EPA-AA-TEB-EF-88-01
Analysis of Impact of Fuel RVP
on Exhaust Emissions
at 75°F Ambient Temperature
by
Celia Shih
May 12, 1988
Test and Evaluation Branch
Emission Control Technology Division
Office of Mobile Sources
U.S. Environmental Protection Agency
-------
TABLE OF CONTENTS
1.0 BACKGROUND
2.0 LIGHT-DUTY GASOLINE-POWERED VEHICLES
2.1 Data Base
2.2 Results from. Paired t-Test
2.3 Regression Results
2.4 Comparisons
2.5 Temperature and RVP Interactions
3.0 OTHER VEHICLE TYPES
4.0 CONCLUSIONS
-------
1.0 BACKGROUND
In the summer of 1985, a study was conducted to evaluate
the potential exhaust emission benefits from fuel volatility
control. In the Emission Factor (EF) programs, vehicles were
tested at three different fuels with Reid Vapor Pressure (RVP)
at 9.0, 10.4, and 11.7 psi. Test data from a total of 207
light-duty gasoline-powered vehicles (LDGVs) were examined
(Ref. 1). These consisted of all 1981 and newer model year
vehicles. It was concluded that vehicles tested with higher
RVP fuels had higher exhaust HC and CO emissions. The effect
of fuel RVP on NOx emissions, however, was found to be
nonsignificant.
The draft Regulatory Impact Analysis (RIA) for volatility
control discussed the impact of fuel volatility on exhaust HC
emissions based on a total of 322 vehicles from EF three-fuel
testing (Ref. 2, pp. 2-110 to 2-120). The relationships
between the exhaust HC and CO emissions and fuel RVP were
assumed to be linear. Regression coefficients were derived for
HC and CO emissions for 1981 and later model year LDGVs.
For pre-1981 LDGVs, the regressions of HC and CO emissions
as a function of fuel RVP were developed from a small data base
of nineteen 1978-80 model year vehicles tested at Automotive
Testing Laboratories (ATL). Sixteen of the vehicles were part
of the test program at ATL under a contract with the American
Petroleum Institute (Ref. 3). Three vehicles were tested at
ATL under an EPA contract (Ref. 4).
In the MOBILE emission prediction model, the basic exhaust
emissions are derived from tests with Indolene fuel (the
certification fuel with RVP at 9.0 psi). To properly account
for the in-use fuels, which are usually higher than 9.0 psi in
RVP, correction factors are used to adjust the estimated
exhaust HC and CO emissions. The RVP correction factors for
LDGVs used in MOBILE3 Version 9 (M3V9) were based on linear
regression coefficients (summarized in Table 1). The same RVP
correction factors were also used for other gasoline-powered
vehicle types in M3V9.
Many more vehicles have been tested in the EF programs
since M3V9. It was also suggested that the relationship
between the exhaust emissions and fuel RVP might not be linear
for certain types of vehicles. Therefore, the impact of fuel
volatility on exhaust emissions was analyzed again so that more
updated RVP correction factors could be used for MOBILE4.
-------
2.0 LIGHT-DUTY GASOLINE-POWERED VEHICLES (LDGVs)
2.1 Data Base
There were no additional data on fuel RVP versus exhaust
emissions from pre-1981 model year vehicles. The correction
factors developed for M3V9 for the pre-1981 LDGVs are to be
used again for MOBILE4.
Between August 1984 and July 1986, a total of 324 LDGVs
were tested in EF programs on three RVP fuels. The three fuels
used were: Indolene, a summer-grade commercial fuel, and a
fuel blended from the other two fuels, with RVPs about 9.0,
11.7, and 10.4 psi, respectively. Many more vehicles,
especially model years 1983+ vehicles, have been tested since
July 1986 with two fuels (Indolene and commercial fuels). The
current analysis was based on all available model years 1981+
LDGV data, including those used in the previous two studies.
Some vehicles tested between October 1983 and July 1984 on
three RVP fuels were excluded from this analysis, since they
were tested with different test procedure.
2.2 Results From Paired t-Test
The statistical tool used to examine the fuel volatility
effect on exhaust emissions was the paired t-test. As every
vehicle in the sample was tested using both the commercial and
Indolene fuels, the emission differences for each pair of tests
were calculated. The paired t-test is a test of the hypothesis
that the mean of the emission differences is zero. Various
technology groups based on vehicle's exhaust emission control
technologies (open loop versus closed loop), certified CO
emission standards (7.0 g/mi or 3.4 g/mi), model years (1981-82
and 1983+), and fuel metering systems (carbureted, ported
fuel-injected, and throttle body fuel-injected) were used as
strata when performing the paired t-test. No comparison was
made for 1981-82 model year throttle body fuel-injected
vehicles that were certified at 3.4 g/mi CO standard because of
the small sample size (N » 2). The results of the paired
t-test are summarized in Table 2. The values listed in Table 2
are the probabilities that the mean differences are due to
random error. Therefore, small numerical values of the
probabilities are equivalent to high significance levels of the
fuel volatility effect on emission differences. The following
observations are noted:
1. The means of the CO emission differences are
significantly different from zero for all technology groups.
2. For HC emissions, the fuel volatility effects are
significant for most of the technology groups. A few
-2-
-------
exceptions are: model year 1981-82 carbureted vehicles (both
open and closed loop) certified at 3.4 CO standard, and 1983 +
throttle body fuel-injected vehicles.
3. The fuel volatility effects on NOx emissions are
significant for all 1983+ vehicles, but are mostly
nonsignificant for model years 1981-82 vehicles.
The paired t-statistic was also used to examine the fuel
volatility effect on exhaust emissions on a bag-by-bag basis.
For the third bag emissions of the FTP (the hot start portion),
the same conclusions as for the FTP composite can be made. For
the emissions from the other two bags (the cold start and the
stabilized portions), no consistent trends were noted.
2.3 Regression Results
The technique of analysis of covariance was used on the
emission versus fuel RVP data, with technology groups being the
covariate. The analysis of covariance tests whether the slope
terms are statistically significant, and also tests whether the
different technology groups have equal slopes and intercepts.
When the hypothesis of equal slopes was accepted, the intercept
values were also examined to see if certain technology groups
could be combined.
Basically, two types of regressions were considered:
non-linear and linear models. Results from a second degree
polynomial model showed both the first and second degree
coefficient terms to be nonsignificant for all three
pollutants. The choice was between a log-linear and linear
model. The predicted HC and CO emissions from a linear model
were very close to the arithmetic means at 10.4 RVP fuel, but
were always lower than the arithmetic means at the 11.7 RVP
fuel. Therefore, a log-linear model was used to describe the
relationships between fuel volatility and HC and CO exhaust
emissions. For NOx emissions, a linear model was found to be
adequate. The results are summarized as follows:
1. Different slopes were to be used for model years
1981-82 and 1983+ vehicles. The slope term for NOx emissions
of the 1981-82 vehicles was found to be nonsignificant. This
is consistent with the paired t-test results discussed above,
which found that there was no fuel volatility effect on model
year 1981-82 NOx emissions.
2. For the HC and CO emissions of the 1981-82 model year
vehicles, common slopes were to be used for all technology
groups.
3. Equal slope and equal intercept terms were to be used
among the technology groups for all three pollutants for the
1983+ carbureted and fuel-injected vehicles.
-3-
-------
The regression coefficients for each technology group are
summarized in Table 3. To obtain the RVP correction factors by
model year group from the regression coefficients for MOBILE4,
the following procedures were used. For the HC and CO
emissions of the model year 1981-82 vehicles, the overall
regression coefficients were derived from the coefficients of
the technology subgroups weighted by their market shares. The
market share values, summarized in Table 4, were based on
actual sales from EPA's Corporate Average Fuel Economy (CAFE)
files. The coefficients of the 1983* vehicles were used
directly as the overall regression coefficients. The RVP
correction factors were then developed by normalizing the
overall regression equations at the Indolene fuel RVP level.
The resulting RVP correction factors for MOBILE4 are summarized
in Table 5. Note that the sales weighted constant terms of the
HC and CO RVP correction factors for model years 1981 and 1982
were very similar. The emission control system technologies of
these two model years were not different. Based on these
reasons, the model years 1981 and 1982 were combined, the
average of the two constant terms were used to represent the
1981-82 model year vehicles.
Note that the above regression results were derived from
fuels with average RVPs between 9.0 and 11.7 psi. Since
vehicles are designed for EPA's certification test using 9.0
psi RVP fuel, this should be used as the lower limit. The
correction factors for fuels at lower than 9.0 psi RVP are
assumed to be 1.0 (that is, no fuel volatility effect on
exhaust emissions). Correction factors for fuel RVPs higher
than 11.7 psi could be calculated from equations in Table 5.
However, the calculated results are extrapolated and may not be
reasonable, particularly for fuels with RVP much higher than
11.7 psi and at high temperatures.
2.4 Comparisons
The following table is a comparison of the LDGV RVP
correction factors for a commercial fuel at 11.7 psi RVP. The
correction factors for M3V9 were calculated from coefficients
listed in Table 1, while those for MOBILE4 were calculated from
coefficients listed in Table 5.
Correction Factor
for RVP -11.7 Fuel
Source Model Year HC CO NOx
MOBILE4 1971-80 1.050 1.089 1.000
1981-82 1.176 1.208 1.000
1983+ 1.241 1.310 1.069
M3V9 1972-80 1/050 1.089 1.000
1981+ 1.111 1.232 1.000
-4-
-------
As can be seen, for a commercial fuel at 11.7 psi RVP the
MOBILE4 correction factors are slightly higher than those from
M3V9 for HC emissions, and for the CO and NOx emissions of the
1983+ vehicles. Correction factors at various RVP levels are
shown in Figures 1 and 2 for the HC and CO emissions. The
differences between the correction factors of M3V9 and MOBILE4
are small. The MOBILE4 correction factors also characterize
the exhaust emission differences between the model years
1981-82 and 1983+ vehicles.
Note that the MOBILE4 correction factors for the model
years 1971-80 are the same as those from M3V9 for the model
years 1972-80. (The model year range of 1972-80 in M3V9 was an
error. Since 1971 was the first year that evaporative emission
standard applied to all Federal region light duty gasoline
powered vehicles, the correction factors should apply to model
year 1971 LDGVs also.)
2.5 Temperature and RVP Interactions
/
The current analysis on the relationship between fuel
volatility and exhaust emissions was based on FTP results,
i.e., at an amb'ient temperature range of 68 to 86° F. From
limited testing done at Ann Arbor and some vehicles tested at
ATL, where a combination of different ambient temperatures and
fuel RVP were used, data suggested that the relationship of
fuel volatility and exhaust emissions at high RVP fuels and
high ambient temperatures may not be linear. At low ambient
temperatures (for example, at 20° F), however, there appears to
be no fuel volatility effect on exhaust emissions.
At the present time, there are not enough data for a
thorough analysis of fuel volatility effect on exhaust
emissions outside of the FTP temperature ranges. This is an
area for future planning of EF test programs. Ideally, there
should be at least forty vehicles tested at three different RVP
levels and at five ambient temperatures (20, SO, 75, 85, and
95° F) so that the relationship among the fuel volatility,
ambient temperature, and exhaust emissions can be quantified
with confidence. This ideal program, of course, would have to
be scaled to resource availability. In addition to fuel RVP,
other fuel properties (e.g., 90% distillation point) should
also be carefully considered in designing this ideal program.
Currently some vehicles are being tested under a contract
with ATL at the ambient temperature of 50° F with 9.0 and 11.7
RVP fuels. A preliminary analysis of these data showed that
there was a fuel volatility effect on exhaust emissions at 50°
F. Based on these available data, an algorithm was developed
to account for the impact of temperature and RVP interaction.
This algorithm was described in a separate technical note as a
part of the MOBILE4 derivation document (Ref. 5).
-5-
-------
3.0 OTHER VEHICLE TYPES
In M3V9, the RVP correction factors for LDGVs were also
used for other gasoline powered vehicle types, such as
light-duty gasoline-powered class 1 and class 2A trucks (LDGTls
and LDGT2s), and heavy-duty gasoline-powered vehicles (HDGVs).
The following shows the mapping of LDGV model year groups to
other vehicle types, according to their equivalent emission
control technologies:
LDGV LDGT1 LDGT2 HDGV
Model Year Model Year Model Year Model Year
1972-80 1972-87 1979-87 1985 +
1981+ 1988+ 1988+
Since August 1984, a total of 92 light-duty
gasoline-powered trucks (LDGTs) were tested in EF programs on
both the Indolene and commercial fuels. The model years of
these trucks ranged from 1982 to 1986. The data were divided
into four subgroups by their emission standards and fuel
metering systems. The sample sizes for each subgroup were
small (22 for model year 1982, 26 for model year 1984, 26 for
1985-86 carbureted, and 18 for 1985-86 fuel-injected). The
fuel volatility effect on CO emissions was found to be
significant. However, the fuel volatility effect on HC and NOx
emissions was mixed with no specific trend. Until enough data
are available to allow the development of a separate set of
LDGT RVP correction factors, the RVP correction factors
developed from LOGVs will also be used for trucks.
There were no HDGV data available. The LDGV RVP
correction factors will be used also for model years 1985 and
later HDGVs. As there were no evaporative emission standards
for the pre-1985 HDGVs, no fuel volatility effect on their
exhaust emissions is assumed.
The mapping of LDGV model year groups to other vehicle
types in MOBILE4 for RVP correction factors is the following:
LDGV LDGT1 LDGT2 HDGV
Model Year Model Year Model Year Model Year
1971-80 1971-83 1979-83 1985+
1981 1984 1984
1982 1985 1985
1983+ 1986+ 1986+
These new mappings of LDGV model year groups to LDGTls and
LDGT2s were based on the similarities in their emission control
technologies, such as open loop vs. closed loop, catalyst
without vs. with air pump. Also, a high percentage of the
1986+ light duty truck fleet was fuel-injected, similar to the
1983+ light duty vehicle fleet.
-6-
-------
4.0 CONCLUSIONS
The effect of fuel volatility on exhaust emissions was
examined. The data base consisted of all model years 1981 and
newer LDGVs tested at different RVP fuels. The results from
the paired t test showed that the effects of fuel volatility on
CO" and HC emissions are significant. The fuel volatility
effects on NOx emissions are significant for all 1983 +
vehicles. The RVP correction factors were developed from the
data for MOBILE4. These RVP correction factors were also to be
used for other gasoline-powered vehicle types.
The impact of temperature and fuel RVP interaction on
exhaust emissions appears to be an area for further analysis,
pending the availability of additional test data.
-7-
-------
Table 1
RVP Correction Factors* for LDGVs
MOBILES Version 9
Model Year
Group
1972-80***
1981 +
Coefficient
Po
llutant
HC
CO
HC
CO
0.
7.
0.
2.
A
56222
16560
59295
54790
0
0
0
0
B
.012512
.334130
.038720
.959900
0
10
0
11
DN**
.67483
.17277
.94143
.18700
* Correction Factor -(At B*RVP)/DN
where DN - (A + B*RVP) at RVP - 9.0.
** In M3V9, DN was erroneously defined at RVP - 11.5.
*** Regression coefficients were derived from data on model
years 1978-80 vehicles.
-8-
-------
Table 2
Paired t-Test Results
Comparison of LDGV Exhaust Emissions Between
Commercial and Indolene Fuels
Technology
Group
N
Probabilities that Differences
Are from Random Error
HC
CO
NOx
All Vehicles
544
0.0000
0.0000
0.0000
Carbureted
296 0.0000
Open Loop
(81,82) 3.4 CO Std
(81,82) 7.0 CO Std
1983 +
Closed Loop
(81,82) 3.4 CO Std
(81,82) 7.0 CO Std
1983 +
Fuel Injected
Ported
(81,82) 3.4 CO Std
1983 +
Throttle Body
(81,82) 3.4 CO Std
(81,82) 7.0 CO Std
1983 +
47
16
19
12
249
60
80
109
248
109
13
96
139
2
21
116
0.0108
0.3150
0.0680
0.0412
0.0000
0.2431
0.0000
0.0002
0.0525
0.0000
0.0005
0.0000
0.4286
0.0468
0.6560
0.0000
0.0019
0.0279
0.0471
0.0101
0.0000
0.0018
0.0000
0.0000
0.0001
0.0000
0.0115
0.0000
0.0074
0.1467
0.0206
0.0000
0.2798
0.6264
0.1873
0.0521
0.0017
0.6409
0.0584
0.0002
0.0000
0.0000
0.3381
0.0000
0.0049
0.5554
0.0000
-9-
-------
Table 3
LDGV Regression Results
Regression Coefficients**
Technology
Group
Model Years 1981-82
HC
A
Vehicles
CO
B
A
B
NOx
A
B
Carbureted Open Loop
3.4 CO std
7.0 CO std
Carbureted Closed
3.4 CO std
7.0 CO Std
Fuel-injected
Ported
Throttle Body
-0.
-0.
Loop
-0.
-0.
-1.
-1.
94
56
91
79
27
25
0.
0.
0.
0.
0.
0.
06
06
06
06
06
06
1
1
1
1
1
1
.14
.96
.59
.78
.10
.22
0
0
0
0
0
0
.07
.07
.07
.07
.07
.07
1.
0.
0.
1.
1.
1.
14
66
92
06
22
18
0.01*
0.01*
0.01*
0.01»
0.01*
0.01*
Model Years 1983+ Vehicles
All -1.88 0.08 0.36 0.10 0.60 0.02
* Coefficient is nonsignificant at 80 percent significance
level.
** Regression models are:
HC or CO emissions - EXP(A + B*RVP),
NOz emissions - A + B'RVP
-10-
-------
Table 4
Market Shares of the Model Year
1981-82 LDGVs
Fuel Catalyst CO Market Share*
System Tech. Standard 1981 1982
Carbureted Open Loop 3.4 0.174 0.125
7.0 0.107 0.205
Carbureted Closed Loop 3.4 0.173 0.072
7.0 0.462 0.427
PFI Closed Loop 0.057 0.062
TBI Closed Loop 0.027 0.109
•Source: Actual sales from EPA's CAFE files.
-11-
-------
Table 5
RVP Correction Factors* for LDGVs
MOBILE4
Model Year
Group
1971-80**
1981
1982
1981-82***
Coefficient
Pollutant
HC
CO
HC
CO
HC
CO
HC
CO
B
DN
0.56222
7.16560
•0.8520
1.6012
-0.8501
1.6200
-0.8511
1.6106
0.012512
0.334130
0.06
0.07
0.06
0.07
0.06
0.07
0.67483
10.17277
0.7320
9.3110
0.7334
9.4877
0.7326
9.3990
1983 +
HC
CO
NOz
•1.88
0.36
0.60
0.08
0.10
0.02
0.3135
3.5254
0.78
* Correction Factor - EXP(A + B*RVP)/DN
where DN - EXP(A + B*RVP) at RVP - 9.0
for 1981+ HC and CO emissions and,
Correction Factor - (A + B*RVP)/DN
where DN - (A + B«RVP) at RVP » 9.0
for pre-1981 HC and CO, and 1983+ NOx emissions.
** The same correction factors used in M3V9.
*** Average of the separate 1981 and 1982 constant terms
were to be used for the 1981-82 model years.
-12-
-------
References
1. "Relationship between Exhaust Emissions and Fuel
Volatility,? EPA memorandum from Thomas L. Darlington to
Charles L. Gray, Jr., June 24, 1985.
2. Draft Regulatory Impact Analysis: Control of Gasoline
Volatility and Evaporative Hydrocarbon Emissions from New Motor
Vehicles, US EPA, July 1987.
3. "A Study of Factors Influencing the Evaporative
Emissions from In-Use Automobiles," Health and Environmental
Sciences Department, API Publication No. 4406, April 1985.
4. Under an EPA contract, the Automotive Testing
Laboratory at East Liberty, Ohio tested a total of 56
light-duty gasoline powered vehicles at three RVP fuels (9.0,
10.4, and 11.7 psi) and three different ambient temperatures to
quantify the effects of fuel RVP and temperature on evaporative
emissions. The exhaust emission results from the three 1980
model year vehicles were used in M3V9 to derive the RVP
correction factors on exhaust emissions.
5. MOBILE4 Derivation Notes, Appendix 8-A, April 1988.
-13-
-------
Figure 1
Effect of Fuel Volatility
on LDGV HC Emissions
1.25-
M4 1983+
O
CO
LL
o
£
o
O
1.15-
1.05-
M4 1981-82
M3V9 198U
1971-80
0.95
8.5
9.5 10.5 11.5
Fuel RVP (in PSI)
12.5
-------
o
£
O
0)
o
o
1.35
1.25-
1.15-
1.05-
Flgure 2
Effect of Fuel Volatility
on LDGV CO Emissions
M4 1983+
M3V9 1981+
M4 1981-82
1971-80
0.95
8.5
9.5
10.5
11.5
12.5
Fuel RVP (in PSI)
------- |