United States       Air and Radiation      EPA420-R-01-022
            Environmental Protection               April 2001
            Agency                     M6.EVP006
vvEPA     Estimating Weighting
            Factors for Evaporative
            Emissions in MOBILE6
                                 > Printed on Recycled Paper

-------
                                                                          EPA420-R-01-022
                                                                                 April 2001
                                                             for
                                                 in

                               M6.EVP.006
                                Larry C. Landman

                         Assessment and Standards Division
                       Office of Transportation and Air Quality
                       U.S. Environmental Protection Agency
                                    NOTICE

    This technical report does not necessarily represent final EPA decisions or positions.
It is intended, to present technical analysis of issues using data which are currently available.
         The purpose in the release of such reports is to facilitate the exchange of
      technical information and to inform the public of technical developments which
        may form the basis for a final EPA decision, position, or regulatory action.

-------
                            ABSTRACT
     In parallel documents (M6.EVP.001,  M6.EVP.002,  and
M6.EVP.005), EPA identified the methods used in MOBILE6 for
estimating resting loss and diurnal emissions based (in part)  on
the performance of the vehicles (i.e.,  pass or fail)  on the purge
and pressure tests.  EPA computes model-year and age specific
average resting loss and diurnal emissions by weighting together
the emissions of passing and failing vehicles according to their
frequency in the in-use fleet.  This document describes this
approach and EPA's predictions of pass and fail rates (i.e.,
weighting factors) as functions of vehicle age.

     This report was originally released (as a draft)  in June
1999.  This current version is the final revision of that draft.
This final revision incorporates suggestions and comments
received from stakeholders during the 60-day review period and
from peer reviewers.

-------
                        TABLE OF CONTENTS
                                                      Page Number
1.0  Introduction   	    1
2.0  Data Sources   	    2
3.0  Analysis   	    6
     3.1  Strata Based on Purge and Pressure Tests  ...    6
           3.1.1 Vehicles  Failing  the  Pressure Test  .  .    6
           3.1.2 Vehicles  Passing  Both the  Pressure
                 and  the Purge  Tests	    9
           3.1.3 Vehicles  Failing  ONLY the  Purge  Test  .   11
           3.1.4 Summary of  Purge  and  the Pressure
                 Failure Rates  	   13
           3.1.5 Vehicles  Failing  Both the  Pressure
                 and  the Purge  Tests	   13
           3.1.6 Effects of  Inspection/Maintenance  ...   16
     3.2  Modeling "Gross Liquid Leakers"    	   16
     3.3  Combining Purge/Pressure Rates with
          Gross Liquid Leaker Rates  	   18
4.0  Modeling Enhanced EVAP Vehicles Equipped with OBD  .   21
5.0  Comparisons with MOBILES   	   25
     5.1  Comparisons of Weighting Factors  	   25
     5.2  Comparisons of Weighted Diurnal  Emissions   .  .   28
     5.3  Comparisons of Diurnal Emissions from Vehicles
          Certified to the Enhanced Evaporative Control
          Standards   	   34
6.0  Summary	   36
7.0  References  	   38
                                11

-------
                    TABLE OF CONTENTS (Continued)
APPENDICES                                             Page Number

 A.  Estimates of Purge/Pressure Strata Size by
     Vehicle Age for 1995 and Older Model Year
     Vehicles for Non-I/M Areas   	   39

 B.  Predicted Frequency of Occurrence of "Gross
     Liquid Leakers" by Emission Type and Vehicle
     Age (for 1995 and Older Model Year Vehicles) ....   40

 C.  Estimates of Purge/Pressure Strata Size by
     Vehicle Age for 1999 and Newer Model Year
     Vehicles for I/M Areas   	   41

 D.  Estimates of Purge/Pressure Strata Size by
     Vehicle Age for 1999 and Newer Model Year
     Vehicles for Non-I/M Areas   	   42

 E.  Predicted Frequency of Occurrence of "Gross
     Liquid Leakers" by Emission Type and Vehicle
     Age (for 1999 and Newer Model Year Vehicles)   ...   43

 F.  Peer Review Comments from Sandeep Kishan   	   44

 G.  Comments from Stakeholders   	   49
                               111

-------
                   Estimating Weighting Factors
               for Evaporative Emissions in MOBILE6
                    Report Number M6.EVP.006

                         Larry C.  Landman
            U.S.  EPA Assessment and Standards Division
1.0   INTRODUCTION

     In three parallel  reports  [1,2,3]*,  the US Environmental
Protection Agency  (EPA) developed methods of estimating the
resting loss and diurnal  emissions from results of real-time
diurnal (RTD) tests of  in-use vehicles in which the ambient
temperature cycled over a 24 degree Fahrenheit range to simulate
in real-time the daily  heating  and cooling that parked vehicles
experience over a 24 hour period.   For many of the vehicles used
in these studies, the recruitment method was designed to recruit
a relatively large number of vehicles  that had problems with
their evaporative control systems.   Specifically,  two tests of
the integrity of each vehicle's evaporative control system were
used to screen the candidate vehicles  (a pressure test** and a
purge test).  This recruitment  bias did not affect the analysis
of these data as described in earlier  reports; since those
analyses were performed within  each purge/pressure grouping, the
selection was random within the purge/pressure and model year
groups.  However, to correctly  represent the entire in-use fleet
the results must be weighted.   In this report, EPA develops
weightings for each stratum to  estimate the emissions of the
entire in-use fleet.  EPA will  use these factors to weight
together the results of the RTD tests,  as well as the results of
hot soak tests and running loss tests  which were also derived
from measurements of a  stratified sample.

     For each of the parallel analyses of resting loss and
diurnal data, the sample  of test vehicles was divided into four
strata.  The first of these strata consisted of several vehicles
having substantial leaks  of liquid gasoline (as opposed to simply
vapor leaks); these vehicles were labeled "gross liquid leakers"
(GLLs)  .  EPA will use the following three definitions [4] (based
*  The numbers in brackets  refer to the references in Section 7 (page 39).

** This pressure  test was  performed by disconnecting the vapor line at  the
   canister and  then pressurizing the tank  from that position with the  gas
   cap in its  normal  position.   This  procedure  differs from  the method
   currently being used in  Inspection and Maintenance  (I/M)  lanes.

-------
                               -2-
on the evaporative emissions test used) for  such vehicles with:

    •  resting loss emissions (i.e., the mean emissions during
       the last six hours of the 24-hour RTD test) were at least
       2.00 grams per hour  (see also reference [1]),  or

    •  hot soak test emissions were at least 10.00 grams per one-
       hour test (see also reference [5]) ,  or

    •  running loss test emissions were at least 7.00 grams per
       mile over a LA-4 driving cycle  (or 137.2 grams per hour)
       (see also reference [6]) .

These three different definitions will identify potentially
different sets of vehicles as being "gross liquid leakers"  (see
Section 3.2).   For the remaining three strata, we used the
results of the purge and pressure tests to match the
stratification of the recruitment process.   This approach
produces the following three additional strata:

   1)   vehicles that pass both the purge and pressure tests,

   2)   vehicles that fail the pressure test  (regardless of their
       performance on the purge test),* and

   3)   vehicles that fail only the purge test.

     This document reports on EPA's proposal to weight those four
strata together to obtain estimates (of running loss, hot soak,
resting loss,  and diurnal emissions) for the entire in-use fleet.


2.0   DATA SOURCES

     To develop the appropriate weighting factor for  the stratum
of vehicles identified as "gross liquid leakers," EPA relied on
five groups of data to estimate the frequency of the  occurrence
of these vehicles  (see also reference  [4]) :

    •  For the "gross liquid leakers"  identified by the RTD test,
       EPA used a sample consisting of 151 vehicles tested by the
       Coordinating Research Council (CRC)  during 1996 as part of
       its real-time diurnal testing program  (Program E-9) [7]
       combined with 119 vehicles tested by  EPA. [1]

    •  For the "gross liquid leakers"  identified by the hot soak
       test, EPA combined the sample of 300 vehicles  tested by
       Auto/Oil during 1993 as part of its real world hot soak
       testing program with 197 vehicles tested by EPA. [5]
   Vehicles failing both purge and pressure are discussed in Section 3.1.4.

-------
                               -3-
    •  For the "gross liquid leakers" identified by the  running
       loss test, EPA used a sample consisting of 150 vehicles
       tested by the CRC during 1997 as part of its running  loss
       testing program  (program E-35) . [6]

    •  The CRC also tested 50 late-model year vehicles during
       1998 as part of its combined hot soak, real-time  diurnal,
       running loss testing program  (E-41) . [8]  (These results
       are used in reference [4] to test the predictions of the
       occurrence of "gross liquid leakers" among newer
       vehicles.)

    •  A fifth source of data consisted of the results of a
       testing program run jointly by the CRC and the American
       Petroleum Institute (API) [9].   This program was designed
       to determine the frequency of vehicles with liquid leaks.
       Actual measurements of evaporative emissions were not
       performed in this program; therefore, we cannot determine
       which of those vehicles identified as having liquid leaks
       would have actually met any of our definitions of a "gross
       liquid leaker."


     To develop the appropriate weighting  factors based  on the
performance on the purge and pressure tests, EPA used data from
an EPA testing contractor, Automotive Testing Laboratories  (ATL),
which performed purge and pressure tests on a random sample  of
13,425 vehicles at its Inspection and Maintenance  (I/M)  lanes in
Indiana and Arizona between the years 1990 and 1995.  Since  the
testing protocols were changing in the early months of the
program, we omitted the first nine months of data.  We then
identified the initial test of each of the test vehicles and
calculated, by vehicle age, the number of pre-1996* model year
vehicles in each of the three purge/pressure categories.

     We combined the results for the I/M lane testing into a
single table  (Table 1 on page 5).  Omitted from all of the
columns in Table 1 are the results on approximately fifteen
percent of the vehicles for which the purge or pressure  tests
were not performed.  The reasons that testing was not performed
varied, and included both periodic problems with the testing
equipment as well as problems related to the vehicle  (e.g.,
presence of check-valves or difficulty accessing the necessary
lines).  All of the subsequent analyses were performed on the
sample of vehicles for which the purge/pressure classification
could be made.  Since all of the subsequent analyses are based on
ratios from Table 1 (e.g., the number of classified pressure
failures divided by the total number of vehicles that were
   Limiting the  analysis to pre-1996 model year  vehicles  is related to the
   phase-in of the enhanced evaporative control vehicles  (see Section 4).

-------
                               -4-
varied, and included both periodic problems with the testing
equipment as well as problems related to the vehicle (e.g.,
presence of check-valves or difficulty accessing the necessary
lines).  All of the subsequent analyses were performed on the
sample of vehicles for which the purge/pressure classification
could be made.  Since all of the subsequent analyses are based on
ratios from Table 1 (e.g., the number of classified pressure
failures divided by the total number of vehicles that were

     In examining the data in Table 1, we note that there were
relatively few vehicles more than 20 years of age.  Since small
sample sizes tend to result in low statistical confidence in the
calculated ratios (i.e., the percent of vehicles at each age that
fall into each of the purge/pressure strata),  those small sample
sizes are an obvious weakness of this analysis.  We will address
that weakness by using the calculated variances in the ratios to
weight the analyses.  (That is,  the ratios from the model years
containing the most vehicles will be counted more heavily than
the ratios from the more sparsely sampled model years.)

     An alternative approach (not being used)  is to smooth the
data from the older vehicles by averaging the results from the 66
vehicles over the age of 20 years (all of which were from the
industry programs) to obtain a sample with:

    •  a mean age of 23.23 years,

    •  19 vehicles (28.8 percent)  passing both the pressure test
       and the purge test,

    •  38 vehicles (57.6 percent)  failing the  pressure test,  and

    •   9 vehicles (13.6 percent)  failing only the purge test.

That averaged failure rate on the pressure test of almost 60
percent among the vehicles over 20 years of age suggests a
substantially higher failure rate among these vehicles than was
predicted in MOBILES (i.e., under 35 percent).  This difference
in estimating the failure rate (on the pressure test) of older
vehicles is due entirely to data recently obtained in the CRC
testing programs.   (The EPA testing used in MOBILES had no data
on vehicles older than 17 years of age.)

-------
                                 -5-
                               Table 1
          Distribution of 14,061 1971 - 95 Model Year Vehicles

Vehicle
Aae*
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Performance
Fail
Pressure
5
48
42
61
81
94
91
94
88
68
46
41
64
49
29
19
13
7
4
12
12
3
7
10
6
6
6
on Purge and
Tests
Fail Only
Purqe
9
29
33
33
42
50
76
74
46
68
44
24
23
20
5
6
3
1
0
1
3
2
0
2
2
3
0
Pressure
Passing
Both
228
1,448
1,302
1,494
1,308
1,475
1,403
1,261
888
682
369
192
152
102
62
34
17
7
2
4
7
7
5
3
2
1
1

Total
242
1,525
1,377
1,588
1,431
1,619
1,570
1,429
1,022
818
459
257
239
171
96
59
33
15
6
17
22
12
12
15
10
10
7
The quantity "Vehicle Age"  is  the  whole number calculated by subtracting
model year from test  year  and then  changing all negative results to zero.

-------
                               -6-



3.0   ANALYSIS

3.1   Modeling Strata Based on Purge and Pressure Tests

     Using the data from Table 1, we calculated the rate at which
vehicles were present  (by age) in the following three categories
determined by the results on the pressure test and the purge
test.

     •  vehicles passing both the pressure test and the purge
       test,

     •  vehicles failing the pressure test, and

     •  vehicles failing only the purge test.

These three categories are not independent.  Given the results
from any two would permit the size of remaining category to be
determined.  EPA chose to model the rates at which vehicles were
present in the first two of those categories.  These rates by
vehicle age (in years) along with the corresponding 90 percent
confidence intervals are given in Tables 2 and 3.  The confidence
intervals were calculated separately for each vehicle age rather
than having an overall calculation for the entire  sample.
Calculating confidence intervals independently  (as if the failure
rate at one age were not related to the failure rates of
neighboring ages) emphasizes the disparity in the  sizes of some
of the samples by age, as the size of the confidence interval is
substantially controlled by the sample size.


3.1.1  Vehicles Failing the Pressure Test

     Calculating (from Table 1) the rates at which vehicles
failed the pressure test (regardless of the performance on the
purge test) produces the data given in Table 2.

     As previously stated,  since the 90 percent confidence
intervals in Table 2 were calculated separately for each age that
was sampled, the confidence intervals are most representative of
the relative sample sizes.   Rather than immediately attempting to
use a regression analysis to obtain an equation relating the rate
of vehicle's failing the pressure test to the vehicle's age, we
first examined the calculated 90 percent confidence intervals in
Table 2.    (Binomial confidence intervals were used since there
were exactly two possible values, namely  "PASS" and "FAIL.")

-------
                                -7-
                              Table 2

           Estimating Rate of Vehicles that Fail the Pressure Test
                 For 14,061 1971-95 Model Year Vehicles
                  With 90 Percent Confidence Intervals
Vehicle
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Sample
Size
242
1,525
1,377
1,588
1,431
1,619
1,570
1,429
1,022
818
459
257
239
171
96
59
33
15
6
17
22
12
12
15
10
10
7
Failure
Rate
2.1%
3.1%
3.1%
3.8%
5.7%
5.8%
5.8%
6.6%
8.6%
8.3%
10.0%
16.0%
26.8%
28.7%
30.2%
32.2%
39.4%
46.7%
66.7%
70.6%
54.5%
25.0%
58.3%
66.7%
60.0%
60.0%
85.7%
90 Percent
Confidence Interval
0.6% - 3.6%
2.4% - 3.9%
2.3% - 3.8%
3.0% - 4.6%
4.7% - 6.7%
4.8% - 6.8%
4.8% - 6.8%
5.5% - 7.7%
7.2% - 10.1%
6.7% - 9.9%
7.7% - 12.3%
12.2% - 19.7%
22.1% - 31.5%
23.0% - 34.3%
22.5% - 37.9%
22.2% - 42.2%
25.4% - 53.4%
25.5% - 67.9%
35.0% - 98.3%
52.4% - 88.8%
37.1% - 72.0%
4.4% - 45.6%
34.9% - 81 .7%
46.6% - 86.7%
34.5% - 85.5%
34.5% - 85.5%
64.0% - 100.0%
     Examining the confidence intervals in Table 2, we found that
some of those confidence  intervals are so large as to be almost
useless.   (For example, for vehicles 17 years of age, knowing
that the actual failure is  most likely between 25 percent and  68
percent is not helpful in predicting the true failure rate.)
However, using both  the sample failure rates and the confidence
intervals, we were able to  make the following four observations
that were then used  to select an appropriate mathematical model:
    •  The pressure failure  rate  appears to start (i.e., for new
       vehicles) between  two and  four percent.

-------
                               -8-
    •  The pressure failure rate increases gradually for the
       first seven years of the vehicle's life.

    •  The pressure failure rate then increases more rapidly for
       the next ten years.

    •  The pressure failure rate then begins to level off
       (approaching 60 percent, according to the data from the
       industry programs)  (see third "bullet" on page 4).

This type of behavior is typical of a logistic growth function.

     Prior to constructing an appropriate logistic growth
function, we first "adjusted" the values of the variable "AGE"
(in Tables 1 through 4) which are based on the test date (since
it is the integer calculated by subtracting the model year from
the test year).  In the MOBILE model, the  (default) age  is the
integer age as of January  1 of the evaluation year.  However, the
approximately 14,000 tests in Tables 1 through 4 occurred over
all 12 months of the testing years.  Thus, the average  (i.e.,
typical)  test date was in  early July (mean date of July  3 and
median date of July 10).   To adjust for that six month difference
(between the default month of January and the average test date
in July), we modified that age value in the tables by adding 0.5
years (six months)  to it prior to the analyses.

     To account for differences in the size of the confidence
intervals (or,  equivalently the sample sizes) the data were
weighted by the reciprocal of the variance.  The "logistic
growth" function that best models the weighted pressure  test
failure rates from Table 2 is given by the following equation:

     Pressure Failure Rate = 1 + !7.733*6x^.01362*(AGE-2)]                      (1)

Estimates based on this equation of failure rates on the pressure
test are given in Appendix A.  These estimates must be adjusted
for the "gross liquid leakers"  (see Section 3.3).

     In Figure 1 (on the following page), we plotted both the
curve described by equation (1) (as  a solid line)  as well as the
90 percent confidence intervals (as dotted lines) for the failure
rate on the pressure test  (from Table 2, shifted by six  months to
compensate for the July testing).    (The small sample of  vehicles
at least 20 years of age was combined to produce reasonably sized
confidence intervals.)  That graph suggests that equation (1) is a
very good fit for the measured failure rates except at ages for
which fewer than 20 vehicles were recruited at each age.  Also,
for vehicles over 20 years of age,  the predicted failure rate on
the pressure test is close to 60 percent which closely
approximates the results of those 66 tests from industry programs
(see third "bullet" on page 4).

-------
                                -9-
     0)
     £
     Q.
     ra
                               Figure 1

        Comparison of Predicted Rates to Confidence Intervals of Measured Rates
                     For Vehicles Failing the Pressure Test
                            (by vehicle age)
        90%
        60%
     £  30%
                            10                20

                              Vehicle Age (years)
30
3.1.2  Vehicles Passing Both  the Pressure and the  Purge  Tests

     As described in the preceding  section,  we first calculated
from Table  1  the  rates at which vehicles passed both the pressure
and the purge tests, yielding the results given in Table 3.

     As in  the case of the failure  rate  on the pressure test, we
were able to  make the following four observations from Table  3:

    •  The  rate at  which vehicles passed both the pressure test
       and  the purge test  starts (i.e.,  for new vehicles)  between
       92 and 96  percent.

    •  The  rate at  which vehicles passed both the pressure test
       and  the purge test  decreases  gradually for the first seven
       years  of the vehicle's life.

    •  The  rate at  which vehicles passed both the pressure test
       and  the purge test  then decreases more rapidly for the
       next ten years.

    •  The  rate at  which vehicles passed both the pressure test
       and  the purge test  then begins to level off (approaching
       20 to  40 percent,  according  to the data from the industry
       programs)  (see second "bullet" on page 4).

-------
                                -10-
                               Table 3

    Estimating Rate of Vehicles that Pass Both the Pressure and Purge Tests
                 For 14,061 1971-95 Model Year Vehicles
                   With 90 Percent Confidence Intervals
Vehicle
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Sample
Size
242
1,525
1,377
1,588
1,431
1,619
1,570
1,429
1,022
818
459
257
239
171
96
59
33
15
6
17
22
12
12
15
10
10
7
Failure
Rate
94.2%
95.0%
94.6%
94.1%
91 .4%
91.1%
89.4%
88.2%
86.9%
83.4%
80.4%
74.7%
63.6%
59.6%
64.6%
57.6%
51 .5%
46.7%
33.3%
23.5%
31 .8%
58.3%
41 .7%
20.0%
20.0%
10.0%
14.3%
90 Percent
Confidence Interval
91 .7% - 96.7%
94.0% - 95.9%
93.5% - 95.6%
93.1% - 95.1%
90.2% - 92.6%
89.9% - 92.3%
88.1% - 90.6%
86.8% - 89.6%
85.2% - 88.6%
81.2% - 85.5%
77.3% - 83.4%
70.2% - 79.2%
58.5% - 68.7%
53.5% - 65.8%
56.6% - 72.6%
47.0% - 68.2%
37.2% - 65.8%
25.5% - 67.9%
1 .7% - 65.0%
6.6% - 40.5%
15.5% - 48.2%
34.9% - 81 .7%
18.3% - 65.1%
3.0% - 37.0%
0.0% - 40.8%
0.0% - 25.6%
0.0% - 36.0%
     As before,  the "logistic  growth" function  appeared to be the
best choice  for modeling the rates at which vehicles passed both
the pressure and the purge  tests.   After adjusting  for age, the
resulting  equation is given below as equation (2):
                                 0.7200
     Rate of Passing Both = 1 - 1 + 13.4o*expi-0.0145*(AGE*2)]
(2)
     Estimates based on equation (2)  of  the rates of vehicles
passing both the purge and pressure  tests are given  in Appendix
A.  These  estimates must be  adjusted for the "gross  liquid
leakers"  (see Section 3.3).

-------
                                -11-
     In Figure  2  (below),  we plotted both the  curve  described by
equation (2)  (as a solid line)  as well as the 90 percent
confidence intervals  (as  dotted lines) for the rate  of vehicles
passing both the purge  and pressure tests on the  pressure test
(from Table  3,  shifted  by six months to compensate  for the July
testing).   (The small  sample of vehicles at least 20 years of age
was again combined  to produce reasonably sized confidence
intervals.)   That graph suggests that equation (2) is  a very good
fit for the  measured rates for ages at which at least 20 vehicles
were sampled.   Also, for  vehicles over 20 years of  age,  the
predicted rate  of vehicles passing both the purge and pressure
tests is close  to 29 percent which closely approximates  the
results of those 66 tests from industry programs  (see second
"bullet" on  page 4).

                               Figure 2

        Comparison of Predicted Rates to Confidence Intervals of Measured Rates
              For Vehicles Passing Both the Purge and Pressure Tests
                            (by vehicle age)
        100%
     J3
     
-------
                                -12-
vehicles older than  16  years of age.  (This effect suggests  that
some of the vehicles that  only failed the purge test  in  a  prior
year would begin  to  also fail the pressure test, thus migrating
into the pressure failure  stratum.)

     This effect  is  illustrated in Figure 3 .   If we combine  this
estimate of the incidence  of vehicles' failing only the  purge
test with the estimate  (from Section 3.1.4) of failing both  the
purge and pressure tests,  we obtain a predicted failure  rate on
the purge test  (regardless of any pressure test result) .   This
purge test failure rate does not exhibit that quirk of a decrease
in failure rate with increasing age.

     The predicted size (in percent) of the stratum of vehicles
that failed only  the purge test is given in Appendix A.  These
estimates must be adjusted for the "gross liquid leakers"  (see
Section 3.3).

     In Figure 3,  for vehicles failing only the purge test,  we
plotted both the  90  percent confidence intervals  (calculated from
Table 1 and shifted  by  six months to compensate for July testing)
and the curve described by subtracting from one hundred  percent
the total of equation (1) plus equation (2).
                               Figure 3

        Comparison of Predicted Rates to Confidence Intervals of Measured Rates
                   For Vehicles Failing ONLY the Purge Test
                            (by vehicle age)
     _  30%
     .2
     0)
     S>  20%
     O
     c
     o
        10%
     "-   0%
                            10               20

                              Vehicle Age (years)
30
The preceding graph  indicates that the combination of  equations
(1) and (2) is a very good fit for the measured rates  for ages at
which at least  100 vehicles were sampled  (i.e., through  age 13).

-------
                                -13-
Also, for vehicles  over 20 years of age,  the  predicted rate of
vehicles failing  only the purge test is close to 12 percent which
closely approximates the results of those 66  tests from industry
programs  (see  fourth "bullet" on page 4).

3.1.4  Summary  of  Purge and the Pressure Failure  Rates

     Combining the  predicted rates from Figures  1,  2,  and 3 (or
Appendix A)  into  a  single area graph produces the following graph
(Figure 4) .
                               Figure 4
               Predicted Distribution of Pressure and Purge Failures
                            (by vehicle age)
       100%
         75%
        50%
        25%
          0%
    Pass Both
Purge and Pressure
                                10        15

                            Vehicle Age (years)
                             20
25
3.1.5  Vehicles Failing  Both the Purge and the Pressure Tests

     When EPA  analyzed the RTD data, it  was  determined that there
were insufficient  test results to distinguish between the diurnal
emissions of vehicles that failed both the purge and pressure
tests from  those that failed only the pressure test.  Therefore,
those vehicles were  combined into the single stratum of vehicles
that failed the pressure test  (regardless of their performance on
the purge test).   Since the purpose of this  study is to develop
factors to  weight  together the estimates of  the individual
stratum to  predict the in-use fleet emissions,  it was not
necessary to model frequency of the stratum  of vehicles that
failed both the purge and pressure tests.

-------
                                -14-
     As a service  to  possible future researchers and modelers who
may require an estimate  of the number of vehicles  failing  both
the purge and pressure tests (and based on the same sample that
produced equations (1) and  (2)) , EPA performed  the  following
analysis.

     The approach  was similar to the one used in Sections  3.1.1
and 3.1.2 in which a  table containing the frequencies with the
corresponding ninety  percent confidence intervals  was created
(see Table 4).


                               Table 4
    Estimating Rate of Vehicles that Fail BOTH the Pressure and Purge Tests
                    For 1971-95 Model Year Vehicles
                  With 90 Percent Confidence Intervals
Vehicle
Aqe
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Sample
Size
242
1,522
1,377
1,587
1,430
1,619
1,568
1,428
1,020
814
458
254
235
169
94
58
33
15
6
17
22
12
12
15
10
10
7
Failure
Rate
0.0%
0.2%
0.0%
0.3%
0.8%
0.4%
0.6%
0.8%
1 .3%
1 .5%
2.6%
2.8%
6.0%
7.1%
7.4%
6.9%
15.2%
26.7%
0.0%
23.5%
4.5%
0.0%
16.7%
53.3%
30.0%
30.0%
57.1%
90 Percent
Confidence Interval
0.0% - 0.5%
0.0% - 0.4%
0.0% - 0.2%
0.0% - 0.5%
0.4% - 1.1%
0.1% - 0.6%
0.3% - 0.9%
0.4% - 1 .2%
0.7% - 1 .9%
0.8% - 2.2%
1 .4% - 3.8%
1.1% - 4.4%
3.4% - 8.5%
3.9% - 10.4%
3.0% - 11.9%
1.4% - 12.4%
4.9% - 25.4%
7.9% - 45.4%
0.0% - 28.5%
6.6% - 40.5%
0.0% - 1 1 .9%
0.0% - 17.7%
0.0% - 34.4%
32.1% - 74.5%
6.2% - 53.8%
6.2% - 53.8%
26.4% - 87.9%

-------
                                -15-
     As the  reader may note, some  of  the sample sizes  in  Table 4
do not match the supposedly same samples in the first  three
tables.   In  the first three tables, vehicles that failed  the
pressure  test but did not have a successful purge test were
included  in  the stratum "fail pressure"  and, thus, included in
the total  as well.  However, those same  vehicles would not  be
included  in  Table 4.

     After adjusting for age  (by adding  0.5), the "logistic
growth" function that best models  the frequency of vehicle's
failing both the pressure and purge tests from Table 4 is given
by the following equation:
                             0.3536
     Rate of Failing Both = 1 + 414.96*exp[-0.32955*AGE]
    (3)
     In Figure  5,  for vehicles  failing both the purge  and
pressure  tests,  we plotted both the  90 percent confidence
intervals  (from Table 4, shifted by  six months to compensate for
the July  testing)  and the curve described by equation  (3).
(Again, the  small  sample of vehicles at least 20 years of age was
combined  to  produce reasonably  sized confidence intervals.)
                               Figure 5

        Comparison of Predicted Rates to Confidence Intervals of Measured Rates
               For Vehicles Failing Both the Pressure and Purge Tests
                            (by vehicle age)
      o
     0)
     o
     m
     c
     o
     0)
     O)

     '(5
        60%
        40%
        20%
                            10                20

                              Vehicle Age (years)
30
Figure 5  suggests that equation (3) is a  very good fit for the
measured  rates  for ages at which  at  least 30 vehicles were

-------
                               -16-
sampled  (i.e., through  age 16).   Also,  for the vehicles over  20
years of age, the data  in Table  4 indicates that 20 out of  66
(30.3%) of those vehicles over 20 years of age  (with a mean age
of 23.2 years) failed both the purge and pressure tests, and
equation (3) predicts that  the  failure  rate would be 29.5 percent.
Thus, equation (3) is also  a very good  fit for the measured  rates
for the older vehicles.
3.1.6 Effects of  Inspection / Maintenance (I/M) Programs

     As part of an  Inspection and Maintenance  (I/M) program,  a
state may choose  to perform a functional test  (e.g., a pressure
test) of each  vehicle's  evaporative control system.  Vehicles
failing the test  would be  required to be repaired.  Thus,  the
presence of such  a  program would alter the number of failing
vehicles.

     The data  used  in  the  analyses in Sections 3.1.1 through
3.1.5 were obtained from two geographic areas  (Hammond,  Indiana
and Phoenix, Arizona)  in which the vehicles were not required to
pass either a  purge or pressure test.  Therefore, those  analyses
and the resulting estimates of failure rates  (i.e., equations
(1), (2),  and  (3)) are  used in MOBILE6 for geographic areas in
which there is no I/M  for  evaporative emissions.

     In a parallel  report  [10],  EPA explains how those rates are
adjusted in MOBILE6 to account for the presence of an I/M  program
for evaporative emissions.


3.2   Modeling the  Stratum  of  "Gross Liquid Leakers" (GLLs)

     The set of vehicles identified as "gross  liquid leakers"
varies depending  upon  which type of evaporative emission is being
considered.  Earlier  (see  "bulleted"  points on page 2),  we
presented three definitions each based on one  type of test (i.e.,
RTD test, hot  soak  test, and running loss test).  EPA developed
these definitions in a parallel report devoted exclusively to the
subject of "gross liquid leakers" [4] .   In that report,   EPA
produced the following two equations to predict the frequency of
"gross liquid  leakers" occurring on the RTD and on the running
loss tests for evaporative emissions:

  Rate of Gross Liquid Leakers
                                         0 08902
     Based on RTD Testing           =  1 + 4i4.613*exp[-0.3684*AGE]             (4>


  Rate of Gross Liquid Leakers
                                        0.06
     Based on Running Loss Testing     =  1 + 12o*exp[-0.4*AGE]                 (5>

-------
                               -17-
     A different approach was necessary in predicting the
frequency of "gross liquid leakers" on the hot soak test because
that (hot soak)  testing had been limited to only newer vehicles
(i.e.,  no vehicles older than 12 years of age).   The approach
developed in that report was based on the untested hypothesis
that the vehicles classified as hot soak "gross liquid leakers"
are the same vehicles identified as gross liquid leakers on
either the running loss or RTD tests.  This hypothesis is based
on the assumptions that:

    •  If a vehicle has a leak of liquid gasoline that is severe
       enough to classify the vehicle as a "gross liquid leaker"
       on the resting loss mode,  then that leak will likely
       result in the vehicle's being classified as a "gross
       liquid leaker" on the hot soak test as well.

    •  If a vehicle has a leak of liquid gasoline that is severe
       enough to classify the vehicle as a "gross liquid leaker"
       on the running loss test,  then there will likely be enough
       liquid gasoline remaining after the engine is shut off to
       classify that vehicle as a "gross liquid leaker" on the
       subsequent hot soak test as well.

That is, the set of vehicles classified as gross liquid leakers
on the hot soak test is the union of the set of vehicles
classified as gross liquid leakers on the RTD test with the set
of vehicles classified as gross liquid leakers on the running
loss test.   (These hypotheses are "untested" because none of the
vehicles classified as "gross liquid leakers" on one test
procedure were tested over either of the other two procedures.)

     Therefore,  the rate of gross liquid leakers as identified on
the hot soak test would be the sum of the two rates for the RTD
testing and the running loss testing minus the number of double
counted vehicles (i.e., the product of those two rates assuming
these two categories are independent of each other).  Using
equations (4) and (5)  plus  the  preceding  assumption, the  predicted
rates of "gross liquid leakers" were calculated  (for each of the
three test types) and are plotted in Figure 6 (on the following
page)  and appear in Appendix B.


     It is important to note that this model of the frequency of
gross liquid leakers is based on the assumption that modern
technology vehicles  (through model year 1995) will show the same
tendency toward gross liquid leaks as do the older technology
vehicles at the same age.  However, if the modern technology
vehicles were to exhibit a lower tendency to leak (due to the
more stringent demands imposed by the new evaporative emissions
certification procedure as well as heightened attention to
safety, e.g., fuel tank protection and elimination of fuel line
leaks), the effect would be to replace each of the three logistic

-------
                                -18-
growth functions with two or three curves specific  to  different
model year  (or  technology)  groups.

                               Figure 6

                 Predicted Occurrences of "Gross Liquid Leakers"
                 On Each of Three Tests of Evaporative Emissions
                            (by vehicle age)
        15%
        10%  --
         5%
         0%
                            10              20

                           Vehicle Age (years)
30
     Since EPA has  no  data to indicate model-year  specific  rates,
EPA will use the model illustrated in Figure 6, to estimate the
occurrence in the in-use fleet of these vehicles that  have
substantial leaks of liquid gasoline  (i.e., "gross liquid
leakers") for vehicles that were not designed to meet  the new
enhanced evaporative test procedure (i.e., vehicles up through
the 1995 model year along with some of the 1996 through 1998
model years).  For  the vehicles designed to meet the new enhanced
evaporative test procedure,  EPA will modify that equation (see
Section 4.0).

3.3   Combining Purge/Pressure Rates with Gross Liquid Leaker Rates

     In Section 3.1 we characterized the three strata  resulting
from the individual vehicle's performance on the purge and
pressure tests.  In Section 3.2, we characterized  the  additional
stratum created for the "gross liquid leakers."  In order to make
these strata non-overlapping (i.e., mutually exclusive),  we must
remove the "gross liquid leakers" from the other three strata.

     To determine the  distribution of the gross liquid leakers
among the other three  strata, we examined the 270  vehicles  in the
combined EPA/CRC RTD testing programs.  Seven vehicles were

-------
                               -19-
identified as "gross liquid leakers" out of those 270 vehicles
that were tested.  Of these seven gross liquid leakers:

    •  four had failed both the purge and pressure tests,

    •  two had failed only the pressure test, and

    •  one had passed both the purge and pressure tests.

This distribution of seven gross liquid leakers proves  that gross
liquid leakers can and do occur within all three of the purge/
pressure strata.  From these statistics, EPA first estimated  the
number of gross liquid leakers within each of the purge/pressure
strata and then removed them to form a fourth stratum consisting
of only the gross liquid leakers.  Rather than attempting  to
estimate the distribution of all gross liquid leakers based on  a
sample of only seven vehicles, MOBILE6 will simply distribute the
liquid leakers proportionately among the three purge/pressure
categories.

     As an example, if we were to take a hypothetical fleet of
10,000,000 vehicles, each 10 years of age, then Appendix B
predicts that 78,000 (0.78 percent) of them will be "gross liquid
leakers" on the RTD test.  Similarly, Appendix A indicates that
1,091,000 (10.91 percent) will fail the pressure test,  647,000
(6.47 percent) will fail only the purge test, and the remaining
8,262,000 (82.62 percent) will pass both tests.  Distributing
those 78,000 "gross liquid leakers" proportionately among  the
three purge/pressure strata predicts the following distribution:
                              Table 5

                     Predicted Distribution on RTD Test
                      (For Vehicles at 10 Years of Age)
Purge/Pressure
Strata
Fail Pressure
Fail ONLY Purge
Pass Both
TOTALS
"Gross Liquid
Leakers"
8,510
5,047
64,444
78,000
Not "Gross Liquid
Leakers"
1,082,490
641,953
8,197,556
9,922,000
TOTALS
1,091,000
647,000
8,262,000
10,000,000
Thus, Table 5 indicates that for 10 year old vehicles on  the RTD
test:

     •  0.78 percent of those vehicles will be "gross liquid
       leakers" on a RTD test,

     •  10.82 percent of those vehicles will fail the pressure
       test, but will not be "gross liquid leakers,"

-------
                                -20-
     •  6.42 percent of those vehicles  will fail only the purge
       test,  but  will not be "gross  liquid leakers," and

     •  81.98  percent of those vehicles will pass both the
       pressure and purge tests,  and will not be "gross liquid
       leakers."

Repeating  this  process for each vehicle age in Appendices  A and B
produces the  estimated size of each  of the four strata  for the
RTD  test for  each age.  The results  are plotted below  (Figure 7):

                               Figure 7

                      Predicted Strata Sizes on RTD Test
                            (by vehicle age)
    100%
     80%  - -
     60%
     40%
     20%
      0%
1 Pass Both
Purge and
Pressure

' Fail Pressure

1 Fail Only
Purge

Leakers
                           10               20

                           Vehicle Age (years)
                                         30
     Repeating this process using  the "running loss"  and "hot
soak" columns  from Appendix B will produce the estimates of the
size of the  four strata for use with each of those two  types of
evaporat ive  emi s s ions.

-------
                               -21-
4.0   MODELING ENHANCED  EVAP VEHICLES  EQUIPPED WITH  OBD

     Beginning with the 1996 model year, manufacturers  were
required to certify at least twenty percent of their vehicles
using a new "enhanced" evaporative testing procedure  (ETP);  that
percentage of ETP vehicles was required  to increase from  the
twenty percent in 1996 up to one hundred percent by 1999.  The
actual phase-in of these ETP vehicles proceeded at a slightly
faster pace (based on EPA's analysis of  data  from the Wisconsin
Inspection/Maintenance program for model years 1996-1999).   The
phase-in rate required by the regulations is  given below  in
Table 6 (copied from 40 CFR 86.096-8)  along with the observed
(actual) phase-in rate.

                           Table 6

                    Phase-In of Vehicles with
                  Enhanced Evaporative Controls

                  Model     Required     Observed
                  Year      Percentage   Percentage
                  1995        0%         0%
                  1996       20%        30%
                  1997       40%        55%
                  1998       90%        90%
                  1999      100%       100%
     To predict the performance of these 1996 and newer vehicles
on the purge and pressure tests, the effects of  two  factors  must
be considered:

     1)    A change in the regulations requires a doubling  of the
          period during which these vehicles must meet the
          evaporative emissions standards  (increased from  5  years
          / 50,000 miles to 10 years / 100,000 miles for light-
          duty vehicles) .

     2)    Most of these ETP vehicles are equipped with an  on-
          board diagnostic  (OBD) system that is  expected to  alert
          each vehicle's owner  (or driver) to most problems  with
          the evaporative control system,  thus permitting  the
          owner to decide whether to repair the  problem.

     In order to meet the increased durability and more stringent
evaporative standards, manufacturers have  implemented a number of
changes, including (but not limited to):

    •  "quick connects" that reduce the possibility  of improper
       assembly when the vehicle is serviced,

-------
                                 -22-
     •  advanced materials  that are less  permeable,  less
       susceptible to puncture, and more durable  (i.e.,
       elastomeric materials used in  hoses and connectors),

     •  improvements made  to the purge system  (to  enable the
       vehicles to pass both the running loss test  and the multi-
       day diurnal test),

     •  tethered gas caps,  and

     •  improved fractional-turn gas caps.

Since  these changes are  expected to result in improved control of
evaporative emissions,  in a parallel  report  (M6.EVP.005),  EPA
decided  to use a separate set of estimates of both  resting loss
and diurnal emissions for these vehicles.   However,  EPA does not
have actual data on the  effects of these changes  in durability
that translates into changes in the purge and pressure failure
rates  estimated in Section 3.1 for the pre-1996 model year
vehicles.

     EPA,  therefore, will  use the doubling in the durability
requirement to modify equations (1), (2), (4), and (5)  (from Sections
3.1.1, 3.1.2,  and 3.2) by replacing the  variable  "AGE" with
"AGE/2" resulting  in:
Pressure Failure Rate of Enhanced Evaporative Control Vehicles

                 	0.6045	
                 ~ 1 + 17.733*exp[-0.003405*(AGEA2)]



Rate of Passing Both for Enhanced Evaporative Control Vehicles

                               0.7200
                                                                    (6)
                                                                    \'I
     Rate of Gross Liquid Leakers on the RTD Test for the Enhanced Evaporative Control
     Vehicles

                                0.08902
                       ~ 1 + 414.613*exp(-0.1842*AGE)

     Rate of Gross Liquid Leakers on the Running Loss Test for the Enhanced Evaporative
     Control Vehicles
                       ~ 1 + 120*exp[-0.2*AGE]

     Using these equations,  we first  generated the  estimated
failure  rates for those (1996 and newer)  vehicles certified to

-------
                               -23-
the enhanced evaporative control standards.  Since these
estimates assume only the benefits of changes in durability with
no estimation of the effect from the OBD system, they are
adjusted in MOBILE6 (Appendices C and D).

     Since the OBD system is designed to alert each vehicle's
owner to problems with the evaporative control system.  The
appearance of a warning light should result in at least some
owners having their malfunctioning vehicles repaired.  Thus, the
OBD system has the potential to affect the relative sizes of the
purge/pressure strata, depending both on its ability to identify
problems in the evaporative control systems and on the owners
inclination to repair such problems.  (In a parallel report  [10],
EPA explains how those rates are adjusted in MOBILE6 to account
for the presence of an I/M program for evaporative emissions.)

     In that parallel report [10], EPA assumes that the effect of
an OBD system will vary, based on the vehicle's warranty and the
presence of an I/M program.  Those assumptions stated that:

    •  The vehicle's malfunction indicator light (MIL) would
       detect/identify 85 percent of the instances of the
       vehicle's failing the purge test or failing the pressure
       test.  (It is assumed that the OBD system would not detect
       the presence of a gross liquid leak.)

    •  While the vehicle is under its full ("bumper to bumper")
       warranty (i.e., up through 3 years / 36,000 miles), 90
       percent of the owners will have the vehicle repaired when
       the MIL indicates a problem.  (These first two assumptions
       suggest that the OBD system combined with the
       manufacturer's warranty program will reduce the incidence
       of vehicles failing either the pressure or purge tests by
       76.5 percent (76.5% = 85% * 90%).)

    •  When the warranty covers only the electronic control
       module and the catalytic converter (i.e., from 36,000
       through 80,000 miles, approximately ages four through six
       years),  only 10 percent of the owners will have the
       vehicle repaired when the MIL indicates a problem.  (This
       assumption suggests that the OBD system combined with the
       manufacturer's warranty program will reduce the incidence
       of vehicles that in their fourth, fifth,  or sixth years
       newly fail either the purge or pressure tests.)  That
       percentage would increase from 10 to 90 percent if that
       geographic area has an I/M program that requires the MIL
       to indicate that there are no problems.

    •  When the vehicle is no longer covered by a manufacturer's
       warranty (i.e., over 80,000 miles or beyond six years of
       age), none (i.e., zero percent)  of the owners will have
       the vehicle repaired when the MIL indicates a problem.

-------
                               -24-
       That percentage would increase from zero to 90 percent if
       that geographic area were to have an I/M program requiring
       the MIL to indicate that there are no problems.

     EPA will adapt those assumptions (modified slightly)  for
evaporative emissions.  Specifically:

    •  The vehicle's malfunction indicator light (MIL) would
       detect/identify 85 percent of the instances of the
       vehicle's failing the purge test or failing the pressure
       test.  (It is assumed that the OBD system would not detect
       the presence of a gross liquid leak.)


    •  While the vehicle is within its full warranty period
       (i.e.,  approximately through the age of three years), 90
       percent of the owners will have the vehicle repaired when
       the MIL indicates a problem.  (These first two assumptions
       suggest that the OBD system combined with the
       manufacturer's warranty program will reduce the incidence
       of vehicles failing either the pressure or purge tests by
       76.5 percent (76.5% = 85% * 90%).)


    •  While the vehicle is under its partial warranty period
       (i.e.,  approximately ages four through six years),  only 10
       percent of the owners will have the vehicle repaired when
       the MIL indicates a problem.  (This assumption suggests
       that the OBD system combined with the manufacturer's
       warranty program will reduce the incidence of vehicles
       (ages four through six) that fail either the purge or
       pressure tests for the first time.)  That percentage would
       increase from 10 to 90 percent if that geographic area
       were to have an I/M program requiring the MIL to indicate
       that there are no problems.


    •  When the vehicle's evaporative control system is no longer
       covered by a manufacturer's warranty (i.e.,  beyond about
       six years of age), none (i.e., zero percent) of the owners
       will have the vehicle repaired when the MIL indicates a
       problem.   That percentage would increase from zero to 90
       percent if that geographic area were to have an I/M
       program requiring the MIL to indicate that there are no
       problems.

  Note:  EPA is continuing to study in-use OBD systems.  The
       assumptions listed here may be revised in future models.

     These MOBILE6 assumptions are different from what was used
in MOBILES.  In that previous model, while the evaporative
emissions of the ETP vehicles were reduced to reflect the more

-------
                               -25-
stringent standards, the weighting factors were not changed to
reflect either the improved durability or the OBD system.


5.0   COMPARISONS WITH MOBILES

5.1   Comparisons of Weighting Factors

     Both the MOBILES model and this proposal weight the
estimated evaporative emissions based on each vehicle's
performance on purge and pressure tests; however there are
several structural differences:

    •  The weighting factors in MOBILES are each functions of a
       continuous variable (i.e.,  each model year's average
       odometer)  while the weighting factors in MOBILE6 are
       functions of a discrete variable (i.e.,  each model year's
       estimated age).

    •  The weighting factors in this proposal are smooth
       functions  (i.e.,  exponential)  of the estimated age.
       Although the weighting factors in MOBILES are continuous,
       they are not smooth functions (they are piece-wise linear
       functions of odometer,  thus almost linear with age).

    •  MOBILES does not have a separate stratum for the vehicles
       classified as "gross liquid leakers" while this proposal
       does.   We will,  therefore,  compare the MOBILES weighting
       factors with the unadjusted factors from Section 3.1 of
       this report.

     The comparisons between the MOBILES weighting factors for
light-duty vehicles for each of the three purge/pressure  strata
and the weighting  factors in this proposal are illustrated in the
following three figures:

    •  Figure 8 compares the estimates of Pre-ETP vehicles (i.e.,
       1995 and older)  passing both the purge test and the
       pressure test.   (Subtracting each of those estimates from
       100 percent will yield the associated rates of vehicles
       failing either the purge test,  or the pressure test,  or
       both tests.)

    •  Figure 9 compares the estimates of Pre-ETP vehicles
       failing the pressure test (regardless of the performance
       on the purge test).

    •  Figure 10 compares the estimates of Pre-ETP vehicles
       failing only the purge test.

-------
                                   -26-
                                  Figure 8

          Comparison of Predictions of Percentage of Pre-ETP Vehicles
                  Passing Both the Purge and Pressure Tests
                              (by vehicle age)
O
   100%
    75%
    50%
    25%
                                  10           15

                                 Vehicle Age (years)
20
25
                                  Figure 9

          Comparison of Predictions of Percentage of Pre-ETP Vehicles
                          Failing the Pressure Test
                              (by vehicle age)
   60%
                                  10            15

                                Vehicle Age (years)
20
25

-------
                               -27-
                              Figure 10
            Comparison of Predictions of Percentage of Pre-ETP Vehicles
                        Failing ONLY the Purge Test
                            (by vehicle age)
        25%
        20%
        15%
     o
        10%
                               10         15

                              Vehicle Age (years)
20
25
     Based on examinations  of  these three graphs, we can make the
following observations  for  the Pre-ETP vehicles:

    •  The most obvious difference  between the weighting factors
       used in MOBILES and  those  developed in this report is that
       the factors in this  report are  capped (around age 20)
       while the MOBILES factors  are not.

    •  Despite the structural  differences  (i.e., smooth function
       of a discrete variable  versus piece-wise linear function
       of a continuous variable)  between these two sets of
       weighting factors, they produce similar results for
       vehicles up to about  12  to 13 years of age.

    •  The estimates developed in this report predict a
       substantially higher  proportion of  vehicles failing the
       pressure test for vehicles older than 13 years of age than
       does the MOBILES model.

     The reader can make similar  comparisons for the ETP vehicles
by replacing the MOBILE6  (dotted) line in each figure with the
values in Appendix D.   (As  stated at  the end of Section 4, while
MOBILE6 uses different  distributions  for the ETP vehicles,
MOBILES uses the same distributions for the Pre-ETP and the ETP.)
This results in MOBILE6 predicting  a  smaller number of
malfunctioning ETP vehicles at any  given age than did MOBILES.

-------
                                -28-
For example, MOBILE6  predicts that at age 25 years with no I/M
program, approximately 26 percent of the ETP vehicles  would fail
either the purge  or the pressure test while MOBILES  predicts that
approximately  56  percent would fail either test.  This much lower
predicted failure rate is illustrated in the following graph
(Figure 11 which  is a modification of Figure 8)  in which the
MOBILE6 curve  (from Appendix A)  is replaced with the
corresponding  curve from Appendix D (i.e., ETP vehicles in a non-
I/M area passing  both the purge test and pressure test).
                              Figure 11

              Comparison of Predictions of Percentage of ETP Vehicles
                   Passing Both the Purge and Pressure Tests
                            (by vehicle age)
     O
        100%
        75%
        50%
                         MOBILES

                        -ETPsin M6
25%
                               10        15

                              Vehicle Age (years)
                                           20
25
5.2   Comparisons of Weighted  Full-Day Diurnal Emissions

     By combining  the  information in this report with  the
information in parallel  reports (i.e., M6.EVP.001  and
M6.EVP.005), we  can estimate the diurnal emissions for the in-use
fleet  (containing  the  25 most recent model years).   To compare
these estimates  with those predicted by the MOBILES  model,  the
MOBILES model was  run  for a fleet with a national  distribution of
model years  (as  of January 1, 1995) with two likely  combinations
of temperature cycle and fuel RVP  (assuming no weathering of the
fuel) :

     •  daily temperatures cycling between 60 and 84  degrees
       Fahrenheit  using  fuel with a 9.0 RVP (see Figure  12)  and

-------
                                -29-
    •  daily temperatures  cycling between 82 and 106  degrees
       Fahrenheit using  fuel  with a 7.0 RVP  (see Figure  13).

Each MOBILES run generates estimated diurnal emissions for the 25
most recent model years,  which for these runs were  the 1971
through 1995 model  years.   In these two examples the  variable
MODEL YEAR can be transformed into AGE using the equation:
                      AGE = 95  -  (MODELYEAR) .

     It is important  to  note that these  (following) estimates are
only of the full one-day diurnal emissions.  They do  include
diurnal emissions from "gross liquid leakers."  But,  they do not
include evaporative emissions from interrupted  (i.e.,  partial-
day) diurnals, nor  from  multi-day diurnals,  nor from  running
loss,  nor from hot  soaks.   Estimates of these excluded sources
are developed in parallel  reports.  A complete comparison between
MOBILE6 and MOBILES requires using activity  data to weight
together all of these individual components  of evaporative
emissions; this can be done by running each  model.
                              Figure 12

         Comparison of Predictions for In-Use Diurnal Emissions by Model Year
                       (Total Grams Per Full-Day Test)
                   60° to 84° F Cycle  - Using 9.0 psi RVP Fuel
                          (As of January 1,1995)
           70
75
80        85

 Model Year
90
95

-------
                                -30-
                              Figure 13

         Comparison of Predictions for In-Use Diurnal Emissions by Model Year
                       (Total Grams Per Full-Day Test)
                  82° to 106° F Cycle - Using 7.0 psi RVP Fuel
                          (As of January 1,1995)
       2
       u>
       o
       ro
           70
75
80         85

  Model Year
90
95
     In Figures  12  and 13,  the typical  (i.e., mean)  24-hour
diurnal emissions  for each vehicle are plotted  against model
year.  Visual  inspections of Figures 12 and  13  suggest that the
estimates of diurnal  emissions resulting from MOBILE6  (for each
of those two combinations of fuel RVP and temperature  cycle) for
each model year  are:

     •  close (within  a gram per vehicle per day) to  MOBILES
       estimates for  the  1986 to 1995 model year vehicles,

     •  typically 25 to 75 percent higher (for 9.0 and  7.0  RVP
       fuels,  respectively)  than MOBILES estimates for 1980 to
       1985 model year vehicles,  and

     •  typically 30 to 40 percent higher than MOBILES  estimates
       for 1972 to  1979 model year vehicles.
     A similar  comparison for the vehicles certified to the new
enhanced evaporative  emission requirements  (i.e.,  1999 and newer
vehicles, along with  some 1996-98 model year vehicles)  is
provided in Section 5.3.

     In Figures 12  and 13,  we weighted each model  year's
estimated diurnal emissions by the relative number of vehicles in

-------
                                -31-
the fleet, we  obtain the estimate of the mean full-day diurnal
emissions for  an  average in-use vehicle subject  to a full-day's
diurnal:

             Estimates of Full-Day Diurnal  Emissions
             For the In-Use Fleet  (As of January  1995)
               In  Units of Grams per Day per Vehicle
     Temperature  Cycle and Fuel	
     60° to   84°  F with 9.0 psi RVP
     82° to  106°  F with 7.0 psi RVP
                MOBILES
                Estimate
                   4.86
                   7.61
MOBILE6
Estimate
  5.42
 10.86
     We repeated  these calculations for  several  dozen
combinations of fuel  RVP and temperature  cycles  and then graphed
the resulting  averages in Figures 14 through  17.   The first three
figures  (Figures  14  through 16) are based on  January 1,  1995
(thus covering model  years 1971 through  1995).   Figure 17 is
based at January  1,  1985 (thus, covering  model years 1961 through
1985).  Therefore,  Figures 16 and 17 differ only by the model
years covered.  Since both the horizontal and vertical scales
vary among these  four graphs, care should be  taken in making
comparisons between  these figures.
                               Figure 14

           Comparison of Predictions of Full-Day Diurnal Emissions by RVP
           For the In-Use Fleet (as of January 1995) with the 60° to 84° F Cycle
                     In Units of Grams per Day per Vehicle
        15
        10
      U)
     o
     I
     "(5
      |
     Q
—MOBIL E5

 -MOBILES
                                            10
                                       12
                               Fuel RVP (psi)

-------
                                  -32-
                                Figure 15

        Comparison of Predictions of Full-Day Diurnal Emissions by RVP
       For the In-Use Fleet (as of January 1995) with the 72° to 96° F Cycle
                    In Units of Grams per Day per Vehicle
                        MOBILES

                      -MOBILES
                                 8            9

                                 Fuel RVP  (psi)
10
11
                                Figure 16

        Comparison of Predictions of Full-Day Diurnal Emissions by RVP
       For the In-Use Fleet (as of January 1995) with the 82° to 106° F Cycle
                    In Units of Grams per Day per Vehicle
O   20
ro
c
3
Q
                     MOBILES

                   -MOBILES
                                        8

                                 Fuel RVP  (psi)
             10

-------
                               -33-
                              Figure 17

           Comparison of Predictions of Full-Day Diurnal Emissions by RVP
          For the In-Use Fleet (as of January 1985) with the 82° to 106° F Cycle
                     In Units of Grams per Day per Vehicle
       60
     i 40
             —MOBILES

             -MOBILE6
                                   8

                              Fuel RVP (psi)
                                                     10
     Based on examinations  of these four graphs, we made  the
following observations  (for Pre-ETP vehicles) :

    •  For the lowest temperature  cycle (i.e., daily temperatures
       cycling between 60°  and 84°F) ,  the MOBILE6 approach to
       predicting fleet full -day diurnal emissions produce
       results quite similar  to MOBILES for the full range of
       fuel RVPs .    (Figure  14.)

    •  For the two higher temperature  cycles, the MOBILE6
       approach to predicting full -day diurnal emissions produce
       results similar to MOBILES  for  fuel RVPs up to 10 psi,
       which is a reasonable  upper bound for in-use fuel RVP at
       those temperatures.   (Figures  15 through 17.)

    •  The MOBILE6 approach (compared  to MOBILES) predicts
       slightly higher full -day diurnal emissions for the lower
       RVP fuels and the reverse (i.e.,  lower diurnal emissions)
       for the higher RVP fuels.   The  range of RVPs for which  the
       new estimates are higher than  the MOBILES estimates varies
       with the temperature cycle:
••
            For  the low temperature cycle  (i.e.,  60°  to 84°F) ,
            the  MOBILE6 approach predicts  slightly  higher diurnal
            emissions for fuel RVPs up through  11 psi.   For fuel
            RVPs near 7 psi,  the higher MOBILE6 predictions are
            within 0.9 grams of HC  (of the MOBILES  estimates)  per

-------
                               -34-
           day per vehicle undergoing  a  full  (24-hour)  diurnal.
           This difference gradually shrinks  to  zero  as the  fuel
           RVP nears  11 psi.   For  fuel RVPs above  11  psi,  the
           MOBILE6 estimates  are slightly  lower  than  the MOBILES
           estimates.

        •• For the moderate temperature  cycle (i.e.,  72° to
           96°F), the new approach predicts slightly  higher
           diurnal emissions  for fuel  RVPs up through 10 psi.
           For fuel RVPs above 10  psi, the new estimates are
           lower than the MOBILES  estimates.   For  RVPs above 11
           psi  (an unlikely occurrence with this temperature
           cycle), the two models  move farther apart.

        •• For the high temperature cycle  (i.e., 82°  to 106°F),
           the new approach predicts slightly higher  diurnal
           emissions  for fuel  RVPs up  through 8.5  psi for the
           in-use fleet as of  January  1995 (Figure 16).   For
           fuel RVPs  above 8.5 psi, the  new estimates are lower
           than the MOBILES estimates.   For RVPs above 10 psi
            (an unlikely occurrence with  this  temperature cycle),
           the two models move farther apart.  A snapshot of the
           in-use fleet as of  January  1985 (Figure 17)  yields
           similar results.

     A more complete comparison of predicted  diurnal  emissions
(i.e., one that takes activity data into account)  can  be obtained
by running the two models.   Similarly  for  the  other evaporative
emissions.
5.3   Comparisons of Diurnal Emissions  from Vehicles Certified to  the

     Enhanced Evaporative  Control  Standards

     In Section 5.2,  we compared, for 1995 and older model year
vehicles, these MOBILE6 estimates of full-day diurnal emissions
with those predicted by the MOBILES model.  In this section, we
perform a similar comparison of the two estimates for the
vehicles certified to the enhanced evaporative standard  (i.e.,
the 1999 and newer along with some 1996 through 1998 model year
vehicles).

     Repeating the process used to create Figures 12 and 13 but
with January 1, 2020 as the evaluation date produced Figures 18
and 19.

-------
                                 -35-
                                Figure 18

             For Enhanced Evaporative Control Vehicles (ETPs)
        Comparison of Predictions for In-Use Diurnal Emissions by Age
                60° to 84° F Cycle - Using 9.0 psi RVP Fuel
                             (Non-l/M Area)
                   In Units of Grams per Day per Vehicle
2   6
u>
O
ro
— MOBILES

 -MOBILES
                               10           15

                              Vehicle Age  (years)
                                   20
                               Figure 19

                For Enhanced Evaporative Control Vehicles
        Comparison of Predictions for In-Use Diurnal Emissions by Age
                82° to 106° F Cycle - Using 7.0 psi RVP Fuel
                             (Non-l/M Area)
                   In Units of Grams per Day per Vehicle
2
u>
O
1_
3
b
                                10           15

                              Vehicle Age (years)
                                   20
25

-------
                               -36-
     Visual inspections of Figures 18 and 19 leads to the
following conclusions:

    •  MOBILE6 predicts substantially lower full-day diurnal
       emissions for these new vehicles than does MOBILES.   This
       difference is due (almost entirely)  to the substantially
       lower failure rate (on the purge and pressure tests)
       predicted in MOBILE6.   (MOBILES uses the same failure
       rates for the ETP vehicles as for the Pre-ETPs while
       MOBILE6 uses Appendices C, D,  and E.  See Figure 11  on
       page 29.)

    •  For vehicles between 10 and 25 years of age,  MOBILE6
       predicts a substantial rise in the full-day diurnal
       emissions of the in-use fleet (on a per vehicle basis),
       but still smaller than the MOBILES predicted increase.

       This increase in MOBILE6 predicted emissions is driven
       primarily by the increasing numbers of gross liquid
       leakers.  This is despite the fact that their predicted
       rate of occurrence among ETP vehicles is very low (less
       than two percent at age 25).

       Thus, modifying the assumptions on the frequency of  gross
       liquid leakers as a function of age among these vehicles
       would significantly affect the graphs for "MOBILE6"  in
       Figures 18 and 19,  with that curve remaining much flatter
       throughout.


6.0   SUMMARY

     Estimates of evaporative emissions in MOBILE6 will be
modeled based on their type:

    •  resting loss emissions,

    •  running loss emissions,

    •  hot soak emissions,  and

    •  diurnal emissions.

Each of these types will be calculated based on whether the
individual vehicles:

    •  are gross liquid leakers,

    •  failed the pressure test,

-------
                               -37-
    •  failed only the purge test,  or

    •  passed both the purge and pressure tests.

Once the estimated evaporative emissions of each sub-strata is
calculated,  they will be weighted together using the equations
developed in this report.

     The analyses suggest that for the full-day diurnal (see
Sections 5.2 and 5.3), MOBILE6:

    •  predicts full-day diurnal emissions lower than does
       MOBILES for the 1999 and  newer vehicles,

    •  predicts full-day diurnal emissions similar to those of
       MOBILES for the 1986-1995 model year vehicles,  and

    •  predicts full-day diurnal emissions higher than does
       MOBILES for the 1985 and  older vehicles.

     Similarly, we can perform a simplified analysis to estimate
the effect  (benefit) of reducing fuel RVP on full-day diurnal
emissions:

    •  In each of the four figures  (Figure 14 through 17),  the
       MOBILES graph is steeper  than the graph of MOBILE6.   The
       lower slopes associated with the MOBILE6  estimates  will
       result in smaller decreases  in predicted full-day diurnal
       emissions associated with a  reduction in fuel RVP than the
       corresponding changes predicted by MOBILES.

    •  Analyses performed to calculate the effect (e.g., either
       the  cost per ton or the benefit)  will make RVP control
       programs slightly less attractive for controlling full-day
       diurnal emissions.

    •  The  difference in the predicted effect of lowering  fuel
       RVP  is small for lower RVP fuels.   Thus,  estimating the
       effects of reducing the fuel RVP from 8 psi (or from a
       lower value)  will be similar under both methods.

-------
                               -38-
7.0   REFERENCES

1)  Larry Landman,  "Evaluating Resting Loss and Diurnal
   Evaporative Emissions Using RTD Tests," Report numbered
   M6.EVP.001, April 2001.

2)  Larry Landman,  "Modeling Hourly Diurnal Emissions and
   Interrupted Diurnal Emissions Based on Real-Time Diurnal
   Data," Report numbered M6.EVP.002, April 2001.

3)  Larry Landman,  "Modeling Diurnal and Resting Loss Emission
   from Vehicles Certified to the Enhanced Evaporative
   Standards," Report numbered M6.EVP.005, April 2001.

4)  Larry Landman,  "Evaporative Emissions of Gross Liquid Leakers
   in MOBILE6," Report numbered M6.EVP.009, April 2001.

5)  Louis Browning, "Update of Hot Soak Emissions Analysis"
   prepared by Louis Browning of ARCADIS Geraghty & Miller, Inc.
   for EPA, Report numbered M6.EVP.004, September 1998

6)  Larry Landman,  "Estimating Running Loss Evaporative Emissions
   in MOBILE6," Report numbered M6.EVP.008, April 2001.

7)  D. McClement, J. Dueck, B. Hall, "Measurements of Diurnal
   Emissions from In-Use Vehicles, CRC Project E-9", Prepared
   for the Coordinating Research Council, Inc. by Automotive
   Testing Laboratories, Inc., June 19, 1998.

8)  D. McClement, "Real World Evaporative Testing of Late Model
   In-Use Vehicles, CRC Project E-41", Prepared for the
   Coordinating Research Council, Inc. by Automotive Testing
   Laboratories, Inc., December 17, 1998.

9)  D. McClement, "Raw Fuel Survey in I/M Lanes", Prepared for
   the American Petroleum Institute and the Coordinating
   Research Council, Inc. by Automotive Testing Laboratories,
   Inc.,  June 10,  1998.

10) Megan Beardsley, "Estimating Benefits of Inspection/
   Maintenance Programs for Evaporative Control Systems," Report
   numbered M6.IM.003, November 1999.

-------
                       -39-
                   Appendix A

Estimates of Purge/Pressure Strata Size by Vehicle Age
      For 1995 and Older Model Years Vehicles
                (For Non-l/M Areas)
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Failing
Pressure
Test
3.23%
3.27%
3.40%
3.62%
3.96%
4.44%
5.10%
5.99%
7.18%
8.78%
10.91%
13.70%
17.30%
21 .79%
27.12%
33.07%
39.19%
44.90%
49.76%
53.50%
56.16%
57.92%
59.02%
59.66%
60.03%
60.24%
Failing
Only
Purge
Test
1 .77%
1 .80%
1 .88%
2.02%
2.23%
2.53%
2.95%
3.51%
4.25%
5.23%
6.47%
8.00%
9.76%
11.61%
13.29%
14.51%
15.06%
14.95%
14.41%
13.70%
13.03%
12.50%
12.13%
1 1 .89%
1 1 .74%
1 1 .65%
Passing
Both
Purge and
Pressure
Tests
95.00%
94.93%
94.72%
94.36%
93.81%
93.03%
91 .96%
90.51%
88.57%
85.99%
82.62%
78.30%
72.94%
66.60%
59.58%
52.42%
45.76%
40.14%
35.84%
32.80%
30.81%
29.58%
28.85%
28.45%
28.23%
28.11%

-------
                         -40-
                     Appendix B

Predicted Frequency of Occurrence of "Gross Liquid Leakers"
            by Emission Type and Vehicle Age
         For 1995 and Older Model Years Vehicles

 (Reproduced  from Report  Number:  M6.EVP.009  [4])
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Resting
Loss/
Diurnal
0.02%
0.03%
0.04%
0.06%
0.09%
0.13%
0.19%
0.27%
0.39%
0.55%
0.78%
1.08%
1 .49%
2.00%
2.63%
3.36%
4.15%
4.97%
5.75%
6.46%
7.05%
7.54%
7.91%
8.19%
8.40%
8.55%
Running
Loss
0.05%
0.07%
0.11%
0.16%
0.24%
0.35%
0.50%
0.72%
1.02%
1 .40%
1.88%
2.43%
3.02%
3.61%
4.16%
4.62%
5.00%
5.29%
5.51%
5.66%
5.77%
5.84%
5.89%
5.93%
5.95%
5.97%
Hot
Soak
0.07%
0.10%
0.15%
0.23%
0.33%
0.48%
0.70%
1.00%
1.41%
1.95%
2.64%
3.48%
4.46%
5.54%
6.67%
7.83%
8.95%
10.00%
10.94%
11.75%
12.42%
12.94%
13.34%
13.63%
13.85%
14.00%

-------
                                -41-
                             Appendix C

         Estimates of Purge/Pressure Strata Size by Vehicle Age
               For 1999 and Later Model Years Vehicles
                           (For I/M Areas*)
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Failing
Pressure
Test
0.76%
0.76%
0.77%
0.78%
0.80%
0.82%
0.85%
0.89%
0.93%
0.98%
1 .04%
1.11%
1 .20%
1 .29%
1.41%
1 .54%
1 .69%
1 .86%
2.06%
2.30%
2.56%
2.87%
3.22%
3.62%
4.07%
4.57%
Failing
Only
Purge
Test
0.42%
0.42%
0.42%
0.43%
0.44%
0.46%
0.47%
0.50%
0.52%
0.56%
0.60%
0.64%
0.69%
0.75%
0.82%
0.91%
1 .00%
1.11%
1 .23%
1 .37%
1 .52%
1 .69%
1 .88%
2.08%
2.29%
2.51%
Passing
Both
Purge and
Pressure
Tests
98.83%
98.82%
98.81%
98.79%
98.76%
98.72%
98.67%
98.62%
98.54%
98.46%
98.36%
98.25%
98.11%
97.95%
97.77%
97.56%
97.31%
97.03%
96.71%
96.34%
95.92%
95.44%
94.90%
94.30%
93.64%
92.92%
This assumes  that  the  I/M program requires repairs to vehicles  with  a  MIL
that indicates that there is a problem.

-------
                                  -42-
                               Appendix D

                   Estimates of Strata Size by Vehicle Age
                  For 1999 and Later Model Years Vehicles
                           (For Non-l/M Areas*)
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Failing
Pressure
Test
0.76%
0.76%
0.77%
0.78%
0.85%
0.94%
1 .06%
1.21%
1 .39%
1.61%
1 .87%
2.18%
2.53%
2.94%
3.42%
3.97%
4.62%
5.36%
6.21%
7.20%
8.34%
9.64%
11.13%
12.82%
14.73%
16.86%
Failing
Only
Purge
Test
0.42%
0.42%
0.42%
0.43%
0.47%
0.53%
0.60%
0.70%
0.81%
0.95%
1.12%
1.31%
1 .53%
1 .79%
2.09%
2.44%
2.83%
3.29%
3.81%
4.40%
5.05%
5.78%
6.58%
7.44%
8.34%
9.27%
Passing
Both
Purge and
Pressure
Tests
98.83%
98.82%
98.81%
98.79%
98.68%
98.53%
98.34%
98.09%
97.79%
97.43%
97.01%
96.52%
95.94%
95.27%
94.49%
93.59%
92.55%
91 .35%
89.98%
88.40%
86.61%
84.57%
82.29%
79.74%
76.92%
73.86%
       Up through the  age of three (3) years,
       are identical to those  in Appendix C.
these  values
*  This assumes that either the geographic area has  no I/M program or that
   the existing I/M program does not include a check of the OBD MIL.

-------
                         -43-
                      Appendix E

Predicted Frequency of Occurrence of "Gross Liquid Leakers"
            by Emission Type and Vehicle Age
         For 1999 and Newer Model Years Vehicles
Vehicle
Age
(years)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Resting
Loss/
Diurnal
0.02%
0.03%
0.03%
0.04%
0.04%
0.05%
0.06%
0.08%
0.09%
0.11%
0.13%
0.16%
0.19%
0.23%
0.27%
0.33%
0.39%
0.47%
0.55%
0.66%
0.78%
0.92%
1.08%
1 .27%
1 .49%
1.73%
Running
Loss
0.05%
0.06%
0.07%
0.09%
0.11%
0.13%
0.16%
0.20%
0.24%
0.29%
0.35%
0.42%
0.50%
0.61%
0.72%
0.86%
1.02%
1 .20%
1 .40%
1 .63%
1.88%
2.14%
2.43%
2.72%
3.02%
3.32%
Hot
Soak
0.07%
0.09%
0.10%
0.13%
0.15%
0.19%
0.23%
0.27%
0.33%
0.40%
0.48%
0.58%
0.70%
0.83%
1.00%
1.19%
1.41%
1 .66%
1.95%
2.28%
2.64%
3.04%
3.48%
3.96%
4.46%
4.99%

-------
                               -44-
                           Appendix F


         Response to Peer Review Comments from Sandeep Kishan


     This report was formally peer reviewed by one peer reviewer
(Sandeep Kishan).   In this appendix,  comments from Sandeep Kishan
are reproduced in plain text, and EPA's responses to those
comments are interspersed in indented italics.  Each of these
comments refer to page numbers in the earlier draft version
(dated June 10, 1999) that do not necessarily match the page
numbers in this final version.

               ************************************

This memorandum provides peer review comments on the EPA draft
report:  "Estimating Weighting Factors for Evaporative Emissions
in MOBILE6," Document No. M6.EVP.006, June 10, 1999.

Overall, this report was clearly written and the general
methodology seems good.  We do not have any recommendations on
alternate datasets.  The part of the report that needs a better
explanation is the last paragraph on Page 16 describing the
relationship among the hot soak, RTD, and running loss gross
liquid leakers.

Document No. M6.EVP.006  (June 10, 1999)

We have the following questions, comments, and recommendations on
this draft report.   For each item we give the page number and
paragraph that the comment refers to, if it is a specific
comment.

1.   Page 3, last paragraph - The text states that approximately
     15% of the vehicles did not have purge or pressure tests
     performed for one reason or another.  The report goes on to
     say that the proposal is to assume that these vehicles were
     distributed proportionately among the three purge/pressure
     strata.  Are there any engineering reasons to believe that
     these omitted vehicles are different from the vehicles
     included in the data analysis with respect to purge/pressure
     fail rates?

     No, we have no reason to assume  (or believe) that these
     omitted vehicles  (that did not have purge and/or pressure
     tests performed) are different from the remaining vehicles.
     Which is why we treated them as we did.

2.   Page 4, last paragraph - The last sentence in this paragraph
     needs to be restated.  We are not sure what the author is
     trying to say with this sentence.  Is the author trying to
     say that the CRC data is unreliable since it is not an EPA

-------
                               -45-
     data source?  Is the author trying to say that the 57.6%
     fail rate is amazingly high and the value cannot possibly be
     biased high by EPA since the data is entirely CRC data?  The
     underlining of the word entirely helps add to the confusion
     that this sentence brings to the mind of the reader.

     The wording (and underlining) has been changed to avoid
     confusion.

3.   Page 6,  last paragraph in Section 3.1 - How did you
     calculate the confidence limits given in Table 2?  We would
     have used binomial confidence intervals.  It appears this is
     what the author has done.  It should be so stated in the
     text.

     The reviewer is correct, EPA did use binomial confidence
     intervals.  This fact has been added to the text just prior
     to Table 2.

4.   Page 6,  last paragraph in Section 3.1 - The author chose to
     select the first two of the three categories given at the
     beginning of the section to model.  Was there any basis for
     the selection?  The only reason we can see is that modeling
     the purge test only failures would appear to be a nightmare
     as shown in Figure 3 on page 12.

     "Nightmare" is probably too strong a word.  We would also
     have had some difficulty if we had attempted to directly
     model the vehicles that failed only the pressure test.
     Since we have three frequency curves that must add to
     exactly 100 percent at each age,  we only need to determine
     two of those curves  (the third being the remainder).  This
     means that there are three different approaches to modeling
     these three curves.  While each approach yields slightly
     different sets of equations, the predicted values are close
     for the ages with large sample sizes.

     Since the logistic growth curve is monotonic (strictly
     increasing or strictly decreasing), EPA chose to model the
     two curves most likely to also be monotonic, specifically
     the vehicles failing the pressure test  (strictly increasing)
     and the vehicles passing both tests  (strictly decreasing).
     (Similarly, the vehicles failing both tests are expected to
     have a strictly increasing curve.)

5.   Page 8,  first full paragraph - This comment concerns the
     discussion of the average test date for the dataset.
     Clearly,  if testing is going on all year long at a
     relatively constant rate, the average date will be near July
     1.  The middle sentence in this paragraph needs to be
     reworded.  The phrase "the typical test" is a little strange
     and gives the impression that 14,000 tests were performed
     during one week in July.

-------
                               -46-
     That paragraph has been rewritten to eliminate the
     confusion.

6.    Page 8,  second full paragraph - We understand that weighting
     the vehicle age average values they have fewer observations
     less than those that have more observations is appropriate
     but why should the reciprocal of the variance be used?  In
     general,  we think that a preferable approach is to use
     logistic  regression on the individual data points, not the
     averages  by age,  to determine the logistic growth function.
     Use of logistic regression would automatically take into
     account the relatively sparser data for older vehicles and,
     in addition,  would provide confidence limits on the logistic
     growth curve.

     That approach was attempted but did not produce satisfactory
     predictions of the failure rates of the older vehicles.
     That approach gave a relatively little weighting to the
     small sample of 66 test vehicles older than 20 years of age.
     EPA's approach allows for models  (equations) that predict
     rates that closely approximated the observed rates listed
     (as the four bullets) on page 4.

7.    Page 8,  second full paragraph - The report states that
     Equation  I is the resulting function.  It would be useful to
     the reader for the report to state how this function was
     obtained  from the data.  Was it obtained using logistic
     regression, non-linear regression, or some other means?

     "Some other means. "  Rather than using a "canned" program
     that minimized the sum of the squares of the residuals,  we
     first divided each residual by the corresponding standard
     deviation.   We then minimized the sum of the squares of
     those fractions.   Included in that sum was the single value
     of the 66 older vehicles  (from page 4).   The graphical
     equivalent of that modified approach is to first take the
     rates at  ages 0 through 20 plus the average of ages 21-26
     (as a single point)  with the corresponding 22 confidence
     intervals.   Then, construct a curve that lies within all of
     the confidence intervals and closely approximates all 22
     observed rates.

8.    Page 16,  last paragraph - We did not follow the reasoning
     behind the concept that the hot soak gross liquid leakers is
     the union of the RTD gross liquid leakers and the running
     loss gross liquid leakers.  We did not understand the
     concept in Reference 4 and the text in this report does not
     help us understand the reasoning either.  In our opinion,
     this paragraph is the weakest part of the report.

     That material has been revised.

-------
                               -47-
9.    Page 18,  last paragraph - In the text,  there is a
     typographical error.   The value 10,910,000 should be
     1,091,000.

     The reviewer is correct.   The number has been corrected.

10.   Page 21,  middle of the page - We understand that there is no
     data for the late model vehicles which have enhanced
     evaporative emissions control systems.   We agree that the
     replacement of age with age/2 in the equations developed
     earlier is  appropriate.  However,  it may be useful to the
     reader to explain the assumption that is being made here.
     To us,  the  deterioration of evaporative emission control
     systems is  contributed to by the effects of age and miles
     which can have different types of effects on systems.  By
     substituting age/2 for age in the equations, the author has
     made the assumption that 2 times the miles is approximately
     equal to 2  times the age.  We suggest a sentence be added in
     this vicinity stating that this assumption has been made.

     This statement is now on page 22.   Point number 1 on page 21
     has been revised to include both age and mileage.

11.   Page 22 - This comment concerns the assumed values for the
     ability of  the MIL to detect or identify the vehicles
     failing the purge or pressure tests.  We understand that you
     have done your best to figure out these values of 85, 10,
     and 90% but we think that, as the car ages the response will
     go down continuously rather than a stepwise fashion.  Maybe
     an estimate of a logistic growth function would be better to
     use here.  We realize that there probably is no data.

     This suggestion will be considered once actual data become
     available.

12.   Page 23,  first paragraph - It would be helpful to put behind
     the value 76.5% the following:  (= 0.85 * 0.90).

     Done.

13.   Page 33 - The first line on the page we believe should read
     "eventually will drive."   If the occurrence of gross liquid
     leakers is  really driving the increase in diurnal
     hydrocarbon emissions in the older vehicle ages for the 2020
     calculation, then we suggest that doing an independent study
     of gross liquid leakers on vehicles using enhanced
     evaporative control standards will be suggested by people
     who are reading this report.

     Done.

14.   Page 35,  top of the page - The report shows preliminary
     analyses for full day diurnals and the effects of the

-------
                          -48-
proposals in this report.  We think it would also be useful
to see comparisons of MOBILES versus the proposed
modifications when applied to running losses, hot soaks, and
resting losses.

At the time of this was peer review, the MOBILE6 model was
not fully operational.  Since it now is fully operational,
the readers of this report can do their own comparisons
using whatever assumptions they wish.

-------
                               -49-
                           Appendix G

            Response to Written Comments from Stakeholders
     The following comments were submitted in response to EPA's
posting a draft of a related report  (M6.IM.003) on the MOBILE6
website.  The full text of these comments is posted on the
MOBILE6 website.
Comment Number:        102

     Name/Affiliation:   David Lax /  API

     Date:              January 25,  2000

     Comment:

     "Pilot Lane Data from Hammond and Phoenix -  ... Problems
     with these data  (and the analysis)  include the fact that all
     pre-1996 model year vehicles are combined, even though it is
     clear that evaporative control systems improved
     significantly between the early-70s and mid-90s."

     "In summary, there are some improvements in  the evaporative
     I/M methodology relative to MOBILES ... However, the
     baseline failure rates are too high for the  mid-80s to mid-
     90s vintage vehicles.  This is because all 1971 to 1995
     model year vehicles were combined to generate baseline
     failure rates."

     EPA's Response:

     This is a familiar situation.  For example,  if during 1995,
     you wish to obtain  (predictions of) test results of 1990
     model year vehicles at the age of 15 years,  you can either
     delay the testing for ten years (i.e., until those vehicles
     actually reach the target age), or you can test vehicles
     that are then at the target age (i.e., 1980  model year
     vehicles) and extrapolate the results.

     API's concern is that rather than there being a single curve
     estimating the occurrence of pressure failures, there are in
     fact two or more curves  that are model year  specific.  While
     this concern is a possibility, the data necessary to test
     this hypothesis are not yet available.  This report
     addresses this possibility in the discussion of "gross
     liquid leakers"  (see the last paragraph on Page 17).

-------
                          -50-
Comment:

"In additional, the canister-side procedure  [for identifying
pressure leaks] can result in failing cars that are not
failures if the technician is not careful.   This can occur
if there is a pressure control valve between the tank and
the canister (which most pre-enhanced evap vehicles have).
In this case, the line between the canister and the control
valve is pressurized to 14 inches of water (inH20)  (which
has a very small vapor space),  and then the pressure bleeds
off through the control valve into the tank.   This appears
to be a failure, but in fact is not.  Unless there is some
procedure for estimating the volume of air introduced in the
system  (and we do not believe there was in this testing),
this can result in false failures."

EPA's Response:

EPA does not believe that a significant number of these
false failures actually occurred.  In Table 1  (Page 5),  we
note that less than three percent of the vehicles under  two
years of age (53 of 1, 767) were identified as having failed
the pressure test.  Even if all of these were false failures
(which EPA believes is highly unlikely), then that would
suggest that the  "true" failure rate at 20 years of age
would be reduced from 56.2 percent  (see Appendix A) down to
53.3 percent.  This change is not significant.  A more
realistic estimate of false failures would produce even  a
smaller, less significant change.

Additionally, if some of the test vehicles were mislabeled
as having  (falsely) failed the pressure test, then the
emissions from these problem-free vehicles would have been
included in the average emissions of this stratum,  creating
an offsetting error.  (That is, predicting a larger number
of failing vehicles, but also predicting lower emissions for
each of those failing vehicles.)

The contractor that performed the actual vehicle testing for
both the EPA and CRC programs  (Automotive Testing
Laboratories) also believes that this type of technician
error was unlikely.  This contractor stated:

     "I  disagree  with  the  example  about  a  "careful"
     technician not being able to discern a restriction
     in the  line  from the canister to  the  tank.   When
     this  line  is blocked,  a  14"  pressure  is achieved
     instantly.   When a  fuel  tank is  completely full
     and the purge  line  is  completely unrestricted,  it
     still  takes  20 or  30  seconds to  fill  the system
     with  enough  gas to  indicate  a 14  inch pressure.

-------
                     -51-
We  deliberately limit  the  maximum pressure   (and
therefore  flow)  to  about 28"  pressure  to  avoid
blowing   any  lines  during   the  pressurization
process.   The  typical  fill  time to perform  the
canister  end pressure test  is  30 to  120 seconds.
The comment about a  "careful" technician implies a
technician would not notice the difference between
a  one  second fill  and  a thirty  second fill.   I
would  agree  that  a  new  technician  might  not
understand there was a problem, but after a single
day  the  tech would  realize  a normal fill  and an
abnormal fill,  like the one cited.

"I do agree  that if the technician is not careful
that errors  can be  made.   I personally feel  the
tank pressure method  is much  more prone to errors
than the canister pressure method. For example, it
is very difficult  to  judge  if the vise  grips used
to pinch  off the  vapor line  are  in  fact sealing
the line.  I have  seen  a number of false failures
that were healed by relocating and reclamping the
line.

"In  either  case,   a  careless  or  hasty  technician
can easily report  a false failure  -  the pressure
test takes great care to perform  accurately.

"During  performance  of the  Hammond and  Phoenix
pilot programs,  we  were  given an  opportunity to
monitor  the  quality  of  the  lane  techs in  that
vehicles  that were  returned to  the lab  were given
a  comprehensive inspection.   We would notice when
false  failures  were   occurring,  and   we  would
initiate  intensive  training   when more  than  an
occasional false failure was recorded for vehicles
returning to the lab.

"The  most   usual   cause  of   false  failure  was
selection of the wrong  line  to apply pressure to.
If  the  technician pressurized  a  purge  line,  the
vehicle would not pass  the test.  Exactly the same
risk would apply  to pressurizing from  the  tank -
if  the  technician  clamps  the  wrong  hose,  the
vehicle will fail.   Valves  and other restrictions
were not found to be a major problem.

"There were restrictions placed the vapor lines of
most  vehicles  before  enhanced  evap was put  in

-------
                     -52-
place.   Many of  these  were  in the canister,  or in
the  last  few  inches  of  vapor  line  before  the
canister.

"The  pressure   test   performed  in  Hammond  and
Arizona    did    not    include   these   underhood
restrictors and  pressure  relief valves - pressure
was  introduced  between   the  valve  and  the  fuel
tank.  Pressure  relief valves located at the fuel
tank could not  be bypassed.   Vehicles that  could
not  be  pressurized  at a normal  rate because  of
such  valves  were   reported  as  untestable,   or
blocked vapor line,  not as pressure failures.

"I do agree the  CRC  data  provide some significant
insight into the two procedures.   E-9,  E-35,  and
E-41 each  include tests  from  the  canister to the
tank and  from  the tank  to  the  canister.  Because
the  test  from  the canister to the  tank includes
the cap and the seal between the tank and the cap,
the failure rate  from  the canister to the  tank is
always higher than from the tank to the canister.
In these three studies, when there is disagreement
between  the two methods,  the  cap is  typically
failed.   Pressure levels  are reported for  the two
methods on each  vehicle  at  initial,  one,  and two
minute periods. "

-------