-------
                                -25-                           DRAFT

                                                          July  13,  1999
     While only the test results  from  the  57  vehicles  that  were
tested over the full range of fuel RVPs and temperature  cycles
were used to determine the coefficient  ("B")  which determines  the
slope of the lines.  The full data set was used only to  solve  for
the individual constant terms  ("A").

     For each of the strata  identified in  Section  6.4, we
calculated the value of "A"  that  would minimize the difference
between the predicted and the actual resting  losses (i.e.,  the
residuals).   If more tests had been conducted at a given
combination of temperature and fuel RVP  (e.g.,  72  °F using  6.8 psi
RVP fuel),  then the average  resting loss emissions at  that
combination was then more heavily weighted in the  process to
calculate the value "A".

     This process produced a regression equation for each of the
18 strata;  however, the predicted results based on the vehicle's
pass/fail status on the purge test were inconsistent.  This
inconsistency is not surprising since  the types of mechanical
problems that would cause a  purge failure are not  likely to
contribute to resting loss emissions.*  To address this  situation,
the population was stratified based simply on whether  the vehicles
pass or fail just the pressure test.  The regression equations  for
each of the 12 resulting strata are given in  Appendix D.  The
regression equations are unique for each stratum in which testing
was performed.  The untested strata of pre-1980 fuel-injected
vehicles used the regression equations of the pre-1980 carbureted
vehicles.

     Using these 12 equations, we calculated  an estimate of the
hourly resting loss emissions at  each hour of the  three
temperature cycles.  Then, adding the hourly  estimates for  the
first 24 hours of each test  produced the daily resting loss
emissions (for each of the 24 strata).  These equations  all
predict that the full day's  resting loss emissions (in grams)
would be 24 times the hourly resting loss  (calculated  at the day's
low temperature)  plus 0.766.

     These equations predict resting loss  emissions of the
carbureted vehicles to be higher  than  for the fuel injected
vehicles.   While these regressions can be used to  calculate
reasonable estimates of resting loss emissions within  the range of
temperature and fuel RVPs that were actually  tested, we  must
determine (see Section 11) how to extrapolate beyond the limits of
the test data.
   In the  previous version of MOBILE, it was  noted that  resting loss
   emissions are independent  of the canister state  (i.e.,  whether the
   canister is  saturated or fully purged).

-------
                                -26-                           DRAFT

                                                          July  13, 1999
9.0    Characterizing  24-Hour  Diurnal  Emissions

     Diurnal evaporative emissions, like most other evaporative
emissions, are functions of both fuel volatility and temperature
which are themselves interdependent.  The RVP is a measure of
vapor pressure  (VP)* at a single temperature, 100°F.   The
Clausius-Clapeyron relationship was used to estimate the  vapor
pressure at each temperature and for each of the fuels  (RVPs of
6.8 and 9.0 psi) used in this testing program.   (See Appendix  B.)

     To characterize the diurnal emissions, we again  (see Section
8.0)  identified the 57 vehicles in the EPA program that were
tested over a wide range of vapor pressures.  These test  vehicles
were distributed among 12 strata (of the 18 potential  strata
identified in Section 6.5).  Within each stratum, we then
attempted to regress the diurnal emissions against combinations of
fuel volatility and temperature.

     A similar approach was attempted to characterize  resting  loss
emissions (see previous section) but had not been successful.
However,  this approach produced more satisfactory results in
characterizing the diurnal emissions even in strata that  were
sparsely tested.  Most likely this difference was due  to  the
effect that the test-to-test variability was substantially larger
relative to the resting loss emissions than to the diurnal
emissions.  Therefore, any test-to-test variability was less
likely to hide patterns evidenced in the diurnal emissions
measurements.

     For each RTD test, the Clausius-Clapeyron relationship was
used to estimate the vapor pressure at both the low and the high
temperatures.  Using these estimates, we calculated both  the
average of the low and the high vapor pressures, as well  as the
difference between the low and the high vapor pressures  (both  in
kPa).   Multiplying these two quantities together produced a single
product term (VP*?VP) that incorporates the parameters of the  RTD
test (i.e.,  both the temperature cycle and the fuel's  RVP).

     The mean diurnal emissions  (calculated in the previous
section by subtracting a daily resting loss value from the RTD
test results) were repeatedly regressed against a polynomial of
that product term of vapor pressures within each stratum.  The
independent variable used in the regressions was either:

     1)   the product term  (i.e., the average vapor pressures
          times the difference of the vapor pressures)  or

     2)   the square of that product term  (to allow for possible
          non-linearity).
   In Appendix B,  we  illustrate how the  Clausius-Clapeyron relationship can
   be used  to  estimate  a  fuel's vapor pressure  at  each temperature if the
   fuel's RVP is known.

-------
                                 -27-                           DRAFT

                                                           July 13,  1999
However, when we  graphed the mean diurnal emissions  against this
"vapor pressure product" term,  we noted that the  affect  of  the RVP
of the test fuel  on diurnal emissions was not being  completely
accounted for by  the "vapor pressure product" term.   We,
therefore, reran  the previous regressions and included RVP  as one
of the independent  variables.  Thus, in each of those 12  strata,
we generated both a nonlinear (i.e., quadratic) model and a linear
model*.  A two step process was used to choose among those  models:

     1)   We performed a visual inspection  of  the data.   (This
          approach, in and of itself, is  not very precise,  but we
          wanted to make certain that the model  selected would be
          both  reasonable and accurately  represent the test data.)

     2)   We compared the statistical parameters  associated with
          each  of those regressions.   (That is, we identified the
          model  that optimized:  the F-ratio,  the statistical
          significance of the independent variable,  and the R-
          squared value.)

The regression analyses performed did not always  (i.e.,  in  all 12
strata) identify  the fuel  RVP as a statistically  significant
variable.  However,  for consistency, RVP was used as an
independent variable in all of the strata regardless of  its
significance level.

     Although the equations that we developed  in  this analysis are
empirical (i.e.,  data driven) models, we did impose  the  following
three restrictions  that were based on engineering experience with
diurnal emissions:

      •   The diurnal emissions should  increase with increasing
          fuel  RVP  (with all other parameters  held constant).

      •   The diurnal emissions should  increase with increasing
          temperature cycles  (with all  other parameters held
          constant).

      •   For each combination of fuel  delivery  system (i.e., fuel
          injected versus carbureted) and purge/pressure category,
          the diurnal emissions should  increase with each
          successively older model year grouping  (for each
          combination of temperature cycle  and fuel  RVP).
 * Theoretically,  in each of those models,  a  zero change in daily temperature
   (hence,  in  ?VP) should result in zero  diurnal emissions.  This  physical
   necessity would result in the constant term in each regression being  zero.
   However, this requirement was dropped because:
    (1)   of the resulting low r-squared values,
    (2)   of the lack of test data having diurnal temperature ranges less than
         24 degrees, and
    (3)   we will require, for any diurnal emissions,  a difference between the
         daily high and low temperatures of at least  five degrees Fahrenheit.

-------
                                -28-                           DRAFT

                                                           July 13, 1999
     Seven  separate  strata required additional effort to meet
these three criteria  (that were based on engineering experience):

       •  the three strata  of  1972-1979 model year carbureted
         vehicles,

       •  the 1980-1985 model  year FI vehicles that passed the
         pressure test,  and

       •  the three strata  of  1986 and newer model year carbureted
         vehicles.

     Basing the  estimates  of  diurnal emissions for the 1972-1979
model year carbureted vehicles  resulted in predicted diurnal
emissions  (for some temperature cycle / RVP combinations) that
were lower than  for the  newer (1980-85 model year) vehicles
(possibly due to the small number of 1972-79 vehicles tested over
different temperature cycles  and with different fuel RVPs).  As  a
result, we used  a modification  of the equations that resulted from
the analysis of  the 1980-85 model year carbureted vehicles.
Specifically, we used the  same  coefficients, but we altered the
constant terms so that when the modified equations were used to
estimate the emissions of  the Pre-1980 vehicles,  the sum of the
residuals  (within each purge  /  pressure stratum)  was zero.

     The strata  of 1980-85 FI vehicle that passed both the purge
and pressure tests was represented by only a single vehicle that
was tested over  the full range  of temperature cycles and fuel
RVPs.  Therefore, the results of those tests were combined with
the tests on the three 1980-85  FI vehicles the failed the purge
test but passed  the pressure  test into a single stratum of
vehicles that passed the pressure test (represented by four
vehicles).  The  regression of these data was used to determine the
coefficients for both the  stratum of 1980-85 FI vehicle the passed
both the purge and pressure tests and the stratum of 1980-85 FI
vehicle the failed only  the purge test.  The coefficient for each
stratum was the  value that would make each sum of residuals zero.

     The last three problem strata were the 1986 and newer
carbureted vehicles.  As is illustrated in Appendix C, only four
combinations of  temperature cycle and fuel RVP were tested (in
each of the three purge/pressure substrata).  The two untested
combinations were the combinations that would have yielded results
at the highest and the lowest VP values.   Having test data over
such a narrow range  (i.e.,  only the four middle values)  of vapor
pressures makes  selecting  the proper regression curve difficult.
   Theoretically,  in each of those models, a zero change in daily temperature
   (hence,  in ?VP) should result in  zero diurnal  emissions.   This physical
   necessity would result in the constant  term in each regression being zero.
   This requirement was dropped because of the resulting low r-squared values
   and because for any diurnal emissions we will require a difference between
   the daily high  and low temperatures of  at least  five degrees Fahrenheit.

-------
                                -29-                           DRAFT

                                                         July 13,  1999

We first, therefore, attempted to enlarge the scope of the data by
estimating the diurnal emissions at the two missing extreme
values.  We did this by observing that the diurnal emissions of
the 1986-95 carbureted vehicles (at the four tested combinations
of fuel RVP and temperature cycle)  were between the corresponding
diurnal emissions of the 1986-95 FI vehicles and the 1980-85
carbureted vehicles for each tested combination of fuel RVP,
temperature cycle, and purge/pressure result. If this pattern were
to hold true for the two untested combinations,  then the diurnal
emissions of the 1986-95 carbureted vehicles would be:

     •   for tests using 6.8 RVP fuel over the 60-86  °F cycle:
          •• between  4.815  and  9.519  for vehicles  failing  the
             pressure  test,
          •• between  4.372  and  5.100  for vehicles  failing  only  the
             purge  test, and
          •• between  0.187  and  2.976  for vehicles  passing  both  the
             pressure  and the purge tests.

     •   for tests using 9.0 RVP fuel over the 82-106  °F cycle:
          •• between  28.26  and  45.456 for vehicles failing the
             pressure  test,
          •• between  21.046 and 50.67 for vehicles failing only
             the  purge test, and
          •• between  9.932  and  36.565 for vehicles passing both
             the  pressure and the purge tests.

We then experimented,  using the tested values for the 1986-95
carbureted vehicles with the coefficients determined for the 1980-
85 carbureted vehicles and for the 1986-95 FI vehicles to
determine which set would most closely predict the preceding
estimates of the untested configurations.   While neither set was
perfect, the coefficients developed for the 1986-95 FI vehicles
came closer and were selected.

     Once the coefficient values of the equation were determined
for each of the 15 strata,  we then transformed the constant term
(for each stratum) to minimize the sum of the differences between
the predicted and calculated diurnal emissions.   The resulting
equations are given in Appendix E.   The statistics associated with
the eight regressions are given in Appendix F.

     In the five strata in which the vehicles passed both  the
purge test and the pressure test,  the data strongly suggest a non-
linear relationship (i.e.,  quadratic)  between the diurnal
emissions and that "vapor pressure product" term.  In the various
strata containing vehicles that failed either the purge or
pressure (or both) tests,  the relationship between diurnal
emissions and the vapor pressure product term was sometimes linear
and sometimes non-linear.

-------
                                -30-                           DRAFT

                                                          July 13,  1999
10.0   Gross  Liquid  Leakers
     Three issues related to vehicles with gross liquid leaks need
to be addressed:

     1)   the  frequency of  the occurrence of gross  liquid  leakers
           (possibly as a function of vehicle age),

     2)   the  magnitude of  the emissions from gross liquid
          leakers, and

     3)   the  effects of changes in vapor pressure  on  the  diurnal
          and  resting loss  emissions of these gross liquid
          leakers.

Analyses of these issues were hampered by a lack of a  substantial
number of identified gross liquid leakers.   We anticipate  revising
the following  initial estimates for MOBILE7 based on additional
data.

10.1  Frequency  of Gross  Liquid  Leakers

     In a concurrent report  (Document Number M6.EVP.006, entitled
"Estimating Weighting Factors for Evaporative Emissions in
MOBILE6"), EPA first uses data from EPA testing programs,   CRC
testing programs, and an American Petroleum Institute  (API)
testing program to estimate the occurrence of the gross liquid
leakers at three different vehicle ages:


                                    Frequency of
                    Vehicle        "Gross Liquid
                      Age             Leakers"
                      5.62              0.20%
                     12.50              2.00%
                     21.29              7.84%


EPA then found a logistic growth curve that closely approximates
these three values:


          ......  „  4     	0.09063	
     Gross L.qu.d Leaker Rate  = 1 +  337 .2*exp[.0 .3625 .  AGE]


     In this analysis,  vehicle age was estimated by subtracting
the model year from the test year.   Since the test dates averaged
(both mean and median)  early July,  the preceding equation  actually
estimates the occurrence of gross liquid leakers as of July of
each given calendar year.  However,  the MOBILE models base their
estimates as of January 1 of each calendar year.  Therefore, EPA
proposes to modify the preceding equation so that its predictions
are based on January first:

                                          0 09063
     Gross Liquid Leaker Rate  =
                             1 + 337.2*exp[-0.3625 * (AGE -  0.5)]

-------
                                -31-
    DRAFT

July  13,  1999
     Plotting both the unmodified curve  (i.e., based on ages as of
July) and the preceding set of three failure rates produces Figure
10-1 below:

                           Figure  10-1

               Frequency  of  Gross  Liquid  Leakers
       >
       u
       0)
       a-
       a>

5% -


0% H











*l •* ^
0 1



<
-p


X
'
I


0 2
/• ^





	





0 3
                           Vehicle Age  (years)
     The dotted line in Figure 10-1 is the logistic growth
function.  The rapidly increasing proportion of gross liquid
leakers in the in-use fleet tends to be offset by the decreasing
number of older vehicles in the in-use fleet.  This graph (or the
preceding equation) predicts:

    •  Fewer than one-half a percent of vehicles (at each age) up
       to eight years of age will be "gross liquid leakers."

    •  "Gross liquid leakers" do not reach one percent of the
       fleet until the vehicles exceed 10 years of age.

    •  "Gross liquid leakers" reach two percent of the fleet for
       vehicles exceeding 13 years of age.

    •  The portion of the fleet that is "gross liquid leakers"
       then rises  (almost linearly)  to about eight percent for
       vehicles that are 22 years old.

    •  The increase in the frequency of "gross liquid leakers"
       then levels off and the frequency approaches just over nine
       percent (about age 30).

-------
                                -32-                           DRAFT

                                                         July 13,  1999
     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
that single logistic growth function with a family of two or three
curves.

     Since EPA has no data to indicate that the multiple curve
scenario is the correct approach, EPA proposes to use the single
curve approach to estimate the occurrence in the in-use fleet of
these vehicles that have substantial leaks of liquid gasoline
(i.e.,  "gross liquid leakers").

10.2  Magnitude  of Emissions  from  Gross  Liquid  Leakers

     In Section 10.1, we concluded that the frequency of gross
liquid leakers is a function of vehicle age.  The question as to
whether the magnitude of the emissions are also a function of age
cannot be answered with the available data.

     Seven vehicles  (five in the CRC study and two in the EPA
study)  have been identified as gross liquid leakers.  However, two
of the five CRC vehicles exhibited questionable results.
Specifically:

     1)    For vehicle number 9111, the RTD test was aborted after
          only 16 hours due to the high evaporative emissions.
          CRC used the emissions measured during the first 16
          hours to estimate the emissions during the final eight
          hours.   (The cumulative HC through 16 hours was 616.71
          grams which was extrapolated to 777.14 for the full 24
          hours.)  Therefore, the calculated resting loss
          emissions  (i.e., the mean of the untested hours 19
          through 24) might be in error.

     2)    Vehicle number 9129 exhibited relatively normal
          emissions  for the about the first nine hours of the RTD
          test, after which the hourly emissions quickly rose then
          stabilized at about 11 grams per hour.  This suggests
          that the leak actually developed during the RTD test
           (around the tenth hour).  Therefore, while this
          vehicle's  resting losses  (i.e., the mean of hours 19
          through 24) were representative of other gross leakers,
          the calculated diurnal emissions are likely not
          representative of other gross leakers.   (The calculated
          resting loss emissions at 72°F  from  this  vehicle were
          10.77 grams per hour.  Had that level of emissions
          simply continued for the full 24 hours, the total
          resting loss emissions would have been 258.48 grams
          compared to the 181.79 grams actually measured for  the

-------
                                -33-
DRAFT
                                                          July  13, 1999
          entire  24-hour  RTD  test.   Computationally,  this would
          result  in  a  substantial  negative estimate of diurnal
          emissions.)

     An additional difficulty is caused by the  two vehicles in the
EPA sample not being tested with the same  fuel  as  the  five CRC
test vehicles.  However,  since the major mechanism driving the
emissions of these vehicles is the leaks of  liquid gasoline, the
effects of changes in temperature or fuel  RVP should be  relatively
small  (see Section 10.3).  If we, therefore, simply average the
emissions of these two vehicles, we obtain the  following table:
Veh No
5002
5082
RVP
9.0
9.0
Temp Cycle
72. to. 96
82. to. 106
Means:
6.3
6.3
9.0
72. to. 96
82. to. 106
72. to. 96
Means:
RTD
91.09
158.80
124.95
54.80
99.35
87.26
80.47
Hourly RstL
1.88
3.81
2.85
1.45
2.88
2.07
2.13
     If we then average the preceding  two  means  with the results
from the five vehicles in the CRC sample  (omitting  the  non-resting
loss data from vehicle 9129), we obtain:
Veh No
9049
9054
9087
91 1 1
9129
5002
5082
Means:
Std Dev:
RTD
181.35
316.59
478.16
777.14
Ignore
122.01
77.58
325.47
264.96
Hourly RstL
4.87
10.58
14.12
16.51
10.77
2.96
2.09
8.84
5.62
     A third complication becomes  apparent  when the hourly
emissions for these tests are examined.*  Several of the  tests
exhibit high and decreasing hourly emissions  for the first three
hours.  (We expected the tests to  exhibit increasing emissions  for
the first few hours.)  EPA believes that the  unexpectedly high
emissions for the first two hours  resulted  from the evaporation of
   A more  thorough analysis of  the hourly emissions  is  contained  in report
   M6.EVP.002.

-------
                                -34-                           DRAFT

                                                          July 13,  1999
gasoline that had leaked prior to the start of the test.
Compensating for that  (hypothesized)  problem results in reducing
the above mean of the RTD emissions from 325.47 grams per day down
to 312.45 (a decrease of 13.01 grams).

     On page 25, we noted that the daily resting loss emissions
(assuming a daily temperature profile similar to those in Appendix
A) would be 24 times the hourly resting loss (at the day's low
temperature) plus 0.766.  Since including the 0.766 term will
increase the day's total resting loss less than 0.4 percent, , we
will assume the resting loss emissions are completely independent
of temperature  (see Section 11.1).   Therefore,  based on the means
in the preceding table, we propose to use,  in MOBILE6, for the
category of gross liquid leakers:

       •  DAILY RESTING Loss   =  ( 24  *  HOURLY RESTING Loss)
                           =  ( 24 * 8.84)
                              212.16   (GRAMS / DAY )
     and
       •  Full Day's DIURNAL =  MEAN RTD - DAILY RESTING Loss
                           =  312.45 - 212.16
                           =  100.29   (GRAMS/DAY)

     Thus, while the occurrence of these gross liquid leakers is
relatively rare among newer vehicles (Section 10.1),  their
presence has a substantial effect on the total resting loss and
diurnal emissions of the in-use fleet.

10.3   Effects  of  Vapor  Pressure Changes on  Gross  Liquid
       Leakers

     As previously discussed, the true vapor pressure is a
function of both the ambient temperature and the Reid vapor
pressure of the fuel.  Since only two of the seven vehicles that
have been identified as gross liquid leakers were tested over a
range of fuel RVPs, there are not enough data to relate changes in
diurnal and resting loss emissions to changes in fuel RVP.
However,  as noted in the preceding section,  changes in fuel RVP
are expected to have only minimal  (proportional)  effects on the
total diurnal and resting loss emissions of vehicles whose primary
mechanism of evaporative emissions is leaking liquid gasoline.
Thus,  until additional data are available,  EPA proposes to treat
the diurnal and resting loss emissions of the gross liquid leakers
as independent of fuel RVP.

     In the previous section, EPA proposed to treat the hourly
resting emissions of these gross liquid leakers as if they are
independent of ambient temperature as well.   In a concurrent
report (document number M6.EVP.002, entitled "Modeling Hourly
Diurnal Emissions and Interrupted Diurnal Emissions Based on
Real-Time Diurnal Data"),  EPA was able to use the hourly diurnal
emissions to estimate the effects of temperature changes on the
diurnal emissions of these gross liquid leakers.   That report
concludes that the full-day's diurnal emissions of gross liquid
leakers is dependent only upon the daily temperature range  (i.e.,

-------
                                -35-                          DRAFT

                                                         July 13,  1999
the difference between the daily high and low temperatures).
Thus, for any of the three temperature cycles in Appendix A,  the
mean of the full-day's diurnal emissions of gross liquid leakers
is the constant 100.29 grams (calculated in the previous section).

     Therefore, EPA is proposing that both the hourly resting loss
emissions and full-day's diurnal emissions of gross liquid leakers
are independent of vapor pressure for any of the three temperature
cycles in Appendix A.


11.0  Other  Topics

     Several topics were not discussed in the preceding analysis
because either:

     1)   They will be discussed in forthcoming reports.

   or

     2)   No changes are planned in how they were handled in
         MOBILES.

11.1  Temperature  Ranges

     All of the tests used in this analysis were performed using
one of the three temperature cycles in Appendix A.   This results
in all of the resting loss data being at only three temperatures
(i.e., 60, 72, and 82 °F).  In Section 8, we developed regression
equations to estimate hourly resting loss emissions at
theoretically any temperature.   We will limit that  potentially
infinite temperature range as we did in the previous version of
MOBILE,  specifically:

  1)  We will assume, for light-duty vehicles other than gross
     liquid leakers, there are no resting loss emissions when the
     temperatures are below or equal to  40°F.   (This  assumption
     was used consistently for all evaporative emissions in
     MOBILES.)

  2)  We will assume, for light-duty vehicles other than gross
     liquid leakers, that when the ambient temperatures are above
     105°F that the resting loss emissions are the  same as those
     calculated at  105°F.

Since vehicles classified as gross liquid leakers were not handled
separately in MOBILES,  we will now make a new assumption
concerning those vehicles'  emission performance as  relates to
temperatures.  Specifically:

  3)  For the vehicles classified as gross liquid leakers,  we will
     assume the resting loss emissions are completely independent
     of temperature, averaging 8.84 grams per hour.

     The equations developed in this report to estimate hourly
diurnal emissions theoretically could also be applied to any

-------
                                -36-                           DRAFT

                                                         July 13,  1999
temperature cycle.  We will limit those functions by assuming that
the 24-hour diurnal emissions will be zero for any temperature
cycle in which the difference between the daily high and low
temperatures (i.e., the "diurnal temperature range")  is less than
five degrees Fahrenheit.

11.2  Heavy-Duty  Vehicles  (HDGVs)

     The analyses in this report were based only on RTD tests of
light-duty gasoline-powered vehicles (LDGVs)  and light-duty
gasoline-powered trucks (LDGTs).  Since the data did not indicate
a significant difference between the RTD emissions from LDGVs and
LDGTs, they were combined in a single group of analyses.

     Since no RTD testing was performed on any HDGVs, we will use
the same approach that was used in the earlier version of MOBILE.
That is, the ratio of diurnal emissions of the HDGVs to those of
the LDGTs is proportional to both the corresponding ratios of the
evaporative emission standards and the corresponding market shares
(under each of the emission standards).   Translating that sentence
into an equation yields:

          DlHDGV  =  DILDGT * [  ( 1.5 * 0.875 )  +  ( 2.0 * 0.125 )  ]
                  =  1.5625  *  DILDGT

                     Where,  D!HDGV is the  full  day's  diurnal
                            emissions from the HDGVs.

                            D!|_DGT is the  full  day's  diurnal
                            emissions from the corresponding
                            LDGTs.


We will use the same formula for resting losses (obviously
changing Dl  to  "hourly resting losses").

11.3  High Altitude  Evaporative  Emissions

     We will continue to use the multiplicative adjustment factor
of 1.30  (from previous version of  MOBILE)  to  adjust both the
resting loss and diurnal emissions for high altitude.

11.4  Motorcycles   (MC)

     RTD evaporative emission tests were not performed on
motorcycles (MC).   In MOBILES, the resting loss and diurnal
emissions from motorcycles were modeled using carbureted vehicles
equipped with open-bottom canisters.   That approach will continue
with MOBILE6.

     We first identified 109 RTD tests of carbureted vehicles
equipped with open-bottom canisters (all 1988 or earlier model
years),  and calculated both the hourly resting loss  (associated
with the test's low temperature) and the full-day's diurnal for
each of those 109 tests.  The diurnal emissions were then regressed

-------
                                -37-                           DRAFT

                                                          July 13, 1999
against both the vapor pressure product term  (developed in Section
9) and the age of each test vehicle.  As illustrated in Table 11-
1, each of those variables is statistically significant.  MOBILE6
will use the linear regression equation generated by that analysis
to calculate the full day's diurnal emissions.


                            Table  11-1

                Regression of  Diurnal  Emissions
                  (Simulated  Motorcycle  Fleet)
Dependent variable
No Selector
is:
R squared = 59.0% R squared (adjusted)
s = 10.20 with 109 - 3 = 106 degrees of
Source
Regression
Residual
Variable
Constant
age
VP_Product
Sum of Squares
15892.9
11024.5
Coefficient s.e.
-36.7971 4
0.855491 0
0.058251 0

= 58.3%
freedom
df
2
106
of Coeff
.5620
.1894
.0051


Mean Square
7946.46
104.005
t-rat i o
-8.07
4.52
1 1 .5
Diurnal

F-ratio
76.4
prob
< 0.0001
< 0.0001
< 0.0001
Translating that regression analysis into an equation yields:

        24-Hour Diurnal  Emissions (grams) of  Motorcycles

           =   -36.7971  +  (  0.855491  * Vehicle_Age  )

                +  ( 0.058251 * VP_Product_Term  )

EPA proposes to use this equation to estimate the 24-hour diurnal
emissions from motorcycles.

     Similarly, the hourly resting  loss emissions were regressed
against both the temperature at which those values were calculated
(i.e.,  the daily low temperature) and the age of each test
vehicle.  As illustrated in Table 11-2, only the vehicle age is
statistically significant.  It is possible that temperature was
not found to be statistically significant simply due to the fact
that most of the resting loss emissions were calculated at the
same temperature  (72 °F).  Since temperature should be an
important factor in determining resting loss emissions, EPA
proposes to use for MOBILE6 the linear regression equation
generated by the analysis (in Table 11-2) that uses both
variables.

-------
                                -38-
                            Table  11-2
    DRAFT

July  13,  1999
         Regression of  Hourly Resting  Loss Emissions
                  (Simulated  Motorcycle  Fleet)
Dependent variable is:
No Selector
Hourly Resting Loss
R squared = 5.6% R squared (adjusted) = 3.8%
s= 0.1346 with 109-3 = 106 degrees of freedom
Source
Regression
Residual
Variable
Constant
age
Daily_Low
Temp
Sum of Squares
0.114078
1.92123
Coefficient
0.044345
0.006134
0.000859
df
2
106
s.e. of Coeff
0.1572
0.0025
0.0022
Mean Square
0.057039
0.018125
t-rat i o
0.282
2.45
0.399
F-ratio
3.15
prob
0.7784
0.0159
0.6909
Translating that regression analysis into an equation yields:

       Hourly Resting Loss Emissions  (grams) of Motorcycles

           =   0.044345   +  ( 0.006134 * Vehicle_Age  )

                 +  ( 0.000859 * Daily_Low_Temperature  )

EPA proposes to use this equation to estimate the hourly resting
loss emissions from motorcycles.

11.5   Pre-Control Vehicles

     Non-California vehicles prior to the 1972 model year were not
required to meet an evaporative emission standard.  These
uncontrolled vehicles would simply vent vapors to the atmosphere
as pressure built up.   Since that situation is similar to that of
a controlled vehicle with a vapor leak,  we hypothesized that the
resting loss and diurnal evaporative emissions of the pre-1972
vehicles would be comparable to the emissions of the pre-1980
vehicles that had failed the pressure test.

     To characterize the hourly resting loss emissions from these
pre-control vehicles,  we proceeded in a similar fashion to the
approach in Section 8.  We first identified the two pre-1980
vehicles in our study that both had failed the pressure test and
were tested over the full range of fuels and temperature cycles.
Possibly due to that small sample size,  a regression of those data
produced a slope of resting loss versus temperature that was not
statistically different from zero.  We,  therefore, decided to use
the same slope (0.002812) that was developed in Section 8.  Since

-------
                                -39-                           DRAFT

                                                          July 13,  1999
most of the RTD tests  (i.e., 37 of 47) that were performed on the
34 candidate vehicles were run over the same temperature cycle
(i.e., 72 to 96 degrees), the variable "temperature" would not
make a useful independent variable to analyze those 47 resting
loss results.  However, the variable  "age" was found to be
statistically significant.  Combining the results of regressing
the data against age with the previously calculated temperature
slope yields the following equation:

    Hourly Resting Loss (grams) =  -0.768438
                                +   ( 0.002812  *  Temperature in °F )
                                +   ( 0.040528  *  Vehicle  Age  in Years )

EPA proposes to use this equation to estimate the hourly resting
loss emissions from pre-control vehicles with the restriction that
the calculated value must be at least the estimated hourly resting
loss of the  (newer) 1972-79 model year vehicles  (as calculated in
Appendix D).

     To characterize the full day's diurnal emissions from these
pre-control vehicles, we proceeded in a similar fashion to the
approach in Section 9.   In the preceding paragraph we noted that
only two of the candidate vehicles  (i.e., pre-1980 vehicles that
failed the pressure test) were tested over the full range of fuels
and temperature cycles.  Attempting to analyze the resting loss
emissions of those two vehicles as a function of temperature
produced only mediocre results.  However, the corresponding
analysis for diurnal emissions as a function of the vapor pressure
product term produced satisfactory results, as shown in Table 11-3:


                            Table  11-3

                Regression of  Diurnal   Emissions
                 (Simulated  Pre-Control  Fleet)
                     (Based on  Two  Vehicles)
Dependent variable
No Selector
R squared = 92.3%
s = 5.503 with 6
Source
Regression
Residual
Variable
Constant
VP_Product
is:


Diurnal
R squared (adjusted) = 90.4%
-2 = 4 degrees of freedom
Sum of Squares
1456.41
121.136
Coefficient s.e.
-6.52265 6.
0.05115 0.
df
1
4
of Coeff
175
0074
Mean Square
1456.41
30.284
t-ratio
-1 .06
6.93
F-ratio
48.1
prob
0.3504
0.0023

-------
                                -40-                           DRAFT

                                                          July 13,  1999
     Similar to the statements in the preceding material  on the
resting loss emissions from these test vehicles, the diurnal
emissions from these tests are almost exclusively from tests
performed over the 72 to 96 degree cycle using a single fuel RVP.
Thus, using a variable for vapor pressure for the full set of 47
tests would not be productive.  However, as with the resting loss
emissions, we used the preceding coefficient  (0.05115) to estimate
diurnal emissions (based on the vapor pressures) and then regress
the calculated residuals against vehicle age.  That regression
analysis yields the following equation:

    24-Hour  Diurnal  (grams)  =  -40.67512
                             +  ( 0.05115  *  VP Product Term )
                             +  (1.41114  *   Vehicle Age in Years  )


EPA proposes to use this equation to estimate the 24-hour diurnal
emissions from pre-control vehicles with the restriction  that the
calculated value must be at least the estimated full-day's diurnal
of the  (newer)  1972-79 model year vehicles  (as calculated in
Appendix E).

11.6  Duration  of  Diurnal  Soak  Period

     The analyses in this report were based on diurnals of exactly
24 hours in length.   In the real-world, the soak period could run
for longer or shorter periods of time.

     Estimating diurnal emissions when the  soak period is less
than 24 hours are analyzed in report number M6.EVP.002 (entitled
"Modeling Hourly Diurnal Emissions and Interrupted Diurnal
Emissions Based on Real-Time Diurnal Data").

     Estimating diurnal emissions when the  soak period is more
than 24 hours are analyzed in report number M6.EVP.003 (entitled
"Evaluating Multiday Diurnal Evaporative Emissions Using RTD
Tests").

11.7  1996  and  Newer  Model  Year  Vehicles

     Starting with the 1996 model year, EPA began certifying some
of the new LDGVs and LDGTs using the RTD test.  Estimating the
resting loss and diurnal emissions from these vehicles will be
analyzed in report number M6.EVP.005.

-------
                               -41-
    DRAFT

July  13, 1999
                           Appendix A

                     Temperature  Cycles   (°F)
Hour
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
— Temperati
60°-84°F
60.0
60.5
63.5
68.3
73.2
77.4
81.1
83.1
83.8
84.0
83.5
82.1
79.7
76.6
73.5
70.8
68.9
67.0
65.2
63.8
62.7
61.9
61.3
60.6
60.0
jres Cycling
72°-96°F*
72.0
72.5
75.5
80.3
85.2
89.4
93.1
95.1
95.8
96.0
95.5
94.1
91.7
88.6
85.5
82.8
80.9
79.0
77.2
75.8
74.7
73.9
73.3
72.6
72.0
Between —
82°-1 06°F
82.0
82.5
85.5
90.3
95.2
99.4
103.1
105.1
105.8
106.0
105.5
104.1
101.7
98.6
95.5
92.8
90.9
89.0
87.2
85.8
84.7
83.9
83.3
82.6
82.0
Change in
Temperature
...
0.5
3.0
4.8
4.9
4.2
3.7
2.0
0.7
0.2
-0.5
-1 .4
-2.4
-3.1
-3.1
-2.7
-1 .9
-1 .9
-1 .8
-1 .4
-1.1
-0.8
-0.6
-0.7
-0.6
  *  The temperature versus time values for the 72-to-96 cycle are
    reproduced from Table 1 of Appendix II of 40CFR86.


These three temperature  cycles are parallel  (i.e., identical
hourly increases/decreases).  The temperatures peak at hour nine.
The most rapid increase  in temperatures occurs during the third
and fourth hours.
For cycles in excess  of  24 hours, the pattern is repeated.

-------
                                -42-
                                           DRAFT

                                       July 13,  1999
                            Appendix B
                          Vapor  Pressure
Using  the  Clausius-Clapeyron  Relationship

     The Clausius-Clapeyron relationship  is  a  reasonable  estimate
of vapor pressure over the moderate  temperature  range (i.e.,  60°
to 106°F)* being considered for adjusting the  diurnal emissions.
This relationship assumes that the logarithm of  the vapor pressure
is a linear function of the reciprocal  (absolute) temperature.

     In a previous EPA work assignment, similar  RVP fuels were
tested, and their vapor pressures  (in kilo Pascals) at three
temperatures were measured.  The results  of  those tests are given
in the following table:
Nominal
RVP
7.0
9.0
Measured
RVP
7.1
8.7
Vapor Pressure (kPa)
75° F
30.7
38.2
100° F**
49.3
60.1
130° F
80.3
96.5
     ** The VPs at  100°  F  are  the fuels'  RVPs (in kilo Pascals).

Plotting these six vapor pressures  (using a logarithm scale  for
the vapor pressure) yields the graph  (Figure B-l) on the  following
page.

For each of those two RVP fuels,  the Clausius-Clapeyron
relationship estimates that, for  temperature in degrees Kelvin,
the vapor pressure  (VP) in kPa will be:

     Ln(VP) = A +  (B / Absolute Temperature), where:

                               A          B
RVP
8.7
7.1
                            13.5791
                            13.7338
-2950.47
-3060.95
   C. Lindhjem and D. Korotney,  "Running Loss Emissions from Gasoline-Fueled
   Motor Vehicles",  SAE Paper 931991,  1993.

-------
                                -43-
                            Ficrure  B -1
                                         DRAFT

                                    July 13,  1999
         Comparison  of Vapor  Pressure  to  Temperature
         100
       re
       a.
       2!
       3
       (A
       (A
       0)
       O
       Q.
       re
          1 0
           0.0030
0.0031
0.0032
0.0033
0.0034
                       Reciprocal of Temp  (1/°K)
We will assume that the specific fuels used  in  the  vehicles that
were tested in this analysis had vapor pressure versus  temperature
curves similar to the curves for these to  two test  fuels.
Extrapolating the trends in either the "A" or "B" values  to fuels
with nominal RVPs of 6.3, 7.0, and 9.0 psi;  and then requiring the
lines  (in log-space) to pass through  the appropriate pressures at
100°F, yields the linear equations with  coefficients:
                   RVP
                    6.3
                    6.8
                    9.0
       13.810
       13.773
       13.554
                     B
      -3121.05
      -3085.79
      -2930.67
We will use the above to estimate vapor pressures  for the  6.3,
6.8, and 9.0 psi RVP fuels.

In general, given the fuel RVP, we  can approximate A and B with
these equations:
     B = -3565.2707   +   (  70.5114  *  RVP )
   and
     A = Ln( 6.89286  *  RVP  )  -  (  B / 310.9

-------
               -44-
    DRAFT

July 13, 1999
           Appendix C
Mean Evaporative Emissions by  Strata
    By Vapor Pressure Products
Strata
Pre-1980 Carburete
Fail Purge/
Fail Pressu
Pre-1980 Carburete
Fail Purge/
Pass Pressure



Pre-1980 Carburete
Pass Purge/
Fail Pressu





Pre-1980 Carburete
Pass Purge/
Pass Pressui



1980-85 Carbureted
Fail Purge/
Fail Pressu
1980-85 Carbureted
Fail Purge/
Pass Pressui





1980-85 Carbureted
Pass Purge/
Fail Pressu





Fuel
RVP
a e.s
:e
3 6.8
6 .8
9 .0
6 .8
9 .0
9 .0
3 6.8
6 .3
re 6 .8
9 .0
6 .3
6 .8
9 .0
9 .0
3 6.8
6 .8
e 9.0
6 .8
9 .0
9 .0
6 .8
re
6 .8
6 .3
e 6.8
9 .0
6 .3
6 .8
9 .0
9 .0
6 .8
6 .3
re 6 .8
9 .0
6 .3
6 .8
9 .0
9 .0
Temp.
Cycle
72 .TO. 96
60 .TO. 8 4
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
72 .TO. 96
82 .TO. 106
60 .TO. 8 4
72 .TO. 96
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
82 .TO. 106
72 .TO. 96
82 .TO. 106
60 .TO. 8 4
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
72 .TO. 96
82 .TO. 106
72 .TO. 96
60 .TO. 8 4
72 .TO. 96
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
82 .TO. 106
72 .TO. 96
82 .TO. 106
60 .TO. 8 4
72 .TO. 96
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
82 .TO. 106
72 .TO. 96
82 .TO. 106
VP
times
?VP
567 . 02
374 .77
567 . 02
655 . 07
789 .30
968 .66
1323 .87
374 .77
489 .32
567 . 02
655 . 07
683 . 98
789 .30
968 .66
1323 .87
374 .77
567 . 02
655 . 07
789 .30
968 .66
1323 .87
567 . 02
374 .77
489 .32
567 . 02
655 . 07
683 . 98
789 .30
968 .66
1323 .87
374 .77
489 .32
567 . 02
655 . 07
683 . 98
789 .30
968 .66
1323 .87
Count
13
1
7
1
1
1
1
2
1
20
3
1
2
3
2
1
11
1
1
1
1
1
3
1
11
4
1
3
4
3
2
1
8
3
1
2
3
2
Mean
Diurna!
25 . Ill
16 .229
21 . 055
17 .511
36 .321
44 .222
76 .801
21 .284
17 .426
24 .385
21 .572
24 .328
42 .799
35 .331
72 .263
7 .861
13 .240
17 .423
32 .292
38 .297
100 . 094
27 .401
8 .834
16 .541
17 .756
16 .823
14 . 962
19 .669
25 .415
55 .324
13 .383
20 .741
16 .508
27 .768
43 .384
31 . 965
45 .319
53 .615
Mean
Hourly
Resting
Loss
0 .452
0 .250
0 .307
0 .218
0 .204
0 .250
0 .259
0 .238
0 . 140
0 .227
0 . 103
0 . 175
0 . 174
0 . 107
0 .274
0 . 167
0 .263
0 .239
0 .293
0 .204
0 . 062
0 .265
0 . 124
0 . 185
0 . 163
0 . 172
0 . 146
0 . 169
0 . 163
0 . 162
0 . 121
0 .253
0 . 139
0 . 127
0 .444
0 .216
0 .276
0 .308

-------
               -45-
DRAFT
                                      July 13,
Mean  Evaporative Emissions by  Strata
By Vapor  Pressure Products   (continued)
    1999
Strata
1980-85 Carbureted
Pass Purge/
Pass Pressui





1986+ Carbureted
Fail Purge/
Fail Pressu
1986+ Carbureted
Fail Purge/
Pass Pressui

1986+ Carbureted
Pass Purge/
Fail Pressu

1986+ Carbureted
Pass Purge/
Pass Pressui

1980-85 Fuel Injei
Fail Purge/
Fail Pressu
1980-85 Fuel Injei
Fail Purge/
Pass Pressui



1980-85 Fuel Injei
Pass Purge/
Fail Pressu



1980-85 Fuel Injei
Pass Purge/
Pass Pressui



Fuel
RVP
6 .8
6 .3
e 6.8
9 . 0
6 .3
6 .8
9 . 0
9 . 0
N/A
re
6 . 8
9 . 0
e 6.8
9 . 0
6 . 8
9 . 0
re 6 .8
9 . 0
6 . 8
9 . 0
e 6.8
9 . 0
:teffl/A
re
:teS. 8
6 .8
e 9.0
6 .8
9 . 0
9 . 0
:teS. 8
6 .8
re 9 . 0
6 .8
9 . 0
9 . 0
:teS. 8
6 .8
e 9.0
6 .8
9 . 0
9 . 0
Temp.
60 .TO. 8 4
72 .TO. 96
72 .TO. 96
60 .TO. 8 4
82 .TO . 106
82 .TO. 106
72 .TO. 96
82 .TO. 106
N/A
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
72 .TO. 96
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
72 .TO. 96
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
72 .TO. 96
N/A
60 .TO. 8 4
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
72 .TO. 96
82 .TO. 106
60 .TO. 8 4
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
72 .TO. 96
82 .TO. 106
60 .TO. 8 4
72 .TO. 96
60 .TO. 8 4
82 .TO. 106
72 .TO. 96
82 .TO. 106
VP
times
?VP
374.77
489.32
567.02
655 . 07
683 .98
789.30
968 .66
1323 .87
N/A
567.02
655 . 07
789.30
968 .66
567.02
655 . 07
789.30
968 .66
567.02
655 . 07
789.30
968 .66
N/A
374.77
567.02
655 . 07
789.30
968 .66
1323 .87
374.77
567.02
655 . 07
789.30
968 .66
1323 .87
374.77
567.02
655 . 07
789.30
968 .66
1323 .87

3
3
38
7
3
4
7
3
0
1
1
1
1
2
2
2
2
10
1
1
1
0
3
3
4
3
4
4
2
3
2
2
2
2
1
4
2
2
2
1
Mean
Diurna!
5.302
16 .308
9.081
11.352
22 . 047
14 .999
21.089
43 .900
N/A
10.230
12 .840
25.720
17 .670
15.865
21 .765
21 .480
26 .265
9 .481
6.440
8 .630
8 . 140
N/A
4 .329
7.910
6 .556
10 . 744
11.506
26 .730
19 .624
19 .482
25.861
39 .424
39.065
50.255
12 .943
8 .541
7 .845
11.861
13 .330
25.503
Mean
Hourly
Resting
Loss
0.065
0.195
0 . 107
0.147
0 . 170
0.169
0.194
0.274
N/A
0.100
0.097
0.155
0.148
0.233
0.342
0 . 124
0.308
0.138
0.092
0.102
0 . 075
N/A
0.010
0.011
0 . 045
0.041
0.086
0.123
0.198
0.206
0.184
0.300
0.231
0.252
0.296
0.080
0.157
0.218
0 .227
0.348

-------
               -46-                       DRAFT
                                      July 13,
Mean  Evaporative Emissions by  Strata
By Vapor  Pressure Products   (continued)
1999
1986+ Fuel Injects
Fail Purge/
Fail Pressu
1986+ Fuel Injectf
Fail Purge/
Pass Pressui
1986+ Fuel Injectf
Pass Purge/
Fail Pressu
1986+ Fuel Injectf
Pass Purge/
Pass Pressui
Fuel
RVP
id N/A
:e
id 6 .3
6 .8
e 6.3
6 .8
9 .0
6 .3
6 .8
9 .0
9 .0
id 6 .3
6 .8
:e 6 .3
6 .8
9 .0
6 .3
6 .8
9 .0
9 .0
id 6 .3
6 .8
e 6.3
6 .8
9 .0
6 .3
6 .8
9 .0
9 .0
Temp.
N/A
60 .TO. 8 4
60 .TO. 8 4
72 .TO . 96
72 .TO . 96
60 .TO. 8 4
82 .TO . 106
82 .TO . 106
72 .TO . 96
82 .TO . 106
60 .TO. 8 4
60 .TO. 8 4
72 .TO . 96
72 .TO . 96
60 .TO. 8 4
82 .TO . 106
82 .TO . 106
72 .TO . 96
82 .TO . 106
60 .TO. 8 4
60 .TO. 8 4
72 .TO . 96
72 .TO . 96
60 .TO. 8 4
82 .TO . 106
82 .TO . 106
72 .TO . 96
82 .TO . 106
VP
times
?VP
N/A
321.73
374.77
489.32
567.02
655 . 07
683 .98
789.30
968 .66
1323 .87
321.73
374.77
489.32
567.02
655 . 07
683 .98
789.30
968 .66
1323 .87
321.73
374.77
489.32
567.02
655 . 07
683 .98
789.30
968 .66
1323 .87
0
3
12
5
18
17
5
15
17
12
1
12
4
19
19
4
16
19
12
2
16
6
69
31
6
24
31
21
Mean
Diurna!
N/A
3 .002
5 .413
6 .027
9.083
7 .802
11.068
14.498
11 . 734
23 .895
5.206
6 .600
10.259
9 .202
8 .611
14 . 842
15 .824
16 .193
32 .116
0.602
1 .611
2 .345
7.166
2 .398
3 .576
5 .487
4.426
13 .640
Mean
Hourly
Resting
Loss
N/A
-0.009
0.011
0.024
0.060
0.034
0.064
0.073
0.056
0.087
0.037
0.042
0.038
0.094
0.053
0.088
0.110
0 . 114
0.129
-0.001
0 . 027
0.032
0.062
0.034
0.049
0.073
0.064
0.123

-------
                              -47-
    DRAFT

July 13,  1999
                           Appendix  D
             Modeling  Hourly  Resting Loss Emissions
                As  Functions  of Temperature   (°F)


 In  each  of the following  12  strata,  resting loss emissions  (j
 per hour)  are modeled  using  a pair of  numbers  (A  and B),  where

  Hourly Resting  Loss (grams) =   A +  ( B * Temperature  in  °F )
          Where
                B =  0.002812  (for ALL  strata)  and

               "A" is given in the following table:
Fuel Delivery
Carbureted


Fuel Injected


Model Year
Range
Pre-1980
1980-1985
1986-1995
Pre-1980*
1980-1985
1986-1995
Pass Pressure
Test
0.05530
-0.05957
-0.07551
0.05530
-0.09867
-0.14067
Fail Pressure
Test
0.07454
-0.02163
0.05044
0.07454
0.02565
-0.10924
    * The untested stratum  (Pre-1980 FI vehicles)  was  represented
      using the Pre-1980 model year carbureted vehicles.

If we use any  temperature profile in which the hourly change  in
temperature  is proportional to the cycles in Appendix A,  we find:

24-Hour Resting  Loss (grams)  =  ( 24 * Hourly_Resting_Loss_at_Low_Temp)

                 + (  0.03193 * Diurnal_Temperature_Range )

Where B  is  given above,  and where the Diurnal_Temperature_Range is
the difference of the daily high temperature minus the daily  low
temperature.

-------
                           -48-
    DRAFT

July 13,  1999
                       Appendix E
             Modeling  24-Hour Diurnal  Emissions
      As  Functions of Vapor Pressure   (kPa)and RVP  (psi)


In each  of the  following 18  strata,  24-hour  diurnal emissions
modeled  using four  constants:

                 A ,
                 B,
                 C, and
                 D.   Where,

 24-Hour Diurnal (grams)  =
              =  A
              + B * RVP (in psi)
              + C * [(Mean VP) * (Change in VP)]
              + D * [(Mean VP)  * (Change in  VP)]2  / 1,000

For each  of the 18  strata,  the  four constants  used to  model
emissions are given  below in  the following table:

Fuel Delivery
Carbureted










Model Year
Ran ge
1972-79*



1980-1985



1986-
1995**


Fail
Pressure
Test
-0.29374
-0.62160
0.039905
0
-1.22213
-0.62160
0.039905
0
18.97709
-1.81237
0
0.017098

Fail Only
Purge Test
21.94883
-2.23907
0
0.02990
16.69934
-2.23907
0
0.02990
13.90647
-2.14898
0.021368
0
Pass Both
Purge and
Pressure
21.13354
-2.42617
0
0.024053
15.50536
-2.42617
0
0.024053
8.37118
-0.767027
0
0.005934
     The B, C,  and D values are based on 1980-85 carbureted
     vehicles.
     The B, C,  and D values are based on 1986-95 FI  vehicles.

-------
                             -49-
    DRAFT
July 13,  1999
                     Appendix E   (continued)
              Modeling  24-Hour  Diurnal  Emissions
              As Functions of  Vapor Pressure   (kPa)

 In  each of  the following  18 strata,  24-hour  diurnal  emissions
 modeled using four  constants:

                   A , B,  C,  D.    Where,

24-Hour Diurnal (grams)   =
                = A
                + B * RVP (in  psi)
                + C * [(Mean VP) * (Change in  VP)]
                + D * [(Mean  VP) * (Change in  VP)]2 / 1,000

Fuel Delivery
Fuel Injected










Model Year
Range
1972-79*



1980-1985



1986-1995



Fail
Pressure
Test
-0.29374
-0.62160
0.039905
0
7.11253
-1.25128
0.036373
0
14.19286
-1.81237
0
0.017098

Fail Only
Purge Test
21.94883
-2.23907
0
0.02990
7.48130
-0.701002
0
0.010466
9.93656
-2.14898
0.021368
0
Pass Both
Purge and
Pressure
21.13354
-2.42617
0
0.024053
5.6211 1
-0.701002
0
0.010466
5.85926
-0.767027
0
0.005934
    * The  three untested strata of Pre-1980 FI vehicles were
      represented using the Pre-1980 model year carbureted
      vehicles (which were themselves based on the 1980-85 model
      year carbureted vehicles).

-------
                        -50-
    DRAFT

July 13,  1999
                    Appendix  F
Regression Analyses of  24-Hour  Diurnal  versus  Fuel RVP
           and Vapor  Pressure  Product  Term

        Regression of Mean  Diurnal Emissions
       Based  on Three 1980-85 Carb Vehicles
       Passing Both  Purge and Pressure Tests
Dependent variable
No Selector
R squared = 97.1%
s = 2.754 with 6
Source
Regression
Residual
Variable
Constant
VP Product
Sqrd / 1,000
Fuel RVP
is:


Diurnal
R squared (adjusted) = 95.2%
-3 = 3 degrees of freedom
Sum of Squares
765.294
22.76
Coefficient s.e.
14.3895 9
0.024053 0
-2.42617 1
df
2
3
of Coeff
.439
0027
.326
Mean Square
382.647
7.58666
t-ratio
1.52
8.78
-1 .83
F-ratio
50.4
prob
0.2248
0.0031
0.1648
        Regression of Mean  Diurnal Emissions
        Based on  Two  1980-85 Carb Vehicles
             Failing the  Pressure  Test
Dependent variable is:
No Selector
R squared = 99
s = 1.307 with
Source
Regression
Residual
Variable
Constant
VP_Product
Term
Fuel RVP
Diurnal
4% R squared (adjusted) = 99.0%
6-3 = 3 degrees of freedom
Sum of Squares
822.877
5.12862
Coefficient
-1 .00903
0.039905

-0.621600
df
2
3
s.e. of Coeff
4.18
0.0023

0.650
Mean Square
41 1.438
1.70954
t-ratio
-0.241
17.0

-0.956
F-ratio
241
prob
0.8250
0.0004

0.4096

-------
                 -51-
         Appendix  F    (continued)
    DRAFT

July 13,  1999
Regression  of  Mean  Diurnal Emissions
Based  on Three  1980-85  Carb Vehicles
     Failing ONLY  the  Purge Test
Dependent variable
No Selector
R squared = 94.7%
s = 4.853 with 6
Source
Regression
Residual
Variable
Constant
VP Product
Sqrd / 1,000
Fuel RVP
is:


Diurnal
R squared (adjusted) = 91.1%
-3 = 3 degrees of freedom
Sum of Squares
1256.90
70.6578
Coefficient s.e.
15.3041 16
0.029900 0
-2.23907 2
df
2
3
of Coeff
.6300
0048
3370
Mean Square
628.449
23.5526
t-ratio
0.920
6.19
-0.958
F-ratio
26.7
prob
0.4253
0.0085
0.4087
Regression  of  Mean  Diurnal Emissions
  Based  on  Four  1980-85  Fl Vehicles
      Passing  the  Pressure Test
Dependent variable is:
No Selector
R squared =
s = 0.4728
Source
Regression
Residual
Variable
Constant
Diurnal
99.6% R squared (adjusted) = 99.3%
with 6-3 = 3 degrees of freedom
Sum of Squares
156.976
0.670742
Coefficient s.e.
7.29846 1
VP Product 0.010466 0
Sqrd /
Fuel RVP
1,000
-0.701002 0
df
2
3
of Coeff
.620
0005
2277
Mean Square
78.4882
0.223581
t-ratio
4.50
22.2
-3.08
F-ratio
351
prob
0.0204
0.0002
0.0542

-------
                 -52-
         Appendix  F   (continued)
    DRAFT

July 13, 1999
 Regression  of Mean  Diurnal  Emissions
  Based  on  Two  1980-85  Fl  Vehicles
       Failing the Pressure  Test
Dependent variable is:
No Selector
R squared = 94
s = 3.511 with
Source
Regression
Residual
Variable
Constant
VP_Product
Term
Fuel RVP
Diurnal
4% R squared (adjusted) = 90.7%
6-3 = 3 degrees of freedom
Sum of Squares
626.019
36.9725
Coefficient s.e.
7.82649 1
0.036373 0

-1.25128 1
df
2
3
of Coeff
1.23
0063

.746
Mean Square
313.009
12.3242
t-ratio
0.697
5.77

-0.717
F-ratio
25.4
prob
0.5361
0.0104

0.5253
 Regression  of Mean  Diurnal  Emissions
  Based on  16  1986-95  Fl  Vehicles
Passing Both  Purge and  Pressure  Tests
Dependent variable is:
No Selector
Diurnal
R squared = 97.1% R squared (adjusted) = 95.2%
s = 0.6560 with 6-3 = 3 degrees of freedom
Source
Regression
Residual
Variable
Constant
VP Product
Sqrd / 1,000
Fuel RVP
Sum of Squares
43.7687
1.29117
Coefficient s.e.
4.70657 2
0.005934 0
-0.767027 0
df
2
3
of Coeff
.248
0007
3159
Mean Square
21.8844
0.43039
t-ratio
2.09
9.09
-2.43
F-ratio
50.8
prob
0.1273
0.0028
0.0935

-------
                 -53-
        Appendix F   (continued)
    DRAFT

July 13, 1999
Regression of Mean Diurnal Emissions
  Based  on 11   1986-95  Fl Vehicles
      Failing  the  Pressure  Test
Dependent variable
No Selector
R squared = 98.9%
s = 1.206 with 6
Source
Regression
Residual
Variable
Constant
VP Product
Sqrd / 1,000
Fuel RVP
is:


Diurnal
R squared (adjusted) = 98.1%
-3 = 3 degrees of freedom
Sum of Squares
382.227
4.36316
Coefficient
14.5718
0.017098
-1 .81237
df
2
3
s.e. of Coeff
4.1330
0.0012
0.5807
Mean Square
191.113
1.45439
t-ratio
3.53
14.2
-3.12
F-ratio
131
prob
0.0388
0.0007
0.0524
Regression of Mean Diurnal Emissions
  Based  on 12   1986-95  Fl Vehicles
     Failing ONLY the Purge  Test
Dependent variable is:
No Selector
R squared = 95
s = 1.578 with
Source
Regression
Residual
Variable
Constant
VP Product
Fuel RVP
Diurnal
7% R squared (adjusted) = 92.8%
6-3 = 3 degrees of freedom
Sum of Squares
164.793
7.47312
Coefficient s.e.
11.0427 5
0.021368 0
-2.14898 0
df
2
3
of Coeff
.050
0028
7849
Mean Square
82.3963
2.49104
t-ratio
2.19
7.54
-2.74
F-ratio
33.1
prob
0.1166
0.0048
0.0715

-------