Technical Report
EPA-AA-SDSB-86-Q1
Factors Influencing Benzene Emissions
from Passenger Car Refueling
By
Paul M. Laing
May 1986
NOTICE
Technical Reports do not necessarily represent final EPA
decisions or positions. They are intended to present
technical analysis of issues using data which are
currently available. The purpose in 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.
Standards Development and Support Branch
Emission Control Technology Division
Office of Mobile Sources
Office of Air and Radiation
U. S. Environmental Protection Agency
-------
-2-
Table of Contents
Fa ge
I. Introduction 3
II. Experimental Approach 3
III. Sampling and Analytical Technique 4
IV. Experimental Results . 7
V. Other Sources of Data 9
VI. Combined Analysis 12
VII. Current Benzene Refueling Emission Rates
and Inventory 16
VIII. Possible Future Benzene Refueling Emission
Rat'es 23
IX. Summary 25
-------
-3-
I. Introduction
Hydrocarbon emissions occur during the refueling of
gasoline-fueled vehicles. These emissions arise from the
displacement of fuel tank vapors and the evaporation of
dispensed and spilled liquid fuel. Benzene, which has been
listed as a hazardous air pollutant based on evidence of
carcinogenicity, is among the hydrocarbons emitted during
refueling. In an effort to characterize this problem, EPA
conducted a test program to measure the amount of benzene
emitted during a refueling event. The test program conducted
by EPA was relatively limited in scope, so, to broaden the data
base for analysis, EPA's data were supplemented with other
sources of similar data. A combined data analysis resulted in
the development of an empirical relationship between benzene
refueling emissions and three parameters that significantly
affect these emissions. This relationship was used to estimate
average national benzene refueling emission rates, and the
effects of lead phase-down and potential fuel RVP controls on
benzene refueling emissions.
II. Experimental Approach
Recommended refueling emissions test procedures were
adapted for use in the EPA test program. [1]* However, to
provide the test program with a reliable overall approach for
measuring uncontrolled benzene refueling emissions, a benzene
emission measuring system was added to the SHED.
The benzene measurement system utilized charcoal
collection tubes to capture hydrocarbons from air samples
pumped from the SHED (using a small Accuhaler Low Flow Pump)
during each test. The benzene content of each tube was
determined by an independent local laboratory (Environmental
Research Group, Inc.) using Standard NIOSH Method #S311.[2]
The mass of benzene captured in each tube and the volume of the
air sampled from the SHED were used to calculate the SHED
benzene concentration (mass/volume).
All EPA tests were run on a 1983 4-door Cutlass Supreme
with an 18rl gallon rear fill fuel tank and a fill height of
about 4 inches.
Two different fuels were used in the test program. These
included an Indolene fuel (9.0 psi RVP and 1.36 wt. percent
benzene) and a commercial unleaded fuel (11.9 psi RVP and 1.49
wt. percent benzene). Liquid fuel benzene concentrations were
determined by the local laboratory that performed the charcoal
tube analyses.
Numbers in brackets designate references at the end of
this report.
-------
-4-
Fuel dispensing rates ranged from 5 to 7 gallons/minute,
with total dispensed volumes ranging from 14.9 to 15.3 gallons
of gasoline (95 percent fill). As is discussed later,
dispensed fuel temperatures ranged from 79.8°F to 93°F, and
tank temperatures ranged from 9l°F to 120°F.
The refueling emission tests were performed in the manner
outlined in Table 1. The test vehicle was pushed into the SHED
"cold," i.e., at ambient temperature. At this point there was
a ten percent fill in the tank. With the SHED open, the
vehicle's fuel tank was then raised by heating blankets to the
desired temperature.
At the end of the heating phase, the heating blankets were
unplugged, the fuel nozzle was inserted into the fuel neck, the
mixing fans were started, and the SHED was sealed. A
background HC reading was taken inside the sealed SHED prior to
the refueling operation using a FID (Flame Ionization
Detector). The refueling was then performed by turning on the
fuel cart from outside the SHED. The benzene sampling pump was
then started to draw an air sample across the charcoal
collection tube. The refueling ended at a 95 percent fill. A
final FID sample reading was taken, and the fuel tank cap was
reinstalled. The benzene sampling pump was operated long
enough to insure that the charcoal collection tube captured
enough benzene to meet analysis detection and quantification
limits (usually about 30 minutes). During the charcoal tube
sampling period the SHED remained closed and sealed.
Temperature sensors were located at three points inside
the tank, in the fuel cart, and at various other places in the
testing setup. Temperature values were recorded approximately
every two minutes throughout the test. Upon completion of the
test, the heating blankets and thermocouples were disconnected,
and the vehicle was removed from the SHED.
Ill. Sampling and Analytical Technique
HC levels were measured before and after each refueling
operation using standard FID techniques. Overall HC emissions
were calculated by taking the difference between these two
measurements.
The following procedure was devised to determine benzene
emissions. A charcoal collection tube was used to capture
hydrocarbons from an air sample pumped out of the SHED. The
tube was then analyzed for benzene content by a local
laboratory using Standard NIOSH Method #S311.[2] The air
volume sampled was calculated by multiplying the calibrated
pump stroke volume by the number of strokes made- during each
run. (The pump stroke volume was calibrated with the sampling
equipment train in place.) The benzene concentration in the
SHED was calculated by dividing the mass of emissions measured
-------
-5-
Table l
Test Secruence
1. Drain and refuel tank to 10 percent of fuel capacity.
2. Push vehicle into SHED.
3. Connect heating blankets and thermocouples.
4. Heat vehicle tank to desired temperature.
5. Insert fuel nozzle.
6. Close SHED and start mixing fans.
7. Take initial HC sample reading (using FID).
8. Begin refueling.
9. Start benzene sampling pump.
10. End refueling when tank is charged to 95 percent of
capacity.
11. Take final FID sample reading, and reinstall the fuel tank
cap.
12. Operate benzene sampling pump until charcoal collection
tubes capture enough benzene to meet analysis
quantification limits (usually about 30 minutes).
13. Shut off benzene sampling pump and disconnect
thermocouples and heating blankets.
-------
-6-
by the air volume sampled. The total mass of benzene refueling
emissions was calculated by multiplying the benzene
concentration in the SHED by the SHED volume. SHED volume was
adjusted to take into account the space occupied by the vehicle.
This procedure is presented more concisely by the
following equation:
grams benzene emitted (Vs)
(Bt) * (1)
refueling event (VA)
Where:
Br = Charcoal tube benzene content, grams
Vs = SHED volume = 41.8 m1
= (Total SHED volume) - (Standard Vehicle Correction
Factor)
= (48.1 m1) - (6.3 mJ) = 41.8 raJ
VA = Volume of air sample removed from SHED, m1
- (Ns) * (VP)
(Ns = Number of pump strokes)
(VP = Calibrated volume of each pump stroke
= 6.61 x 10"s mJ)
Equation (1) is based on the following four assumptions:
i) Benzene levels are relatively constant throughout the SHED.
ii) The charcoal tube is an efficient trap, allowing no
significant amount of benzene to escape.
iii) No significant amount of benzene exists in the SHED prior
to refueling.
iv) Temperature and pressure corrections to standard
conditions were not necessary.
Benzene refueling emissions (BRE) on a grams
benzene/gallon basis were obtained by dividing the results from
Equation (1) by the volume of fuel dispensed into the tank
during each test. The' concentration of benzene emitted from
the fillneck in units of parts per million (ppm) can be
approximated by the following conversion.
Benzene Cone.(ppm) = 82,700 * BRE(grams benzene/gallon) (2)
There is one final note about, the use of this technique to
measure benzene emissions. Very low benzene concentrations
were experienced throughout the test program. Prior to
refueling, it was not possible to measure a benzene background
level with this technique. Benzene levels were only detectable
after refueling. The local lab's analytical equipment had a
minimum quantification level of 12 ng (1.25 ppm for a 3 liter
-------
-7-
sample) so sampling times (and air volumes) on the order of 30
minutes (1-3 liters) were needed to capture the minimum mass of
benzene which could be accurately quantified. EPA's
experiments produced benzene mass levels that ranged between 22
and 32 ng.
IV. Experimental Results
EPA determined in earlier work[3] that the following three
factors significantly influence overall HC refueling emissions:
1. Dispensed fuel temperature: T0
2. Difference between tank fuel temperature (Tt) and
dispensed fuel temperature (T0): AT
3. Fuel volatility: expressed as RVP
As a result of this earlier work, measurements were made
throughout the benzene test program to determine what effect
(if any) these three parameters would have on benzene
emissions. A fourth factor, liquid fuel benzene concentration,
was also studied to determine its impact on benzene emissions.
Table 2 contains the experimental conditions and results
of the fourteen tests run by EPA. Fuel specifications along
with tank and dispensed temperature conditions are shown for
each run. Overall HC emissions in all fourteen runs were
determined using standard FID techniques. Benzene emissions
were calculated using Equation (1) for all but two of the
fourteen runs. In the runs labeled 5A and 5B, unusually high
HC background levels (over 1600 ppm) existed in the SHED prior
to refueling. It would seem likely that benzene levels were
also unusually high at the beginning of these two test runs.
(Although it not possible to know this for certain because
background benzene measurements were not taken.) Equation (1),
which is based on insignificant benzene background levels, was
therefore not applicable to Runs 5A and 5B. However, it was
possible to modify Equation (1) to provide a reasonably good
estimate of the benzene emissions occurring in these two runs.
The modified equation that follows is based upon the assumption
that the benzene mass fraction of the background hydrocarbons
was equal to that found in the refueling vapors.
adjusted grams benzene (Vs) (HC2 - HCt)
refueling event (VA) (HC*)
Where:
Bt, Vs , and VA are the same as in Equation (1).
HCi = SHED hydrocarbon level prior to refueling, ppm
HC* = SHED hydrocarbon level after refueling, ppm
-------
Table 2
EPA Teat Condition* and Results
wt. t
Fuel
Gallons
Tt - To
Tube
~ of
SHED
Benzene
SHED
Ratio:
Test
Fuel
Benzene
RVP
of Fuel
To
Tt
Bs (uq) Strokes
(a)
(q/qal)
lsl_
96.4
9 HC
1A
IND
1.36
9.0
14.9
90.5
92.0
1.5
25
391.0
.404
.0271
.00419
IB
IND
1.36
9.0
15.4
90.5
92.0
1.5
28
365.0
.485
.0315
105.2
.00461
1C
IND
1.36
9.0
15.1
90.0
92.2
2.2
26
432.0
.381
.0252
88.9
.00429
2A
IND
1.36
9.0
15.1
80.5
92.5
12.0
32
498.0
.406
.0269
74.5
.00545
2B
IND
1.36
9.0
*15.1
83.3
91.0
7.7
25
503.0
.314
.0208
74.5
.00421
2D
IND
1.36
9.0
15.1
79.8
91.7
11.9
25
508.0
.311
.0206
77.4
.00402
3 A
Coram
1.49
11.9
14.9
90.5
93.0
2.5
22
200.1
.695
.0467
149.9
.00464
3B
Cocm
1.49
11.9
15.3
93.0
92.0
-1.0
24
237.5
.639
.0418
149.4
.00428
4A
Comm
1.49
11.9
15.2
82.0
92.0
10.0
27
364.4
.469
.0308
97.2
.00483
4B
Comb
1.49
11.9
15.0
80.8
91.9
11.1
27
435.0
.392
.0262
89.2
.00439
5A
Conn
1.49
11.9
15.2
93.0
118.9
25.9
22
330.9
.211*
.0139
61.8
.00341
5B
Comm
1.49
11.9
15.1
91.2
120.0
28.8
22
341.1
.268*
.0177
72.8
.00368
6A
Comm
1.49
11.9
is.o
83.8
101.0
17.2
22
415.2
.335
.0223
79.9
.00419
6B
Comm
1.49
11.9
14.9
81.8
99.8
18.0
22
401.0
.347
.0233
74.6
.00465
Value has been adjusted to account foe high HC background levels measured prior to.experiment run.
-------
-9-
This correction appears reasonable given the cause of the
background emissions (excess emissions related to high tank
temperatures), the consistency of the final results, and the
fact that there were no significant differences between
statistical analyses performed on datasets which did and did
not include Runs 5A and 5B.
At the conclusion of the planned effort, it soon became
apparent that the EPA test program was too limited to produce a
meaningful relationship among the test parameters. Too few
tests were run, and the fuel benzene and temperature ranges
evaluated were too limited. Therefore, other sources of data
were sought to supplement the EPA data.
V. Other Sources of Data
Benzene refueling emission information was found in the
following five published sources: "Onboard Control of Vehicle
Refueling Emissions, Demonstration of Feasibility," API
Publication #4306;[4] "Benzene Emissions from Motor Vehicles,"
Clarke;[5] "Factors Influencing the Composition and Quantity
of Passenger Car Refueling Emissions - Part I," Lemmons and
Gabele;[6] "Composition of Vapor Emitted from a Vehicle Gas
Tank During Refueling," SAE Technical Paper Series #860086;[7]
and "Benzene Emissions From Passenger Cars," CONCAWE Report
#12/83.[8] As is discussed further below, three of these five
sources (API #4306, Clarke, and Lemmons and Gabele) contained
data that were compatible with EPA's data and were used in the
final analysis of benzene rates. Table 3 contains the data
taken from these three other sources. The following five
paragraphs describe the procedures and the compatibility of the
experiments performed in each of these reports.
The data from API report #4306C4] were collected using
standard SHED techniques and two different vehicles. The test
fuel was spiked to 4.99 wt. percent benzene to make measurement
of benzene concentration by gas chromatography possible. The
vehicles were charged with a 90 percent fill of test fuel at
85°F following a diurnal test. This procedure is comparable to
the approach used in the EPA test program. Two tests at
identical temperature conditions were performed on each vehicle
providing four data points with units of grams benzene/gallon.
The Clarke report[5] involved six tests on two California
specification vehicles. The tests were performed using
standard SHED refueling procedures with a fuel containing 0.83
wt. percent benzene. This report contained a total of six data
points in units of grams benzene/gallon.
The data cited in the Lemmons and Gabele report[6] were
also obtained using standard SHED techniques. This report was
obtained in draft form with a caution from the authors about
the applicability of the data to the vehicle fleet in general
-------
-10-
Table 3
Benzene Refueling Emission Data From Other Sources
Wt. %
Tt - Td
Fuel
grams Hz
Benzene
To
(AT)
RVP
per
Source
in Fuel
{°F)
(F°)
(psi)
qallon
API
4.99
85.0
0.0
9.0
. 1533
API
4.99
85.0
0.0
9.0
. 1663
API
4.99
85.0
0.0
9.0
. 1580
API
4.99
85.0
0.0
9.0
. 1750
Clarke
0.83
85.0
0.0
9.0
.02165
Clarke
0.83
85.0
0.0
9.0
.01948
Clarke
0.83
70.0
0.0
11.0
.02563
Clarke
0.83
70.0
0.0
11.0
.02350
Clarke
0.83
70.0
0.0
11.0
.02350
Clarke
0.83
70.0
0.0
11.0
.01709
L & G*
1.236
45.0
-30.0
11.4
.03197
L & G
1.236
45.0
+5.0
11.4
.02691
LH
1.236
59.0
-44.0
11.4
.04278
L & G
1.236
59.0
-9,0
11.4
.03588
L & G
1.236
59.0
+ 13.0
11.4
.02900
L & G
1.236
59.0
+26.0
11.4
.02300
L & G
1.236
76.0
-26.0
11.4
.04991
L & G
1.236
76.0
+9.0
11.4
.03588
L & G
1.236
76.0
+44.0
11.4
.02507
LH
1.236
92.0
-7.0
11.4
.05612
(L & G) is Lemmons and Gabele.
-------
-11-
due to the extreme temperature conditions at which some of the
experiments were run. Therefore, these data were inspected for
consistency with theoretical expectations and other comparable
data. In total, eleven test points were available from this
source. Based upon trends in this and other data sources,
ideal vapor-liquid behavior, past experience, and engineering
judgment, most of the data was found to be reasonable and
consistent. However, one of the experiments involving extreme
temperature conditions did produce unexpected results. The
experiment involving the combination of vehicle tank and
dispensed temperatures which should have resulted in the
highest total HC emissions based on earlier EPA work,[3]
actually had the lowest overall HC emission level in the
Lemmons and Gabele work. This "abnormal" result could be due
to excessive fuel weathering at the high temperature prior to
refueling. This one data point was eliminated. In the end,
ten data points remained on a grams benzene/gallon basis.
The data presented in SAE Paper #860086[7] were obtained
from four test runs using standard SHED techniques. These
tests differed from the other sources in two general ways.
First, they were bench tests performed on a side fill fuel tank
mounted on a stand. Also, no attempt was made during these
experiments to determine the quantity of the refueling
emissions; only the composition of these emissions was
measured. Therefore, these data were only available on a mass
fraction basis (in units of grams benzene/grams HC) and could
not be used in the final analysis because they were not
available in the proper format (grams benzene/gallon).
The CONCAWE report[8] obtained data using fuels and
vehicles of European specifications. The experiments were run
using modified SHED techniques to measure refueling emissions.
These data were not compatible with the other sources because
an unusual test procedure was used in these experiments. The
fuel tank was preconditioned with a 40 percent charge of test
fuel prior to the refueling test to saturate the tank's vapor
space. This initial charge was drained from the tank, and then
the tank was refueled to 40 percent. The drain of this initial
charge is inconsistent with standard procedures and has
indeterminable effects on the results. One likely problem in
this test approach is that the vapor space in the fuel tank
would not be saturated prior to refueling. Therefore, the
CONCAWB data were not used in this report.
The end result of combining the EPA data with the three
compatible sources was the formation of a combined data set
obtained from a broader set of test conditions. Tables 2 and 3
show these data sets which contain a total of 34 data points
and are in units of grams benzene/gallon.
-------
-12-
VI. Combined Analysis
A primary goal of this study was the development of an
empirical relationship between benzene refueling emissions and
fuel/temperature conditions. This relationship has several
important uses such as the estimation of a national benzene
emission rate and the prediction of fuel alteration effects. A
relationship between benzene refueling emissions and three
significant parameters was developed by fitting a multivariate
linear regression model to the data in Tables 2 and 3. This
relationship has the following form:
grams benzene/gallon = (.035 * %Bz) - (1.60 x 10*4 * T0)
- (4.24 x 10"4 * AT) (4)
Where:
%Bz = Fuel benzene concentration (wt.%), expressed as
percent not decimal fraction.
To = Dispense fuel temperature (°F).
AT = Difference between fuel tank temperature and
dispensed fuel temperature (F°)
Rl = 0.9607
SE = 0.0092
Each of the parameters included in the regression model is
statistically significant at a confidence of at least 99
percent, i.e., the probability that any of the three parameters
has no effect upon benzene refueling emissions is less than one
percent. The correlation coefficients among the predictor
variables ranged from 0.04 to 0.30 which indicates a high
degree of predictor variable independence.
A regression model that included fuel RVP as a factor was
also fitted to this data and showed RVP to have a relatively
insignificant effect . on benzene refueling emissions
(significanfce =» 0.40). Also, for reasons not fully understood,
this regression model showed T0 to have a fairly
insignificant effect on refueling emissions (significance =
0.46). This model is not recommended for use, but it is
included in Appendix A for those who may be interested.
Of the three equation parameters, fuel benzene content
expressed in weight percent (%Bz) has by far the greatest
impact on benzene refueling emissions. Stated in another way,
small changes in %Bz greatly affect the amount of benzene
emitted during refueling. The degree to which this phenomenon
occurs can be seen graphically in Figure 1. Figure 1 is a plot
of Equation 4. Each of the three lines in this figure is the
-------
-13-
result of varying one of the three equation parameters (%Bz, -
To, or AT), while keeping the other two equation parameters
constant at current national annual average conditions. This
figure clearly shows that fuel benzene content has the greatest
influence on benzene refueling emissions. This is also obvious
by a comparison of the %Bz, T0, and AT coefficients in
equation 4.
It is also interesting to note the effect that T0 has on
benzene refueling emissions. Figure 1 shows a weak inverse
relationship between TD and benzene refueling emissions.
While it is expected that T0 should have only a minor effect
on benzene refueling emissions because all of the refueling
tests were performed at temperatures well below the 175°F
boiling point of benzene, it is counterintuitive to expect that
benzene refueling emissions should decrease with increasing
dispensed fuel temperature. However, this result is easily
explained. It is important to remember that the effect TD
has on benzene refueling emissions (i.e., the sign and
magnitude of the coefficient) was determined by a multivariate
statistical analysis. Such an analysis attempts to account for
the effects of several different influences by determining the
coefficients of the parameters in the prediction equation which
are best able- to predict the actual results of the emission
tests. Sometimes the effects of the individual influences are
not completely separable, and the regression analysis tends to
distort their individual influences especially when one of the
parameters has only limited effects. With all of the refueling
tests being performed at temperatures well below the boiling
point of benzene, it is possible that other, more substantial,
variability exists which is not taken into account by the
prediction equation (such as fuel delivery rate and vehicle
fuel system configuration). These other effects have the
potential to distort the contribution of T0 and cause a
negative T0 coefficient to provide the best fit line.
Figure 2 shows the degree to which the prediction equation
fits the actual data. In this figure, the residuals (actual
benzene refueling emissions minus predicted benzene refueling
emissions) are plotted against the predicted benzene refueling
emissions. Most of the predicted values are seen to fall
within 0.01 grams benzene/gallon of the actual values.
Although this range is quite reasonable, it is also apparent
that a certain amount of variability exists in the results.
Much of this variability can be explained by influences such as
differences in vehicle fuel system configuration, fuel delivery
rate, analytical measurement variability, and a certain degree
of inherent test to test and vehicle to vehicle test
variability.
It is interesting to note that the two largest residuals
(+0.025 and -0.016 grams benzene/gallon), both from the Lemmons
and Gabele report,[6] are the result of experiments which
involved dispensed fuel temperatures at the outer limits of
-------
Figure 1
Benzene-Refueling Emissions Fran Prediction Equation 4
(Displacement Loss only)
0.14
0.12
0.10
a o.o8
S3
0.06
with T &aT constant .§ nat'n avg. conditions
with % BziT0 constant @ nat'n avg. oonditicns
with % BziaT constant § nat'n avg. conditi
0.04
0.02
Wt. % Bz 0
(in fuel)
0.5
1.0
1.5
2.0
2.5
3.0
3.5
| T (F*)
-ib
¦tt
+fe 4s +i
Tn (*P) |
52
¦ h
62
J"
- - h
72
J
h-
82
8?
-------
+ 0.025 _l
4-
FI1OT2 2
Residual Values Vs. Predicted Values
+ 0.020 -
~ 0.015 -
~ 0.010
~ 0.005 -
0.005
0.010
0.015
¦* 4-
4-
*
+
~
0.03
~i
0.05
l I i «
0.07 0.09 0.11 0.13 0.15
0.17
Predicted B*iasion Cgrams benzene/gal ion)
i
0.19
0.01
-------
-16-
values tested (92 and 45°F, respectively). The large residual
at 45°F suggests that the assumption of linearity is not valid
at such low temperatures. This deviation from linearity is not
unique. Similar results have been observed in studies of total
hydrocarbon refueling emissions. For example, in "A Study of
Variables that Effect the Amount of Vapor Emitted During the
Refueling of Automobiles,"[9] refueling emissions are seen to
deviate from linearity as AT grows increasingly negative. In
fact, this study showed that at large negative ATs, actual
refueling emissions were far less than the prediction of a
linear model. This result compares closely with the large
negative residual of -0.016 grams benzene/gallon (when T0 =
45°F and AT = -30F°) found in this report. The other large
residual mentioned above (+0.025 grams benzene/gallon when TD
= 92°F) is due possibly to the prediction equation's potential
distortion of T0' s contribution to overall benzene refueling
emissions. As was discussed earlier, it is likely that as T0
increases, benzene refueling emissions will also increase.
However, this is in opposition to the prediction equation
estimate, and therefore, at high dispensed fuel temperatures
(when the actual contribution of TD is expected to be
substantially positive) a large residual results.
It is also interesting to compare Figure 2 with a similar
figure (Figure 3) that depicts the total HC refueling emission
prediction equation residuals from "Refueling Emissions from
Uncontrolled Vehicles".[3] Both figures show similar point
scatters with the majority of predicted emission rates falling
within approximately 25 percent of their actual value. The
biggest difference between the two figures is that Figure 2
lacks data points in the mid-range of its predicted emission
values (the result of having only a limited number of different
gasoline benzene concentrations). In general, Figure 2 shows
the prediction equation to fit the data quite well.
At this point it should be noted that because Equation (4)
is an empirical relationship developed specifically for the
range of test fuel and temperature conditions used in the
various experimental studies, the accuracy of its use outside
this test range has not been evaluated, and caution should be
used when applying this equation to conditions outside of the
following parameter ranges:
%Bz: 0.8 to 5.0 wt.%,
T0 : 50 to 90 °F,
AT: -15 to +20F°,
Fuel RVP: 9 to 12 psi;
VII. Current Benzene Refueling Emission Rates and Inventory
An important use of Equation (4) is the estimation of
current national benzene refueling emission rates. This is
accomplished by obtaining average national values for the three
parameters in this equation.
-------
figure 3 *
USIMRL VALUES ₯6. MffilCTffl VUJDBS
I.HM
I
~
.
* "
*
~
*
2 «
* *
41 |
2
~
«
*
a
~ i|
*
-
~
*
* *~
a
* II
* mm 1 ,NMM* «.mm * **** m.mm , Mm
HDICIM (grams HC/gallon)
Refueling Emissions fron Uncontrolled Vehicles," Rothman and Johnson, u.
I
S. EPA, OAR, CMS, 1985.
-------
-18-
NIPER gasoline surveystio] and gasoline sales reports from-
"Petroleum Marketing Monthly"[11] can be used to calculate a
sales-weighted national average benzene content in gasoline.
Table 4 shows leaded and unleaded gasoline sales volumes for
the winter of 1984-1985 and the summer of 1985. Table 5
combines this sales information with data from NIPER gasoline
surveys to calculate average national gasoline benzene
concentrations (seasonal and annual).
National averages for the other two equation parameters
(Td and AT) were developed in an earlier EPA report on
refueling emissions[3] and are shown below. This data,
together with the fuel benzene content averages, provide the
information needed to calculate current benzene refueling
emission rates for summer, winter, and an overall twelve month
average using Equation (4). The results of these calculations
are shown in Table 6.
National Average Parameter Values
Parameter Summer Winter Annual
T0 (°F) 76.2 60.3 68.9
AT (F°) +8.8 -0.8 +4.4
In the "Evaluation of Air Pollution Regulatory Strategies
for Gasoline Marketing Industry,"[12] theoretical benzene
emission rates were developed from the ideal gas law without
the use of empirical data. It is interesting to compare the
empirical benzene refueling emission rate calculated in this
report (0.0428 grams benzene/gallon) with the theoretical rate
developed in the Gasoline Marketing Study.[12] This comparison
is slightly more complicated than it first seems because the
two benzene refueling emission rates are in different units
(the emission rate in the Gasoline Marketing Study[12] is
expressed as 0.0066 grams benzene/grams HC). This complication
can be eliminated by converting the units of the emission rate
in this report to the units of the emission rate in [12]
through the use of a national average HC refueling emission
rate in units of grams HC/gallon. The current national average
HC refueling emission rate is estimated at 5.4 grams HC/gallon
by substituting the following three current national average
parameter values into the EPA HC refueling emission prediction
equation[3]: 1) Fuel RVP = 11.6 psi,[10] 2) AT = 4.4F°,[3]
and 3) T0 = 68.9°F.[3] Therefore, the national average
benzene refueling emission rate developed in this report can
also be expressed as (0.0428/5.4) or approximately 0.0079 grams
benzene/grams HC. These two rates (0.0079 and 0.0066 grams
benzene/grams HC), which only differ by twenty percent, are
quite similar. As a matter of fact, much of the twenty percent
difference can be explained by 1) the difference in gasoline
benzene concentrations used to estimate these rates (1.59 wt. %
in this report, and 1.54 wt. % in [12]), and 2) the fact that
-------
-19-
Table 4
Gasoline Sales Volumes*
WINTER SALES Percent of
Fuel Type (millions of gallons) Winter Sales
Leaded Regular: 21,177.6 (38.8%)
Unleaded Regular: 25,054.3 (45.9%)
Premium: 8,345.8 (15.3%)
Total: 54,577.7
Fuel Type
Leaded Regular:
Unleaded Regular
Premium:
Total:
SUMMER SALES
(millions of gallons)
22,042.7
27,645.1
9,019 . 5
58,707.3
Percent of
Summer Sales
(37.5%)
(47.1%)
(15.4%)
Fuel Type
Leaded Regular:
Unleaded Regular
Premium:
Total:
ANNUAL SALES
(millions of gallons)
43.220.3
52.699.4
17,365.3
113,285.0
Percent of
Annual Sales
)
(38. 2
(46.5%)
(15.3%)
"Petroleum Marketing Monthly," Energy
Administration, October 1984 to September 1985
Information
-------
-20-
Table 5
National Average Gasoline Benzene Concentrations
SUMMER
WINTER
Fuel Type
Regular unleaded
Premium unleaded
Leaded regular
Percent
Sales*
47, 1
15.4
37.5
Avg. Bz
Content
(Wt. %)»»
1.57
1.59
1.60
Seasonal Average Bz Content: 1.58 wt. 1
Seasonal Sales Volumes: 51.8 %
Annual Average Bz Content: 1.59 wt. %
Percent
Sales***
45.9
15.3
38.8
Avg. Bz
Content
(Wt. %)*»*»
1.38
1.68
1.84
1.60 wt. %
48.2 %
* "Petroleum Marketing Monthly," Energy Information Administration,
April 1985 to September 1985.
** "Motor Gasolines, Summer 1984," NIPER, Shelton and Dickson,
February, 1985.
*** "Petroleum Marketing Monthly," Energy Information Administration,
October 1984 to March 1985.
**** "Motor Gasolines, Winter 1984-85," NIPER, Shelton and Dickson, June,
1985.
-------
-21-
Table 6
Predicted Benzene Refueling Emission Rates
(Grams Benzene/Gallon)
Sumner*
Winter**
Annua1* * *
Displacement
Loss Only
0.0394
0.0467
0.0428
Total Loss
(Includes Spillage)
0.0441
0.0515
0.0475
*
* *
* * *
April to September.
October to March.
Twelve months.
-------
-22-
the Gasoline Marketing Study[12] rate was developed under the
assumption that static conditions apply, and therefore, it does
not account for possible dynamic effects such as liquid/vapor
turbulence or vapor growth/shrinkage effects associated with
non-zero ATs. This comparison shows that theory and
experiment compare quite closely.
It is also interesting to calculate the average
concentration of benzene emitted during refueling and compare
it with benzene concentrations reported in the recent SAE Paper
"Vehicle Evaporative and Exhaust Emissions as Influenced by
Benzene Content of Gasoline".C13] An average benzene refueling
emission concentration can be calculated in the following
manner through the use of Equation (2): 0.0428 x 82,700 = 3,540
ppm or 0.354 percent by volume. The SAE Paper[13] shows that
the concentration of benzene in the head space of a vehicle's
fuel tank averages 0.324 percent by volume for a liquid fuel
benzene concentration of 1.5 volume percent. This head space
concentration (0.324 volume percent) compares quite closely
with the benzene refueling emission concentration calculated in
this report (0.354 volume percent).
The benzene refueling emission values developed thus far
are for displacement losses only. A total benzene refueling
emission factor must also consider spillage losses. From EPA's
emission factor document (AP-42)[14], refueling spillage losses
are estimated to be about 0.3 grams HC/gallon of gasoline
dispensed. Assuming the spilled gasoline completely
evaporates, benzene spillage emissions can be calculated by
multiplying the liquid fuel's benzene weight content (%Bz from
above divided by 100) by the HC spillage emission rate (0.3
grams HC/gallon). Total benzene refueling emission rates are
obtained from Equation 5 (below) which adds the spillage factor
(.003 * %Bz) to the displacement emission equation. This
calculation was performed using national average summer,
winter, and annual fuel/temperature conditions, and the results
are shown in Table 6.
grams benzene/gallon = (.038 * %Bz) - (1.60 x 10"4 * TD)
- (4.24 X 10~4 * AT) (5)
Equation (5) predicts a national annual benzene refueling
emission rate of 0.048 grams benzene/gallon. Using this
average and a 10 gallon fill, a national average value of 0.48
grams benzene/refueling event was calculated.
A current annual benzene refueling emission inventory was
obtained by multiplying the average annual emission rate by the
total gallons of gasoline sold in this period. At present, it
is estimated that 5,440 megagrams of benzene are emitted each
year as a result of refueling. In a GAO (U.S. General
Accounting Office) report on air pollution,[15] the total
benzene emission inventory is estimated as 253,600 megagrams.
-------
-23-
Therefore, benzene refueling emissions account for roughly two
percent of the total benzene emission inventory.
Lastly, it should be noted that the NIPER fuel surveys
show that gasoline benzene content varies significantly from
sample to sample ranging in value from 0.1 to 5.0 wt. percent.
This range, when incorporated into Equation (5), indicates that
at current national average temperature conditions benzene
refueling emissions can range from near zero up to 0.177 grams
benzene/gallon.
VIII. Possible Future Benzene Refueling Emission Rates
Another important use of benzene emission prediction
equation is the estimation of future emission rates. In order
to use Equation (5) for this purpose, future fuel and
temperature conditions must be evaluated. Although temperature
conditions have inherent variability, future seasonal averages
should be similar to current averages. The only parameter that
has a potential to significantly affect future benzene
refueling emissions is fuel composition. Future fuel can be
affected by lead phase-down and potential RVP control. The
possible effects of these fuel changes on benzene refueling
emissions are discussed below.
In an EPA final regulatory impact analysis, "Costs and
Benefits of Reducing Lead in Gasoline,"C16], a DOE refining
model predicted that lead phase-down would not change the
benzene content of the entire gasoline pool (leaded and
unleaded combined), but it would cause a shift in benzene
content between the leaded and unleaded pools. The DOE model
predicted that lead phase-down would increase benzene content
in unleaded gasoline and decrease benzene content in leaded
gasoline. Lead phase-down would therefore not have any
significant effect on overall benzene refueling emissions, but
it would create a greater benzene exposure potential to the
users of unleaded gasoline. Also, it should be mentioned that
lead phase-out is a future possibility. With lead phase-out is
the potential for an increase in the entire gasoline pool's
benzene content as a means to increase octane.[17] Therefore,
lead phase-out could possibly increase benzene refueling
emissions, but whether or not any significant increase occurs
depends on the refiner's blending decisions.
Potential fuel RVP control is another-way in which future
fuel may be altered. Currently, EPA is considering whether
fuel volatility should be regulated as a means of controlling
HC emissions. It has been estimated that a fuel's benzene
content increases slightly with decreasing RVP as shown in
Table 7. [18] Using these potential changes in weight percent
with the national average temperature conditions, it is
estimated that the most stringent RVP control would increase
benzene refueling emissions by about 0.002 grams benzene/gallon
-------
-24-
Table 7
Potential RVP Effects on Gasoline Benzene Content*
and Benzene Refueling Emissions
Refueling Emission
RVP (psi) Benzene (wt.%) Rate (g Benz/qallon)*»
11.5 1.59 .0475
10.5 1.60 .0479
9.5 1.64 .0494
9.0 1.65 .0498
* "Study of Gasoline Volatility and Hydrocarbon Emissions
From Motor Vehicles," EPA-AA-SDSB-85-5, November, 1985.
** Calculated using Equation (5)
-------
-25-
(4.9 percent). The effects of RVP control on benzene refueling
emissions are shown in Table 7. Therefore, RVP control could
possibly increase benzene refueling emissions slightly.
IX. Summary
To characterize uncontrolled benzene refueling emissions,
an in-house EPA test program was conducted to measure benzene
refueling emissions for fuels with different RVPs and benzene
contents under varying temperature conditions. To broaden the
data base for analysis, three other sources of benzene
refueling emission data were combined with the "in-house"
data. Two other sources were studied but determined
incompatible with the other sources due to test procedure
differences.
A regression model was fitted to the combined data set
producing the benzene refueling emission prediction equation
shown below. The model found the following parameters
significantly influence benzene refueling emissions: fuel
benzene content (%Bz), dispensed fuel temperature (T0), and
the difference between tank fuel temperature and dispensed fuel
temperature (AT). Fuel RVP was not found to be a
statistically significant influence.
grams benzene/gallon = (.038 * %Bz) - 1.60 x 10'4 * T0)
- (4.24 X 10~4 * AT)
At current national average conditions including the
effects of spillage, the prediction equations yields an average
benzene refueling emission of 0.048 grams/gallon.
Lead phase-down should not alter overall benzene refueling
emissions but could create a greater benzene exposure to the
users of unleaded fuel. Lead phase-out could possibly increase
overall benzene refueling emissions. Lastly, potential fuel
RVP control may slightly increase benzene refueling emissions.
-------
Ac know 1 edcpnent s
The author wishes to recognize Tony Barth and Carl Fuller
for the special effort they contributed to make the test
procedure a succeessful undertaking.
The author also wishes to thank Fred Holbert for his
technical guidance and Glenn Passavant for his continual
support and special effort to make this project a success.
-------
-27-
References
1. "Instrumentation and Techniques for Vehicle
Refueling Emissions Measurement," SAE Recommended Practice
J1045, August, 1973.
2. NIOSH Manual of Analytical Methods, Volume 3, Second
Edition, 1977.
3. "Refueling Emissions from Uncontrolled Vehicles,"
Rothman and Johnson, U.S. EPA,OAR,OMS, 1985, Public Docket No.
A-84-7.
4. "Onboard Control of Vehicle Refueling Emissions,
Demonstration of Feasibility," API Publication No. 4306,
October, 1978.
5. "Benzene Emissions from Motor Vehicles," P.J.
Clarke, Exxon, April 21, 1978, Public Docket No. A-84-7.
6. "Factors Influencing the Composition and Quantity of
Passenger Car Refueling Emissions - Part I," J. N. Braddock,
T.J. Lemmons, and P.A. Gabele, U.S. EPA, RTP, N.C. (Submitted
to SAE for October 1986 publication).
7. "Composition of Vapor Emitted From a Vehicle
Gasoline Tank During Refueling," SAE Paper # 860086, Furey and
Nagel, General Motors Research Laboratories, February, 1986.
8. "Benzene Emissions from- Passenger Cars," CONCAWE
Report No. 12/83, August 21, 1984.
9. "A Study of Variables that Effect the Amount of
Vapor Emitted During the Refueling of Automobiles," Edward M.
Liston, API Report CEA-21, May, 1975.
10. "Motor Gasolines, Summer 1984," and "Motor
Gasolines, Winter 1984 - 85," NIPER, Shelton and Dickson,
February and June, 1985.
11. ''Petroleum Marketing Monthly," Energy Information
Administration, October, 1984 to September, 1985.
12. "Evaluation of Air Pollution Regulatory Strategies
for Gasoline Marketing Industry," EPA-45O/3-84-0l2a, July, 1984.
13. "Vehicle Evaporative and Exhaust Emissions as
Influenced by Benzene Content of Gasoline," SAE Technical Paper
Series #860531, Seizinger, Marshall, Cox, and Boyd, NIPER,
February, 1986.
14. "Compilation of Air Pollutant Emission Factors,"
(AP-42), U.S. EPA, OAWM, OAQPS, 1977.
-------
-28-
15. "AIR POLLUTION - EPA's Strategy to Control Emissions
of Benzene and Gasoline Vapor," GAG/RCED-86-6, December, 1985.
16. "Costs and Benefits of Reducing Lead in Gasoline,"
EPA Pinal Regulatory Impact Analysis, February, 1985.
17. Lead Phase-Out and Octane Enhancement," V.E, Pierce
and B.B. Bansal, The M.W. Kellogg Co., Chemical Engineering
Processes, March, 1986.
18. "Study of Gasoline Volatility and Hydrocarbon
Emissions From Motor Vehicles," EPA-AA-SDSB-85-5, November,
1985.
-------
-29-
Appendix A
Regression Model with RVF as a Parameter
(Using the data from Tables 2 and 3)
grams Bz/gallon = (.0346 * Bz) - (7.71 x 10"5 * T0)
- (4.38 x 10"4 * AT) - (5.77 x 10"4 * RVP)
R2 = 0.9616
SE = .00927
Parameter Significance
Bz
0.0000
To
0.4653
AT
0.0002
RVP
0.4026
------- |