United States Air and Radiation EPA420-R-01-026
Environmental Protection April 2001
Agency M6.EVP004
vvEPA Update of Hot
Soak Emissions
> Printed on Recycled Paper
-------
EPA420-R-01-026
April 2001
of Hot
M6.EVP.004
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
Preapred for EPA by:
ARCADIS Geraghty & Miller, Inc.
EPA Contract No. 68-C6-0068
Work Assignment No. 1-01
NOTICE
This technical, report does not necessarily represent final EPA decisions or positions.
It is intended to present technical analysis of issues using data which are currently available.
The purpose in the release of such reports is to facilitate the exchange of
technical information and to inform the public of technical developments which
may form the basis for a final EPA decision, position, or regulatory action.
-------
I. INTRODUCTION
The U.S. Environmental Protection Agency's (EPA) highway emission factor model,
currently MOBILESa, calculates in-use emission factors for exhaust and evaporative emissions using
national average values as supplemented by user-supplied input (e.g., temperature, fuel volatility,
etc.). EPA is currently working to develop a new version of the model (MOBILE6) to further
improve its accuracy and include more "real world" data.
Evaporative "hot soak" (trip-end) emissions represent one area where data now exist to better
characterize conditions observed during "real world" driving conditions. A hot soak is defined as
the evaporative losses produced as fuel evaporates from the carburetor and fuel tank in carbureted
vehicles, or from the fuel tank in fuel inj ected vehicles, as a result of heating of the fuel tank and fuel
system above ambient temperatures. Average temperatures that occur during a hot soak event are
shown in Figure 1. As can be seen from this figure, fuel system temperatures greatly exceed
ambient temperatures during a hot soak event.
Hot soak emissions generally occur during the one-hour period1 after the engine is shut down
and are measured in a sealed housing for evaporative emission determination (SHED). Results from
SHED tests are in grams per one-hour test (g/test). Level of emissions during a hot soak is a function
of fuel volatility (Reid Vapor Pressure [RVP]) and ambient temperature, as well as other variables.
In previous versions of the MOBILE model, hot soak emissions were characterized using
data derived from laboratory testing of light-duty vehicles and trucks. Testing was conducted under
EPA-derived fuel RVP and temperature criteria. These criteria stem from the EPA certification test
procedure, and specifically consisted of a certification test fuel with a fuel volatility level of 9.0 psi,
a fuel tank fill level of 40%, and an ambient temperature of approximately 82°F. Two additional
fuels with RVP levels of 10.4 and 11.7 psi were also used during testing performed in 1984 through
1989. In 1990, data from testing in Hammond, Indiana was also added to the emission factor
database. This test program involved the procurement of vehicles tested in Indiana Inspection and
Maintenance (I/M) program lanes, where vehicles were driven on an EVI240 transient test cycle, and
testing of the evaporative emissions control systems was performed to see if either failures existed
due to improper pressure and/or improper purging of vapors. This testing is known as
pressure/purge tests. Some vehicles that failed either test were also tested for their diurnal and hot
soak emissions, in an attempt to assess whether failure of pressure/purge testing could be correlated
to high diurnal and hot soak emissions.
The MOBILE model contains correction factors for the effects of RVP and temperature on
hot soak emissions that allow the user to adjust these conditions to correspond to local values. These
correction factors have been developed though statistical analysis of the EPA hot soak emissions
data.
1 The majority of hot soak emissions occur within 10 minutes after engine shut-off, but
are measured during a hot soak test for a 1 hour period.
1
-------
Since the development of the latest version of MOBILE, EPA has recognized the need to
incorporate additional hot soak data into its modeling efforts. The data used to generate hot soak
curve fits for MOBILESa did not incorporate low RVP fuels, which are now regulated in various
parts of the country. Furthermore, the data did not fully represent "real world" conditions, as "real
world" vehicles never have certification fuel in their fuel tanks and are operated under a wide range
of ambient temperature conditions.
22O
2OO
I I!
Air Under Hood
Main Fuel Jet
Carburetor Bowl
Fuel-Pump Inlet
Atmospheric Temperature
Approx. 6O"F
15 IO 5 O
Time Before Idle
5 IO 15 2O
Time After Shutdown to Idle, min.
Figure 1. Temperatures occurring during a hot soak event
To this end, several studies have recently been conducted characterizing hot soak emissions
at higher ambient temperatures and over a wider range of fuel RVPs than those contained in previous
EPA testing. The two most significant studies were conducted by the Auto/Oil Air Quality
Improvement Research Program (AQIRP) and by EPA, both under contract with the Automotive
Testing Laboratories (ATL). Both studies recruited vehicles from Arizona Inspection and
Maintenance (I/M) testing lanes, and the testing was performed under comparable conditions. Table
1 lists the testing conditions and average results of emissions testing for both studies. In addition,
two other EPA work assignments (Contract 68-C3-0006, Work Assignments 0-07 and 0-11) contain
hot soak testing on smaller numbers of vehicles. This report details an analysis of these "real world"
databases and develops correction factors for RVP and ambient temperature based upon this data.
-------
These databases encompass vehicles with "real world" gasolines in their fuel tanks, have a variety
of tank fill levels, are tested as received and at a variety of ambient temperatures.
Table 1. Comparison of testing criteria for "real world" hot soak studies
Testing dates
Number of test vehicles
Type of vehicles
Model years represented
Location of Testing
Daytime temperature range
Hot soak cut-point
As-received average HC
emissions across fleet
Range of emissions
No. of high emitters
Percent high-emitters
Auto/Oil
June 15, 1993 to
September 15, 1993
299
In-use LDVs and LDTs
1983 - 1993
Automotive Testing
Laboratory, Mesa, Arizona
82°Fto 112°F
2.0 grams/test
1.53 grams/test
0.04 to 49.39 grams/test
46 out of 299
15.3%
EPA
July?, 1995 to
September 29, 1995
181
In-use LDVs and LDTs
1981 - 1994
Automotive Testing
Laboratory, Mesa, Arizona
at least 80 °F
2.0 grams/test
1.76 grams/test
0.06 to 46.95 grams/test
28 out of 181
15.6%
The approach described below, developed under direction by EPA, attempts only to replace
the existing MOBILES estimates for adjustment of hot soak emissions for RVP levels below 9
pounds per square inch (psi) with new estimates based on the new data. The baseline estimates of
hot soak emissions and the effects of RVP values above 9 psi would be retained for MOBILE6.
II.
DATA ANALYSIS
There are several variables which directly affect hot soak emissions. Hot soak emissions in
vehicles (with newer evaporative emissions control systems) are usually due to small leaks in the
evaporative emission control system (j oints, lines, valves) and permeation of the fuel hoses and tank.
These fuel vapor leaks are generally driven by the heating of the fuel system above ambient
conditions during a hot soak event As seen in Figure 1, fuel system temperatures greatly exceed
-------
ambient temperatures during a hot soak event. Fuel tank temperature is usually close to ambient but
can increase in fuel injected vehicles due to fuel returning from the hot engine compartment.
Typically, tank temperatures in fuel injected vehicles can exceed ambient temperatures by 5 to 15°F.
Thus hot soak emissions are not a direct function of ambient temperature.
Data from the "real world" data set are characterized in Table 2. In addition to the categories
of the data used in MOBILESa, two new strata were added. The first is "gross liquid leaker." This
refers to vehicles which produce abnormally high evaporative emissions as a result of a fuel leak and
which have hot soak emissions of over 10 grams per test. The second is the addition of two new
model year groupings (1981-1985, and 1986 and newer) for vehicles that passed both the pressure
and purge tests. This stratification of model year groups was used to capture the significant
improvement of evaporative emissions systems in most automobiles that occurred beginning with
the 1986 model year. Other strata used in MOBILESa were continued, i.e., hot soak data from
vehicles that passed both the pressure and purge tests were stratified by fuel system type (carbureted
[Carb], throttle body fuel injected [TBI], and port fuel injected [PFI]) and by vehicle type
classification (passenger cars [LDV] and light-duty trucks [LDT]).
Hose permeation can also be a large source of hot soak emissions, particularly in fuel inj ected
vehicles. PFI systems typically run at pressures of 40-50 psi, while TBI systems run around 10 psi.
Permeation of fuel through elastomers in the fuel and evaporative control system can be very
temperature sensitive and can be a large source of hot soak emissions in newer vehicles. Injector
leaks in fuel injected systems can result in very high hot soak emissions (liquid leakers).
A further factor in real world hot soak emissions is the different molecular weight of the fuel
vapor. Because different fuels have different light ends and there is some weathering (loss of light
ends over time) of fuel components in fuel tanks, hot soak emissions can vary in molecular weight
by up to 50%. In the Auto/Oil Air Quality Improvement Research Program (AQIRP), over 50
current vehicles were tested on a variety of fuels. Hot soak emission molecular weight varied from
65.6 g/g mole to 92.3 g/g mole with an average molecular weight of 84.1 g/g mole. These large
fluctuations in molecular weight can significantly affect the mass of emissions emitted during a hot
soak event.
With these factors in mind, it was not surprising to find significant scatter in the real world
hot soak data. Furthermore, most of the data represented RVPs of 5 to 7 psi. Extrapolation of this
data past 9 psi is not recommended due to the narrow range of RVPs in the data set. Therefore curve
fits using real world data were only generated up to 9 psi RVP. For RVPs over 9 psi, previous
MOBILESa curves were used. This presented an additional challenge to the regression analyses,
making sure that the two curves met at 9 psi RVP at all temperatures. This required that the
functional form of the equation be identical to those presently used in MOBILESa and that the
temperature coefficient in those equations be the same. Thus regression analyses were performed
only on RVP using the real world data.
In some cases, the data produced a negative coefficient with regard to RVP (i.e. as RVP
-------
increased, predicted hot soak emissions decreased). As this is intuitively incorrect, additional data
points were generated using the MOBILESa curve fits at 9 psi RVP and added to the data sets that
produced a negative RVP coefficient until the resulting curve produced by regression analysis had
a positive RVP coefficient. Further discussion is found in each of the sections below.
The following sections describe the stratification of the hot soak data sets, the methodology
used to determined curve fits of that data, and a discussion of the results of the curve fits.
III. GROSS LIQUID LEAKERS
Liquid fuel leaks from fuel systems can result in very high hot soak emissions. This is
particularly true for fuel injection systems that operate at high pressure (40-45 psi). If an injector is
leaking due to damage or incorrect position, pressure built up in the fuel system will bleed off
through that injector. Liquid leaks can also exist in carbureted fuel systems as a result of leaking
carburetor gaskets or a defective fuel shut off at the carburetor bowl. The real world data set included
17 liquid leakers, 9 of which fell into the gross liquid leaker category. Gross liquid leakers were
defined as those vehicles with liquid fuel leaks that were measured at over 10 grams per test of hot
soak emissions. Since the set of liquid leaker data was so small, all that could be defined was an
average value for two different fuel systems, namely carbureted (Carb) and port fuel injected (PFI).
Carb vehicles had an average gross liquid leaker hot soak value of 14.60 g/test, while PFI vehicles
had an average liquid leaker hot soak value of 57.79 g/test. It is reasonable that fuel inj ected systems
would have much higher liquid leak emissions as they are usually under higher fuel pressure. While
there is no data on TBI liquid leakers in the data sets, Bernoulli's equation indicates that the leak
rate for TBI systems would be about one half that for PFI systems (the square root of the ratio of
operating pressures). Therefore, without further data, the author suggests assuming that TBI liquid
leakers might emit approximately half the emissions of PFI systems. These estimates are further
revised in the report, "Evaporative Emissions of Gross Liquid Leakers in MOBILE6,"
(M6.EVP.009).
IV. PRESSURE TEST FAIL VEHICLES
Of the 630 vehicles tested, 80 vehicles that did not fall into the gross liquid leaker category
failed the pressure test. Data within this strata had significant scatter and in several cases there was
not enough data to support further stratification by fuel system type, so all pressure fail data were
aggregated together similar to what was done for MOBILESa. The MOBILESa curve fit for
pressure failed vehicles was in the form of:
Hot Soak = exp(A*(RVP-9.0) +B*(Temp - 82) + C) (IV. 1)
Since the data was at various ambient temperatures, each hot soak test value was adjusted to 95 °F
using the MOBILESa temperature correction as shown below:
Adjusted Hot Soak = Hot Soak * exp(1.774+0.05114*(95-82))/exp(1.774+0.05114*(Temp-82)) (IV.2)
-------
In addition, to be consistent with MOBILESa, all fuel injected test data (TBI and PFI) were
divided by 0.88, the fuel system adjustment factor in MOBILESa (in MOBILESa, this factor is
multiplied by all fuel inj ected vehicle results to adjust for the difference between in-use and FTP fuel
tank levels). A regression analysis was run on the adjusted hot soak versus RVP data to determine
coefficient B in equation IV. 1 above. A t-statistic of 2.88 resulted for the coefficient with a P-value
of 0.0051. The coefficient C was determined so that the calculated hot soak results using the real
world curve fit matched the MOBILESa curve fit at 9 psi RVP at all temperatures. This still resulted
in a P-value of 0.0195 for coefficient C. The equation for all pressure fail vehicles for less than 9
psi RVP is:
Hot Soak = exp(0.413356*(RVP - 9.0) + 0.05114*(Temp - 82) + 1.774) (IV.3)
This may be compared to the MOBILESa equation for RVP less than 9.0 psi, which is:
Hot Soak = exp (0.4443*(RVP - 9.0) + 0.05114*(Temp - 82) +1.774) (IV.4)
For fuel injected vehicles, the fuel system adjustment factor of 0.88 should be multiplied by both
equations IV.3 and IV.4 to obtain hot soak emission results.
Predicted hot soak emissions calculated using equation IV.3 are shown in Table 3 for
pressure fail vehicles before application of the fuel system adjustment factor. MOBILESa estimates
calculated using equation IV.4 are also included for reference. Figure 2 shows the real world and
MOBILESa curve fits as well as the real world data for TBI vehicles. Figure 3 shows the real world
and MOBILESa curve fits as well as real world data for PFI vehicles. Figure 4 shows the real world
and MOBILESa curve fits as well as real world data for Carb vehicles. Figures 2 and 3 show both
real world and MOBILESa curve fits with the fuel system adjustment factor applied. While the
curve fits are the same in all three figures (except for application of the fuel system adjustment
factor), real world data were shown divided by fuel system type in Figures 2 through 4 so that the
reader could see how the real world data compared against the new and MOBILESa curve fits.
As seen in Table 3 and Figures 2, 3 and 4, the new pressure fail curve fits predict slightly
higher emissions in the 5 to 7 psi RVP range than the previous MOBILESa curve fits. This indicates
that real world pressure test fail data shows slightly higher levels of hot soak emissions than
previously estimated from the laboratory data used to generate the curve fits for MOBILESa.
V. PURGE TEST FAIL VEHICLES
Of the 630 vehicles tested, 47 vehicles that did not fall into the gross liquid leaker category
failed the purge test. Data within this strata had significant scatter and in several cases there were
not enough data to support further stratification by fuel system type, so all purge fail data were
aggregated together similar to what was done for MOBILESa. The MOBILESa curve fit for purge
failed vehicles was in the form of:
-------
Hot Soak = exp(A*(RVP-9.0) +B*(Temp - 82) + C) (V. 1)
Since the data was at various ambient temperatures, each hot soak test value was adjusted to 95 °F
using the MOBILESa temperature correction as shown below:
Adjusted Hot Soak = Hot Soak *exp(1.76223+0.05114*(95-82))/exp(1.76223+0.05114*(Temp-82)) (V.2)
In addition, to be consistent with MOBILESa, all fuel inj ected test data were divided by 0.88, the fuel
system adjustment factor in MOBILESa. A regression analysis was run on the adjusted hot soak
versus RVP data to determine coefficient B in equation V. 1 above. A t-statistic of 2.37 resulted for
the coefficient with a P-level of 0.0222. The coefficient C was determined so that the calculated hot
soak results using the real world curve fit matched the MOBILESa curve fit at 9 psi RVP at all
temperatures. This coefficient resulted in a P-value of 0.0261. The equation for all purge fail
vehicles for less than 9 psi RVP is:
Hot Soak = exp(0.552175*(RVP-9.0) +0.05114*(Temp-82)+ 1.76223) (V.3)
This may be compared to the MOBILESa equation for RVP less than 9.0 psi, which is:
Hot Soak = exp (0.4443*(RVP - 9.0) + 0.05114*(Temp - 82) +1.76223) (V.4)
For fuel injected vehicles, the fuel system adjustment factor of 0.88 should be multiplied by both
equations V.3 and V.4 to obtain hot soak emission results.
Predicted hot soak emissions calculated using equation V.3 are shown in Table 4 for purge
fail vehicles before application of the fuel system adjustment factor. MOBILESa estimates
calculated using equation V.4 are also included for reference.
Figure 5 shows the real world and MOBILESa curve fits as well as the real world data for
TBI vehicles. Figure 6 shows the real world and MOBILESa curve fits as well as real world data for
PFI vehicles. Figure 7 shows the real world and MOBILESa curve fits as well as real world data for
Carb vehicles. Figures 5 and 6 show both real world and MOBILESa curve fits with the fuel system
adjustment factor applied. Again, real world data have been stratified by fuel system type in Figures
5 through 7 for comparison purposes only.
As seen by Table 4 and Figures 5, 6 and 7, the new purge fail curve fits predict slightly lower
emissions in the 5 to 7 psi RVP range than the previous MOBILESa curve fits. This indicates that
real world purge test fail data shows slightly lower levels of hot soak emissions than previously
estimated from the laboratory data used to generate the curve fits for MOBILESa.
-------
VI. PASSING VEHICLES
Of the 630 vehicles tested, 494 vehicles which did not fall into the gross liquid leaker
category passed both the pressure and purge tests. The hot soak data set of vehicles that passed both
the pressure and purge tests allow some disaggregations, although some of the disaggregations did
not produce statistically significant curves due to large data scatter. In some cases, the data even
produced negative coefficients for RVP and additional data calculated from MOBILESa curve fits
at 9 psi RVP had to be added to produce reasonable trends over the RVP range (5 to 9 psi). The
functional form of current MOBILESa pass/pass vehicle equations and strata were used to perform
regression analyses of the real world data. An additional stratification was added to each set,
however. The data were divided into two model year groupings for each vehicle type. Since
manufacturers became more aware of the need to 'fine tune' evaporative emission systems during
the 1981 through 1985 model years, the data was stratified into two model year groupings, namely
1981-1985 and 1986+. Discussion of the methodology used and the results of the regression analysis
are contained within each subsection below.
A. TBI
Of the 494 vehicles that passed both the pressure and purge tests, 102 vehicles had TBI fuel
systems. A curve fit similar to that used in MOBILESa for TBI vehicles was used:
Hot Soak = (A + B*RVP)*(C + D*Temp2)/E (VI. 1)
Since the data were at various ambient temperatures, each hot soak test value was adjusted to 95 °F
using the temperature correction factor defined in MOBILESa curve fits for TBI pass/pass vehicles:
Adjusted Hot Soak = Hot Soak * (-2.4636+0.00056161*952))/(-2.4636+0.0005616l*Temp2) (VI.2)
In addition, to be consistent with MOBILESa, all fuel inj ected test data were divided by 0.88,
the fuel system adjustment factor used in MOBILESa. A regression analysis was run on the adjusted
hot soak versus RVP data to determine coefficients A and B in equation VI. 1 above for each strata.
Of the 102 TBI tests, 17 corresponded to LDVs with model years between 1981 and 1985,
56 were LDVs model years 1986+, and 29 were LDTs model years 1986+. There were no test data
for LDTs model years 1981-1985.
For the TBI LDVs with model years (MY) between 1981 and 1985, t-statistics of-0.44 and
0.87 and P-values of 0.666 and 0.396 resulted for coefficients A and B, respectively, indicating that
neither produced statistically significant curves due to significant data scatter. It did, however,
produce a reasonable trend with regard to RVP. For TBI LDVs with MY 1986+, t-statistics of -0.44
and 0.65 and P-values of 0.661 and 0.516 resulted for coefficients A and B, respectively, indicating
that neither produced statistically significant curves due to significant data scatter. It did, however,
also produce a reasonable trend with regard to RVP. For TBI LDTs with MY 1986+, t-statistics of
-------
-3.96 and 6.76 and P-values of 4.9E-04 and 3.0E-07 resulted for coefficients A and B, respectively,
indicating that these coefficients were statistically significant. Coefficients C and D were retained
from the MOBILESa curve fits for TBI pass/pass vehicles and coefficient E was determined so that
the calculated hot soak results using the real world curve fits matched the MOBILESa curve fits at
9 psi RVP at all temperatures.
The equations for the two model year groupings of LDVs are as follows:
1981-1985 MY LDVs
Hot Soak = (-0.52111 + 0.159322*RVP)*(-2.4636 + 0.00056161 *Temp2)/l.898 (VI.3)
1986+MY LDVs
Hot Soak = (-1.27508 + 0.28853*RVP)*(-2.4636 + 0.00056161 *Temp2)/2.748 (VI.4)
For comparison, the MOBILESa equation for MY 1981+ LDV TBI vehicles that pass both
the pressure and purge tests is:
Hot Soak = (0.258327 + 0.041297*RVP)*(-2.4636 + 0.00056161 *Temp2)/l.31 (VI.5)
As explained in Section IV, equations VI.3, VI.4, and VI.5 should be multiplied by the fuel system
adjustment factor of 0.88 to obtain hot soak emission results.
Predicted hot soak emissions for TBI LDVs calculated using equations VI.3 and VI.4 are
shown in Table 5 (with the fuel system adjustment factor applied) along with MOBILESa TBI LDV
estimates (calculated using equation VI.5). Plots of hot soak emissions at 95 °F are shown in Figure
8. Curves shown in Figure 8 also have the fuel system adjustment factor applied.
As seen in Table 5 and Figure 8, the new TBI LDV curve fits predict slightly lower emissions
in the 5 to 7 psi RVP range than the previous MOBILESa curve fits. While no conclusions can be
drawn from these curve fits (as they are not statistically significant), one might assume that the real
world vehicle set used to define these curve fits had lower hot soak emissions in the 5 to 7 psi RVP
range than that estimated from MOBILESa (which was produced from an extrapolation of higher
laboratory data). Furthermore, curve fits for MY 86+ vehicles showed lower hot soak emissions than
the MY 81-85 group, which is reasonable assuming an improvement in evaporative control system
design.
The equation for LDTs derived from the real world data is:
1986+MYLDTs
Hot Soak = (-0.71055 + 0.17803*RVP)*(-2.4636 + 0.00056161 *Temp2)/2.596 (VI.6)
The MOBILESa equation for MY 1981+ LDT TBI vehicles that pass both the pressure and
purge tests is:
-------
Hot Soak = (0.078327 + 0.041297*RVP)*(-2.4636 + 0.00056161 *Temp2)/l.31 (VI.7)
Equations VI.6 and VI.7 should be multiplied by the fuel system adjustment factor of 0.88 to obtain
hot soak emission results.
Predicted hot soak emissions for TBI LDTs calculated using equation VI.6 are shown in
Table 6 along with MOBILESa TBI LDT estimates (calculated using equation VI.7) with the fuel
system correction factor applied in both cases. Plots of hot soak emissions for TBI LDT pass/pass
vehicles at 95 °F are shown in Figure 9 with the fuel system correction factor applied to the curve
fits.
Predicted hot soak emissions using the real world curve fits are lower in the 5 to 7 psi RVP
range than predicted using previous MOBILESa curve fits. The real world data shows lower levels
of hot soak emissions in this region than previously extrapolated from the laboratory data at higher
RVPs used to generate the curve fits for MOBILESa.
B. PFI
Of the 494 vehicles that passed both the pressure and purge tests, 279 vehicles had PFI fuel
systems. A curve fit similar to that used in MOBILESa for PFI vehicles was used:
Hot Soak = (A + B*RVP)*(C*Temp)/D (VI.8)
Since the data was at various ambient temperatures, each hot soak test value was adjusted to 95 °F
using the temperature correction factor defined in MOBILESa curve fits for PFI pass/pass vehicles:
Adjusted Hot Soak = Hot Soak * 95/Temp (VI.9)
In addition, as explained in Section IV, to be consistent with MOBILESa, all fuel injected
test data were divided by 0.88, the fuel system adjustment factor in MOBILESa. A regression
analysis was run on the adjusted hot soak versus RVP data to determine coefficients A and B in
equation VI.8 above for each strata.
Of the 279 PFI tests, 15 corresponded to LDVs with model years between 1981 and 1985,
225 were LDVs model years 1986+, and 39 were LDTs model years 1986+. There were no test data
for LDTs model years 1981-1985.
For the PFI LDVs with model years (MY) between 1981 and 1985, the first regression
analysis resulted in a negative B coefficient, implying a decrease in emissions with increasing fuel
RVP, which is intuitively incorrect. To correct this situation, 15 additional data points calculated
using the MOBILESa curve fit at 9 psi RVP and 95 °F were added to the 15 real world data points.
This produced a positive B coefficient with a t-statistic of 2.02 with a P-value of 0.0528. Real world
data for PFI LDVs with MY 1986+ also produced a negative B coefficient. By adding 25
MOBILESa calculated data points at 9 psi RVP and 95 °F to the 223 real world data points, a positive
10
-------
B coefficient resulted with a t-statistic of 1.74 and a P-value of 0.084. For PFI LDTs with MY
1986+, a positive B coefficient was achieved without addition of MOBILESa points, but the t-
statistic and P-value were only 0.29 and 0.77, respectively, indicating that it was not statistically
significant. Coefficient C was retained from the MOBILESa curve fits for PFI pass/pass vehicles
and coefficient D was determined so that the calculated hot soak results using the real world curve
fits matched the MOBILESa curve fits at 9 psi RVP at all temperatures.
The equations for the two model year groupings of LDVs are as follows:
1981-1985 MY LDVs
Hot Soak = (-0.058967 + 0.100658*RVP)*(0.0055541*Temp)/0.749 (VI. 10)
1986+MY LDVs
Hot Soak = (-0.0097563+0.082809*RVP)*(0.0055541*Temp)/0.651 (VI. 11)
For comparison purposes, the MOBILESa equation for MY 1981+ LDV PFI vehicles that
pass both the pressure and purge tests is:
Hot Soak = (-0.40673 + 0.10297*RVP)*(0.0055541*Temp)/0.46 (VI. 12)
Equations VI. 10, VI. 11, and VI. 12 should be multiplied by the fuel system adjustment factor of 0.88
to obtain hot soak emission results.
Predicted hot soak emissions for PFI LDVs calculated using equations VI. 10 and VI. 11 are
shown in Table 7 (with the fuel system adjustment factor applied) along with MOBILESa PFI LDV
estimates (calculated using equation VI. 12). Plots of hot soak emissions at 95 °F are shown in Figure
10. Curves shown in Figure 10 also have the fuel system adjustment factor applied.
Predicted hot soak emissions using the real world curve fits are generally higher in the 5 to
7 psi RVP range than predicted using previous MOBILESa curve fits. While this could indicate that
real world data shows higher levels of hot soak emissions in this region than previously estimated
from the laboratory data used to generate the curve fits for MOBILESa, it could also be a artifact of
the significant data scatter.
The equation for LOT hot soak emissions derived from the real world data is:
1986+MY LDTs
Hot Soak = (0.3456+0.04906*RVP)*(0.0055541*Temp)/0.805 (VI. 13)
For comparison purposes, the MOBILESa equation for MY 1981+ LDT PFI vehicles that
pass both the pressure and purge tests is:
Hot Soak = (0.078327 + 0.041297*RVP)*(0.0055541*Temp)/0.46 (VI. 14)
11
-------
Equations VI. 13 and VI. 14 should be multiplied by the fuel system adjustment factor of 0.88 to
obtain hot soak emission results.
Predicted hot soak emissions for PFI LDTs calculated using equation VI. 13 are shown in
Table 8 along with MOBILESa PFI LDT estimates (calculated using equation VI. 14) with the fuel
system correction factor applied in both cases. Plots of hot soak emissions for PFI LDT pass/pass
vehicles at 95 °F are shown in Figure 11 with the fuel system correction factor applied to the curve
fits.
Predicted hot soak emissions using the real world curve fit are only slightly higher than
previous MOBILESa estimates in the 5 to 7 psi RVP range.
C. Carb
Of the 494 vehicles that passed both the pressure and purge tests, 113 vehicles had Carb fuel
systems. A curve fit similar to that used in MOBILESa for Carb vehicles was used:
Hot Soak = (A + B*RVP)*(C + D*Temp2)/E (VI. 15)
Since the data was at various ambient temperatures, each hot soak test value was adjusted to 95 °F
using the temperature correction factor defined in MOBILESa curve fits for Carb pass/pass vehicles:
Adjusted Hot Soak = Hot Soak * (-2.4636+0.00056161*952))/(-2.4636+0.0005616l*Temp2) (VI. 16)
A regression analysis was run on the adjusted hot soak versus RVP data to determine coefficients
A and B in equation VI. 15 above for each disaggregation.
Of the 113 Carb tests, 45 corresponded to LDVs with model years between 1981 and 1985,
38 were LDVs model years 1986+, 14 were LDTs with model years between 1981 and 1985 and 16
were LDTs model years 1986+.
For Carb LDVs with MYs 1981-1985, the first regression analysis resulted in a negative B
coefficient. To correct this situation, 4 additional data points were calculated using the MOBILESa
curve fit at 9 psi RVP and 95 °F and added to the 43 real world data points. This produced a positive
B coefficient with a t-statistic of 2.29 and a P-value of 0.0266. Real world data for LDVs with MYs
1986+ also produced a negative B coefficient. An additional 4 MOBILESa calculated points at 9
psi RVP and 95 °F were added to the 38 real world points. This gave a positive B coefficient with
a t-statistic of 1.56 and a P-value of 0.127, indicating that it was not statistically significant at a 95%
confidence level but produced a reasonable trend. For Carb LDTs, a similar trend was found.
Twelve MOBILESa data points (at 9 psi and 95 °F) had to be added to the 11 real world data points
for MYs 1981-1985 and 15 MOBILESa calculated data points had to be added to the 16 real world
data points for the MYs 1986+ to obtain a positive B coefficient. The t-statistic for the MY 1981-
1985 B coefficient was 0.56 indicating that is was not statistically significant (P-value of 0.57), but
the MY 1986+ B coefficient t-statistic was 5.16 indicating it was statistically significant (P-value of
12
-------
1.6E-05). Coefficients C and D were retained from the MOBILESa curve fits for Carb pass/pass
vehicles and coefficient E was determined so that the calculated hot soak results from using the real
world curve fits matched the MOBILESa curve fits at 9 psi RVP and all temperatures.
The equations for the two model year groupings of LDVs are as follows:
1981-1985 MY LDVs
Hot Soak = (-1.13591 + 0.39098*RVP)*(-2.4636 + 0.00056161 *Temp2)/2.081 (VI. 17)
1986+MY LDVs
Hot Soak = (-1.7318 + 0.45214*RVP)*(-2.4636 + 0.00056161 *Temp2)/2.041 (VI. 18)
For comparison purposes, the MOBILESa equation for MY 1981+ LDV Carb vehicles that
pass both the pressure and purge tests is:
Hot Soak = (0.25593+0.13823*RVP)*(-2.4636+0.00056161*Temp2)/l.31 (VI. 19)
Predicted hot soak emissions for Carb LDVs calculated using equations VI. 17 and VI. 18 are
shown in Table 9 along with MOBILESa Carb LDV estimates (calculated using equation VI. 19).
Plots of hot soak emissions at 95 °F are shown in Figure 12.
Predicted hot soak emissions using the real world curve fits are lower than the previous
MOBILESa estimates in the 5 to 7 psi RVP range. While no conclusions can be drawn from these
curve fits (as additional data needed to be added to make the curves show a positive trend with
RVP), one might assume that the real world vehicle set used to define these curve fits had lower hot
soak emissions in the 5 to 7 psi RVP range than that estimated from MOBILESa (which was
produced from an extrapolation of higher laboratory data). Furthermore, curve fits for MY 86+
vehicles showed lower hot soak emissions than the MY 81-85 group, which is reasonable assuming
an improvement in evaporative control system design.
The equations for LDTs derived from the real world data are:
1981-1985 MY LDTs
Hot Soak = (1.29368+0.08904*RVP)*(-2.4636 + 0.00056161 *Temp2)/2.541 (VI.20)
1986+MY LDTs
Hot Soak = (-1.8687+0.43908*RVP)*(-2.4636 + 0.00056161 *Temp2)/2.527
(VI.21)
For comparison purposes, the MOBILESa equation for MY 1981+ LOT Carb vehicles that
pass both the pressure and purge tests is:
Hot Soak = (-0.164070 + 0.13823*RVP)*(-2.4636+0.00056161*Temp2)/1.31 (VI.22)
Predicted hot soak emissions for Carb LDTs calculated using equations VI.20, and VI.21
13
-------
are shown in Table 10 along with MOBILESa Carb LDT estimates (calculated using equation
VI.22). Plots of hot soak emissions for Carb LDT pass/pass vehicles at 95 °F are shown in Figure
13.
As can be seen from Figure 13 and Table 10, real world data for Carb LDTs with MYs 81-85
predict higher hot soak emissions than MOBILESa and MYs 86+ predict lower hot soak emissions
than MOBILESa. It would be expected that newer model Carb LDTs would have significantly lower
emissions than older model LDTs due to improvements in the evaporative emission control system.
VII. DISCUSSION OF RESULTS
The new "real world" curve fits provided reasonable trends in hot soak emissions relative to
RVP and temperature. For RVPs between 5.0 and 9.0, the "real world" curve fits provided a more
accurate picture of "real world" hot soak emissions for MY 1986+ vehicles. However, the data sets
analyzed contained no data over 9.0 RVP and thus extrapolations beyond 9 psi RVP could not be
developed. This created some dilemma as to meeting the MOBILESa curves at 9 psi RVP. The
methodology used in this report provides a better real world curve fit for lower RVPs and still allows
using MOBILESa curve fits above 9 psi RVP without a discontinuity.
In most cases the curve fits provided reasonable agreement with previous data. The addition
of a liquid leaker category adds better definition of the real world conditions. In addition, the
additional stratification of model year groups provides a better picture of hot soak emissions as
technology improves.
To improve the curve fits developed in this report, additional data are needed, particularly
in the 9 psi RVP range and higher. Previous data in this region were generated using laboratory tests
and may not be indicative of real world conditions. Furthermore, new vehicles now entering the
market have significantly improved evaporative emission control systems. These vehicles should
also be tested to give a more accurate picture of in-use emissions from the current and future U.S.
vehicle fleet.
VIII. HIGH ALTITUDE AND HEAVY DUTY VEHICLES
High altitude hot soak emissions are determined by a multiplicative adjustment to the low
altitude estimates. The adjustment factor (1.3) is the same as was used in MOBILES.
No heavy duty vehicles were tested to measure hot soak emissions for MOBILE6. MOBILE6
will use the same technique for calculating hot soak emissions trucks with a GVWR over 8,500
pounds (all heavy duty classes and busses) as was used in MOBILES. This technique assumes that
the difference between the evaporative emission standards for heavy duty vehicles and light duty
vehicles reflects a difference in the actual uncontrolled emission rates of these vehicles. This has
been determined to be a factor of 1.5 for heavy duty trucks up to 14,000 pounds GVWR and a factor
of 2.0 for heavy duty trucks over 14,000 pounds GVWR. These adjustments are the same as were
used in MOBILES.
14
-------
The resulting heavy duty truck hot soak emissions will also be affected by the differences in
the distribution of fuel delivery systems and any differences in the rate of pressure/purge failures.
These base hot soak emission rates are further adjusted by the effects of the new evaporative test
procedure and the introduction of new emission standards, such as Tier 2. These effects are
discussed in the report, "Modeling Diurnal and Resting Loss Emissions from Vehicles Certified to
Enhanced Evaporative Standards, Including OBD Assumptions," (M6.EVP.005, EPA420-P-99-009).
Table 2. Data strata
Fuel
System
Carb
FI
TBI
PFI
Carb
TBI
PFI
Carb
TBI
TBI
TBI
TBI
PFI
PFI
PFI
PFI
Carb
Carb
Carb
Carb
Pressure
Test
All
All
Fail
Fail
Fail
All
All
All
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Purge
Test
All
All
All
All
All
Fail
Fail
Fail
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Pass
Leaker
Category
Liquid
Liquid
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vapor
Vehicle
Type
All
All
All
All
All
All
All
All
LDV
LDV
LOT
LOT
LDV
LDV
LOT
LOT
LDV
LDV
LOT
LOT
Model
Year
All
All
All
All
All
All
All
All
81-85
86+
81-85
86+
81-85
86+
81-85
86+
81-85
86+
81-85
86+
Total Vehicles/ Average Emissions
Sample
Size
2
7
19
40
21
12
23
12
17
56
0
29
15
225
0
39
45
38
14
16
630
Average Standard
Hot Soak Deviation
(g/test) (g/test)
14.60
57.79
5.30
2.50
6.39
1.71
10.69
4.52
0.54
0.61
-
0.48
0.51
0.66
-
1.17
2.27
1.35
3.68
1.29
2.50
0.09
26.71
9.52
2.80
3.93
2.48
9.90
2.95
0.37
1.24
-
0.35
0.43
2.37
-
2.54
3.50
1.67
4.18
1.42
2.99
15
-------
Table 3. Pressure fail hot soak emission estimates*
(g/test)
RVP
psi
5.0
6.0
7.0
8.0
9.0
Temp
°F
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
New
Curve Fit
0.79
1.70
3.66
7.88
1.19
2.57
5.53
11.91
1.80
3.88
8.36
18.00
2.73
5.87
12.64
27.22
4.12
8.87
19.11
41.15
MOBILESa
Curve Fit
0.70
1.50
3.23
6.96
1.09
2.34
5.04
10.85
1.69
3.65
7.86
16.92
2.64
5.69
12.25
26.39
4.12
8.87
19.11
41.15
* unadjusted for fuel system
Figure 2. Estimated TBI Pressure Fail Hot Soak Emissions and Real World Data
15.0
Real World Data
TBI Curve Fit
•MOBILESa
Test Points at Varying T, Curve Fits atT=95 deg F
-- 12.5
10.0
-- 7.5
-- 5.0
-- 2.5
in
S
O)
ro
O
V)
16
-------
Rgure 3. Estimated PR Pressure Fail Hot Soak Emissions and Real World Data
15.0
Real World Data
PFI Curve Fit
• MOBILESa
Test Points at Varying T, Curve Fits at T=95 deg F
0.0
9.0
Rgure 4. Estimated Carb Pressure Fail Hot Soak Emissions and Real World Data
Real World Data
Carb Curve Fit
• MOBILESa
Test Points at Varying T, Curve Fits at T=95 deg F
15.0
12.5
10.0
-- 7.5
5.0
2.5
&
J£
8
CO
0.0
5.0 5.5 6.0 6.5 7.0
RVP(psi)
7.5
8.0
8.5
9.0
17
-------
Table 4. Purge fail hot soak emission estimates*
(g/test)
RVP
psi
5.0
6.0
7.0
8.0
9.0
Temp
°F
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
New
Curve Fit
0.45
0.96
2.07
4.47
0.78
1.67
3.60
7.76
1.35
2.91
6.26
13.48
2.34
5.05
10.87
23.41
4.07
8.77
18.89
40.67
MOBILESa
Curve Fit
0.69
1.48
3.19
6.88
1.07
2.31
4.98
10.73
1.67
3.61
7.77
16.73
2.61
5.62
12.11
26.08
4.07
8.77
18.89
40.67
* unadjusted for fuel system
Figure 5. Estimated TBI Purge Fail Hot Soak Emissions and Real World Data
Real World Data
TBI Curve Fit
• MOBILESa
Test Points at Varying T, Curve Fits atT=95 deg F
15.0
-- 12.5
10.0
0.0
8.5
9.0
in
s
O)
ro
O
OT
18
-------
Rgure 6. Estimated PR Purge Fail Hot Soak Emissions and Real World Data
Real World Data
PFI Curve Fit
• MOBILESa
Test Points at Varying T, Curve Fits at T=95 deg F
15.0
-- 12.5
10.0
-- 7.5
-- 5.0
-- 2.5
ro
o
V)
0.0
5.0
5.5
6.0
6.5
7.0
RVP(psi)
7.5
8.0
8.5
9.0
Rgure 7. Estimated Garb Purge Fail Hot Soak Emissions and Real World Data
15.0
Real World Data
Garb Curve Fit
• MOBILESa
Test Points at Varying T, Curve Fits at T=95 deg F
- 12.5
-- 10.0 _
- 7.5
- 5.0
-- 2.5
8
CO
0.0
5.0
5.5
6.0
6.5
7.0
RVP(psi)
7.5
8.0
8.5
9.0
19
-------
Table 5. TBILDV pass/pass hot soak emission estimates*
(g/test)
RVP
psi
5.0
6.0
7.0
8.0
9.0
Temp
op
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
MY 81-85
LDV
0.09
0.27
0.48
0.72
0.14
0.42
0.75
1.13
0.19
0.57
1.03
1.55
0.24
0.73
1.30
1.96
0.29
0.88
1.58
2.38
MY 86+
LDV
0.04
0.11
0.20
0.30
0.10
0.30
0.54
0.82
0.17
0.50
0.89
1.34
0.23
0.69
1.23
1.86
0.29
0.88
1.58
2.38
MY 81+ LDV
MOBILESa
0.22
0.65
1.16
1.76
0.24
0.71
1.27
1.91
0.26
0.77
1.37
2.07
0.28
0.82
1.47
2.22
0.29
0.88
1.58
2.38
* adjusted for fuel system
Figure 8. Estimated TBI LDV Pass/Pass Hot Soak Emissions and Real World Data
Test Points at Varying T, Curve Fits at T=95 deg F
• Real World 81-85
D Real World 86+
LDV 81-85 Curve Fit
.... LDV 86+ Curve Fit
MOBILESa
5.0
5.5
2.50
2.25
2.00
1.75
1.50
1.25
20
-------
Table 6. TBILDT pass/pass hot soak emission estimates*
(g/test)
RVP
|)SI
5.0
6.0
7.0
8.0
9.0
Temp
op
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
MY 86+
LDTs
0.04
0.13
0.23
0.34
0.08
0.25
0.45
0.68
0.13
0.38
0.68
1.02
0.17
0.50
0.90
1.36
0.21
0.63
1.13
1.70
MY 81+ LDTs
MOBILESa
0.13
0.40
0.71
1.08
0.15
0.46
0.82
1.23
0.17
0.51
0.92
1.39
0.19
0.57
1.02
1.54
0.21
0.63
1.13
1.70
* adjusted for fuel system
Figure 9. Estimated TBI
D Real World 86+
.... LDT 86+ Curve Fit
MOBILESa
LDT Pass/Pass Hot Soak Emissions and Real World Data
Test Points at Varying T, Curve Fits at T=95 deg F
B
n [
D
1 1
0 5.5 6.0
^^ . -_,..- ' •Dr
• • T^'uT n n
1 1
l 1 1 1 1
6.5 7.0 7.5 8.0 8.5 9
- 2.50
- 2.25
- 2.00
- 1.75
-1.50 «"
-1.25 |
8
-1.00 W
4-1
o
- 0.75 Z
- 0.50
- 0.25
- 0.00
0
RVP(psi)
21
-------
Table 7. PFILDV pass/pass hot soak emission estimates*
(g/test)
RVP
psi
5.0
6.0
7.0
8.0
9.0
Temp
op
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
MY 81-85
LDVs
0.22
0.26
0.30
0.35
0.27
0.32
0.37
0.43
0.32
0.38
0.44
0.51
0.37
0.44
0.51
0.58
0.41
0.50
0.58
0.66
MY 86+
LDVs
0.23
0.27
0.32
0.36
0.27
0.33
0.38
0.44
0.32
0.39
0.45
0.51
0.37
0.44
0.51
0.59
0.41
0.50
0.58
0.66
MY 81+ LDVs
MOBILESa
0.09
0.10
0.12
0.14
0.17
0.20
0.24
0.27
0.25
0.30
0.35
0.40
0.33
0.40
0.47
0.53
0.41
0.50
0.58
0.66
* adjusted for fuel system
Figure 10. Estimated PR LDV Pass/Pass Hot Soak Emissions and Real World Data
1.50
D
D
• Real World 81-85
D Real World 86+
LDV 81-85 Curve Fit
... LDV 86+ Curve Fit
MOBILESa
-- 1.25
-- 1.00
Test Points at Varying T,
Curve Fits atT=95 deg F
22
-------
Table 8. PFILDT pass/pass hot soak emission estimates*
(g/test)
RVP
psi
5.0
6.0
7.0
8.0
9.0
Temp
op
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
MY 86+
LDTs
0.27
0.32
0.38
0.43
0.29
0.35
0.41
0.47
0.31
0.38
0.44
0.50
0.34
0.40
0.47
0.51
0.36
0.43
0.50
0.57
MY 81+ LDTs
MOBILESa
0.23
0.27
0.32
0.36
0.26
0.31
0.36
0.42
0.29
0.35
0.41
0.47
0.33
0.39
0.46
0.52
0.36
0.43
0.50
0.57
* adjusted for fuel system
Figure 11. Estimated PFI LDT Pass/Pass Hot Soak Emissions and Real World Data
D Real World 86+
.... LDT 86+ Curve Fit
MOBILESa
Test Points at Varying T,
Curve Fits atT=95 deg F
5.0 5.5 6.0 6.5 7.0
RVP (psi)
7.5
8.0
8.5
1.50
-- 1.25
-- 1.00
-- 0.75
-- 0.50
-- 0.25
8
V)
0.00
9.0
23
-------
Table 9. Carb LDV pass/pass hot soak emission estimates
(g/test)
RVP
psi
5.0
6.0
7.0
8.0
9.0
Temp
°F
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
MY 81-85
LDVs
0.27
0.82
1.47
2.21
0.40
1.21
2.17
3.27
0.54
1.60
2.87
4.33
0.67
2.00
3.57
5.38
0.80
2.39
4.27
6.44
MY 86+
LDVs
0.18
0.54
0.97
1.46
0.33
1.00
1.79
2.70
0.49
1.46
2.62
3.95
0.64
1.93
3.44
5.19
0.80
2.39
4.27
6.44
MY 81+ LDVs
MOBILESa
0.50
1.51
2.70
4.07
0.58
1.73
3.09
4.66
0.65
1.95
3.48
5.25
0.72
2.17
3.88
5.85
0.80
2.39
4.27
6.44
Figure 12. Estimated Carb LDV Pass/Pass Hot Soak and Real World Data
5.0
4.5
4.0
3.5
Test Points at Varying T, Curve Fits atT=95
• Real World 81-85
D Real World 86+
LDV 81-85 Curve Fit
.... LDV 86+ Curve Fit
MOBILESa
8.5
9.0
24
-------
Table 10. Carb LDT pass/pass hot soak emission estimates
(g/test)
RVP
psi
5.0
6.0
7.0
8.0
9.0
Temp
op
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
75
90
105
120
MY 81-85
LDTs
0.48
1.43
2.55
3.85
0.50
1.50
2.68
4.05
0.52
1.57
2.81
4.24
0.55
1.65
2.94
4.44
0.57
1.72
3.07
4.64
MY 86+
LDTs
0.09
0.27
0.48
0.73
0.21
0.63
1.13
1.70
0.33
0.99
1.78
2.68
0.45
1.36
2.43
3.66
0.57
1.72
3.07
4.64
MY 81+ LDTs
MOBILESa
0.28
0.84
1.50
2.26
0.35
1.06
1.89
2.86
0.43
1.28
2.29
3.45
0.50
1.50
2.68
4.04
0.57
1.72
3.07
4.64
Figure 13. Estimated Carb LDT Pass/Pass Hot Soak and Real World Data
5.0
• Real World 81-85
D Real World 86+
. — LDT 81-85 Curve Fit
- - - LDT 86+ Curve Fit
MOBILESa
Test Points at Varying T, Curve Fits at T=95 deg F
-- 4.5
-- 4.0
-- 3.5
-- 3.0
-- 2.5
8.5
9.0
25
-------
Appendix
Response to Stakeholder Comments
API/Sierra Comments:
The American Petroleum Institute (API) and their contractor Sierra Research (Sierra)
reviewed and commented on the analysis done by Arcadis Geraghty and Miller (Arcadis) under
contract to EPA. The basic recommendations of API to EPA are that:
1. The analysis prepared by Arcadis to support revisions to MOBILE6 hot soak emissions
estimates should be discarded;
2. The Auto/Oil real-world data be used to develop revised baseline hot soak emission rates;
and
3. Additional data collected since the release ofMOBILES be used to develop revised RVP and
temperature correction factors which are applied to the baseline emission rates developed
from the real-world data.
The API approach proposes that the new, and relatively small (480 vehicles), Auto/Oil data
sample be used to replace the baseline estimate of hot soak emissions. Most of the Auto/Oil data
measurements have RVP values less than 8 psi. The baseline estimate of hot soak emissions in
MOBILES is based on a data sample containing over 3,500 hot soak measurements at 9 psi RVP
alone. These measurements include both passenger cars and light duty trucks cover a large range
of model years and a range of temperatures. In addition, the API approach proposes that the small
EPA sample of 181 vehicle tests be used to determine the adjustment of hot soak emissions for
gasoline Reid vapor pressures (RVP) and temperatures for MOBILE6. The Arcadis approach,
developed under direction by EPA, attempts only to replace the existing MOBILES estimates for
adjustment of hot soak emissions for RVP levels below 9 pounds per square inch (psi) with new
estimates based on the new data. The baseline estimates of hot soak emissions and the effects of
RVP values above 9 psi would be retained for MOBILE6 with the EPA approach.
EPA did commit at workshops and in planning for MOBILE6 to reassess the effects of RVP
values below 9 psi in light of the new Auto/Oil data. It was clear that the new data alone, from both
Auto/Oil and EPA, did not have sufficient data at RVP levels at 9 psi and above to determine all of
the effects of RVP and temperature on hot soak emissions needed for the MOBILE6 model from the
new data. EPA had stated clearly at all the MOBILE6 workshops that only hot soak emissions
adjustments for RVP levels below 9 psi would be updated using new data and that MOBILES hot
soak emission rates would be used for all cases with RVP values of 9 psi and above. EPA had never
proposed replacing the baseline hot soak emission estimates used for MOBILES. In the analysis of
the new data, EPA specifically directed their contractor, Arcadis, to only consider the effects of RVP
below 9 psi on baseline hot soak emissions.
26
-------
EPA anticipated that hot soak emission levels in the new data would differ to some degree
from those in MOBILES at 9 psi. It can be expected that emission results from any two data samples
describing the same condition might differ to some extent. EPA specifically directed Arcadis to
assure that the estimate of hot soak emission levels were consistent with the existing estimate for hot
soak emissions at 9 psi RVP to avoid a discontinuity in the hot soak emission levels at 9 psi in
MOBILE6. Both Arcadis and EPA were concerned with the magnitude of the differences actually
observed as the analysis progressed.
EPA decided that resolution of the observed differences between the MOBILES estimates
at 9 psi RVP and the new data results would have to wait for a new (future) analysis and additional
data. Since the hot soak values used to estimate the baseline 9 psi RVP case in MOBILES had been
based on a large samples of vehicles at a variety of temperatures, the existing MOBILES hot soak
values at 9 psi RVP would be retained for use in MOBILE6 and the new analysis of RVP values less
than 9 psi would need to account for any differences with the MOBILES estimates. EPA believes
that the large amount of existing data at RVP values of 9 psi and above should not be ignored in the
development of hot soak estimates for MOBILE6.
Once this decision had been made, the options available to EPA and the contractor for
completion of the analysis were limited. The final choice of analytical approach chosen by the
contractor and approved by EPA is reasonable, given the circumstances. New data, perhaps using
data collection methods similar to that used by Auto/Oil, may be able to resolve the differences for
future versions of the MOBILE model. Values are needed both above and below the 9 psi RVP level
and at a variety of temperatures. One alternative, not explicitly proposed by API, would be to reject
the idea that any changes can be made to hot soak emission estimates for use in MOBILE6 using the
currently available data. In this case, the MOBILES hot soak estimates would be used for MOBILE6
at all temperatures and RVP levels.
In addition to their proposal, API had specific criticisms of the work done by Arcadis:
1. The use of an approach in which Arcadis fabricated data points calculated by MOBILESa
hot soak equations and added those to the "real world" hot soak database because the
emissions versus Reid vapor pressure (RVP) trends observed in the "real world" data for
some model-year and technology groups were "intuitively incorrect. "
As Auto/Oil found in their own analysis of the data, the correlation of the Auto/Oil/EPA hot
soak emissions data with fuel vapor pressure did not yield any significant relationships. The Arcadis
method assured that any relationships between the hot soak emissions data with fuel vapor pressure
would yield a hot soak value at 9 psi RVP which would match the MOBILES results. Their
approach included the "known" MOBILES value in the regression equation. This approach is similar
to forcing an equation through a known value (i.e., zero). Matching the MOBILES value at 9 psi was
an EPA requirement of the analysis results. The alternate approach suggested by API does not match
the MOBILES values at 9 psi for any temperature other than 90 degrees Fahrenheit.
27
-------
2. The application of a constraint that the revised factors had to match the MOBILESa factors
at 9. OpsiRVP (at all temperatures). This prevented any meaningful revision to the hot soak
emission rates, even though the real-world data support significant revisions for a number
of technology and model year groups.
This criticism assumes that the "real world" sample of 480 vehicles tested as part of the
Auto/Oil study is sufficient to replace the baseline MOBILES hot soak emission estimate based on
a much larger sample. Most of the Auto/Oil measurements were at RVP values less than 9 psi.
Only three measurements were at RVP values greater than or equal to 9 psi and only an additional
two measurements are greater than 8 psi. The average RVP measurement for the Auto/Oil sample
is 6.6 psi. Although the MOBILES estimate is based on results which did not use real world
preconditioning, API has not provided sufficient argument that the large sample of thousands of hot
soak data used for MOBILES is faulty and should be discarded. The measurement techniques for
both the original MOBILES sample and the API data are very similar and equally valid. The new
real world data certainly is a better estimate of hot soak emissions at the average RVP of 6 to 7 psi,
but EPA believes that these data values are insufficient to replace the baseline hot soak emission
estimate from MOBILES.
3. Data from vehicles tested multiple times over a range of RVP s and temperatures were
treated as separate data points in the Ar cadis analysis. (These vehicles were not part of the
real-world hot soak programs.) This has the effect of assigning more influence to vehicles
with multiple test scores relative to vehicles that were tested only once. It would have been
more appropriate to generate a mean hot soak score for vehicles tested multiple times (or
only use the score best matching the RVP and temperature conditions of the real-world hot
soak programs) before generating regression equations that are used to estimate fleet-
average emissions.
This is an unfortunate result of mis-communication between Arcadis and EPA. If there were
time and resources to repeat the analysis, these vehicles would likely be handled differently.
However, the EPA data in the sample does have the major advantage over the real world
measurements in that the same vehicles were tested with fuels of various RVP values; a design
desired for inferring RVP effects on hot soak emissions. Removing the EPA vehicles from the
analysis would not have improved the validity of the result.
4. Hot soak emission results from fuel-injected vehicles in the real-world databases were
corrected to account for in-use fuel tank level using the adjustment from MOBILESa. The
MOBILESa-based adjustment is applied to translate the 40% fill level required in the FTP
to a nominal 55% fill level observed in-use. However, Arcadis apparently failed to
recognize that the real-world hot soak programs tested vehicles with the fuel level they had
when recruited for testing. Hence, an adjustment for fuel tank level is not necessary in this
case, provided the distribution of fuel levels in the test programs adequately reflects in-use
conditions. (That was not investigated in the Arcadis report.)
28
-------
Arcadis believed (incorrectly) that the fuel level adjustment was necessary and EPA
supervision of the contractor work failed to note this problem during the analysis. Since the error
only affects the real world data, the error will likely make the estimate of hot soak emissions at RVP
values less than 9 psi lower than they should be and make the effects of reducing in-use RVP levels
from 9 psi higher than the data actually shows. It is not clear how this oversight can be corrected
without repetition of the analysis. This cannot be done because of time and resource limitations.
The estimate of hot soak emissions is clearly an area which should be revisited to resolve this and
the other issues presented by API.
5. In the report M6.RTD. 001, EPA acknowledges that ". . . neither the purge test nor the
pressure test is a perfect identifier of vehicles that have problems with their evaporative
control systems. " We agree with this observation. In the report, "Analysis of Real-Time
Evaporative Emissions Data, " Sierra Research demonstrates that the pressure/purge test
is clearly not very reliable in identifying excess hot soak and diurnal emissions. For
example, Sierra Research shows that well over half of the excess hot soak emissions
measured in a recent program sponsored by the Coordinating Research Council (CRC) and
in one conducted for EPA were not identified by the purge/pressure test procedures. Sierra
Research also evaluated the data from a recent CRC diurnal emissions test program and
concluded that (a) the pressure/purge test was unsuccessful in identifying a substantial
fraction of late-model vehicles with high evaporative emissions, and (b) resting loss
emissions were not strongly influenced by pressure/purge test status.
EPA agrees that the pressure/purge test pass/fail status of vehicles is not necessarily needed
to determine the base hot soak emission levels. The original intention of separating hot soak
emissions into pass/fail categories was intended to better quantify the effects of evaporative system
failures by using a stratified random sampling technique. This is similar to the sampling approach
used for exhaust emissions, which often targets high emitting vehicles using a screening test. In
MOBILE6 the hot soak emission rates themselves do not depend on vehicle age. By adjusting the
fraction of pressure/purge failures by age, MOBILE6 accounts for the increase in evaporative
emissions that is expected as the vehicles age. In addition, the pass/fail status is used in the
MOBILE model to allow for the evaluation of in-use control strategies. If the need to evaluate these
programs disappears or alternate methods are devised to determine program benefits, and other
methods are used to account for the effects of age on evaporative emissions, then the need to test and
evaluate hot soak data by pass/fail status will not necessarily be needed in future versions of the
model.
6. It is important to note that if EPA moves any existing evaporative emission factors
components in MOBILE to a new liquid leaks category, the effect of lowering RVP on the
remaining evaporative (hot soak, diurnal, and running loss) emissions will be increased (in
percentage terms). This is because leaks are not expected to be influenced by lower RVP
while the decrement in the volatility-driven evaporative emissions will be spread over a
lower baseline.
29
-------
Since the effect of RVP is integral to the calculation of the hot soak emission levels.
Vehicles which have liquid leaks are removed from the sample before the base hot soak emission
rates are determined. As a result, the hot soak emission estimates appropriately account for the
effects of changes in RVP. The emission impact of liquid leaks, which do not include an adjustment
for changes in RVP, is added to the evaporative emission estimates. This issue will be of more
concern in data sets where liquid leakers are not removed from the base emission estimates.
FACA Comments:
An Auto/Oil hot soak pilot study has been conducted. The results of these analyses should
be reviewed when available to provide insight into evaporative emissions deterioration.
Although EPA did use these data along with the data collected under EPA sponsorship, the
data did not show any significant evaporative emission deterioration. It should be pointed out,
however, that the available studies on hot soak emissions were not designed to specifically answer
the question of emission deterioration overtime.
AAMA Comments:
The American Automobile Manufacturers Association had the following comment regarding
hot soak emissions:
EPA did not address how it will estimate hot soak emissions from vehicles that either do or
do not have onboard refueling vapor control or how it intends to estimate hot soak emissions
from vehicles certified to the enhanced evaporative requirements or onboard diagnostic
requirements.
EPA did not intend to change the estimate for the basic hot soak emission rates from those
used in MOBILES to be used in MOBILE6. The effects of onboard refueling vapor recovery
(ORVR) systems and the effects of the enhanced evaporative requirements were addressed in the
development of MOBILES. These effects have not been changed for MOBILE6. The onboard
diagnostic requirements are addressed in a separate document, "Modeling Diurnal and Resting Loss
Emissions from Vehicles Certified to Enhanced Evaporative Standards, Including OBD
Assumptions," (M6.EVP.005, EPA420-P-99-009).
AIR Comments:
Air Improvement Resource, Inc., has five areas of comment on the update to the hot soak
emission rates:
1. The RVP regression analysis, which includes the restriction that the curve equals the
MOBILES value at 9 psi, is inappropriate and should be revised.
30
-------
The plots showing the new test data, which include theRVP regression curves, are provided
in the EPA documentation and clearly show that the new test data fall well below the
MOBILES values assumed at 9psi. I/regressions were developed without the fixed point at
9psi, the hot soak predictions would be significantly less. There is not a valid reason to have
the new regressions fixed to the MOBILES value at 9psi other than for convenience sake for
EPA (i.e., it eliminates the need to reinvestigate the hot soak regression over 9 psi in
MOBILE).
The reasons for the significant difference between the new test data and MOBILES at 9 psi
may be due to differences in test conditions and/or due to the passage of time. The MOBILES
hot soak data were measured at standard conditions (e.g., tank fill level of 40%) with
certification fuel (i.e., Indolene). The MOBILES data may also be outdated, as more than 90
percent of hot soak tests were carried over from MOBILE4.1.
An improved statistical approach would be to combine the new data with the raw data used
to develop MOBILES. Once the data were combined, the development ofRVP regression
coefficients would be completed with a complete set of all the hot soak test data.
Alternatively, EPA could throw out the older data and use only the data from the recent test
programs. Since 1992, summer gasoline volatility limits restrict summer gasoline RVP to
below 9 psi, thus there is little need for the older, MOBILES data over 9 psi. Given that the
documentation describes the new test data as an improvement over that in MOBILES, the
best approach for MOBILE6 would be simply to rely solely on the new test data.
The intention of EPA's analysis of hot soak emissions was only to update the estimate for
Reid vapor pressures (RVP) below 9 psi. EPA does not feel that the newer data alone is sufficient
to estimate base emission levels for all RVP levels and temperatures. Although the idea of
combining the new data with the existing data has much merit, this level of effort was not anticipated
for addressing hot soak emissions for MOBILE6. This approach can be used for future updates to
the hot soak emission estimates.
2. The inclusion of additional data points (i. e., "dummy " data) in the cases for which the R VP
regression produced negative coefficients is statistically inappropriate. In this instance, a
negative coefficient signifies an increase in emissions with lowering RVP. The additional
data points need to be removed from the analysis.
As an example, the regression for 1986-and-later LDGVs passing both pressure and purge
tests is based on 223 actual test records combined with 25 additional data points. The 25
data points all equal the MOBILES prediction at 9 psi for 1981-and-later LDGVs passing
the pressure and purge tests. The documentation states that the MOBILES 9 psi data points
were added until the resulting regression coefficients became positive. Thus for this case,
10 percent of the regression data did not originate from either test program, but was taken
from MOBILES. In another, worst-case example, half of the data used in the regression was
based on the MOBILES prediction (i.e., 1981 to 1985 model year PFILDGV passing both
31
-------
pressure and purge tests). The inclusion of these data points biases the resulting statistical
analysis toward an over estimation of the RVP effect on hot soak emissions. EPA needs to
redo the RVP analysis without the addition of MOBILES data points at 9 psi (and with the
raw data used for MOBILES), otherwise the proposed model will over state the impacts of
reduced RVP on hot soak emissions.
API had a similar comment. As Auto/Oil found in their own analysis of the data, the
correlation of the hot soak emissions data with fuel vapor pressure did not yield any significant
relationships. Arcadis was attempting to assure that any relationships between the hot soak
emissions data with fuel vapor pressure would yield a hot soak value at 9 psi RVP which would
match the MOBILES results. Their approach was intended to include the "known" MOBILES value
in the regression equation. This approach is similar to forcing an equation through a known value
(i.e., zero). Matching the MOBILES value at 9 psi was an EPA requirement of the analysis results.
Without an alternate method, it is not clear how the results from the new data could have been made
to match the existing estimate at 9 psi.
3. The documentation does not address how EPA intends to estimate the proportion of the fleet
falling in each of the hot soak emission categories (i.e., liquid leakers, pressure/purge test
complying, and pressure/purge test failing).
For each model year and vehicle class there are four categories of pressure and purge test
complying, pressure test failing, and purge test failing, and liquid leakers. The estimation
of model year fleet-average hot soak emission rates will be highly dependent upon what
proportion of the fleet is assumed to fall into each of these categories.
In the case of 1986-and-later LDGVs, for instance, the average hot soak emission rates
reported for the two test programs combined are 0.66, 2.50, 10.69 and 57.79 grams for
pressure and purge test complying, pressure test failing, and purge test failing, and liquid
leakers, respectively. These data show a factor of 100 difference in emission rates between
the lowest and highest emitters.
The method EPA intends to use to estimate what portion of the fleet is pressure test failing,
purge test failing and liquid leakers needs to be provided for comment. This method needs
to address the impacts of vehicle age andl/M on the estimated proportions.
The derivation of the pressure test and purge test pass/fail rates is addressed in a separate
report, "Estimating Weighting Factors for Evaporative Emissions in MOBILE6," (M6.EVP.006,
EPA420-P-99-023).
4. The documentation does not include how MOBILE6 will model hot soak emissions for
vehicles subject to enhanced evaporative test procedures.
EPA has not yet documented how hot soak emission rates will be estimated for vehicles
subject to enhanced evaporative test procedures and standards. In MOBILES, EPA assumed
32
-------
a 50 percent reduction for pressure and purge test complying vehicles and a 30 percent
reduction for pressure and purge failing vehicles. It may be that the same hot soak
reductions assumed in MOBILES will still be applied in MOBILE6. This shouldbe clarified.
In addition, the impacts of enhanced evaporative standards on liquid leakers, which were
not modeled in MOBILES, also needs to be documented.
EPA is using in MOBILE6 the same assumptions about the effect of the new enhanced test
procedure effects on hot soak emissions as were used in MOBILES. However, in addition to the
effect on the emission rates of vehicles with properly operating emission controls, MOBILE6 now
assumes that the new enhanced test procedure will affect the rate of emission control system failures.
The derivation of the base pressure test and purge test pass/fail rates is addressed in a separate report,
"Estimating Weighting Factors for Evaporative Emissions in MOBILE6," (M6.EVP.006,
EPA420-P-99-023). The effect of enhanced evaporative standards on the base pressure test and
purge test pass/fail rates is addressed in a separate document, "Modeling Diurnal and Resting Loss
Emissions from Vehicles Certified to Enhanced Evaporative Standards, Including OBD
Assumptions," (M6.EVP.005, EPA420-P-99-009).
Evaporative emissions due to liquid leaks is discussed in a separate report, "Evaporative
Emissions of Gross Liquid Leakers in MOBILE6," (M6.EVP.009, EPA420-P-99-025).
5. EPA has not yet documented how hot soak emission rates will be estimated for heavier
LDGT(6001 to 8500 Ibs. GVWR) or for HDGV (over 8500 Ibs. GVWR) vehicle classes. In
the updated hot soak analysis, EPA did not state whether the two test programs include any
vehicles over 6000 Ibs. GVWR. The hot soak methodology for the heavier LDGT and HDGV
needs to be explicitly stated, even if EPA expects to continue to use MOBILES data and
methods for these vehicle classes.
Section VIE has been added to this report to address the calculation of hot soak emissions
for gasoline fueled heavy duty vehicles.
California Comments:
The California Inspection and Maintenance Review Committee had the following comment:
In the MOBILE6 evaporative emissions module, EPA classifies vehicles based on whether
they fail the pressure or purge test. MOBILE6 assumes that gross liquid leaks occur only
among vehicles that fail one of these tests. However, EPA 's paradigm for evaporative
emissions is at odds with the results of real-world studies. For example, an Auto/Oil study
of hot soak emissions from 300 vehicles ("Real World Hot Soak Evaporative Emissions-A
Pilot Study, " Brooks, D. et al. (1995), SAE Paper 951007) found that more than half of
excess evaporative emissions come from cars that do not fail either the pressure or purge
tests.
Failure of the pressure or purge tests is thus a poor surrogate for actual evaporative
33
-------
emission rates. Furthermore, no I/Mprogram includes the pressure or purge tests so failure
of these tests does not appear to be a relevant factor in assessing actual I/M programs.
Several studies, including recent CRC-sponsoredstudies, have measured actual evaporative
emission rates of vehicles. EPA should not continue to base evaporative emissions estimates
on an errant paradigm. Instead, EPA should simply use the real world data directly (with
appropriate attention to sample validity issues of course) to determine evaporative emission
rates.
MOBILE6 does assume that liquid leaks occur for all categories of pass/fail. Evaporative
emissions due to liquid leaks is discussed in a separate report, "Evaporative Emissions of Gross
Liquid Leakers in MOBILE6," (M6.EVP.009, EPA420-P-99-025).
EPA agrees that the pressure/purge test pass/fail status of vehicles is not necessarily needed
to determine the base hot soak emission levels. The original intention of separating hot soak
emissions into pass/fail categories was intended to better quantify the effects of evaporative system
failures by using a stratified random sampling technique. This is similar to the sampling approach
used for exhaust emissions, which often targets high emitting vehicles using a screening test. In
MOBILE6 the hot soak emission rates themselves do not depend on vehicle age. By adjusting the
fraction of pressure/purge failures by age, MOBILE6 accounts for the increase in evaporative
emissions that is expected as the vehicles age. In addition, the pass/fail status is used in the
MOBILE model to allow for the evaluation of in-use control strategies. If the need to evaluate these
programs disappears or alternate methods are devised to determine program benefits, and other
methods are used to account for the effects of age on evaporative emissions, then the need to test and
evaluate hot soak data by pass/fail status will not necessarily be needed in future versions of the
model.
34
------- |