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%

-------