containing the frequencies with the
corresponding ninety percent confidence intervals was created
(see Table 4).
Table 4
Estimating Rate of Vehicles that Fail BOTH the Pressure and Purge Tests
For 1971-95 Model Year Vehicles
With 90 Percent Confidence Intervals
Vehicle
Aqe
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
Sample
Size
242
1,522
1,377
1,587
1,430
1,619
1,568
1,428
1,020
814
458
254
235
169
94
58
33
15
6
17
22
12
12
15
10
10
7
Failure
Rate
0.0%
0.2%
0.0%
0.3%
0.8%
0.4%
0.6%
0.8%
1 .3%
1 .5%
2.6%
2.8%
6.0%
7.1%
7.4%
6.9%
15.2%
26.7%
0.0%
23.5%
4.5%
0.0%
16.7%
53.3%
30.0%
30.0%
57.1%
90 Percent
Confidence Interval
0.0% - 0.5%
0.0% - 0.4%
0.0% - 0.2%
0.0% - 0.5%
0.4% - 1.1%
0.1% - 0.6%
0.3% - 0.9%
0.4% - 1 .2%
0.7% - 1 .9%
0.8% - 2.2%
1 .4% - 3.8%
1.1% - 4.4%
3.4% - 8.5%
3.9% - 10.4%
3.0% - 11.9%
1.4% - 12.4%
4.9% - 25.4%
7.9% - 45.4%
0.0% - 28.5%
6.6% - 40.5%
0.0% - 1 1 .9%
0.0% - 17.7%
0.0% - 34.4%
32.1% - 74.5%
6.2% - 53.8%
6.2% - 53.8%
26.4% - 87.9%
-------
-15-
As the reader may note, some of the sample sizes in Table 4
do not match the supposedly same samples in the first three
tables. In the first three tables, vehicles that failed the
pressure test but did not have a successful purge test were
included in the stratum "fail pressure" and, thus, included in
the total as well. However, those same vehicles would not be
included in Table 4.
After adjusting for age (by adding 0.5), the "logistic
growth" function that best models the frequency of vehicle's
failing both the pressure and purge tests from Table 4 is given
by the following equation:
0.3536
Rate of Failing Both = 1 + 414.96*exp[-0.32955*AGE]
(3)
In Figure 5, for vehicles failing both the purge and
pressure tests, we plotted both the 90 percent confidence
intervals (from Table 4, shifted by six months to compensate for
the July testing) and the curve described by equation (3).
(Again, the small sample of vehicles at least 20 years of age was
combined to produce reasonably sized confidence intervals.)
Figure 5
Comparison of Predicted Rates to Confidence Intervals of Measured Rates
For Vehicles Failing Both the Pressure and Purge Tests
(by vehicle age)
o
0)
o
m
c
o
0)
O)
'(5
60%
40%
20%
10 20
Vehicle Age (years)
30
Figure 5 suggests that equation (3) is a very good fit for the
measured rates for ages at which at least 30 vehicles were
-------
-16-
sampled (i.e., through age 16). Also, for the vehicles over 20
years of age, the data in Table 4 indicates that 20 out of 66
(30.3%) of those vehicles over 20 years of age (with a mean age
of 23.2 years) failed both the purge and pressure tests, and
equation (3) predicts that the failure rate would be 29.5 percent.
Thus, equation (3) is also a very good fit for the measured rates
for the older vehicles.
3.1.6 Effects of Inspection / Maintenance (I/M) Programs
As part of an Inspection and Maintenance (I/M) program, a
state may choose to perform a functional test (e.g., a pressure
test) of each vehicle's evaporative control system. Vehicles
failing the test would be required to be repaired. Thus, the
presence of such a program would alter the number of failing
vehicles.
The data used in the analyses in Sections 3.1.1 through
3.1.5 were obtained from two geographic areas (Hammond, Indiana
and Phoenix, Arizona) in which the vehicles were not required to
pass either a purge or pressure test. Therefore, those analyses
and the resulting estimates of failure rates (i.e., equations
(1), (2), and (3)) are used in MOBILE6 for geographic areas in
which there is no I/M for evaporative emissions.
In a parallel report [10], EPA explains how those rates are
adjusted in MOBILE6 to account for the presence of an I/M program
for evaporative emissions.
3.2 Modeling the Stratum of "Gross Liquid Leakers" (GLLs)
The set of vehicles identified as "gross liquid leakers"
varies depending upon which type of evaporative emission is being
considered. Earlier (see "bulleted" points on page 2), we
presented three definitions each based on one type of test (i.e.,
RTD test, hot soak test, and running loss test). EPA developed
these definitions in a parallel report devoted exclusively to the
subject of "gross liquid leakers" [4] . In that report, EPA
produced the following two equations to predict the frequency of
"gross liquid leakers" occurring on the RTD and on the running
loss tests for evaporative emissions:
Rate of Gross Liquid Leakers
0 08902
Based on RTD Testing = 1 + 4i4.613*exp[-0.3684*AGE] (4>
Rate of Gross Liquid Leakers
0.06
Based on Running Loss Testing = 1 + 12o*exp[-0.4*AGE] (5>
-------
-17-
A different approach was necessary in predicting the
frequency of "gross liquid leakers" on the hot soak test because
that (hot soak) testing had been limited to only newer vehicles
(i.e., no vehicles older than 12 years of age). The approach
developed in that report was based on the untested hypothesis
that the vehicles classified as hot soak "gross liquid leakers"
are the same vehicles identified as gross liquid leakers on
either the running loss or RTD tests. This hypothesis is based
on the assumptions that:
If a vehicle has a leak of liquid gasoline that is severe
enough to classify the vehicle as a "gross liquid leaker"
on the resting loss mode, then that leak will likely
result in the vehicle's being classified as a "gross
liquid leaker" on the hot soak test as well.
If a vehicle has a leak of liquid gasoline that is severe
enough to classify the vehicle as a "gross liquid leaker"
on the running loss test, then there will likely be enough
liquid gasoline remaining after the engine is shut off to
classify that vehicle as a "gross liquid leaker" on the
subsequent hot soak test as well.
That is, the set of vehicles classified as gross liquid leakers
on the hot soak test is the union of the set of vehicles
classified as gross liquid leakers on the RTD test with the set
of vehicles classified as gross liquid leakers on the running
loss test. (These hypotheses are "untested" because none of the
vehicles classified as "gross liquid leakers" on one test
procedure were tested over either of the other two procedures.)
Therefore, the rate of gross liquid leakers as identified on
the hot soak test would be the sum of the two rates for the RTD
testing and the running loss testing minus the number of double
counted vehicles (i.e., the product of those two rates assuming
these two categories are independent of each other). Using
equations (4) and (5) plus the preceding assumption, the predicted
rates of "gross liquid leakers" were calculated (for each of the
three test types) and are plotted in Figure 6 (on the following
page) and appear in Appendix B.
It is important to note that this model of the frequency of
gross liquid leakers is based on the assumption that modern
technology vehicles (through model year 1995) will show the same
tendency toward gross liquid leaks as do the older technology
vehicles at the same age. However, if the modern technology
vehicles were to exhibit a lower tendency to leak (due to the
more stringent demands imposed by the new evaporative emissions
certification procedure as well as heightened attention to
safety, e.g., fuel tank protection and elimination of fuel line
leaks), the effect would be to replace each of the three logistic
-------
-18-
growth functions with two or three curves specific to different
model year (or technology) groups.
Figure 6
Predicted Occurrences of "Gross Liquid Leakers"
On Each of Three Tests of Evaporative Emissions
(by vehicle age)
15%
10% --
5%
0%
10 20
Vehicle Age (years)
30
Since EPA has no data to indicate model-year specific rates,
EPA will use the model illustrated in Figure 6, to estimate the
occurrence in the in-use fleet of these vehicles that have
substantial leaks of liquid gasoline (i.e., "gross liquid
leakers") for vehicles that were not designed to meet the new
enhanced evaporative test procedure (i.e., vehicles up through
the 1995 model year along with some of the 1996 through 1998
model years). For the vehicles designed to meet the new enhanced
evaporative test procedure, EPA will modify that equation (see
Section 4.0).
3.3 Combining Purge/Pressure Rates with Gross Liquid Leaker Rates
In Section 3.1 we characterized the three strata resulting
from the individual vehicle's performance on the purge and
pressure tests. In Section 3.2, we characterized the additional
stratum created for the "gross liquid leakers." In order to make
these strata non-overlapping (i.e., mutually exclusive), we must
remove the "gross liquid leakers" from the other three strata.
To determine the distribution of the gross liquid leakers
among the other three strata, we examined the 270 vehicles in the
combined EPA/CRC RTD testing programs. Seven vehicles were
-------
-19-
identified as "gross liquid leakers" out of those 270 vehicles
that were tested. Of these seven gross liquid leakers:
four had failed both the purge and pressure tests,
two had failed only the pressure test, and
one had passed both the purge and pressure tests.
This distribution of seven gross liquid leakers proves that gross
liquid leakers can and do occur within all three of the purge/
pressure strata. From these statistics, EPA first estimated the
number of gross liquid leakers within each of the purge/pressure
strata and then removed them to form a fourth stratum consisting
of only the gross liquid leakers. Rather than attempting to
estimate the distribution of all gross liquid leakers based on a
sample of only seven vehicles, MOBILE6 will simply distribute the
liquid leakers proportionately among the three purge/pressure
categories.
As an example, if we were to take a hypothetical fleet of
10,000,000 vehicles, each 10 years of age, then Appendix B
predicts that 78,000 (0.78 percent) of them will be "gross liquid
leakers" on the RTD test. Similarly, Appendix A indicates that
1,091,000 (10.91 percent) will fail the pressure test, 647,000
(6.47 percent) will fail only the purge test, and the remaining
8,262,000 (82.62 percent) will pass both tests. Distributing
those 78,000 "gross liquid leakers" proportionately among the
three purge/pressure strata predicts the following distribution:
Table 5
Predicted Distribution on RTD Test
(For Vehicles at 10 Years of Age)
Purge/Pressure
Strata
Fail Pressure
Fail ONLY Purge
Pass Both
TOTALS
"Gross Liquid
Leakers"
8,510
5,047
64,444
78,000
Not "Gross Liquid
Leakers"
1,082,490
641,953
8,197,556
9,922,000
TOTALS
1,091,000
647,000
8,262,000
10,000,000
Thus, Table 5 indicates that for 10 year old vehicles on the RTD
test:
0.78 percent of those vehicles will be "gross liquid
leakers" on a RTD test,
10.82 percent of those vehicles will fail the pressure
test, but will not be "gross liquid leakers,"
-------
-20-
6.42 percent of those vehicles will fail only the purge
test, but will not be "gross liquid leakers," and
81.98 percent of those vehicles will pass both the
pressure and purge tests, and will not be "gross liquid
leakers."
Repeating this process for each vehicle age in Appendices A and B
produces the estimated size of each of the four strata for the
RTD test for each age. The results are plotted below (Figure 7):
Figure 7
Predicted Strata Sizes on RTD Test
(by vehicle age)
100%
80% - -
60%
40%
20%
0%
1 Pass Both
Purge and
Pressure
' Fail Pressure
1 Fail Only
Purge
Leakers
10 20
Vehicle Age (years)
30
Repeating this process using the "running loss" and "hot
soak" columns from Appendix B will produce the estimates of the
size of the four strata for use with each of those two types of
evaporat ive emi s s ions.
-------
-21-
4.0 MODELING ENHANCED EVAP VEHICLES EQUIPPED WITH OBD
Beginning with the 1996 model year, manufacturers were
required to certify at least twenty percent of their vehicles
using a new "enhanced" evaporative testing procedure (ETP); that
percentage of ETP vehicles was required to increase from the
twenty percent in 1996 up to one hundred percent by 1999. The
actual phase-in of these ETP vehicles proceeded at a slightly
faster pace (based on EPA's analysis of data from the Wisconsin
Inspection/Maintenance program for model years 1996-1999). The
phase-in rate required by the regulations is given below in
Table 6 (copied from 40 CFR 86.096-8) along with the observed
(actual) phase-in rate.
Table 6
Phase-In of Vehicles with
Enhanced Evaporative Controls
Model Required Observed
Year Percentage Percentage
1995 0% 0%
1996 20% 30%
1997 40% 55%
1998 90% 90%
1999 100% 100%
To predict the performance of these 1996 and newer vehicles
on the purge and pressure tests, the effects of two factors must
be considered:
1) A change in the regulations requires a doubling of the
period during which these vehicles must meet the
evaporative emissions standards (increased from 5 years
/ 50,000 miles to 10 years / 100,000 miles for light-
duty vehicles) .
2) Most of these ETP vehicles are equipped with an on-
board diagnostic (OBD) system that is expected to alert
each vehicle's owner (or driver) to most problems with
the evaporative control system, thus permitting the
owner to decide whether to repair the problem.
In order to meet the increased durability and more stringent
evaporative standards, manufacturers have implemented a number of
changes, including (but not limited to):
"quick connects" that reduce the possibility of improper
assembly when the vehicle is serviced,
-------
-22-
advanced materials that are less permeable, less
susceptible to puncture, and more durable (i.e.,
elastomeric materials used in hoses and connectors),
improvements made to the purge system (to enable the
vehicles to pass both the running loss test and the multi-
day diurnal test),
tethered gas caps, and
improved fractional-turn gas caps.
Since these changes are expected to result in improved control of
evaporative emissions, in a parallel report (M6.EVP.005), EPA
decided to use a separate set of estimates of both resting loss
and diurnal emissions for these vehicles. However, EPA does not
have actual data on the effects of these changes in durability
that translates into changes in the purge and pressure failure
rates estimated in Section 3.1 for the pre-1996 model year
vehicles.
EPA, therefore, will use the doubling in the durability
requirement to modify equations (1), (2), (4), and (5) (from Sections
3.1.1, 3.1.2, and 3.2) by replacing the variable "AGE" with
"AGE/2" resulting in:
Pressure Failure Rate of Enhanced Evaporative Control Vehicles
0.6045
~ 1 + 17.733*exp[-0.003405*(AGEA2)]
Rate of Passing Both for Enhanced Evaporative Control Vehicles
0.7200
(6)
\'I
Rate of Gross Liquid Leakers on the RTD Test for the Enhanced Evaporative Control
Vehicles
0.08902
~ 1 + 414.613*exp(-0.1842*AGE)
Rate of Gross Liquid Leakers on the Running Loss Test for the Enhanced Evaporative
Control Vehicles
~ 1 + 120*exp[-0.2*AGE]
Using these equations, we first generated the estimated
failure rates for those (1996 and newer) vehicles certified to
-------
-23-
the enhanced evaporative control standards. Since these
estimates assume only the benefits of changes in durability with
no estimation of the effect from the OBD system, they are
adjusted in MOBILE6 (Appendices C and D).
Since the OBD system is designed to alert each vehicle's
owner to problems with the evaporative control system. The
appearance of a warning light should result in at least some
owners having their malfunctioning vehicles repaired. Thus, the
OBD system has the potential to affect the relative sizes of the
purge/pressure strata, depending both on its ability to identify
problems in the evaporative control systems and on the owners
inclination to repair such problems. (In a parallel report [10],
EPA explains how those rates are adjusted in MOBILE6 to account
for the presence of an I/M program for evaporative emissions.)
In that parallel report [10], EPA assumes that the effect of
an OBD system will vary, based on the vehicle's warranty and the
presence of an I/M program. Those assumptions stated that:
The vehicle's malfunction indicator light (MIL) would
detect/identify 85 percent of the instances of the
vehicle's failing the purge test or failing the pressure
test. (It is assumed that the OBD system would not detect
the presence of a gross liquid leak.)
While the vehicle is under its full ("bumper to bumper")
warranty (i.e., up through 3 years / 36,000 miles), 90
percent of the owners will have the vehicle repaired when
the MIL indicates a problem. (These first two assumptions
suggest that the OBD system combined with the
manufacturer's warranty program will reduce the incidence
of vehicles failing either the pressure or purge tests by
76.5 percent (76.5% = 85% * 90%).)
When the warranty covers only the electronic control
module and the catalytic converter (i.e., from 36,000
through 80,000 miles, approximately ages four through six
years), only 10 percent of the owners will have the
vehicle repaired when the MIL indicates a problem. (This
assumption suggests that the OBD system combined with the
manufacturer's warranty program will reduce the incidence
of vehicles that in their fourth, fifth, or sixth years
newly fail either the purge or pressure tests.) That
percentage would increase from 10 to 90 percent if that
geographic area has an I/M program that requires the MIL
to indicate that there are no problems.
When the vehicle is no longer covered by a manufacturer's
warranty (i.e., over 80,000 miles or beyond six years of
age), none (i.e., zero percent) of the owners will have
the vehicle repaired when the MIL indicates a problem.
-------
-24-
That percentage would increase from zero to 90 percent if
that geographic area were to have an I/M program requiring
the MIL to indicate that there are no problems.
EPA will adapt those assumptions (modified slightly) for
evaporative emissions. Specifically:
The vehicle's malfunction indicator light (MIL) would
detect/identify 85 percent of the instances of the
vehicle's failing the purge test or failing the pressure
test. (It is assumed that the OBD system would not detect
the presence of a gross liquid leak.)
While the vehicle is within its full warranty period
(i.e., approximately through the age of three years), 90
percent of the owners will have the vehicle repaired when
the MIL indicates a problem. (These first two assumptions
suggest that the OBD system combined with the
manufacturer's warranty program will reduce the incidence
of vehicles failing either the pressure or purge tests by
76.5 percent (76.5% = 85% * 90%).)
While the vehicle is under its partial warranty period
(i.e., approximately ages four through six years), only 10
percent of the owners will have the vehicle repaired when
the MIL indicates a problem. (This assumption suggests
that the OBD system combined with the manufacturer's
warranty program will reduce the incidence of vehicles
(ages four through six) that fail either the purge or
pressure tests for the first time.) That percentage would
increase from 10 to 90 percent if that geographic area
were to have an I/M program requiring the MIL to indicate
that there are no problems.
When the vehicle's evaporative control system is no longer
covered by a manufacturer's warranty (i.e., beyond about
six years of age), none (i.e., zero percent) of the owners
will have the vehicle repaired when the MIL indicates a
problem. That percentage would increase from zero to 90
percent if that geographic area were to have an I/M
program requiring the MIL to indicate that there are no
problems.
Note: EPA is continuing to study in-use OBD systems. The
assumptions listed here may be revised in future models.
These MOBILE6 assumptions are different from what was used
in MOBILES. In that previous model, while the evaporative
emissions of the ETP vehicles were reduced to reflect the more
-------
-25-
stringent standards, the weighting factors were not changed to
reflect either the improved durability or the OBD system.
5.0 COMPARISONS WITH MOBILES
5.1 Comparisons of Weighting Factors
Both the MOBILES model and this proposal weight the
estimated evaporative emissions based on each vehicle's
performance on purge and pressure tests; however there are
several structural differences:
The weighting factors in MOBILES are each functions of a
continuous variable (i.e., each model year's average
odometer) while the weighting factors in MOBILE6 are
functions of a discrete variable (i.e., each model year's
estimated age).
The weighting factors in this proposal are smooth
functions (i.e., exponential) of the estimated age.
Although the weighting factors in MOBILES are continuous,
they are not smooth functions (they are piece-wise linear
functions of odometer, thus almost linear with age).
MOBILES does not have a separate stratum for the vehicles
classified as "gross liquid leakers" while this proposal
does. We will, therefore, compare the MOBILES weighting
factors with the unadjusted factors from Section 3.1 of
this report.
The comparisons between the MOBILES weighting factors for
light-duty vehicles for each of the three purge/pressure strata
and the weighting factors in this proposal are illustrated in the
following three figures:
Figure 8 compares the estimates of Pre-ETP vehicles (i.e.,
1995 and older) passing both the purge test and the
pressure test. (Subtracting each of those estimates from
100 percent will yield the associated rates of vehicles
failing either the purge test, or the pressure test, or
both tests.)
Figure 9 compares the estimates of Pre-ETP vehicles
failing the pressure test (regardless of the performance
on the purge test).
Figure 10 compares the estimates of Pre-ETP vehicles
failing only the purge test.
-------
-26-
Figure 8
Comparison of Predictions of Percentage of Pre-ETP Vehicles
Passing Both the Purge and Pressure Tests
(by vehicle age)
O
100%
75%
50%
25%
10 15
Vehicle Age (years)
20
25
Figure 9
Comparison of Predictions of Percentage of Pre-ETP Vehicles
Failing the Pressure Test
(by vehicle age)
60%
10 15
Vehicle Age (years)
20
25
-------
-27-
Figure 10
Comparison of Predictions of Percentage of Pre-ETP Vehicles
Failing ONLY the Purge Test
(by vehicle age)
25%
20%
15%
o
10%
10 15
Vehicle Age (years)
20
25
Based on examinations of these three graphs, we can make the
following observations for the Pre-ETP vehicles:
The most obvious difference between the weighting factors
used in MOBILES and those developed in this report is that
the factors in this report are capped (around age 20)
while the MOBILES factors are not.
Despite the structural differences (i.e., smooth function
of a discrete variable versus piece-wise linear function
of a continuous variable) between these two sets of
weighting factors, they produce similar results for
vehicles up to about 12 to 13 years of age.
The estimates developed in this report predict a
substantially higher proportion of vehicles failing the
pressure test for vehicles older than 13 years of age than
does the MOBILES model.
The reader can make similar comparisons for the ETP vehicles
by replacing the MOBILE6 (dotted) line in each figure with the
values in Appendix D. (As stated at the end of Section 4, while
MOBILE6 uses different distributions for the ETP vehicles,
MOBILES uses the same distributions for the Pre-ETP and the ETP.)
This results in MOBILE6 predicting a smaller number of
malfunctioning ETP vehicles at any given age than did MOBILES.
-------
-28-
For example, MOBILE6 predicts that at age 25 years with no I/M
program, approximately 26 percent of the ETP vehicles would fail
either the purge or the pressure test while MOBILES predicts that
approximately 56 percent would fail either test. This much lower
predicted failure rate is illustrated in the following graph
(Figure 11 which is a modification of Figure 8) in which the
MOBILE6 curve (from Appendix A) is replaced with the
corresponding curve from Appendix D (i.e., ETP vehicles in a non-
I/M area passing both the purge test and pressure test).
Figure 11
Comparison of Predictions of Percentage of ETP Vehicles
Passing Both the Purge and Pressure Tests
(by vehicle age)
O
100%
75%
50%
MOBILES
-ETPsin M6
25%
10 15
Vehicle Age (years)
20
25
5.2 Comparisons of Weighted Full-Day Diurnal Emissions
By combining the information in this report with the
information in parallel reports (i.e., M6.EVP.001 and
M6.EVP.005), we can estimate the diurnal emissions for the in-use
fleet (containing the 25 most recent model years). To compare
these estimates with those predicted by the MOBILES model, the
MOBILES model was run for a fleet with a national distribution of
model years (as of January 1, 1995) with two likely combinations
of temperature cycle and fuel RVP (assuming no weathering of the
fuel) :
daily temperatures cycling between 60 and 84 degrees
Fahrenheit using fuel with a 9.0 RVP (see Figure 12) and
-------
-29-
daily temperatures cycling between 82 and 106 degrees
Fahrenheit using fuel with a 7.0 RVP (see Figure 13).
Each MOBILES run generates estimated diurnal emissions for the 25
most recent model years, which for these runs were the 1971
through 1995 model years. In these two examples the variable
MODEL YEAR can be transformed into AGE using the equation:
AGE = 95 - (MODELYEAR) .
It is important to note that these (following) estimates are
only of the full one-day diurnal emissions. They do include
diurnal emissions from "gross liquid leakers." But, they do not
include evaporative emissions from interrupted (i.e., partial-
day) diurnals, nor from multi-day diurnals, nor from running
loss, nor from hot soaks. Estimates of these excluded sources
are developed in parallel reports. A complete comparison between
MOBILE6 and MOBILES requires using activity data to weight
together all of these individual components of evaporative
emissions; this can be done by running each model.
Figure 12
Comparison of Predictions for In-Use Diurnal Emissions by Model Year
(Total Grams Per Full-Day Test)
60° to 84° F Cycle - Using 9.0 psi RVP Fuel
(As of January 1,1995)
70
75
80 85
Model Year
90
95
-------
-30-
Figure 13
Comparison of Predictions for In-Use Diurnal Emissions by Model Year
(Total Grams Per Full-Day Test)
82° to 106° F Cycle - Using 7.0 psi RVP Fuel
(As of January 1,1995)
2
u>
o
ro
70
75
80 85
Model Year
90
95
In Figures 12 and 13, the typical (i.e., mean) 24-hour
diurnal emissions for each vehicle are plotted against model
year. Visual inspections of Figures 12 and 13 suggest that the
estimates of diurnal emissions resulting from MOBILE6 (for each
of those two combinations of fuel RVP and temperature cycle) for
each model year are:
close (within a gram per vehicle per day) to MOBILES
estimates for the 1986 to 1995 model year vehicles,
typically 25 to 75 percent higher (for 9.0 and 7.0 RVP
fuels, respectively) than MOBILES estimates for 1980 to
1985 model year vehicles, and
typically 30 to 40 percent higher than MOBILES estimates
for 1972 to 1979 model year vehicles.
A similar comparison for the vehicles certified to the new
enhanced evaporative emission requirements (i.e., 1999 and newer
vehicles, along with some 1996-98 model year vehicles) is
provided in Section 5.3.
In Figures 12 and 13, we weighted each model year's
estimated diurnal emissions by the relative number of vehicles in
-------
-31-
the fleet, we obtain the estimate of the mean full-day diurnal
emissions for an average in-use vehicle subject to a full-day's
diurnal:
Estimates of Full-Day Diurnal Emissions
For the In-Use Fleet (As of January 1995)
In Units of Grams per Day per Vehicle
Temperature Cycle and Fuel
60° to 84° F with 9.0 psi RVP
82° to 106° F with 7.0 psi RVP
MOBILES
Estimate
4.86
7.61
MOBILE6
Estimate
5.42
10.86
We repeated these calculations for several dozen
combinations of fuel RVP and temperature cycles and then graphed
the resulting averages in Figures 14 through 17. The first three
figures (Figures 14 through 16) are based on January 1, 1995
(thus covering model years 1971 through 1995). Figure 17 is
based at January 1, 1985 (thus, covering model years 1961 through
1985). Therefore, Figures 16 and 17 differ only by the model
years covered. Since both the horizontal and vertical scales
vary among these four graphs, care should be taken in making
comparisons between these figures.
Figure 14
Comparison of Predictions of Full-Day Diurnal Emissions by RVP
For the In-Use Fleet (as of January 1995) with the 60° to 84° F Cycle
In Units of Grams per Day per Vehicle
15
10
U)
o
I
"(5
|
Q
MOBIL E5
-MOBILES
10
12
Fuel RVP (psi)
-------
-32-
Figure 15
Comparison of Predictions of Full-Day Diurnal Emissions by RVP
For the In-Use Fleet (as of January 1995) with the 72° to 96° F Cycle
In Units of Grams per Day per Vehicle
MOBILES
-MOBILES
8 9
Fuel RVP (psi)
10
11
Figure 16
Comparison of Predictions of Full-Day Diurnal Emissions by RVP
For the In-Use Fleet (as of January 1995) with the 82° to 106° F Cycle
In Units of Grams per Day per Vehicle
O 20
ro
c
3
Q
MOBILES
-MOBILES
8
Fuel RVP (psi)
10
-------
-33-
Figure 17
Comparison of Predictions of Full-Day Diurnal Emissions by RVP
For the In-Use Fleet (as of January 1985) with the 82° to 106° F Cycle
In Units of Grams per Day per Vehicle
60
i 40
MOBILES
-MOBILE6
8
Fuel RVP (psi)
10
Based on examinations of these four graphs, we made the
following observations (for Pre-ETP vehicles) :
For the lowest temperature cycle (i.e., daily temperatures
cycling between 60° and 84°F) , the MOBILE6 approach to
predicting fleet full -day diurnal emissions produce
results quite similar to MOBILES for the full range of
fuel RVPs . (Figure 14.)
For the two higher temperature cycles, the MOBILE6
approach to predicting full -day diurnal emissions produce
results similar to MOBILES for fuel RVPs up to 10 psi,
which is a reasonable upper bound for in-use fuel RVP at
those temperatures. (Figures 15 through 17.)
The MOBILE6 approach (compared to MOBILES) predicts
slightly higher full -day diurnal emissions for the lower
RVP fuels and the reverse (i.e., lower diurnal emissions)
for the higher RVP fuels. The range of RVPs for which the
new estimates are higher than the MOBILES estimates varies
with the temperature cycle:
For the low temperature cycle (i.e., 60° to 84°F) ,
the MOBILE6 approach predicts slightly higher diurnal
emissions for fuel RVPs up through 11 psi. For fuel
RVPs near 7 psi, the higher MOBILE6 predictions are
within 0.9 grams of HC (of the MOBILES estimates) per
-------
-34-
day per vehicle undergoing a full (24-hour) diurnal.
This difference gradually shrinks to zero as the fuel
RVP nears 11 psi. For fuel RVPs above 11 psi, the
MOBILE6 estimates are slightly lower than the MOBILES
estimates.
For the moderate temperature cycle (i.e., 72° to
96°F), the new approach predicts slightly higher
diurnal emissions for fuel RVPs up through 10 psi.
For fuel RVPs above 10 psi, the new estimates are
lower than the MOBILES estimates. For RVPs above 11
psi (an unlikely occurrence with this temperature
cycle), the two models move farther apart.
For the high temperature cycle (i.e., 82° to 106°F),
the new approach predicts slightly higher diurnal
emissions for fuel RVPs up through 8.5 psi for the
in-use fleet as of January 1995 (Figure 16). For
fuel RVPs above 8.5 psi, the new estimates are lower
than the MOBILES estimates. For RVPs above 10 psi
(an unlikely occurrence with this temperature cycle),
the two models move farther apart. A snapshot of the
in-use fleet as of January 1985 (Figure 17) yields
similar results.
A more complete comparison of predicted diurnal emissions
(i.e., one that takes activity data into account) can be obtained
by running the two models. Similarly for the other evaporative
emissions.
5.3 Comparisons of Diurnal Emissions from Vehicles Certified to the
Enhanced Evaporative Control Standards
In Section 5.2, we compared, for 1995 and older model year
vehicles, these MOBILE6 estimates of full-day diurnal emissions
with those predicted by the MOBILES model. In this section, we
perform a similar comparison of the two estimates for the
vehicles certified to the enhanced evaporative standard (i.e.,
the 1999 and newer along with some 1996 through 1998 model year
vehicles).
Repeating the process used to create Figures 12 and 13 but
with January 1, 2020 as the evaluation date produced Figures 18
and 19.
-------
-35-
Figure 18
For Enhanced Evaporative Control Vehicles (ETPs)
Comparison of Predictions for In-Use Diurnal Emissions by Age
60° to 84° F Cycle - Using 9.0 psi RVP Fuel
(Non-l/M Area)
In Units of Grams per Day per Vehicle
2 6
u>
O
ro
MOBILES
-MOBILES
10 15
Vehicle Age (years)
20
Figure 19
For Enhanced Evaporative Control Vehicles
Comparison of Predictions for In-Use Diurnal Emissions by Age
82° to 106° F Cycle - Using 7.0 psi RVP Fuel
(Non-l/M Area)
In Units of Grams per Day per Vehicle
2
u>
O
1_
3
b
10 15
Vehicle Age (years)
20
25
-------
-36-
Visual inspections of Figures 18 and 19 leads to the
following conclusions:
MOBILE6 predicts substantially lower full-day diurnal
emissions for these new vehicles than does MOBILES. This
difference is due (almost entirely) to the substantially
lower failure rate (on the purge and pressure tests)
predicted in MOBILE6. (MOBILES uses the same failure
rates for the ETP vehicles as for the Pre-ETPs while
MOBILE6 uses Appendices C, D, and E. See Figure 11 on
page 29.)
For vehicles between 10 and 25 years of age, MOBILE6
predicts a substantial rise in the full-day diurnal
emissions of the in-use fleet (on a per vehicle basis),
but still smaller than the MOBILES predicted increase.
This increase in MOBILE6 predicted emissions is driven
primarily by the increasing numbers of gross liquid
leakers. This is despite the fact that their predicted
rate of occurrence among ETP vehicles is very low (less
than two percent at age 25).
Thus, modifying the assumptions on the frequency of gross
liquid leakers as a function of age among these vehicles
would significantly affect the graphs for "MOBILE6" in
Figures 18 and 19, with that curve remaining much flatter
throughout.
6.0 SUMMARY
Estimates of evaporative emissions in MOBILE6 will be
modeled based on their type:
resting loss emissions,
running loss emissions,
hot soak emissions, and
diurnal emissions.
Each of these types will be calculated based on whether the
individual vehicles:
are gross liquid leakers,
failed the pressure test,
-------
-37-
failed only the purge test, or
passed both the purge and pressure tests.
Once the estimated evaporative emissions of each sub-strata is
calculated, they will be weighted together using the equations
developed in this report.
The analyses suggest that for the full-day diurnal (see
Sections 5.2 and 5.3), MOBILE6:
predicts full-day diurnal emissions lower than does
MOBILES for the 1999 and newer vehicles,
predicts full-day diurnal emissions similar to those of
MOBILES for the 1986-1995 model year vehicles, and
predicts full-day diurnal emissions higher than does
MOBILES for the 1985 and older vehicles.
Similarly, we can perform a simplified analysis to estimate
the effect (benefit) of reducing fuel RVP on full-day diurnal
emissions:
In each of the four figures (Figure 14 through 17), the
MOBILES graph is steeper than the graph of MOBILE6. The
lower slopes associated with the MOBILE6 estimates will
result in smaller decreases in predicted full-day diurnal
emissions associated with a reduction in fuel RVP than the
corresponding changes predicted by MOBILES.
Analyses performed to calculate the effect (e.g., either
the cost per ton or the benefit) will make RVP control
programs slightly less attractive for controlling full-day
diurnal emissions.
The difference in the predicted effect of lowering fuel
RVP is small for lower RVP fuels. Thus, estimating the
effects of reducing the fuel RVP from 8 psi (or from a
lower value) will be similar under both methods.
-------
-38-
7.0 REFERENCES
1) Larry Landman, "Evaluating Resting Loss and Diurnal
Evaporative Emissions Using RTD Tests," Report numbered
M6.EVP.001, April 2001.
2) Larry Landman, "Modeling Hourly Diurnal Emissions and
Interrupted Diurnal Emissions Based on Real-Time Diurnal
Data," Report numbered M6.EVP.002, April 2001.
3) Larry Landman, "Modeling Diurnal and Resting Loss Emission
from Vehicles Certified to the Enhanced Evaporative
Standards," Report numbered M6.EVP.005, April 2001.
4) Larry Landman, "Evaporative Emissions of Gross Liquid Leakers
in MOBILE6," Report numbered M6.EVP.009, April 2001.
5) Louis Browning, "Update of Hot Soak Emissions Analysis"
prepared by Louis Browning of ARCADIS Geraghty & Miller, Inc.
for EPA, Report numbered M6.EVP.004, September 1998
6) Larry Landman, "Estimating Running Loss Evaporative Emissions
in MOBILE6," Report numbered M6.EVP.008, April 2001.
7) D. McClement, J. Dueck, B. Hall, "Measurements of Diurnal
Emissions from In-Use Vehicles, CRC Project E-9", Prepared
for the Coordinating Research Council, Inc. by Automotive
Testing Laboratories, Inc., June 19, 1998.
8) D. McClement, "Real World Evaporative Testing of Late Model
In-Use Vehicles, CRC Project E-41", Prepared for the
Coordinating Research Council, Inc. by Automotive Testing
Laboratories, Inc., December 17, 1998.
9) D. McClement, "Raw Fuel Survey in I/M Lanes", Prepared for
the American Petroleum Institute and the Coordinating
Research Council, Inc. by Automotive Testing Laboratories,
Inc., June 10, 1998.
10) Megan Beardsley, "Estimating Benefits of Inspection/
Maintenance Programs for Evaporative Control Systems," Report
numbered M6.IM.003, November 1999.
-------
-39-
Appendix A
Estimates of Purge/Pressure Strata Size by Vehicle Age
For 1995 and Older Model Years Vehicles
(For Non-l/M Areas)
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Failing
Pressure
Test
3.23%
3.27%
3.40%
3.62%
3.96%
4.44%
5.10%
5.99%
7.18%
8.78%
10.91%
13.70%
17.30%
21 .79%
27.12%
33.07%
39.19%
44.90%
49.76%
53.50%
56.16%
57.92%
59.02%
59.66%
60.03%
60.24%
Failing
Only
Purge
Test
1 .77%
1 .80%
1 .88%
2.02%
2.23%
2.53%
2.95%
3.51%
4.25%
5.23%
6.47%
8.00%
9.76%
11.61%
13.29%
14.51%
15.06%
14.95%
14.41%
13.70%
13.03%
12.50%
12.13%
1 1 .89%
1 1 .74%
1 1 .65%
Passing
Both
Purge and
Pressure
Tests
95.00%
94.93%
94.72%
94.36%
93.81%
93.03%
91 .96%
90.51%
88.57%
85.99%
82.62%
78.30%
72.94%
66.60%
59.58%
52.42%
45.76%
40.14%
35.84%
32.80%
30.81%
29.58%
28.85%
28.45%
28.23%
28.11%
-------
-40-
Appendix B
Predicted Frequency of Occurrence of "Gross Liquid Leakers"
by Emission Type and Vehicle Age
For 1995 and Older Model Years Vehicles
(Reproduced from Report Number: M6.EVP.009 [4])
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Resting
Loss/
Diurnal
0.02%
0.03%
0.04%
0.06%
0.09%
0.13%
0.19%
0.27%
0.39%
0.55%
0.78%
1.08%
1 .49%
2.00%
2.63%
3.36%
4.15%
4.97%
5.75%
6.46%
7.05%
7.54%
7.91%
8.19%
8.40%
8.55%
Running
Loss
0.05%
0.07%
0.11%
0.16%
0.24%
0.35%
0.50%
0.72%
1.02%
1 .40%
1.88%
2.43%
3.02%
3.61%
4.16%
4.62%
5.00%
5.29%
5.51%
5.66%
5.77%
5.84%
5.89%
5.93%
5.95%
5.97%
Hot
Soak
0.07%
0.10%
0.15%
0.23%
0.33%
0.48%
0.70%
1.00%
1.41%
1.95%
2.64%
3.48%
4.46%
5.54%
6.67%
7.83%
8.95%
10.00%
10.94%
11.75%
12.42%
12.94%
13.34%
13.63%
13.85%
14.00%
-------
-41-
Appendix C
Estimates of Purge/Pressure Strata Size by Vehicle Age
For 1999 and Later Model Years Vehicles
(For I/M Areas*)
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Failing
Pressure
Test
0.76%
0.76%
0.77%
0.78%
0.80%
0.82%
0.85%
0.89%
0.93%
0.98%
1 .04%
1.11%
1 .20%
1 .29%
1.41%
1 .54%
1 .69%
1 .86%
2.06%
2.30%
2.56%
2.87%
3.22%
3.62%
4.07%
4.57%
Failing
Only
Purge
Test
0.42%
0.42%
0.42%
0.43%
0.44%
0.46%
0.47%
0.50%
0.52%
0.56%
0.60%
0.64%
0.69%
0.75%
0.82%
0.91%
1 .00%
1.11%
1 .23%
1 .37%
1 .52%
1 .69%
1 .88%
2.08%
2.29%
2.51%
Passing
Both
Purge and
Pressure
Tests
98.83%
98.82%
98.81%
98.79%
98.76%
98.72%
98.67%
98.62%
98.54%
98.46%
98.36%
98.25%
98.11%
97.95%
97.77%
97.56%
97.31%
97.03%
96.71%
96.34%
95.92%
95.44%
94.90%
94.30%
93.64%
92.92%
This assumes that the I/M program requires repairs to vehicles with a MIL
that indicates that there is a problem.
-------
-42-
Appendix D
Estimates of Strata Size by Vehicle Age
For 1999 and Later Model Years Vehicles
(For Non-l/M Areas*)
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Failing
Pressure
Test
0.76%
0.76%
0.77%
0.78%
0.85%
0.94%
1 .06%
1.21%
1 .39%
1.61%
1 .87%
2.18%
2.53%
2.94%
3.42%
3.97%
4.62%
5.36%
6.21%
7.20%
8.34%
9.64%
11.13%
12.82%
14.73%
16.86%
Failing
Only
Purge
Test
0.42%
0.42%
0.42%
0.43%
0.47%
0.53%
0.60%
0.70%
0.81%
0.95%
1.12%
1.31%
1 .53%
1 .79%
2.09%
2.44%
2.83%
3.29%
3.81%
4.40%
5.05%
5.78%
6.58%
7.44%
8.34%
9.27%
Passing
Both
Purge and
Pressure
Tests
98.83%
98.82%
98.81%
98.79%
98.68%
98.53%
98.34%
98.09%
97.79%
97.43%
97.01%
96.52%
95.94%
95.27%
94.49%
93.59%
92.55%
91 .35%
89.98%
88.40%
86.61%
84.57%
82.29%
79.74%
76.92%
73.86%
Up through the age of three (3) years,
are identical to those in Appendix C.
these values
* This assumes that either the geographic area has no I/M program or that
the existing I/M program does not include a check of the OBD MIL.
-------
-43-
Appendix E
Predicted Frequency of Occurrence of "Gross Liquid Leakers"
by Emission Type and Vehicle Age
For 1999 and Newer Model Years Vehicles
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Resting
Loss/
Diurnal
0.02%
0.03%
0.03%
0.04%
0.04%
0.05%
0.06%
0.08%
0.09%
0.11%
0.13%
0.16%
0.19%
0.23%
0.27%
0.33%
0.39%
0.47%
0.55%
0.66%
0.78%
0.92%
1.08%
1 .27%
1 .49%
1.73%
Running
Loss
0.05%
0.06%
0.07%
0.09%
0.11%
0.13%
0.16%
0.20%
0.24%
0.29%
0.35%
0.42%
0.50%
0.61%
0.72%
0.86%
1.02%
1 .20%
1 .40%
1 .63%
1.88%
2.14%
2.43%
2.72%
3.02%
3.32%
Hot
Soak
0.07%
0.09%
0.10%
0.13%
0.15%
0.19%
0.23%
0.27%
0.33%
0.40%
0.48%
0.58%
0.70%
0.83%
1.00%
1.19%
1.41%
1 .66%
1.95%
2.28%
2.64%
3.04%
3.48%
3.96%
4.46%
4.99%
-------
-44-
Appendix F
Response to Peer Review Comments from Sandeep Kishan
This report was formally peer reviewed by one peer reviewer
(Sandeep Kishan). In this appendix, comments from Sandeep Kishan
are reproduced in plain text, and EPA's responses to those
comments are interspersed in indented italics. Each of these
comments refer to page numbers in the earlier draft version
(dated June 10, 1999) that do not necessarily match the page
numbers in this final version.
************************************
This memorandum provides peer review comments on the EPA draft
report: "Estimating Weighting Factors for Evaporative Emissions
in MOBILE6," Document No. M6.EVP.006, June 10, 1999.
Overall, this report was clearly written and the general
methodology seems good. We do not have any recommendations on
alternate datasets. The part of the report that needs a better
explanation is the last paragraph on Page 16 describing the
relationship among the hot soak, RTD, and running loss gross
liquid leakers.
Document No. M6.EVP.006 (June 10, 1999)
We have the following questions, comments, and recommendations on
this draft report. For each item we give the page number and
paragraph that the comment refers to, if it is a specific
comment.
1. Page 3, last paragraph - The text states that approximately
15% of the vehicles did not have purge or pressure tests
performed for one reason or another. The report goes on to
say that the proposal is to assume that these vehicles were
distributed proportionately among the three purge/pressure
strata. Are there any engineering reasons to believe that
these omitted vehicles are different from the vehicles
included in the data analysis with respect to purge/pressure
fail rates?
No, we have no reason to assume (or believe) that these
omitted vehicles (that did not have purge and/or pressure
tests performed) are different from the remaining vehicles.
Which is why we treated them as we did.
2. Page 4, last paragraph - The last sentence in this paragraph
needs to be restated. We are not sure what the author is
trying to say with this sentence. Is the author trying to
say that the CRC data is unreliable since it is not an EPA
-------
-45-
data source? Is the author trying to say that the 57.6%
fail rate is amazingly high and the value cannot possibly be
biased high by EPA since the data is entirely CRC data? The
underlining of the word entirely helps add to the confusion
that this sentence brings to the mind of the reader.
The wording (and underlining) has been changed to avoid
confusion.
3. Page 6, last paragraph in Section 3.1 - How did you
calculate the confidence limits given in Table 2? We would
have used binomial confidence intervals. It appears this is
what the author has done. It should be so stated in the
text.
The reviewer is correct, EPA did use binomial confidence
intervals. This fact has been added to the text just prior
to Table 2.
4. Page 6, last paragraph in Section 3.1 - The author chose to
select the first two of the three categories given at the
beginning of the section to model. Was there any basis for
the selection? The only reason we can see is that modeling
the purge test only failures would appear to be a nightmare
as shown in Figure 3 on page 12.
"Nightmare" is probably too strong a word. We would also
have had some difficulty if we had attempted to directly
model the vehicles that failed only the pressure test.
Since we have three frequency curves that must add to
exactly 100 percent at each age, we only need to determine
two of those curves (the third being the remainder). This
means that there are three different approaches to modeling
these three curves. While each approach yields slightly
different sets of equations, the predicted values are close
for the ages with large sample sizes.
Since the logistic growth curve is monotonic (strictly
increasing or strictly decreasing), EPA chose to model the
two curves most likely to also be monotonic, specifically
the vehicles failing the pressure test (strictly increasing)
and the vehicles passing both tests (strictly decreasing).
(Similarly, the vehicles failing both tests are expected to
have a strictly increasing curve.)
5. Page 8, first full paragraph - This comment concerns the
discussion of the average test date for the dataset.
Clearly, if testing is going on all year long at a
relatively constant rate, the average date will be near July
1. The middle sentence in this paragraph needs to be
reworded. The phrase "the typical test" is a little strange
and gives the impression that 14,000 tests were performed
during one week in July.
-------
-46-
That paragraph has been rewritten to eliminate the
confusion.
6. Page 8, second full paragraph - We understand that weighting
the vehicle age average values they have fewer observations
less than those that have more observations is appropriate
but why should the reciprocal of the variance be used? In
general, we think that a preferable approach is to use
logistic regression on the individual data points, not the
averages by age, to determine the logistic growth function.
Use of logistic regression would automatically take into
account the relatively sparser data for older vehicles and,
in addition, would provide confidence limits on the logistic
growth curve.
That approach was attempted but did not produce satisfactory
predictions of the failure rates of the older vehicles.
That approach gave a relatively little weighting to the
small sample of 66 test vehicles older than 20 years of age.
EPA's approach allows for models (equations) that predict
rates that closely approximated the observed rates listed
(as the four bullets) on page 4.
7. Page 8, second full paragraph - The report states that
Equation I is the resulting function. It would be useful to
the reader for the report to state how this function was
obtained from the data. Was it obtained using logistic
regression, non-linear regression, or some other means?
"Some other means. " Rather than using a "canned" program
that minimized the sum of the squares of the residuals, we
first divided each residual by the corresponding standard
deviation. We then minimized the sum of the squares of
those fractions. Included in that sum was the single value
of the 66 older vehicles (from page 4). The graphical
equivalent of that modified approach is to first take the
rates at ages 0 through 20 plus the average of ages 21-26
(as a single point) with the corresponding 22 confidence
intervals. Then, construct a curve that lies within all of
the confidence intervals and closely approximates all 22
observed rates.
8. Page 16, last paragraph - We did not follow the reasoning
behind the concept that the hot soak gross liquid leakers is
the union of the RTD gross liquid leakers and the running
loss gross liquid leakers. We did not understand the
concept in Reference 4 and the text in this report does not
help us understand the reasoning either. In our opinion,
this paragraph is the weakest part of the report.
That material has been revised.
-------
-47-
9. Page 18, last paragraph - In the text, there is a
typographical error. The value 10,910,000 should be
1,091,000.
The reviewer is correct. The number has been corrected.
10. Page 21, middle of the page - We understand that there is no
data for the late model vehicles which have enhanced
evaporative emissions control systems. We agree that the
replacement of age with age/2 in the equations developed
earlier is appropriate. However, it may be useful to the
reader to explain the assumption that is being made here.
To us, the deterioration of evaporative emission control
systems is contributed to by the effects of age and miles
which can have different types of effects on systems. By
substituting age/2 for age in the equations, the author has
made the assumption that 2 times the miles is approximately
equal to 2 times the age. We suggest a sentence be added in
this vicinity stating that this assumption has been made.
This statement is now on page 22. Point number 1 on page 21
has been revised to include both age and mileage.
11. Page 22 - This comment concerns the assumed values for the
ability of the MIL to detect or identify the vehicles
failing the purge or pressure tests. We understand that you
have done your best to figure out these values of 85, 10,
and 90% but we think that, as the car ages the response will
go down continuously rather than a stepwise fashion. Maybe
an estimate of a logistic growth function would be better to
use here. We realize that there probably is no data.
This suggestion will be considered once actual data become
available.
12. Page 23, first paragraph - It would be helpful to put behind
the value 76.5% the following: (= 0.85 * 0.90).
Done.
13. Page 33 - The first line on the page we believe should read
"eventually will drive." If the occurrence of gross liquid
leakers is really driving the increase in diurnal
hydrocarbon emissions in the older vehicle ages for the 2020
calculation, then we suggest that doing an independent study
of gross liquid leakers on vehicles using enhanced
evaporative control standards will be suggested by people
who are reading this report.
Done.
14. Page 35, top of the page - The report shows preliminary
analyses for full day diurnals and the effects of the
-------
-48-
proposals in this report. We think it would also be useful
to see comparisons of MOBILES versus the proposed
modifications when applied to running losses, hot soaks, and
resting losses.
At the time of this was peer review, the MOBILE6 model was
not fully operational. Since it now is fully operational,
the readers of this report can do their own comparisons
using whatever assumptions they wish.
-------
-49-
Appendix G
Response to Written Comments from Stakeholders
The following comments were submitted in response to EPA's
posting a draft of a related report (M6.IM.003) on the MOBILE6
website. The full text of these comments is posted on the
MOBILE6 website.
Comment Number: 102
Name/Affiliation: David Lax / API
Date: January 25, 2000
Comment:
"Pilot Lane Data from Hammond and Phoenix - ... Problems
with these data (and the analysis) include the fact that all
pre-1996 model year vehicles are combined, even though it is
clear that evaporative control systems improved
significantly between the early-70s and mid-90s."
"In summary, there are some improvements in the evaporative
I/M methodology relative to MOBILES ... However, the
baseline failure rates are too high for the mid-80s to mid-
90s vintage vehicles. This is because all 1971 to 1995
model year vehicles were combined to generate baseline
failure rates."
EPA's Response:
This is a familiar situation. For example, if during 1995,
you wish to obtain (predictions of) test results of 1990
model year vehicles at the age of 15 years, you can either
delay the testing for ten years (i.e., until those vehicles
actually reach the target age), or you can test vehicles
that are then at the target age (i.e., 1980 model year
vehicles) and extrapolate the results.
API's concern is that rather than there being a single curve
estimating the occurrence of pressure failures, there are in
fact two or more curves that are model year specific. While
this concern is a possibility, the data necessary to test
this hypothesis are not yet available. This report
addresses this possibility in the discussion of "gross
liquid leakers" (see the last paragraph on Page 17).
-------
-50-
Comment:
"In additional, the canister-side procedure [for identifying
pressure leaks] can result in failing cars that are not
failures if the technician is not careful. This can occur
if there is a pressure control valve between the tank and
the canister (which most pre-enhanced evap vehicles have).
In this case, the line between the canister and the control
valve is pressurized to 14 inches of water (inH20) (which
has a very small vapor space), and then the pressure bleeds
off through the control valve into the tank. This appears
to be a failure, but in fact is not. Unless there is some
procedure for estimating the volume of air introduced in the
system (and we do not believe there was in this testing),
this can result in false failures."
EPA's Response:
EPA does not believe that a significant number of these
false failures actually occurred. In Table 1 (Page 5), we
note that less than three percent of the vehicles under two
years of age (53 of 1, 767) were identified as having failed
the pressure test. Even if all of these were false failures
(which EPA believes is highly unlikely), then that would
suggest that the "true" failure rate at 20 years of age
would be reduced from 56.2 percent (see Appendix A) down to
53.3 percent. This change is not significant. A more
realistic estimate of false failures would produce even a
smaller, less significant change.
Additionally, if some of the test vehicles were mislabeled
as having (falsely) failed the pressure test, then the
emissions from these problem-free vehicles would have been
included in the average emissions of this stratum, creating
an offsetting error. (That is, predicting a larger number
of failing vehicles, but also predicting lower emissions for
each of those failing vehicles.)
The contractor that performed the actual vehicle testing for
both the EPA and CRC programs (Automotive Testing
Laboratories) also believes that this type of technician
error was unlikely. This contractor stated:
"I disagree with the example about a "careful"
technician not being able to discern a restriction
in the line from the canister to the tank. When
this line is blocked, a 14" pressure is achieved
instantly. When a fuel tank is completely full
and the purge line is completely unrestricted, it
still takes 20 or 30 seconds to fill the system
with enough gas to indicate a 14 inch pressure.
-------
-51-
We deliberately limit the maximum pressure (and
therefore flow) to about 28" pressure to avoid
blowing any lines during the pressurization
process. The typical fill time to perform the
canister end pressure test is 30 to 120 seconds.
The comment about a "careful" technician implies a
technician would not notice the difference between
a one second fill and a thirty second fill. I
would agree that a new technician might not
understand there was a problem, but after a single
day the tech would realize a normal fill and an
abnormal fill, like the one cited.
"I do agree that if the technician is not careful
that errors can be made. I personally feel the
tank pressure method is much more prone to errors
than the canister pressure method. For example, it
is very difficult to judge if the vise grips used
to pinch off the vapor line are in fact sealing
the line. I have seen a number of false failures
that were healed by relocating and reclamping the
line.
"In either case, a careless or hasty technician
can easily report a false failure - the pressure
test takes great care to perform accurately.
"During performance of the Hammond and Phoenix
pilot programs, we were given an opportunity to
monitor the quality of the lane techs in that
vehicles that were returned to the lab were given
a comprehensive inspection. We would notice when
false failures were occurring, and we would
initiate intensive training when more than an
occasional false failure was recorded for vehicles
returning to the lab.
"The most usual cause of false failure was
selection of the wrong line to apply pressure to.
If the technician pressurized a purge line, the
vehicle would not pass the test. Exactly the same
risk would apply to pressurizing from the tank -
if the technician clamps the wrong hose, the
vehicle will fail. Valves and other restrictions
were not found to be a major problem.
"There were restrictions placed the vapor lines of
most vehicles before enhanced evap was put in
-------
-52-
place. Many of these were in the canister, or in
the last few inches of vapor line before the
canister.
"The pressure test performed in Hammond and
Arizona did not include these underhood
restrictors and pressure relief valves - pressure
was introduced between the valve and the fuel
tank. Pressure relief valves located at the fuel
tank could not be bypassed. Vehicles that could
not be pressurized at a normal rate because of
such valves were reported as untestable, or
blocked vapor line, not as pressure failures.
"I do agree the CRC data provide some significant
insight into the two procedures. E-9, E-35, and
E-41 each include tests from the canister to the
tank and from the tank to the canister. Because
the test from the canister to the tank includes
the cap and the seal between the tank and the cap,
the failure rate from the canister to the tank is
always higher than from the tank to the canister.
In these three studies, when there is disagreement
between the two methods, the cap is typically
failed. Pressure levels are reported for the two
methods on each vehicle at initial, one, and two
minute periods. "
------- |