United States       Air and Radiation       EPA420-P-99-023
            Environmental Protection                June 1999
            Agency                     M6.EVP.006
vvEPA     Estimating Weighting
            Factors for Evaporative
            Emissions in MOBILE6
                                 > Printed on Recycled Paper

-------
                                     EPA420-P-99-023

                    -  Draft  -


        Estimating  Weighting Factors
   for Evaporative  Emissions in  MOBILE6

                 Larry C. Landman

           Document  Number M6.EVP.006
                    June  10,  1999
                      U.S.  EPA
           Assessment  and Modeling  Division
     National  Vehicle  Fuel  and  Emissions  Laboratory
               2000  Traverwood  Drive
           Ann  Arbor, Michigan  48105-2425
                 734-214-4939  (fax)
                   mobile@epa.gov
                       NOTICE


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

-------
                             ABSTRACT
     In previous documents  (M6.EVP.001, M6.EVP.002, and
M6.EVP.005), EPA proposed methods of estimating vehicles' resting
loss and diurnal emissions based (in part) on the vehicles'
performance (pass or fail) on the purge and pressure tests.  EPA
plans to compute 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
proposed estimates of pass and fail rates as functions of vehicle
age.

     Please note that EPA is seeking any  input from stakeholders
and reviewers that might aid us in modeling any aspect of resting
loss or diurnal evaporative emissions.

     Comments on this report and its proposed use in MOBILE6
should be sent to the attention of Larry Landman.  Comments may be
submitted electronically to mobile@epa.gov, or by fax to (734)
214-4939, or by mail to "MOBILE6 Review Comments", US EPA
Assessment and Modeling Division, 2000 Traverwood Drive, Ann
Arbor,  MI  48105.  Electronic submission of comments is preferred.
In your comments, please note clearly the document that you are
commenting on, including the report title and the code number
listed.  Please be sure to include your name, address,
affiliation, and any other pertinent information.

     This document is being released and posted. Comments will be
accepted for sixty (60)  days.  EPA will then review and consider
all comments received and will provide a summary of those
comments, and how we are responding to them.

-------
                        TABLE OF CONTENTS
                                                     Page Number
1.0  Introduction  	   1
2.0  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
     3.2  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  .  .   20
5.0  Comparisons with MOBILES	24
     5.1  Comparisons of Weighting Factors 	   24
     5.2  Comparisons of Weighted Diurnal Emissions ...   27
     5.3  Comparisons of Diurnal Emissions from Vehicles
          Certified to the Enhanced Evaporative Control
          Standards	32
6.0  Summary	34
7.0  References	36
                                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 	   37

 B.  Predicted Frequency of Occurrence of "Gross
     Liquid Leakers" by Emission Type and Vehicle
     Age  (for 1995 and Older Model Year Vehicles) ....   38

 C.  Estimates of Purge/Pressure Strata Size by
     Vehicle Age for 1999 and Newer Model Year
     Vehicles  for I/M Areas	39

 D.  Estimates of Purge/Pressure Strata Size by
     Vehicle Age for 1999 and Newer Model Year
     Vehicles for Non-I/M Areas 	   40

 E.  Predicted Frequency of Occurrence of "Gross
     Liquid Leakers" by Emission Type and Vehicle
     Age  (for 1999 and Newer Model Year Vehicles) ....   41
                                111

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

                          Larry C.  Landman
             U.S. EPA Assessment and Modeling Division
1 .0  Introduction

     In  three  recently released draft reports  [1,2,3]*,  the US
Environmental Protection Agency  (EPA) proposed 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  (a pressure test**  and a purge test) were used to screen
the candidate vehicles.  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 proposes
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 earlier 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
  * The numbers  in brackets refer to the references  in Section 7 (page 36).

* * 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-                           DRAFT
substantial leaks of liquid gasoline  (as opposed to simply vapor
leaks); these vehicles were labeled "gross liquid leakers."  EPA
proposed [4] using the following  three definitions (based 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 test
       (see also reference  [5]) ,  or

    •  running loss test emissions were at least 7.00 grams per
       mile  (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)
       combined with  119 vehicles tested by EPA. [1]

    •  For the "gross liquid leakers"  identified by the hot soak
       test, EPA used a sample consisting of 300 vehicles tested
  * Vehicles failing both purge and pressure  are discussed in Section 3.1.4.

-------
                                -3-                           DRAFT
       by Auto/Oil during 1993 as part of its real world hot  soak
       testing program combined with 197 vehicles tested by EPA.
       [5]

    •  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) .  [7]   (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) .  [8]  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
  * 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-                           DRAFT
number of classified pressure failures divided by the total number
of vehicles that were classifiable),  EPA proposes to treat the
results of those analyses as if they applied to the entire in-use
fleet.  This proposal is equivalent to assuming that those 15
percent of unclassifiable vehicles were distributed
proportionately among the three purge/pressure strata.

     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 is due
entirely to data recently obtained in the CRC testing programs.

-------
                                -5-
DRAFT
                             Table  1

   Distribution  of  14,061    1971-95  Model  Year Vehicles

Vehicle
Age*
0
1
2
3
4
5
6
7
8
9
1 0
1 1
1 2
1 3
1 4
1 5
1 6
1 7
1 8
1 9
20
21
22
23
24
25
26
Performance on Purge and
Pressure Tests
Fail
Pressure
5
48
42
61
81
94
91
94
88
68
46
41
64
49
29
1 9
1 3
7
4
1 2
1 2
3
7
1 0
6
6
6
Fail Only
Purge
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
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
1 7
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
1 5
6
1 7
22
1 2
1 2
1 5
1 0
1 0
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-                           DRAFT


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.

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

          Estimating  Rate of  Failing 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
1 0
1 1
1 2
1 3
1 4
1 5
1 6
1 7
1 8
1 9
20
21
22
23
24
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
1 5
6
1 7
22
1 2
1 2
1 5
1 0
1 0
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,  knowing that seven of 15 of the vehicles
17 years of age failed the pressure test indicates that the actual
failure is most likely between 25 percent and 68 percent.   A range
that large is not helpful in predicting the true failure rate.)
However, using both the sample failure rates and the confidence

-------
                                                              DRAFT
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.

    •  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  variable "AGE"  (in Tables 1,  2,
and 3)  which is based on the test date  (since it is the integer
calculated by subtracting the model year from the test year).
However,  since the typical test was performed in early July (mean
date of July 3 and median date of July 10),  we modified that
variable by adding 0.5 so that the predicted rate of vehicles
failing the pressure test would be based on the age of the
vehicles as of the first of January which coincides with the date
used in MOBILE.

     To account for differences in the size of the confidence
intervals (or, equivalently the sample size) 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:

             _ .,   _ 4      	0.6045	
     Pressure Fa.lure Rate  =  1  +  1 7.733.exp[_0.01362*(AG EA2)]        (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
measured failure rate on the pressure test  (from Table 2, shifted
by six months to compensate for the July testing)  and the curve
described by  equation (1).   That graph  suggests that equation (1)
is a very good fit for the measured failure rates except at ages
18, 19 and 21 years (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-
   DRAFT
                               Figure  1

               Comparison  of Measured and Predicted  Rates
                  For Vehicles Failing the Pressure Test
                           (by  vehicle  age)
          90%
      0)
      4-*
      TO
      (A
      0)
      0)
      3
      (A
      (1)
          60% --
          30% --
           0%
                             1 0             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 I 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-
DRAFT
                              Table 3

  Estimating  Rate of Passing  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
1 0
1 1
1 2
1 3
1 4
1 5
1 6
1 7
1 8
1 9
20
21
22
23
24
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
1 5
6
1 7
22
1 2
1 2
1 5
1 0
1 0
7
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):
     Rate of Passing Both =  1 -
                                         0.7200
                                  13.40*exp[-0.0145*(AGEA2)]
   (2)

-------
                                -11-
                                                    DRAFT
     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).

     In Figure  2, for vehicles passing  both the  purge and pressure
tests,  we plotted both the measured rates  (from  Table 2,  with age
shifted by six months to compensate for the July testing)  and the
curve described by  equation (2).   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  Measured and  Predicted Rates
          For Vehicles Passing Both the Purge  and Pressure Tests
                           (by vehicle age)
          100%
      (A
      0)
      o
      CD
      O)
      c
      'in
      (A
      TO
      D.
50% --
 0%
                             1 0             20

                             Vehicle Age  (years)
                                                 30
3.1.3   Vehicles  Failing  ONLY the  Purge Test

     The third (and final)  stratum  based on vehicles'  performance
on the purge and pressure tests is  that  containing  vehicles  that
failed only the purge test.  EPA proposes to estimate  that  stratum
by subtracting from one hundred percent the total  of  equation (1)
plus equation (2)  (prior to  adjusting for the gross liquid
leakers, as discussed in Section 3.3).   Since,  for  some  vehicle
ages, the rate of  decline in equation (2) is greater that  the rate
of growth in equation (1),  this approach has  the  effect  of
predicting a decrease in the rate of  purge  only failures for

-------
                                -12-
                                                    DRAFT
vehicles older than 16 years of  age.   (This  effect suggests that
some of the vehicles that had  failed  only the purge test 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 measured rates  (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 Measured and Predicted Rates
                 For Vehicles Failing ONLY the  Purge Test
                           (by vehicle age)
          30%
          20% --
      TO
      o:
          10%--
      (A
      0)
      0)
      O)
      D
      Q_
0%
                             1 0             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-
                                         DRAFT
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:


                              Figure 4

                   Predicted Pressure and Purge  Rates
                           (by vehicle  age)
       1 0 0 %
        7 5 % •
     >  5 0 %
     c
     0)
     3
     .   25 % H
          0 %
                                       Fail ONLY Purge
    Pass Both
Purge and Pressure
                                10       15

                            Vehicle Age  (years)
                             2 0
2 5
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.

     As a service to possible  future  researchers  and modelers who
may require an estimate of  the  number of  vehicles failing both the

-------
                               -14-
DRAFT
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 Failing  BOTH  the Pressure  and Purge  Tests
                For  1971-95 Model  Year Vehicles
              With  90  Percent  Confidence  Intervals
Vehicle
Age
0
1
2
3
4
5
6
7
8
9
1 0
1 1
1 2
1 3
1 4
1 5
1 6
1 7
1 8
1 9
20
21
22
23
24
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
1 5
6
1 7
22
1 2
1 2
1 5
1 0
1 0
7
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% - 11.9%
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-
                                                     DRAFT
     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, 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:
     Rate of Failing Both  =
                                    0.3536
                          1 + 414.96*exp[-0.32955*AGE]
     In Figure  5,  for vehicles failing both the purge  and  pressure
tests,  we plotted  both the measured rates (from Table  4, shifted
by six months to compensate for the July testing) and  the  curve
described  by equation (3).
                               Figure 5

               Comparison of Measured and Predicted Rates
           For Vehicles Failing Both the Pressure and Purge  Tests
                           (by vehicle age)
      0)
      O)
      D
      Q_

      T3
      c
      TO
      D
      (A
      (A
      0)
      O
      CD
          60%
40% --
          20% --
           0%
                             1 0              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  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

-------
                                -16-                           DRAFT
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.

3.2     Modeling  the  Stratum of  "Gross Liquid  Leakers"

     The set of vehicles identified as  "gross  liquid leakers"
varies depending upon which type of evaporative  emission is  being
considered.  Earlier  (see "bulletted" 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).  These
definitions were developed in a recent 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

    Based on  RTD Testing         =   1 + ^ 4^ 3,°e^l°Q236Q4 . A G E j   ( 4 )


  Rate of Gross Liquid  Leakers

    Based on  Running  Loss Testing  =   1 + 120 * exp^O .4 * A G  E ]        ( 5 >


See reference [4] for the details  of how these equations were
derived.

     In that same report, EPA proposed  that the  vehicles
classified as "gross liquid leakers" on the hot  soak test are the
same vehicles identified as gross liquid leakers on either  the
running loss or RTD tests.    (That is, the set  of vehicles
classified as gross liquid leakers on the hot  soak test is  the
union of the set of vehicles classified as gross liquid leakers on
the RTD test with the set of vehicles classified as gross liquid
leakers on the running loss test.)  Therefore, the rate of  gross
liquid leakers as identified on the hot soak test would be  the  sum
of the two rates for the RTD testing and the running loss 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.

-------
                                -17-
   DRAFT
                              Figure  6

              Predicted Occurrences of "Gross Liquid  Leakers"
              On Each of Three Tests of  Evaporative Emissions
                           (by  vehicle  age)
                                          Running Loss Test
                            1 0              20

                            Vehicle Age  (years)
30
     It is important  to  note  that this model of the frequency of
gross liquid leakers  is  based on  the  assumption that modern
technology vehicles will  show the same tendency toward gross
liquid leaks as do the older  technology vehicles at the same age.
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 growth functions with two or three
curves specific to different  model year groups.

     Since EPA has no data to indicate model-year specific rates,
EPA proposes to 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
1996 model year along with some of the 1997 and 1998 model years).
For the vehicles designed to  meet the new enhanced evaporative
test procedure, EPA proposes  to modify that equation (see Section
4.0) .

-------
                                -18-
DRAFT
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
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 the all three of the
purge/pressure strata.  EPA proposes to first estimate the number
of gross liquid leakers within each of the purge/pressure strata
and then to remove 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,  EPA proposes that MOBILE6 will simply
distribute the liquid leakers proportionately among the three
purge/pressure categories.

     For example, if we were  to take  a hypothetical fleet of
10,000,000 vehicles 10 years of age, then Appendix B predicts that
780,000 (7.8 percent) of them will be "gross liquid leakers" on
the RTD test.  Similarly,  Appendix A indicates that  10,910,000
(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 780,000 "gross liquid leakers" proportionately among the
three purge/pressure strata predicts the following distribution:
                              Table 5

                   Predicted Distribution on RTD  Test
Purge/Pressure
Strata
Fail Pressure
Fail ONLY Purge
Pass Both
TOTALS
"Gross Liquid
Leakers"
85,098
50,466
644,436
780,000
Not "Gross
Liquid Leakers"
1,005,902
596,534
7,617,564
9,220,000
TOTALS
1,091,000
647,000
8,262,000
10,000,000

-------
                                -19-
                                           DRAFT
Thus, Table 5 indicates  that  for 10 year old vehicles on the RTD
test:

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

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

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

     •  76.18 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

                   Predicted Strata Sizes  on RTD  Test
                           (by vehicle age)
    1 0 0 %
     80% --
     60% -
     40% -
     20% -
      0 % -I
Pass Both
Purge and
Pressure
                           10               20

                           Vehicle Age   (years)
                                         3 0

-------
                                -20-
                                                      DRAFT
     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.
4.0
Modeling  Enhanced  EVAP Vehicles Equipped  with  OBD
     Beginning with the 1996 model year, manufacturers were
required to certify twenty percent of their vehicles using a new
"enhanced" evaporative testing procedure and to equip those
vehicles with an on-board diagnostic  (OBD)  system that would
detect pressure leaks and malfunctions in the purge system; that
percentage is scheduled to increase to one hundred percent by the
1999 model year.  The phase-in percentages are prescribed in
40 CFR 86.096-8 and shown below in Table 6.


                              Table  6
                       Phase-In  of Vehicles with
                     Enhanced Evaporative  Controls
Model Year
1995
1996
1997
1998
1999
Percentage
0%
20%
40%
90%
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 50,000
          to  100,000 miles  for  light-duty vehicles).
     2)   The OBD system 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):

-------
                                 -21-                            DRAFT
       •    "quick connects" that  reduce the possibility of  improper
           assembly when the vehicle is serviced,
       •    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,  EPA proposed  in a separate report
(M6.EVP.005)  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, proposes to 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)
                                 13.40*exp[-0.003625*(AGEA2)]
                                                                   (7)
     Rate of Gross Liquid  Leakers on the RTD Test for the Enhanced Evaporative
     Control Vehicles

                                    0.08902                       (8)
                          1 + 414.613*exp(-0.1842*AGE)
     Rate of Gross Liquid Leakers on the Running Loss Test for the Enhanced
     Evaporative Control Vehicles

                       =   	P-06	
                          1 + 1 20*exp[-0.2 * A G E ]                  v

-------
                                -22-                           DRAFT
     Using these equations, we generated the estimated failure
rates in Appendix C for those (1996 and newer)  vehicles certified
to the enhanced evaporative control standards.   These estimates
assume only the benefits of changes in durability, with no
estimation of the effect from the OBD system.

     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 (in Appendix C),  depending both on its
ability to identify problems in the evaporative control systems
and on the owners inclination to repair such problems.

     In an separate analysis  (M6.IM.001) of exhaust  (not
evaporative)  emissions,  EPA proposed that the effect of an OBD
system will vary, based on the vehicle's warranty and the presence
of an I/M program.  That proposal stated that:

      •   The vehicle's malfunction indicator light  (MIL) would
          detect/identify 85 percent of the instances of the
          vehicle's exceeding twice the exhaust emission standard.

      •   While  the vehicle is under its full  ("bumper to bumper")
          warranty  (i.e., up through 36,000 miles),  90 percent of
          the owners will have the vehicle repaired when the MIL
          indicates a problem.
      •   When the warranty covers only the electronic control
          module and the catalytic converter (i.e.,  from 36,000
          through 80,000 miles), only 10 percent  of  the owners
          will have the vehicle repaired when the MIL indicates a
          problem.  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), none
           (i.e., zero percent) of the owners will have the  vehicle
          repaired when the MIL indicates a problem.  That
          percentage will increase from zero to 90 percent  if that
          geographic area has an I/M program that requires  the MIL
          to indicate that there are no problems.

     EPA proposes to 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.)

-------
                                -23-                           DRAFT
       •   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.)

       •   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 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 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 currently studying in-use OBD systems.  The
          assumptions listed here are working assumptions and
          subject to change.

     Applying these assumptions to the strata sizes in Appendix  C
produces the tables in Appendices D and E that cover the 1999 and
later model year vehicles,  for the I/M and non-I/M areas
respectively.

     For the vehicles produced during the phase-in years of the
enhanced evaporative standard (i.e.,  model years 1996 through
1998),  EPA proposes to use a three-step approach:

    •   First, the fraction of the fleet not manufactured to the
       new enhanced evaporative standard will be treated the same
       as the 1986-95 model year vehicles.   That is,  the emissions
       for each of the four purge/pressure/leaker strata will be
       calculated and then weighted together (using Appendix A)
       for each of the three years (1996-1998) .

    •   Then, for the vehicles manufactured to the new enhanced
       evaporative standard,  we will calculate the predicted

-------
                                -24-                           DRAFT
       emissions for each of the four purge/pressure/leaker
       strata, and then weight them together using the rates in
       either Appendix D (I/M area)  or Appendix E (non-I/M area).

    •  Finally,  we will weight those two sets of results together
       using the phase-in percentages in Table 6.


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 this proposal 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).

    •  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 vehicles 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 vehicles failing the
       pressure test (regardless of the performance on the purge
       test).

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

-------
                                -25-
                                           DRAFT
                              Figure  8

        Comparison of Predictions for Frequency  of  Vehicles
             Passing Both the Purge and  Pressure Tests
                          (by  vehicle  age)
   1 0 0 %
>  5 0 %
c
0)
3
of  25%
      0 %
 'MOBILES

 Proposal
                               10        15

                           Vehicle Age  (years)
                              2 0
2 5
                              Figure  9

        Comparison of Predictions for Frequency  of  Vehicles
                      Failing  the  Pressure Test
                          (by vehicle  age)
   60%
=^ 4 0 % - -
    0 %
•MOBILES

 Proposal
                              10         15

                           Vehicle Age   (years)
                              2 0
2 5

-------
                                -26-
            DRAFT
                              Figure  10

            Comparison of Predictions  for Frequency of Vehicles
                      Failing ONLY the Purge  Test
                           (by  vehicle  age)
       40%
       2 0 % - -
     0)
     3
     0)
        0 %
                         •MOBILES
                 ~ ~  ~ Proposal
                               10        15

                            Vehicle Age  (years)
2 0
2 5
     Based on examinations  of  these three graphs, we can make the
following observations:

    •   The most obvious difference  between the weighting factors
       used in MOBILES and  those  being proposed in this report is
       that the factors proposed  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 being  proposed in this report predict
       substantially a higher  proportion of vehicles failing the
       pressure test for vehicles older than 13 years of age than
       does the MOBILES model.

-------
                                 -27-                           DRAFT
5.2     Comparisons  of Weighted  Diurnal Emissions

     By combining the information in this  report  with the
information in  earlier reports,  we can estimate the  diurnal
emissions  for the in-use fleet.*  To compare these proposed
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 11)  and

     •  daily temperatures  cycling between  82 and  106 degrees
       Fahrenheit using fuel with a 7.0 RVP  (see  Figure 12).

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 Figures 11 and 12, the typical  (i.e., mean)  24-hour
diurnal emissions for each vehicle are plotted against model year.
Visual inspections of Figures 11 and 12 suggest that the estimates
of diurnal emissions  resulting from this new proposal (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.

In these two examples the variable  MODEL YEAR can be  transformed
into AGE  using the  equation: AGE = 95 - (MODEL YEAR).
   It is  important  to note that these  (following) proposed 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  being developed.  A true  comparison between
   this proposal  and MOBILES  will require  using  activity data  to  weight
   together all of  these individual components of evaporative emissions.

-------
                                -28-
                                                                DRAFT
                             Figure  11

Comparison  of  Predictions  for In-Use Diurnal  Emissions by  Model  Year
             60° to 84°  F  Cycle  --  Using 9.0 RVP Fuel
                      (As  of January  1, 1995)
     4 0
in
E

o>

O
I
  re
  c
     20--
                                                      MOBILE5
                                             ~  ~  ~  Proposed
       7 0
                 7 5
80         85

 Model Year
9 0
9 5
                             Figure  12

Comparison  of  Predictions  for In-Use Diurnal  Emissions by  Model  Year
             82° to 106° F Cycle  --  Using 7.0  RVP  Fuel
                      (As of January 1, 1995)
     4 0
  in
  E
  re
  o>
  O
  X
  c
     20--
         V
       7 0
                 7 5
80         85

 Model Year
                        MOBILES

                        Proposed
9 0
9 5

-------
                                -29-
                                                        DRAFT
     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 11  and  12,  we weighted each model year's estimated
diurnal emissions by the relative number of vehicles in the fleet,
we obtain the estimate  of  the mean daily diurnal emissions for an
average in-use vehicle  subject to a full day's diurnal:
        Temperature  Cycle and Fuel
        60°  to  84°  F with 9.0 RVP
        82°  to 106°  F with 7.0 RVP
                                MOBILES
                                Estimate
                                  4.86
                                  7.61
  New
Proposal
  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 13  through  16.   The first three figures
(Figures 13 through 15) are based on January 1,  1995 (thus
covering model years 1971  through 1995) .   Figure 16 is based at
January 1, 1985  (thus,  covering model years 1961 through 1985) .
Therefore, Figures 15 and  16 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   13

       Comparison  of Predictions for Mean 24-Hour Diurnal Emissions
                60° to  84° F Cycle --   Per In-Use Vehicle
                        (As of January 1,  1995)
         1 5
(A

|  1 0 -'
O)

O
X
-   5 4-
re
c
3
Q
                      •MOBILES

                       Proposed
                            8               1 0

                            Fuel RVP  (psi)
                                                    1 2

-------
                             -30-
                                                          DRAFT
                          Figure  14

Comparison of Predictions for Mean 24-Hour  Diurnal Emissions
           72° to 96° F Cycle  --  Per In-Use Vehicle
                   (As of January 1,  1995)
   3 0
(A

I*
O)

O
I

"re
c
3
Q
0 --
0 --
             •MOBILES

             Proposed
                           8

                         Fuel RVP  (psi)
                                            1 0
                          Figure  15

Comparison of Predictions for Mean 24-Hour  Diurnal Emissions
          82°  to  106°  F  Cycle   --   Per In-Use  Vehicle
                   (As of January 1,  1995)
   4 0
(A
E
re
o>
O
I

"re
   20--
                                 8
                                                      1 0
                         Fuel RVP  (psi)

-------
                                -31-
                                                    DRAFT
                              Figure 16

       Comparison of Predictions for  Mean  24-Hour Diurnal Emissions
               82°  to 106°  F  Cycle  --  Per In-Use Vehicle
                       (As of January 1, 1985)
         6 0
       (A

       re  4
       o>

       O
       I
       -  2
       re
       c
       3
       Q
0 --
0 --
           •MOBILES

            Proposed
                                   8

                             Fuel RVP   (psi)
                                                1 0
     Based on examinations of  these  four graphs,  we made the
following observations:

    •  For the lowest temperature cycle  (i.e.,  daily temperatures
       cycling between  60° and 84°F),  the newly proposed methods
       of predicting fleet diurnal emissions  produce results quite
       similar to MOBILES for  the full range  of fuel RVPs.
       (Figure 13.)

    •  For the two higher temperature cycles,  these newly  proposed
       methods of predicting 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  14 through 16.)
       The new approach  (compared to MOBILES)  predicts slightly
       higher 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
            new approach predicts slightly higher diurnal emissions
            for fuel  RVPs up through 11 psi.  For fuel  RVPs  near  7

-------
                                -32-                          DRAFT
            psi,  the  new approach  predicts  less  than  0.9 grams of
            HC  (above the MOBILES  estimates) per 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 new  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 15).   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 16)  yields  similar
            results.

5.3     Comparisons   of Diurnal  Emissions from  Vehicles
        Certified to  the Enhanced  Evaporative  Control Standards

     In the preceding  section, we compared, for 1995 and older
model year vehicles,  these proposed estimates  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 11 and 12 but
with January 1, 2020  as the base produced the  two following
figures  (Figures 17 and 18).  Visual inspections of Figures 17  and
18 leads to the following conclusions:

    •  This proposal  predicts substantially lower diurnal
       emissions for  these new vehicles than does MOBILES.

    •  For vehicles between 10 and 25 years of age, this  proposal
       predicts a substantial rise in the diurnal emissions of  the
       in-use fleet  (on a per vehicle basis).

       This increase  in emissions is driven primarily by the
       increasing numbers of gross liquid leakers.  With
       improvements in  the durability of evaporative emission
       control systems, discussed in Section 4.0, the high

-------
                          -33-
                                                      DRAFT
 emissions associated with "gross liquid leakers" eventually
 drive the overall fleet  average diurnal emissions curve.
 This is despite the fact that  their predicted rate of
 occurrence among vehicles certified to the enhanced
 evaporative standards  is very  low (less than two percent at
 age 25).

 Thus, modifying assumptions  on the frequency of gross
 liquid leakers as a function of age among these vehicles
 would significantly affect the graphs for "Proposal" in
 Figures 17 and 18, with  that curve remaining much flatter
 throughout.  As we have  no information of the impact of the
 new enhanced evaporative control requirements on the rate
 of occurrence of "gross  liquid leakers," we are especially
 interested in comments in this area.
                        Figure  17

          For  Enhanced Evaporative  Control  Vehicles
 Comparison of Predictions  for In-Use Diurnal Emissions by Age
         60° to 84° F Cycle  --  Using  9.0 RVP Fuel
                     (Non-l/M  Area)
n
E
TO
^
O)
o
TO
c
D
b
6 --
   3 --
                        10        15

                     Vehicle Age  (years)
                                         20
2 5

-------
                                 -34-
                                                        DRAFT
                              Figure  18

                For Enhanced Evaporative Control Vehicles
       Comparison  of Predictions for In-Use Diurnal  Emissions by Age
               82° to 106°  F  Cycle  --  Using 7.0 RVP Fuel
                            (Non-l/M   Area)
       n
       E
       TO
       ^
       O)
      o
       TO
       c
       D
       b
 6 --
         3 --
                              10        15

                           Vehicle  Age  (years)
                                          20
2 5
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,
    •  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.

-------
                                -35-                           DRAFT
     Very preliminary analyses suggest that for the full-day
diurnal (see the footnote on page 27),  the proposals for MOBILE6
would:

    •  predict lower diurnal emissions than does MOBILES for the
       1999 and newer vehicles,

    •  predict similar diurnal emissions than does MOBILES for the
       1986-1995 model year vehicles, and

    •  predict higher diurnal emissions than does MOBILES for the
       1985 and older vehicles.

     Based on this proposal, we can perform a simplified analysis
to estimate the effect (benefit)  on diurnal emissions of reducing
fuel RVP:

    •  In each of the four figures (Figure 13 through 16),  the
       MOBILES graph is steeper than the graph of the new
       approach.  The lower slopes associated with the MOBILE6
       proposal will result in smaller decreases in predicted
       diurnal emissions associated with a change 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.

    •  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.

-------
                                -36-                           DRAFT
7.0  References

1 ) Larry Landman, "Evaluating Resting Loss and Diurnal Evaporative
   Emissions Using RTD Tests," Report numbered M6.EVP.001, July
   1999.

2 ) Larry Landman, "Modeling Hourly Diurnal Emissions and
   Interrupted Diurnal Emissions Based on Real-Time Diurnal Data,"
   Report numbered M6.EVP.002, July 1999.

3 ) Larry Landman, "Modeling Diurnal and Resting Loss Emission from
   Vehicles Certified to the Enhanced Evaporative Standards,"
   Report numbered M6.EVP.005, November 1998.

4 ) Larry Landman, "Evaporative Emissions of Gross Liquid Leakers
   in MOBILE6," Report numbered M6.EVP.009, June 1999.

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, June 1999.

7 ) 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.

8 ) 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.

-------
                          -37-
DRAFT
                      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
1 0
1 1
1 2
1 3
1 4
1 5
1 6
1 7
1 8
1 9
20
21
22
23
24
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%
11.89%
11.74%
11.65%
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%

-------
                            -38-
DRAFT
                         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
1 1
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Resting
Loss /
Diurnal
0.02%
0.03%
0.04%
0.06%
0.09%
0.13%
0.19%
0.27%
0.39%
0.55%
0.78%
1 .08%
1 .49%
2.00%
2.63%
3.36%
4.15%
4.97%
5.75%
6.46%
7.05%
7.54%
7.91%
8.19%
8.40%
8.55%
Running
Loss
0.05%
0.07%
0.11%
0.16%
0.24%
0.35%
0.50%
0.72%
1.02%
1 .40%
1.88%
2.43%
3.02%
3.61%
4.16%
4.62%
5.00%
5.29%
5.51%
5.66%
5.77%
5.84%
5.89%
5.93%
5.95%
5.97%
Hot
Soak
0.07%
0.10%
0.15%
0.23%
0.33%
0.48%
0.70%
1.00%
1 .41 %
1.95%
2.64%
3.48%
4.46%
5.54%
6.67%
7.83%
8.95%
10.00%
10.94%
1 1 .75%
12.42%
12.94%
13.34%
13.63%
13.85%
14.00%

-------
                              -39-
DRAFT
                          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
1 0
1 1
1 2
1 3
1 4
1 5
1 6
1 7
1 8
1 9
20
21
22
23
24
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.

-------
                              -40-
                 DRAFT
                          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
1 0
1 1
1 2
1 3
1 4
1 5
1 6
1 7
1 8
1 9
20
21
22
23
24
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,
   identical to those in Appendix  C.
the values are
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.

-------
                            -41-
DRAFT
                        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
1 1
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%

-------