United States Air and Radiation EPA420-P-99-025
Environmental Protection June 1999
Agency M6.EVP.009
vvEPA Evaporative Emissions of
Gross Liquid Leakers in
MOBILE6
> Printed on Recycled Paper
-------
EPA420-P-99-025
- Draft -
Evaporative Emissions of
Gross Liquid Leakers in MOBILE6
Larry C. Landman
Document Number M6.EVP.009
June 30, 1999
U.S. EPA
Assessment and Modeling Division
National Vehicle Fuel and Emissions Laboratory
2000 Traverwood Drive
Ann Arbor, Michigan 48105-2425
734-214-4939 (fax)
mobile@epa.gov
NOTICE
These reports do not necessarily represent final EPA
decisions or positions. They are intended to present
technical analysis of issues using data which are currently
available. The purpose in release of these 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 six previous documents (M6.EVP.001, M6.EVP.002,
M6.EVP.004, M6.EVP.005, M6.EVP.006, and M6.EVP.008), EPA noted
that a potentially significant portion of evaporative emissions
may be the result of a small number of vehicles leaking liquid
gasoline (rather than gasoline vapors). This document describes
this approach and EPA's proposed estimates of both the frequency
of occurrence vehicles with significant leaks of liquid gasoline
and the magnitude of the emissions from those leaks.
Please note that EPA is seeking any input from stakeholders
and reviewers that might aid us in modeling any aspect of resting
loss or diurnal evaporative emissions.
Comments on this report and its proposed use in MOBILE6
should be sent to the attention of Larry Landman. Comments may be
submitted electronically to mobile@epa.gov, or by fax to (734)
214-4939, or by mail to "MOBILE6 Review Comments", US EPA
Assessment and Modeling Division, 2000 Traverwood Drive, Ann
Arbor, MI 48105. Electronic submission of comments is preferred.
In your comments, please note clearly the document that you are
commenting on, including the report title and the code number
listed. Please be sure to include your name, address,
affiliation, and any other pertinent information.
This document is being released and posted. Comments will be
accepted for sixty (60) days. EPA will then review and consider
all comments received and will provide a summary of those
comments, and how we are responding to them.
-------
TABLE OF CONTENTS
Page Number
1.0 Introduction 1
2.0 Characterizing "Gross Liquid Leakers" 1
2.1 "Gross Liquid Leakers" on RTD Test 3
2.2 "Gross Liquid Leakers" on Hot Soak Test ... 8
2.3 "Gross Liquid Leakers" on Running Loss Test . . 12
2.4 Summary of Magnitudes of Evaporative Emissions . 14
3.0 Frequency of Occurrence of "Gross Liquid Leakers" . 15
3.1 First Approach to Estimating Frequency .... 16
3.1.1 On the RTD Test 16
3.1.2 On the Running Loss Test 18
3.1.3 On the Hot Soak Test 19
3.2 Second Approach to Estimating Frequency .... 21
3.3 Selection of Approach to Estimating Frequency . 24
3.4 Overall Occurrence of "Gross Liquid Leakers"
in the In-Use Fleet 27
4.0 References 29
APPENDICES
A. RTD Emissions of 11 Vehicles with Liquid Leaks ... 30
B. Hot Soak Emissions of 14 Vehicles with Liquid Leaks . 31
C. Running Loss Emissions of 10 Vehicles with
Liquid Leaks 32
D. Predicted Frequency of Occurrence of
"Gross Liquid Leakers" 33
11
-------
*** DRAFT ***
Evaporative Emissions of
Gross Liquid Leakers in MOBILE6
Report Number M6.EVP.009
Larry C. Landman
U.S. EPA Assessment and Modeling Division
1 .0 Introduction
In four recently released draft reports [1,2,3,4]* the US
Environmental Protection Agency (EPA) noted that for some
vehicles, the primary mechanism of evaporative emissions was the
substantial leakage of liquid gasoline (as opposed to simply vapor
leaks). In each of those reports, such vehicles were referred to
as "gross liquid leakers." One consistent feature of these
vehicles is that their evaporative emissions far exceed the
evaporative emissions of the vehicles that were not gross liquid
leakers. In this report, EPA will:
• develop a set of criteria to define "gross liquid leakers,"
• determine the evaporative emissions produced by these
"gross liquid leakers," and
• determine the occurrence (i.e., frequency) of these "gross
liquid leakers" as a function of vehicle age.
2.0 Characterizing "Gross Liquid Leaker"
The term "gross liquid leaker" identifies vehicles having
substantial leaks of liquid gasoline, as opposed to simply vapor
leaks. But, this term has been used in different contexts and it
is, therefore, likely that some vehicles that behave as "gross
liquid leakers" based on one type of evaporative emissions test
might not behave as "gross liquid leakers" on another type of
test. In this analysis, EPA makes use of four different types of
testing programs to identify those vehicles with substantial
liquid leaks:
• a real-time diurnal (RTD) test [1,2] in which evaporative
emissions are measured for stabilized test vehicles that
The numbers in brackets refer to the references in Section 4 (page 25).
-------
-2- DRAFT
are enclosed in a sealed housing with the temperatures
cycling over a 24-hour period to simulate the pressure-
driven evaporative HC emissions that result from the daily
increase in ambient temperature,
• a hot soak test [3] in which evaporative emissions are
measured for one hour following a driving cycle for test
vehicles that are enclosed in a sealed housing,
• a running loss test [4] in which evaporative emissions are
measured during a driving cycle for test vehicles that are
enclosed in a sealed housing, and
• a visual inspection [5].
In this report, EPA first estimates the mean evaporative
emissions of these "gross liquid leakers" for each type of test
(Section 2), and then estimates the likelihood of those types of
leaks occurring (Section 3).
Generally, when EPA predicts evaporative emissions (either
resting loss, diurnal, hot soak, or running loss*) these two
variables are critical:
1 ) the ambient temperature and
2 ) the fuel volatility as measured by the Reid vapor pressure
(RVP) of the test fuel.
However, for vehicles that are classified as "gross liquid
leakers," most (but, not necessarily all) of the evaporative
emissions are the result of the leak of liquid gasoline. Since it
is unlikely the rate of leakage is a function of either the
temperature or the fuel volatility, EPA proposes treating the
evaporative emissions of these vehicles as independent of ambient
temperature and RVP.
An additional source of data was a 1998 test program
conducted for the Coordinating Research Council (CRC) in which 50
late-model year vehicles (1992 through 1997, with a mean age of
4.5 years) were tested using the hot soak, running loss, and RTD
tests.[6] However, none of those 50 vehicles had detected liquid
leaks. Thus, the results from these tests were not used in the
analyses in Section 2. The observation that no "gross liquid
leakers" were identified among this sample of 50 vehicles will be
considered in the analysis in Section 3.
* MOBILES will not consider "gross liquid leakers" in its estimates of
evaporative emissions from crankcase losses or refueling. The methodology
for estimating these emissions has not changed from that in MOBILES.
-------
-3- DRAFT
2.1 "Gross Liquid Leakers" on the RTD Test
The category of vehicles identified as "gross liquid leakers"
was first discussed in a report dealing with evaporative emissions
during resting losses and diurnals. In that report, the term
"gross liquid leaker" was used to refer to vehicles which had
resting loss emissions of at least 2.0 grams per hour. The
analyses in that report were based on tests in which the ambient
temperature cycled over 24 hours to simulate (in real-time) a full
day's temperature pattern. The results of those real-time diurnal
(RTD) tests were used to estimate both resting loss and diurnal
emissions. Those analyses were performed on 119 vehicles tested
in various EPA programs plus 151 vehicles tested for the
Coordinating Research Council (CRC) . [1]
Since the 151 vehicles in the CRC program were randomly
recruited (within each of three model year ranges), EPA proposes
to use that random sample to estimate the means of the resting
loss and diurnal emissions of vehicles that had liquid leaks of
gasoline. The mechanics who inspected the test vehicles
identified 32 of those vehicles as having evidence of some fuel
leakage (from damp hoses and connectors to visible leaks).
Since our intention is to only estimate the mean of the
emissions of the vehicles having only substantial leaks (i.e.,
"gross liquid leakers"), we first limited our sample to vehicles:
1.) whose resting loss emissions (i.e., the mean emissions
during the last six hours of the 24-hour RTD test) were at
least 0.25 grams per hour and
2.) whose total RTD emissions were at least 30 grams per day.
These limitations produced a set of vehicles whose gasoline leaks
had an observable effect on the evaporative emissions (even if
that effect was not sufficient to create a "gross liquid leaker").
Eleven such vehicles were found among the 32 having identified
liquid leaks. The emissions from those 11 vehicles are given in
Appendix A. It is important to note that while all of these
vehicles leaked liquid gasoline, less than half of them were
eventually classified as "gross liquid leakers" (i.e., having
resting loss emissions of at least 2.0 grams per hour). All of
these 11 vehicles are carbureted. In the absence of evidence to
the contrary, EPA proposes to treat fuel injected and carbureted
vehicles with liquid leaks the same for the purposes of resting
loss and diurnal emissions.
The usual approach that EPA has followed in estimating
emission levels is to simply calculate the mean of the sample of
applicable test results. However, the number of vehicles
identified as "gross liquid leakers" (i.e., having resting loss
emissions of at least 2.0 grams per hour) is relatively small, and
the range of their emissions is relatively large. From a
-------
-4- DRAFT
statistical standpoint, the combination of these two conditions
may lead to a high degree of uncertainty in the calculated mean.
An alternate approach is to fit an assumed type of distribution
curve to those limited number of observations. The type of
distribution that has historically been used for emissions is the
lognormal distribution [7] (i.e., the logarithms of the emissions,
rather than the emissions themselves, are assumed to be normally
distributed). EPA proposes to use this approach.
Prior to modeling the estimated diurnal emissions, we
reexamined the data in Appendix A. Since our intent was to model
the distribution of diurnal emissions from vehicles with the
severest leaks, we dropped from the analysis the results of
vehicle number 9042 due to its relatively low diurnal emissions
(suggesting that it was not a "gross liquid leaker" relative to
its diurnal emissions). Additionally, we assumed that if a valid
estimate of the diurnal emissions from vehicle 9129 been
obtained*, then that estimated diurnal would have been less than
the emissions from the two highest emitting vehicles but higher
than the emissions from the remaining eight vehicles. Using these
two assumptions, we ranked the diurnal emissions and assigned a
percentile to each. The plot of those percentiles versus the
corresponding diurnal emissions is given in Figure 2-1, on the
following page. The solid line in that figure is the graph of the
cumulative distribution obtained by assuming that the logarithms
of the emissions are normally distributed. (The mean of the
logarithms of the emissions is 3.812; the corresponding standard
deviation is 1.075.) (Distributions other than the lognormal were
examined, but none came as close to approximating the observed
distribution.) We then used that lognormal distribution to
estimate the frequency associated with each possible diurnal
emission level.
In Reference [1], EPA noted that the hourly diurnal emissions from vehicle
number 9129 suggest that the leak actually developed around the tenth hour
of the test. Hence, that vehicle was a "gross liquid leaker" for only the
second half of the RTD test. Trying to precisely estimate the emissions
during the first half of the RTD test, assuming the vehicle had been a
"gross liquid leaker" for the entire test, is questionable. However,
based on the vehicle's emissions for the last 14 hours of the RTD, it
appears that its 24-hour RTD emissions would have fallen between vehicles
number 9054 and 9087.
-------
-5-
DRAFT
Figure 2-1
Cumulative Distribution of Estimated Diurnal Emissions
For Vehicles Exhibiting Liquid Fuel Leaks
With Diurnal Emissions Over 15 grams per day
c
o
—
3
E
3
O
1 0 0 %
75%
50%
25%
0%
1 00 200 300
Diurnal Emissions (grams / day)
400
Although the lognormal distribution predicts that a small
number of vehicles would have impossibly high diurnal emissions,
EPA chose to limit the maximum emissions based on the assumption
that a truly severe leak would result in the vehicle being quickly
repaired. Since one (real world) test vehicle (in our sample) had
diurnal emissions of almost 400 grams per day, EPA assumed that
the limit of the maximum emissions should be higher than that
value. EPA proposes using 1,000 grams per day as the maximum for
the purpose of estimating fleet averages.
The lognormal distribution also predicts that some leaking
vehicles will have diurnal emissions of close to zero. To
separate the "gross liquid leakers" from vehicles having only
minor or moderate leaks, we again examined the estimated diurnal
emissions in Appendix A. A visual inspection of those data
indicated a relatively large discontinuity (i.e., a break) from
24.86 to 62.64 grams per day. Based on that observation, EPA
proposes using 25 grams per day as the minimum value. For a group
of leaking vehicles whose diurnal emissions were between 25 and
1,000 grams per day, the lognormal distribution predicts that the
mean diurnal emissions of that group of leakers would be 104.36
grams per day. (Doubling the maximum possible diurnal to 2,000
-------
-6-
DRAFT
grams per day would result in increasing the estimated group
average only to 107.41 grams daily.)*
EPA proposes to use 104.36 grams per day as the average full-
day's diurnal emissions from "gross liquid leakers" over a day for
which the maximum daily temperature is exactly 24°F above the
daily low temperature. (See report number M6.EVP.002 to use
temperature cycles with ranges other than 24°F.) Earlier versions
of MOBILE limited the pressure driven leaks (i.e., diurnal
emissions) to times when the ambient temperature was at least
40°F. However, we suspect that, at temperatures below 40°F, the
diurnal emissions would still continue. However, at those low
temperatures, the likelihood of ozone exceedences would be small.
The preceding approach was repeated (using the data in
Appendix A) for resting loss emissions. The resting loss
emissions from the 11 vehicles in Appendix A are plotted below in
Figure 2-2.
Figure 2-2
Cumulative Distribution of Resting Loss Emissions
For 11 Vehicles Exhibiting Liquid Fuel Leaks
And Having Resting Loss Emissions Over 0.25 grams / hour
1 0 0 %
0 5 10 15
Resting Loss Emissions (grams / hour)
2 0
* The more traditional approach would have been to simply average the
diurnal emissions of the four vehicles in Appendix A having RTD emissions
of at least 100 grams with the diurnal emissions of two other leakers from
the EPA testing programs. The mean of those six diurnals is 100.29 grams
per day, which corresponds to using the lognormal distribution with the
maximum diurnal emissions set to 675 grams per day.
-------
-7- DRAFT
As with the previous figure (Figure 2-1), the solid line in Figure
2-2 is the graph of the cumulative distribution obtained by
assuming that the logarithms of the resting loss emissions are
normally distributed. (The mean of the logarithms of the resting
loss emissions is 0.841; the corresponding standard deviation is
1.528.) A visual inspection of that figure suggests that the
lognormal model does not fit the resting loss emissions of leaking
vehicles as well as it fit the diurnal emissions. In fact, a
straight line (i.e., a "uniform" distribution) is the curve that
best fits the resting loss emissions for vehicles having at least
1.0 grams per hour.
In previous analyses (see M6.EVP.001), EPA determined that
the lower bound of the resting loss emissions of the gross liquid
leakers would be 2.0 grams per hour. Since one (real world) test
vehicle (in our sample) had resting loss emissions of about 16
grams per hour, EPA assumed that the limit of the maximum
emissions should be higher than that value. EPA proposes using 50
grams per hour as the maximum for the purpose of estimating fleet
averages. For a group of leaking vehicles whose hourly resting
loss emissions were between 2.0 and 50 grams, the lognormal
distribution predicts that the mean resting loss emissions of that
group of leakers would be 9.163 grams per hour.* (Doubling the
maximum possible resting loss to 100 grams per hour would result
in increasing the estimated group average only to 10.875 grams
hourly.) The linear fit (i.e., uniform distribution) predicts the
mean of the resting losses from vehicles emitting at least 2.0
grams per hour would be 10.518 grams per hour. Thus, all of those
approaches produce similar estimates of the average hourly resting
loss emissions from "gross liquid leakers."
Although the uniform distribution produces a superior
estimate of the observed data compared to the lognormal
distribution, both approached produce similar estimates of the
mean resting loss emissions. Therefore, EPA proposes to use the
lognormal distribution for consistency among the various
evaporative models in this report. EPA proposes to use the
estimate based on the lognormal model (i.e., 9.16 grams per hour)
as the average hourly resting loss emissions from "gross liquid
leakers." Since the mechanism responsible for the vast majority
of the resting loss emissions from these vehicles is the fuel
leaking out of the vehicle, and since this process is not
dependent upon the ambient temperature or fuel volatility, EPA had
The more traditional approach would have been to simply average the
resting loss emissions of the five vehicles in Appendix A having resting
loss emissions of at least 2.0 grams per hour with the resting loss
emissions of two other leakers from the EPA testing programs. The mean of
those seven resting losses is 8.84 grams per hour, which corresponds to
using the lognormal distribution with the maximum hourly resting loss
emissions set to 45.2 grams per hour.
-------
DRAFT
proposed (see reference [1]) considering resting loss emissions
from "gross liquid leakers" as independent of fuel volatility and
temperature.
2.2 "Gross Liquid Leakers" on the Hot Soak Test
The category of vehicles identified as "gross liquid leakers"
based on evaporative emissions during a hot soak, was discussed in
a report prepared for EPA by one of its contractors (see the third
footnote on page 1) . In that report, the term "gross liquid
leaker" was used to refer to "vehicles which produce abnormally
high evaporative emissions as a result of a fuel leak and which
have hot soak emissions of over 10 grams per test." Since the hot
soak test is one hour in duration, "grams per test" is equivalent
to "grams per hour" for the hot soak. (See reference [8] to
calculate hot soak emissions for time periods less than an hour.)
In the analyses for that report, hot soak test results on 493
vehicles were used. Of those 493 vehicles, the mechanics
identified 14 as having evidence of some fuel leakage (from damp
hoses and connectors to visible leaks). Those 14 vehicles (along
with their hot soak test results) are listed in Appendix B. The
hot soak emissions of those 14 leaking vehicles ranged from 2.00
to 88.57 grams per test (averaging 22.47 grams). For the
remaining 479 vehicles that did not have detected liquid leaks,
their hot soak emissions ranged from 0.04 to 88.35 grams per test
(averaging 1.77 grams).
A quick inspection of the emissions listed in Appendix B
suggests that the port fuel injected (PFI) vehicles that have
leaks exhibit higher hot soak emissions that the carbureted (CARB)
vehicles that have leaks. Since the fuel delivery systems in the
PFI vehicles operate at a higher pressure than do the systems in
the carbureted vehicles, a hole in the fuel system of a PFI
vehicle will leak more fuel than a hole of the same size in a
carbureted vehicle.* Therefore, the observation that the PFIs
with liquid leaks have (on average) higher hot soak emissions than
the corresponding carbureted vehicles is reasonable. There was an
insufficient sample of leaking vehicles with throttle body
injection (TBI) systems to analyze. Therefore, the hot soak
emissions from this technology grouping will be estimated using a
theoretical rather than statistical approach.
In Figure 2-3 (on the following page), we plotted the hot
soak emissions (in grams per test) of the six carbureted vehicles
(from Appendix B) versus the corresponding percentiles. The solid
line in that figure is the graph of the cumulative distribution
obtained by assuming that the logarithms of the emissions are
Bernoulli's equation indicates that the leak rate will be proportional to
the square root of the ratio of operating pressures.
-------
-9-
DRAFT
normally distributed. (The mean of the logarithms of the hot soak
emissions is 1.9644; the corresponding standard deviation is
0.6963.) As was done in Section 2.1 with diurnal emissions, that
lognormal distribution was used to estimate the frequency
associated with each possible hot soak emission level. Although
the lognormal distribution predicts that a small number of
carbureted vehicles would have impossibly high hot soak emissions,
EPA chose to limit the maximum emissions based on the assumption
that a truly severe leak would result in the vehicle being quickly
repaired. In Appendix B, we can see that one owner tolerated a
vehicle having hot soak emissions of almost 90 grams per test.
Based on that observation, EPA will assume that, for the purpose
of estimating the mean hot soak emissions, the hot soak emissions
of the "gross liquid leakers" range between 10 and 300 grams per
test.
Figure 2-3
Cumulative Distribution of Hot Soak Emissions
For 6 Carbureted Vehicles Exhibiting Liquid Fuel Leaks
1 0 0 %
75% --
50% --
o
'•?
3
'C
+J
Q
o
-Z 25% --
3
E
3
o o%
5 10 15
Hot Soak Emissions (grams / Test)
20
Using the lognormal distribution in Figure 2-3, we can
predict the mean hot soak emissions for the "gross liquid leaking"
carbureted vehicles assuming hot soak emissions ranging between 10
and 300 grams per test. The mean hot soak emissions of that group
of leakers would be 16.9549 grams per test (or per hour). (That
average emission level was not very sensitive to the assumption of
the emissions of the highest possible leaker. Lowering the
-------
-10-
DRAFT
assumed level of the highest emitting carbureted vehicle to 50
grams reduced the average only to 16.5503. Similarly, raising the
assumed level of the highest emitting vehicle to 1,000 grams
increased the average only to 16.9550.) EPA, therefore, proposes
using 16.95 grams per test as the estimate of hot soak emissions
from "gross liquid leaker" carbureted vehicles.
To estimate the mean of the hot soak emissions from the PFI
vehicles that had liquid leaks, we proceeded in the same fashion
that we employed for the carbureted vehicles. In Figure 2-4 (on
the following page), we plotted the hot soak emissions (in grams
per test) of the seven PFI vehicles (from Appendix B) versus the
corresponding percentiles.
Figure 2-4
Cumulative Distribution of Hot Soak Emissions
For 7 PFI Vehicles Exhibiting Liquid Fuel Leaks
1 0 0 %
25 50 75
Hot Soak Emissions (grams / Test)
1 oo
The solid line in Figure 2-4 (above) is the graph of the
cumulative distribution obtained by assuming that the logarithms
of the emissions are normally distributed. (The mean of the
logarithms of the hot soak emissions is 2.8830; the corresponding
standard deviation is 1.5822.) A visual inspection of that figure
suggests that the lognormal model does not fit the hot soak
emissions of leaking PFI vehicles as well as it fit the carbureted
vehicle. In fact, a straight line (i.e., a "uniform"
distribution) provides almost as good a fit to the hot soak
-------
-11- DRAFT
emissions for the six PFI vehicles having at least 2.25 grams per
test. (We are considering the lognormal distribution to be a
better fit because the sum of the squares of the residuals is
lower than for the linear fit.) EPA proposes to use the lognormal
distribution because it is the better fit and for consistency
among the various evaporative models in this report.
Using the lognormal distribution in Figure 2-4, we can
predict the mean hot soak emissions for the "gross liquid leaking"
PFI vehicles assuming hot soak emissions ranging between 10 and
300 grams per test. The mean hot soak emissions of that group of
leakers would be 57.1425 grams per test (or per hour). (That
average emission level is only slightly sensitive to the
assumption of the emissions of the highest possible leaker.
Lowering the assumed level of the highest emitting carbureted
vehicle to 250 grams reduces the average to 53.3468. Similarly,
raising the assumed level of the highest emitting vehicle to 400
grams increases the average only to 63.0990.) The linear fit
(i.e., uniform distribution) predicts the mean of the hot soak
emissions for PFI vehicles emitting at least 10 grams per test
would be 52.2481 grams per test. Thus, all of those approaches
produce similar estimates of the mean hourly resting loss
emissions from "gross liquid leakers." EPA, therefore, proposes
using 57.14 grams per test as the estimate of hot soak emissions
from "gross liquid leaker" PFI vehicles.
Due to a lack of data (see Appendix B), we were not able to
perform a similar analysis for the TBI vehicles. This situation
was addressed in the report on hot soak emissions (M6.EVP.004), in
which the author stated:
"While there is no data on TBI liquid leakers in the
data sets, Bernoulli's equation indicates that the leak
rate for TBI systems would be about one half that for
PFI systems (the square root of the ratio of operating
pressures). Therefore, without further data, the author
suggests assuming that TBI liquid leakers might emit
approximately half the emissions of PFI systems."
EPA proposes to assume that the frequency of having a hole of
a given size is the same for both the TBI and PFI vehicles. Based
on that assumption, Bernoulli's equation predicts that at each
frequency in the cumulative distribution curve for PFIs (i.e.,
Figure 2-4), the corresponding TBI curve would predict only one-
half the hot soak emissions. Thus, if a TBI vehicle were to have
leaks in its fuel system sufficient to produce hot soak test
emissions between 10 and 300 grams per test, then holes of the
same size would result in hot soak test emissions for PFI vehicles
ranging between 20 and 600 grams per test. The lognormal
distribution predicts that for a sample of PFI leakers whose hot
soak test emissions range between 20 and 600 grams per test, the
mean hot soak test emissions would be 89.9979 grams per test.
Thus, the mean TBIs would have hot soak emissions of one-half of
-------
-12- DRAFT
that predicted value which is 44.9990 grams per test. Therefore,
EPA proposes using 45.00 grams per test as the estimate of hot
soak emissions from "gross liquid leaker" TBI vehicles. [As a
test of this approach, we note that the lognormal distribution
predicts the median (i.e., 50 percentile) hot soak test emissions
of leaking (but not necessarily "gross liquid leaking") PFI
vehicles to be 17.8678 grams per test. This would suggest that
the corresponding median value for leaking TBI vehicles would be
half or 8.9339 grams per test which is quite similar to the actual
test result of 8.28 from Appendix B.]
2.3 "Gross Liquid Leakers" on the Running Loss Test
In 1997, running loss tests were performed on 150 vehicles as
part of a testing program conducted for the Coordinating Research
Council (CRC). The mechanics who inspected those test vehicles
identified 40 of those vehicles as having evidence of some fuel
leakage (from damp hoses and connectors to visible leaks). The
running loss emissions for these vehicles were measured over a
single LA-4 driving cycle, using tank fuel (RVP about 6.8 psi),
and ambient temperature about 95 degrees Fahrenheit. [9]
Since our intention is to estimate the mean of the emissions
of the vehicles having only substantial leaks, we first limited
our sample to leaking vehicles whose running loss emissions were
at least 5.0 grams per mile over the single LA-4 driving cycle.
(Five grams per mile appears to be a reasonable break point since
the next highest running loss emissions for a leaking vehicle was
only 3.52 grams per mile.) Ten such vehicles were found among
those 40 having identified liquid leaks. The emissions from those
10 vehicles (reported as grams per mile, grams per test, and grams
per hour) are given in Appendix C. It is important to note that
while all of these vehicles leaked liquid gasoline, not all of
them are classified as "gross liquid leakers" (using the criteria
developed in this section). All of these 10 vehicles are
carbureted. (Two of the original 40 leaking vehicles were fuel
injected; however, their running loss emissions were each less
than 0.4 grams per mile.)
The approach used in the preceding sections (for diurnal,
resting loss, and hot soak) was repeated for running loss
emissions (using the data in Appendix C). The running loss
emissions from the 10 vehicles in Appendix C are plotted (on the
following page) in Figure 2-5. As with the previous figures, the
solid line is the graph of the cumulative distribution obtained by
assuming that the logarithms of the emissions are normally
distributed. (The mean of the logarithms of the emissions is 4.2;
the corresponding standard deviation is 0.88.)
-------
-13-
DRAFT
Figure 2-5
Cumulative Distribution of Running Loss Emissions
For Vehicles Exhibiting Liquid Fuel Leaks
With Running Loss Emissions Over 5 grams per mile
1 0 0 %
0 10 20 30 40
Running Loss Emissions (grams / mile)
50
To determine the appropriate range of running loss emissions
for these "gross liquid leakers," we reexamined the running loss
test results on all 150 vehicles. All of the vehicles that did
not have an identified liquid leak had running loss emissions (for
the single LA-4 cycle) of less than 4.2 grams per mile. EPA
selected 7.0 grams per mile as the value that distinguished
between vehicles that have liquid leaks and those defined as
"gross liquid leakers." Since one (real world) test vehicle (in
the CRC sample) had emissions on the running loss test of about
almost 43 grams per mile, EPA assumed that the limit of the
maximum emissions should be higher than that value. EPA proposes
using 200 grams per hour as the maximum for the purpose of
estimating fleet averages. For a group of leaking vehicles whose
running loss emissions were between 7.0 and 200 grams per mile,
the lognormal distribution predicts that the mean running loss
emissions of that group of leakers would be 17.649 grams per mile.
(As with the emissions on the hot soak and diurnal tests, that
average emission level was not very sensitive to the assumption of
the emissions of the highest possible leaker. Lowering the
assumed level of the highest emitting carbureted vehicle to 90
grams/mile reduced the average only to 17.181. Similarly, raising
the assumed level of the highest emitting vehicle to 500 grams/
mile increased the average only to 17.696.) As previously stated,
-------
-14-
DRAFT
this analysis of running loss emissions of "gross liquid leakers"
is based solely on carbureted vehicles. Using the logic (and
Bernoulli's equation) from Section 2.2, it could be argued that
the running loss emissions from PFI gross liquid leakers would be
four times that amount. However, it does not seem reasonable to
assume such a high emissions rate based on no data. Therefore, in
the absence of evidence to the contrary, (for the purposes of
running loss emissions of "gross liquid leakers") EPA proposes to
treat fuel injected and carbureted vehicles the same.
Thus, EPA proposes using 17.65 grams per mile as the estimate
of the emissions from a running loss test from ALL "gross liquid
leakers" over a single LA-4 driving cycle.
2.4 Summary of Magnitudes of Evaporative Emissions
For the full day diurnal emissions (based on the temperatures
cycling over a 24 degree Fahrenheit range) of "gross liquid
leaking" vehicles, EPA proposes to use 104.36 grams per day. (See
report number M6.EVP.002 to use other temperature cycles or to
estimate hourly diurnal emissions.)
For the resting loss emissions of all "gross liquid leaking"
vehicles, EPA proposes to use 9.16 grams per hour.
To estimate the result of a hot soak test on "gross liquid
leaking" vehicles:
• EPA proposes to use 16.95 grams per test for carbureted
vehicles,
• EPA proposes to use 45.00 grams per test for TBI vehicles,
and
• EPA proposes to use 57.14 grams per test for PFI vehicles.
To calculate the actual hot soak emissions per hour, the resting
loss emissions must be subtracted from the hot soak test
emissions.
To estimate the result of a running loss test on all "gross
liquid leaking" vehicles, EPA proposes to use 17.65 grams per
mile. To calculate the actual running loss emissions, the resting
loss emissions must be subtracted from the running loss test
emissions.
These proposals are summarized in the Table 2-1 on the
following page.
-------
-15-
DRAFT
Table 2-1
Summary of Emissions from "Gross Liquid Leakers"
Type of Emissions (in grams)
• Hot Soak Test (per hour)*
• Resting Loss (per hour)
• Diurnal (per day)
• Running Loss Test (per mile)*
— Fue
Carbureted
1 6.95
Delivery S
TBI
45.00
9.16
ystem —
PFI
57.14
1 04.36
1 7.65
3.0
Both the hot soak and running loss test emissions include resting loss
emissions; therefore, the resting loss emissions must be subtracted.
Frequency of Occurrence of "Gross Liquid Leaker"
In Section 2, the magnitude of each type of evaporative
emissions from liquid leakers was estimated independently using
lognormal distributions; however, EPA believes the data can be
linked when estimating the frequency of the "gross liquid
leakers." Specifically, EPA proposes to make the following two
basic assumptions in predicting the frequency of gross liquid
leakers:
1.) For each test of evaporative emissions (i.e., RTD, hot
soak, and running loss tests), the frequency of gross
liquid leakers increases as a function of age only. This
model of the frequency is based on the assumption that
modern technology vehicles will show the same tendency
toward gross liquid leaks as do the older technology
vehicles at the same age.* In reference number [10], EPA
modifies this assumption for the 1996 and newer vehicles
certified to the new enhanced evaporative standard.
2.) The vehicles classified as gross liquid leakers on the hot
soak test are the same vehicles identified as gross liquid
leakers on either the running loss or RTD tests. (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
* An alternative approach that EPA is not proposing (due to lack of data)
assumes that the modern technology vehicles 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, such as, fuel tank protection and elimination of fuel line leaks).
This approach would result in replacing each single logistic growth
function with a family of two or more curves.
-------
-16-
DRAFT
running loss test.) 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
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).
EPA considered two different approaches to predict the
occurrence of "gross liquid leakers." (See footnote on page 20.)
3.1 First Approach to Estimate Frequency
The first approach involved two basic steps:
1.) Find two logistic growth functions that separately predict
the rate of "gross liquid leakers" on the RTD test and on
the running loss test, respectively.
2.) Verify that the union of those two functions approximate
the results observed on the hot soak test.
3.1.1 First Approach Estimating Frequency of Gross Liquid
Leakers on the RTD Test
In the report dealing with evaporative emissions measured
during the RTD tests (M6.EVP.001), EPA used the results from a
test fleet of 270 vehicles (i.e., the combined EPA and CRC
samples) to estimate the occurrence of gross liquid leakers within
each of the three model year ranges used in the recruitment
process (the pre-1980, 1980-85, and 1986-95 vehicles). The
estimated rate of occurrence of the "gross liquid leakers" is
reproduced in Table 3-1 (below). The large confidence intervals
are the result of the relatively small sample sizes.
Table 3-1
Frequency of Gross Liquid Leakers
Based on RTD Testing
Vehicle
Age (years)
6.12
13.00
21.79
Sample
Size
85
50
51
Frequency
0.20%
2.00%
7.84%
Standard
Deviation
1.41%
1.98%
3.76%
90% Confide
Lower
0.00%
0.00%
1.65%
mce Interval
Upper
2.52%
5.26%
14.03%
"Vehicle Age" was calculated by subtracting the model year
from the test year and then adding one-half to simulate the
rate as of January first.
-------
-17-
DRAFT
In that earlier report (M6.EVP.001), EPA then derived a
logistic growth curve that exactly fit those three data points
(from Table 3-1). The equation of that function is given below:
Rate of Gross Liquid Leakers
Based on RTD/Resting Loss Testing =
0.08902
1 + 414.613*exp[-0.3684*AGE]
The predicted occurrences of "gross liquid leakers" based on
this equation are given in Appendix D. The frequencies from Table
3-1 are plotted in Figure 3-1 (below). Also graphed in that
figure are the 90 percent confidence intervals (as dotted lines)
from Table 3-1 and the predicted frequencies (as the solid line)
from Appendix D (or from the preceding equation).
Figure 3-1
Predicted Frequency of Gross Liquid Leakers
Based on RTD Testing
15%
55 10% --
o
c
0)
3
o-
0)
5% --
0%
1 0 20
Vehicle Age (years)
30
After EPA had obtained the equation at the top of this page,
additional test data were provided by CRC (project number E-41).
Specifically, a test program run during 1998 found no "gross
liquid leakers" on the RTD test in a sample of 50 late-model year
vehicles (1992 through 1997, with a mean age of 4.5 years). (See
reference [6].) Those test results are consistent with the
preceding equation.
-------
-18-
DRAFT
3.1.2 First Approach Estimating Frequency of Gross Liquid
Leakers on the Running Loss Test
For the 150 vehicles in the CRC running loss testing
program, the occurrence of "gross liquid leakers" (i.e., the six
vehicles in Appendix B whose running loss emissions exceeded 7.0
grams/mile), the occurrence of gross liquid leakers was calculated
within each of the three model year ranges used in the recruitment
process (the same model year ranges used in the RTD testing).
Those estimated rates of occurrence of the "gross liquid leakers"
appear below (in Table 3-2). The large confidence intervals are
again the result of the relatively small sample sizes.
Table 3-2
Frequency of Gross Liquid Leakers
Based on Running Loss Testing
Vehicle
Age (years)
8.84
14.24
22.48
Sample
Size
50
39
61
Frequency
2.00%
5.13%
4.92%
Standard
Deviation
1.98%
3.53%
2.77%
90% Confide
Lower
0.00%
0.00%
0.36%
mce Interval
Upper
5.26%
10.94%
9.47%
It was not possible to exactly fit the frequencies in Table 3-2
with an increasing function (since the observed frequency seem to
drop after age 14.24 years). EPA derived a logistic growth curve
that best fit those three data points. The equation of that
function is:
Rate of Gross Liquid Leakers
Based on Running Loss Testing
0.06
1 + 1 20 * exp[-0.4 * A G E ]
The predicted occurrences of "gross liquid leakers" based on
that equation are also given in Appendix D. The frequencies from
Table 3-2 are plotted below in Figure 3-2. Also graphed in that
figure are the 90 percent confidence intervals (as dotted lines)
from Table 3-2 and the predicted frequencies (as the solid line)
from Appendix D (or from the preceding equation). As can be seen
(either in Figure 3-2 or by comparing Table 3-2 with Appendix D),
the logistic growth curve is within one percentage point of the
observed occurrences at each of the three age points. (Also, the
predicted frequencies are within 40 percent of the standard
deviation of the observed frequencies at each of the three
points.)
-------
-19-
DRAFT
Again, the newly acquired data (noted at the end of Section
3.1.1) in which no "gross liquid leakers" were found during
running loss testing in a sample of 50 late-model year vehicles
(mean age of 4.5 years) are consistent with that preceding
equation.
Figure 3-2
Predicted Frequency of Gross Liquid Leakers
Based on Running Loss Testing
15%
10%
>
u
o>
a-
a>
1 0 20
Vehicle Age (years)
30
3.1.3 First Approach Estimating Frequency of Gross Liquid
Leakers on the Hot Soak Test
To estimate the rate of occurrence of "gross liquid leakers"
on the hot soak test, we first referred to the second assumption
on page 15, which states that the collection of vehicles that are
"gross liquid leakers" on the hot soak test is the union of the
collection of vehicles identified as "gross liquid leakers" on the
running loss test with the collection of vehicles identified as
"gross liquid leakers" on the RTD test. Thus, we were able to
estimate the rate of "gross liquid leakers" on the hot soak test
based solely on the rates of "gross liquid leakers" on the running
loss and RTD tests. In the last column of Appendix D, the rate of
"gross liquid leakers" on the hot soak was calculated by adding
the two preceding columns and then subtracting the product of
those two columns.
To test the reasonableness of the results of that assumption,
we identified the six vehicles (in the hot soak testing program of
-------
-20-
DRAFT
300 vehicles conducted for Auto Oil) that had hot soak test
emissions in excess of 10 grams per test. In this testing
program, the test fleet was again stratified into three model year
ranges, but they were different groupings (1983-85, 1986-90, and
1991-93) . This resulted in a sample of newer vehicles than were
used in the RTD or running loss testing programs.* Those
estimated rates of occurrence of the "gross liquid leakers" within
each of the three new model year ranges appear below in Table 3-3.
The large confidence intervals are again the result of the
relatively small sample sizes. We then compared those observed
rates (in Table 3-3) with the predicted rates in Appendix D.
Table 3-3
Frequency of Gross Liquid Leakers
Based on Hot Soak Testing
Vehicle
Age (years)
1.98
5.55
9.38
Sample
Size
66
166
64
Frequency
1.04%
1.20%
6.25%
Standard
Deviation
1.25%
0.85%
3.03%
90% Confide
Lower
0.00%
0.00%
1.27%
mce Interval
Upper
3.10%
2.60%
11.23%
The observed frequencies from Table 3-3 are plotted in Figure 3-3
(on the following page). Also graphed in that figure are the 90
percent confidence intervals (as dotted lines) from Table 3-3 and
the predicted frequencies (as the solid line) from Appendix D.
Those predicted occurrences from Appendix D are based not on hot
soak test results, but on results of running loss tests and RTD
tests.
Comparing, in Figure 3-3, the predicted rates of "gross
liquid leakers" occurring with the observed rates of "gross liquid
leakers" on the hot soak test, we observe:
• the predicted rates are all lower than the observed rates
which were based on relatively small samples, but
• the predicted rates are all within the 90 percent
confidence intervals of the observed rates (at each of the
three points).
* Since none of the mean ages in Table 3-3 exceeded 10 years, EPA chose
approaches different from those used with the diurnal or running loss
emissions. Rather than predicting the occurrence on the hot soak test of
"gross liquid leakers" among older vehicles based only on data from newer
vehicles, EPA proposes to estimate those rates based on the rates of
"gross liquid leakers" on both the RTD an running loss tests.
-------
-21-
DRAFT
These differences between the predicted and observed rates may
simply be the result of the small sample sizes.
Figure 3-3
Predicted Frequency of Gross Liquid Leakers
On the Hot Soak Test
Based on RTD and Running Loss Testing
u
0)
3
0)
15%
10%
5%
0%
1 0 20
Vehicle Age (years)
30
Again, the newly acquired data (noted at the end of Sections
3.1.1 and 3.1.2) in which no "gross liquid leakers" were found
during hot soak testing in a sample of 50 late-model year vehicles
(mean age of 4.5 years) are consistent with the preceding hot soak
predictions.
3.2 Second Approach to Estimate Frequency
The second approach employed by EPA was to use all of the
observations (in Tables 3-1 through 3-3) to find logistic
functions that optimize (simultaneously) all of the predictions.
This approach produced the following two equations:
Rate of Gross Liquid Leakers
Based on RTD/Resting Loss Testing =
0.0865
1 + 55 * exp[-0.259 * A G E ]
Rate of Gross Liquid Leakers
-------
-22-
DRAFT
Based on Running Loss Testing
0.058
1 + 70 * exp[-0.48 * A G E ]
These two equations (and their union which estimates "gross
liquid leakers" on hot soak tests) predict rates of occurrence
that are all within one-half of the corresponding standard
deviations at each of the nine observations (in Tables 3-1 through
3-3). We can again graph those data (i.e., observed rates and
confidence intervals) from Tables 3-1 through 3-3, but now in
figures with curves from these new predictions (Figures 3-4
through 3-6). The only differences between the three figures in
Section 3.1 and these new corresponding figures are the solid
lines designating the predicted frequencies.
Figure 3-4
Predicted Frequency of Gross Liquid Leakers
Using Second Approach
Based on RTD Testing
15%
0%
1 0 20
Vehicle Age (years)
30
-------
-23-
DRAFT
Figure 3-5
Predicted Frequency of Gross Liquid Leakers Using Second Approach
Based on Running Loss Testing
15%
0%
1 0 20
Vehicle Age (years)
30
Figure 3-6
Predicted Frequency of Gross Liquid Leakers Using Second Approach
On the Hot Soak Test
Based on RTD and Running Loss Testing
15%
0%
1 0 20
Vehicle Age (years)
30
-------
-24-
DRAFT
A visual inspection of these three figures (3-4 through 3-6)
indicates that this approach produces predicted rates (of the
occurrence of "gross liquid leakers") that are all well within the
90 percent confidence intervals of the observed rates (at each of
the nine points). In fact (as noted earlier in this section), all
nine predict rates are within one-half of the corresponding
standard deviations at each of the observations.
3.3
Selection of Approach to Estimate Frequency
In choosing between these two methods (which in EPA's opinion
are the two best candidates) of predicting the frequency of "gross
liquid leakers," we first observed that the greatest difference
between these two methods was in estimating the rate of "gross
liquid leakers" on the hot soak test. In the following graph
(Figure 3-7), we reproduced the estimated frequency curves from
Figures 3-3 and 3-6. In this figure, the "dashed" line is the
estimate produced using the first method (i.e., from Figure 3-3 in
Section 3.1.3), and the solid line is the estimate produced using
the second method (i.e., from Figure 3-6 in Section 3.2).
Figure 3-7
Comparing Predicted Frequency of Gross Liquid Leakers
On the Hot Soak Test
15%
0%
1 0 20
Vehicle Age (years)
30
A visual inspection of this figure indicates that:
1 The two predicted rates are similar for vehicles at least
17 years of age or older.
-------
-25-
DRAFT
• For vehicles newer than 17 years of age, the second method
predicts a substantially higher occurrence of "gross liquid
leakers." (For vehicles up through the age of 10, the
second method predicts more than twice as many "gross
liquid leakers" as does the first method.)
To decide between these two models, EPA made use of a recent
testing program run jointly by the CRC and the American Petroleum
Institute (API). This program was specifically designed to
determine the frequency of vehicles with liquid leaks. Since
actual measurements of evaporative emissions were not performed in
this program, we cannot determine which of those vehicles
identified as having liquid leaks would have met our criteria for
"gross liquid leakers." [5]
In that API/CRC program, 1,000 vehicles were inspected for
any signs of leaks with the engine operating (during at least a
portion of the visual inspection). (This protocol was expected to
permit identification of vehicles exhibiting fuel leaks on the
RTD, hot soak, or running loss tests.) The vehicles were then
classified by the mechanic according to the severity of the
observed leaks. The visible liquid leaks were classified as
either:
• small liquid leaks (e.g., single drops) or
• larger leaks (e.g., steady flow of drops).
This classification was based on a visual inspection rather than
on the results of a test of the actual evaporative emissions. The
results of that study are summarized in the following table:
Table 3-4
Frequency of Leaking Vehicles
In API/CRC Testing Program
Model
Year
Range
Pre-80s
80-85
86-91
92-98
Mean
Age
(years)
22.329
14.394
9.429
3.979
Sample
Sizes
70
155
352
423
Vehicles
with
Small
Leaks
5
1 0
2
0
Vehicles
with
Larger
Leaks
2
1
2
0
Total
with
Any
Leaks
7
1 1
4
0
90% Conf Interval
Lower
4.10%
3.70%
0.21%
0.00%
Upper
15.90%
10.49%
2.07%
0.49%
The 90 percent confidence intervals in Table 3-4 are based on the
(total) number of vehicles with either small or large visible
leaks. Those vehicles which were identified as having large
visible liquid fuel leaks were almost certainly "gross liquid
-------
-26-
DRAFT
leakers," and many of the vehicles which were identified as
having small visible liquid fuel leaks were possibly "gross liquid
leakers" as well. Thus, EPA considers the upper bound of the
confidence intervals as a conservative estimate of the occurrence
of the "gross liquid leakers." If we reproduce Figure 3-7, and
include the 90 percent confidence intervals from Table 3-4 (as
dotted lines), we produce Figure 3-8:
Figure 3-8
Comparing Predicted Frequency of Gross Liquid Leakers
On the Hot Soak Test
15%
0%
1 0 20
Vehicle Age (years)
30
A visual inspection of Figure 3-8 strongly suggests the
second method for predicting the frequency of "gross liquid
leakers" over predicts the actual occurrence of "gross liquid
leakers" for vehicles under the age of 13 years. (The conclusion
that the second method "OVER PREDICTS" the frequency is based on the
results of the API/CRC testing program, primarily the relatively
large sample sizes in Table 3-4 compared with those in Table 3-3.)
Therefore, EPA proposes to use the first method (Section 3.1)
to estimate the frequencies of the occurrence of "gross liquid
leakers" on the three types of tests for evaporative emissions.
The results of that method are given in Appendix D.
-------
-27- DRAFT
3.4 Overall Occurrence of "Gross Liquid Leakers" in the In-
Use Fleet
The equations in Section 3.1 (or the results in Appendix D)
predict the occurrence of "gross liquid leakers" identified on the
RTD test to range between 0.02 to 8.55 percent by vehicle age, and
for those identified on the running loss test to range between
0.05 and 5.97 percent by vehicle age. It is reasonable to ask
what is the overall percentage of these vehicles in the entire
in-use fleet. To answer that question, we referred to another
report which provides an estimate of the national distribution by
age of light-duty vehicles (LDVs) and light-duty trucks (LDTs).
(See reference [11].) Applying the percentages from Appendix D to
those estimated vehicle counts produces Table 3-5 on the following
page. The predicted total counts in Table 3-5 suggest that "gross
liquid leakers" represent approximately 1.2 to 1.6 percent of the
entire in-use fleet.
-------
-28-
DRAFT
Table 3-5
Predicted Occurrence of Gross Liquid Leakers
In the National In-Use Fleet of LDVs and LDTs
(as of January 1995)
Calendar
Year Minus
Model Year
0
1
2
3
4
5
6
7
8
9
1 0
1 1
1 2
1 3
1 4
1 5
1 6
1 7
1 8
1 9
20
21
22
23
24 and older
TOTALS:
Vehicle
Counts
9,581,160
12,690,223
12,595,718
12,479,871
12,328,489
12,124,815
11,850,006
11,484,1 10
11,007,677
10,404,139
9,663,040
8,783,860
7,508,980
6,076,245
4,896,767
3,929,300
3,140,650
2,503,094
2,030,454
1,710,242
1,451,096
1,240,664
1,069,132
928,705
3,724,043
175,202,480
"Gross Liquid Leakers"
Identified on:
RTD
2,052.19
3,924.61
5,621.77
8,033.14
11,433.59
16,178.24
22,702.78
31,499.85
43,050.78
57,685.81
75,350.55
95,286.08
111,677.02
121,573.75
128,727.45
131,947.97
130,511.75
124,468.78
116,862.35
110,464.75
102,385.03
93,514.33
84,580.52
76,080.74
312,764.31
2,018,378
Running Loss
4,750.99
9,349.56
13,760.93
20,159.55
29,321.45
42,197.16
59,817.53
83,045.60
112,104.66
145,891.73
181,302.31
213,090.08
226,678.25
219,360.63
203,502.92
181,708.71
157,112.78
132,479.39
11 1,810.15
96,801.22
83,696.51
72,483.90
63,008.21
55,054.76
221,641.23
2,740,130
-------
-29- DRAFT
4.0 References
1) Larry Landman, "Evaluating Resting Loss and Diurnal
Evaporative Emissions Using RTD Tests," Report numbered
M6.EVP.001, July 1999.
2) Larry Landman, "Modeling Hourly Diurnal Emissions and
Interrupted Diurnal Emissions Based on Real-Time Diurnal
Data," Report numbered M6.EVP.002, July 1999.
3) 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
4) Larry Landman, "Estimating Running Loss Evaporative Emissions
in MOBILE6," Report numbered M6.EVP.008, June 1999.
5) 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.
6) 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.
7) Melvin Ingalls, "Mobile Source Exposure Estimation," prepared
by Southwest Research Institute for EPA, EPA Report Number
EPA460/3-84-008, March 1984, Appendix A.
8) Edward L. Glover, "Hot Soak Emissions as a Function of Soak
Time," Report numbered M6.EVP.007.
9) D. McClement, "Measurement of Running Loss Emissions from In-
Use Vehicles (CRC Project E-35)", CRC Report No. 611, Prepared
for the Coordinating Research Council, Inc. by Automotive
Testing Laboratories, Inc., February 1998.
1 0) Larry Landman, "Modeling Diurnal and Resting Loss Emissions
from Vehicles Certified to the Enhanced Evaporative
Standards," Report numbered M6.EVP.005.
1 1 ) Tracie R. Jackson, "Fleet Characterization Data for MOBILE6:
Development and Use of Age Distributions, Average Annual
Mileage Accumulation Rates, and Projected Vehicle Counts for
Use in MOBILE6," Report numbered M6.FLT.007.
-------
-30-
DRAFT
Appendix A
RTD Emissions of 11 Vehicles with Liquid Leaks
With RTD > 30 and Resting Loss > 0.25
(Arranged in Increasing Order of Estimated Resting Losses)
(ALL of the Leaking Vehicles Were Carbureted)
Vehicle
Number
9095
9037
9046
9042
9098
9148
9049
9054
9129
9087
9111
Real-Time
Diurnal
(RTD)
(grams / day)
32.26
33.44
33.76
30.88
45.21
47.97
181.35
316.59
181.79
478.16
777.14
Estimated
Rst Loss
(at 72°F)
(hourly)
0.28
0.47
0.62
0.89
0.90
1.27
4.87
10.58
10.77
14.12
16.51
Estimated
Diurnal
(grams / day)
24.85
21.47
18.21
8.83
22.91
16.63
64.55
62.64
Ignore*
139.22
380.79
An examination of the hourly RTD data from this vehicle
(in reference [1]) suggests that the leak actually
developed around the tenth hour of the 24-hour test.
While the resting loss estimate (based on hours 19
through 24) is most likely valid, the estimate of diurnal
emissions is unreliable.
Note that while aM. 11 of these vehicles are liquid leakers
most of them do not qualify as "gross liquid leakers."
-------
-31-
DRAFT
Appendix B
Hot Soak Emissions of 14 Vehicles with Liquid Leaks
(With Hot Soak Emissions At Least 2.0 grams / test)
Sorted by Fuel Delivery System
In Increasing Order of Emissions
Program
Auto Oil
EPA
EPA
Auto Oil
EPA
EPA
Vehicle
Number
134
177
122
79
173
97
Fuel
System
GARB
GARB
GARB
GARB
GARB
GARB
Temp
(°F)
94
95
105
92
92
110
RVP
(psi)
6.0
6.1
6.1
7.0
6.7
6.7
Hot Soak
(grams HC)
2.54
4.63
5.53
9.49
14.53
14.66
Program
EPA
Vehicle
Number
143
Fuel
System
TBI
Temp
(°F)
94
RVP
(Psi)
6.4
Hot Soak
(grams HC)
8.28
Program
Auto Oil
Auto Oil
Auto Oil
EPA
Auto Oil
EPA*
EPA*
Vehicle
Number
35
199
47
33
276
372
266
Fuel
System
PFI
PFI
PFI
PFI
PFI
PFI
PFI
Temp
(°F)
104
96
93
113
87
106
105
RVP
(psi)
6.7
6.5
6.1
6.0
6.3
9.0
9.0
Hot Soak
(grams HC)
2.00
2.26
11.56
46.95
49.39
54.18
88.57
* These two vehicles were tested using a substantially
more volatile fuel.
-------
-32-
DRAFT
Appendix C
Running Loss Emissions of 10 Vehicles with Liquid Leaks
(With Running Loss Emissions At Least 5.0 grams / mile)
(Arranged in Increasing Order of Estimated Resting Losses)
(ALL of the Leaking Vehicles Were Carbureted)
Vehicle
Number
35044
35125
35099
35085
35045
35071
35047
35129
35054
35091
Running
Loss HC
(grams / mile)
5.009
5.297
5.649
6.880
7.469
9.175
13.480
13.566
24.841
42.973
Running
Loss HC
(grams / LA-4)
37.47
39.44
42.17
51.18
55.79
68.84
100.19
100.72
184.96
318.90
Running
Loss HC
(grams / hour)
98.32
103.49
110.65
134.29
146.39
180.63
262.89
264.28
485.32
836.76
-------
-33-
DRAFT
Appendix D
Predicted Frequency of Occurrence of "Gross Liquid Leakers"
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
1 1
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%
1 1 .75%
12.42%
12.94%
13.34%
13.63%
13.85%
14.00%
------- |