United States Air and Radiation EPA420-R-01-023
Environmental Protection April 2001
Agency M6.EVP008
vvEPA Estimating Running Loss
Evaporative Emissions in
MOBILES
> Printed on Recycled Paper
-------
EPA420-R-01-023
April 2001
in
M6.EVP.008
Larry C. Landman
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
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.
-------
ABSTRACT
This report documents the method used in MOBILE6 for
estimating the running loss emissions from vehicles.
In earlier versions of EPA's MOBILE model, running loss
emissions (defined as evaporative hydrocarbons that are emitted
when the vehicle is in operation) were calculated as functions of
ambient temperature, fuel volatility, driving cycle, and vehicle
parameters (i.e., fuel delivery system, model year ranges, and
functionality of the evaporative control system). This report is
not a complete re-analysis of the older data used in those
previous versions of MOBILE. Rather, this report incorporates
the effects of "gross liquid leakers" (see report M6.EVP.009)
with the MOBILES running loss estimates, and then verifies that
this approach is consistent with the results of recent running
loss testing (while the MOBILES estimates alone are not).
This report was originally released (as a draft) in June
1999. This current version is the final revision of that draft.
This final revision incorporates suggestions and comments
received from stakeholders during the 60-day review period and
from peer reviewers.
-------
Estimating Running Loss Evaporative
Emissions in MOBILES
Report Number M6.EVP.008
Larry C. Landman
U.S. EPA Assessment and Standards Division
1.0 INTRODUCTION
Running loss emissions are defined as evaporative
hydrocarbons that are emitted when the vehicle is in operation.
Since the MOBILE4 computer model, the US Environmental Protection
Agency (EPA) has estimated running loss emissions based on
analyses of testing performed by one of its contractors
(Automotive Testing Laboratories, Inc.). In MOBILE6, the running
loss emissions are calculated separately for each hour of the
day, based on the vehicle activity for that hour. The hourly
emissions are then weighted together (to form a daily composite
value) proportional to the number of miles driven each hour.
The test programs were designed to test in-use vehicles with
three different driving cycles:
• The New York City Cycle (NYCC) features low speed stop-
and-go traffic conditions with an average speed of 7.1
mph. Details on this cycle can be found on EPA's website
(at http://www.epa.gov/oms/emisslab/methods/nycccol.txt) .
• The EPA Urban Dynamometer Driving Schedule (UDDS) is
commonly called the "LA-4" or "the city test" and
represents city driving conditions. It is used for
light-duty vehicle testing and has an average speed of
19.6 mph. Details on this cycle can be found on EPA's
website (at http://www.epa.gov/oms/emisslab/methods/uddscol.txt) .
• The Highway Fuel Economy Driving Schedule (HWFET or HFET)
represents highway driving conditions under 60 mph with
an average speed of 47.9 mph. Details on this cycle can
be found on EPA's website (at
http://www.epa.gov/oms/emisslab/methods/hwycol.txt) .
The duration of the running loss test is approximately one hour
for each of those three driving cycles. Therefore, the NYC
driving cycle is repeated six times (6 bags), the two portions of
the LA-4 cycle are repeated three times (6 bags), and the HFET
driving cycle is repeated five times (5 bags).
-------
-2-
The running loss emissions test programs were designed to
collect data at four levels of fuel volatility (7.0, 9.0, 10.4,
11.7 psi in Reid Vapor Pressure [RVP]) and at three levels of
ambient temperature (80, 95, and 105° F). Not all vehicles were
tested for all combinations of fuel RVPs and ambient
temperatures, however. There was usually no testing at extreme
conditions, such as the combinations of high RVP fuel and high
ambient temperature (e.g., 11.7 psi/105° F), and low RVP fuel and
low ambient temperature (e.g., 7.0 psi/80° F), because of their
less likely occurrences in the real world. Also, if the running
loss emission results from a test vehicle were low (less than 0.5
grams) at certain fuel and temperature combination (for example,
9.0 psi/95° F), it was assumed that at the combinations of lower
fuel volatility and/or lower ambient temperatures (i.e., 7.0
psi/95° F, 9.0 psi/80° F, and, 7.0 psi/80° F), this vehicle would
have emissions at a similarly low level. Therefore, to save
resources, the vehicle was not tested for the combinations of
lower fuel volatility and lower ambient temperatures. Further,
there have been no tests on 11.7 psi RVP fuel shortly after the
issuance of MOBILE4 in 1989.
In MOBILE4 model, when the test data were not available at
certain combinations of fuel volatility and ambient temperature,
the gram per mile (g/mi) running loss emissions were estimated
from a variable called "True Vapor Pressure (TVP)." In the
MOBILE4.1 model, this TVP was used to correlate with the running
loss emissions from failed vehicles. These TVPs by bag are
expressed as functions of fuel volatility and fuel tank
temperature. The TVP values were calculated for all combinations
of fuel volatility (7.0, 9.0, 10.4, and 11.7 psi RVP) and tank
temperature profiles (with the initial tank temperatures at 80,
87, 95, and 105° F).
In recent years, industry sources have performed running
loss testing programs in which random samples of in-use vehicles
were tested (see Section 2). In this analysis, we compared these
new data to the MOBILES predictions to determine whether changes
need to be made for MOBILE6.
2.0 NEW RUNNING LOSS TEST DATA
During the summer of 1997, running loss tests were performed
on 150 vehicles as part of a testing program (project number
E-35) conducted for the Coordinating Research Council (CRC) . [1]*
The running loss emissions for these vehicles were measured over
a single LA-4 driving cycle, using tank fuel (RVP about 6.8 psi),
and ambient temperature about 95 degrees Fahrenheit. The
following summer (1998), CRC conducted a testing program in which
The numbers in brackets refer to the references in Section 6 (page 8).
-------
-3-
running loss tests were performed on 50 late-model year vehicles
(1992 through 1997, with a mean age of 4.5 years) (project number
E-41) . [2] These 50 newer vehicles were again tested using tank
fuel (RVP about 6.8 psi) and with an ambient temperature of about
95 degrees Fahrenheit; however, a longer driving cycle was used
consisting of an LA-4 followed by two NYCC cycles followed by a
second LA-4. A summary of the results from those two programs
are given below in Table 1. Within each age range, the mean
running loss test emissions were calculated as well as the 90
percent confidence intervals. The value "Mean Age" was calculated
by subtracting the model year from the test year (either 1997 or
1998) .
Table 1
Summary of CRC Running Loss Testing
CRC
Project
E-35
E-41**
MdYr
Range
Pre-80
80-85
86-91
92-97
Mean
Age
(years)
21.984
13.744
8.340
4.320
Sample
Size
61
39
50
50
Runing
Loss
(gram/mile)
2.3044
1.3800
0.4678
0.3351
90 Percent
Confidence Interval
0.9730
0.5745
0.1497
0.0901
3.6358
2.1855
0.7859
0.5801
** The running loss results of the vehicles tested in Project E-41 are based on a longer
driving cycle but at a slower average speed than the cycle used in E-35.
3.0 MOBILES PREDICTIONS OF RUNNING LOSS EMISSIONS
The MOBILES model was run to generate predictions of the
running loss emissions in the CRC project E-35, that is:
• the ambient temperature was set equal to 95° F,
• the driving cycle was set to a single LA-4, and
• the fuel RVP was set to 6.8 psi.
MOBILES estimates were calculated for each model year within each
of the three purge/pressure strata from reference [4]. Then,
using the weighting factors from Appendix A of that reference,
revised (i.e., re-weighted) MOBILES predictions were produced for
the running loss emissions. Since the CRC testing measured all
evaporative emissions that occurred during the test, those
results (in Table 1) include both running loss and resting loss
emissions. Therefore, resting loss emissions (from reference [3])
were calculated and added to the re-weighted MOBILES estimates.
Since most of the CRC testing was performed during the
summer of 1997, two separate MOBILES runs were necessary (one on
January 1, 1997 and the second on January 1, 1998). The two
-------
-4-
MOBILE5 runs were averaged together to estimate the running loss
emissions of the in-use fleet (by vehicle age) measured during
summer 1997. Those predictions are given below in Table 2.
Table 2
Re-Weighted MOBILES Predictions of Fleet Running Loss
(At CRC Test Conditions)
Age
(years)
0
1
2
3
4
5
6
7
Predicted
Run Loss
(g/mi)
0.1028
0.1220
0.1421
0.1456
0.1507
0.1576
0.1672
0.1800
Age
(years)
8
9
10
11
12
13
14
15
Predicted
Run Loss
(g/mi)
0.1975
0.2199
0.2498
0.2863
0.3334
0.3850
0.4403
0.4945
Age
(years)
16
17
18
19
20
21
22
23
24
Predicted
Run Loss
(g/mi)
0.5434
0.5836
0.6193
0.6409
0.6554
0.6648
0.6706
0.6742
0.6759
The comparison (between the data in Tables 1 and 2]
illustrated by the following graph (Figure 1).
is
Even the most cursory comparison between the average running
loss emissions in Table 1 and the re-weighted MOBILES predicted
running loss emissions in Table 2 (or simply between the data in
Figure 1) suggests that not only do the predicted values
underestimate the observed mean values, but they also do not even
fall within those rather large 90 percent confidence intervals.
This underestimation is most significant for vehicles over the
age of 10 years. There are a number of possible explanations for
those differences; however, EPA believes that the most likely
explanation is the presence of vehicles identified as "gross
liquid leakers" (GLLs) (see reference [5]) in the CRC sample.
In reference [5], EPA used the term "gross liquid leaker" to
identify vehicles having substantial leaks of liquid gasoline, as
opposed to simply vapor leaks. In that report, EPA stated that
the running loss emissions from such a vehicle tested over a
single LA-4 driving cycle would be at least 7.0 grams per mile.
When we examine the running loss test data used in the analysis
for MOBILES, it is questionable whether any of the test vehicles
would meet EPA's definition of a GLL.* In the upcoming section
(Section 4.0), we will consider the effect of adding the
emissions from the GLLs to the (preceding) MOBILES estimate.
The possible absence of "gross liquid leakers" in the data set used for
MOBILES is not unreasonable considering the relatively small number of
such vehicles in the in-use fleet.
-------
-5-
Figure 1
Comparing Re-Weighted MOBILES Predictions with
CRC Running Loss Emissions
E 3
5
2 2
CRC Means
90%Conf. Interval
MS Estimate
12
Vehicle Age (years)
18
24
4.0 EFFECT OF "GROSS LIQUID LEAKERS" ON RUNNING LOSS
EMISSIONS
In reference [5], EPA defined for running loss testing,
"gross liquid leakers" to be vehicles with both liquid leaks of
gasoline and running loss test emissions of at least 7.0 grams
per mile. Using that definition, we note that six (6) of the
vehicles in the CRC testing programs met those criteria. We can
then revise Table 2 by including the estimated running loss
emissions of the "gross liquid leakers" (from reference [5]) . The
revised values are in Table 3 (on the following page).
When we compared the CRC running loss test results (from
Table 1) with these MOBILES predictions that were modified to
include the effects of GLLs (from Table 3), we obtained the graph
in Figure 2 (on the following page).
-------
-6-
Table 3
Re-Weighted MOBILES Predictions of Fleet Running Loss Including GLLs
(At CRC Test Conditions)
Age
(years)
0
1
2
3
4
5
6
7
Predicted
Run Loss
(g/mi)
0.1170
0.1411
0.1697
0.1821
0.2014
0.2279
0.2642
0.3163
Age
(years)
8
9
10
11
12
13
14
15
Predicted
Run Loss
(g/mi)
0.3877
0.4778
0.5940
0.7292
0.8843
1 .0426
1.1978
1.3365
Age
(years)
16
17
18
19
20
21
22
23
24
Predicted
Run Loss
(g/mi)
1 .4559
1.5509
1 .6335
1 .6846
1.7210
1 .7447
1 .7609
1 .7727
1.7788
Figure 2
Comparing CRC Running Loss Emissions with
Re-Weighted MOBILES Predictions Including GLLs
at
E 3
E
5
2 2
-------
-7-
though the difference grows to almost 0.58 grams per mile for the
oldest vehicles:
• From a statistical standpoint, those larger differences
are actually relatively small, less than 10 percent of a
standard deviation.
And,
• The differences between the CRC averages and the
predicted results are the largest in the portion of the
in-use fleet that contributes the least to the total
emissions due to the small number of in-use vehicles
involved. For example, fewer that one-tenth of the fleet
is composed of vehicles older than 15 years of age,
thereby reducing the effect of any potential offset.
5.0 CONCLUSIONS
EPA proposes, for MOBILE6, to use the MOBILES model to
estimate the running loss emissions from that portion of the
fleet that does not contain vehicles that are "gross liquid
leakers." For the portion of the fleet composed (entirely) of
vehicles that are "gross liquid leakers," EPA proposes to use
report M6.EVP.009 (i.e., reference [5]) to both estimate and
weight the emissions. The mean running loss emissions of "gross
liquid leakers" was estimated to be 336.78 grams per hour
(divided by the average speed to obtain units of "grams per
mile") .
In that same report, the estimated frequency of "gross
liquid leakers" in the in-use fleet (as a function of the
vehicle's age) is given by the equation:
Rate of Gross Liquid Leakers
„_,„.. T ^ 0.06
Based on Running Loss Testing = 1 + -i20*exp[-0.4*AGE]
Although this analysis concentrates on light-duty vehicles,
this approach shall be used in MOBILE6 for all vehicle types.
-------
-8-
6.0 REFERENCES
1) D. McClement, "Measurement of Running Loss Emissions from In-
Use Vehicles (CRC Project E-35)", CRC Report No. 611,
Prepared for the Coordinating Research Council, Inc. by
Automotive Testing Laboratories, Inc., February 1998.
2) D. McClement, "Real World Evaporative Testing of Late Model
In-Use Vehicles, CRC Project E-41", Prepared for the
Coordinating Research Council, Inc. by Automotive Testing
Laboratories, Inc., December 17, 1998.
3) Larry Landman, "Evaluating Resting Loss and Diurnal
Evaporative Emissions Using RTD Tests," Report numbered
M6.EVP.001, April 2001.
4) Larry Landman, "Estimating Weighting Factors for Evaporative
Emissions in MOBILE6," Report numbered M6.EVP.006, April
2001.
5) Larry Landman, "Evaporative Emissions of Gross Liquid Leakers
in MOBILE6," Report numbered M6.EVP.009, April 2001.
-------
-9-
Appendix A
Response to Peer Review Comments from Sandeep Kishan
This report was formally peer reviewed by one peer reviewer
(Sandeep Kishan). In this appendix, comments from Sandeep Kishan
are reproduced in plain text, and EPA's responses to those
comments are interspersed in indented italics. Each of these
comments refer to page numbers in the earlier draft version
(dated July 1, 1999) that do not necessarily match the page
numbers in this final version.
************************************
This memorandum provides peer review comments on two EPA
documents: "Estimating Running Loss Evaporative Emissions in
MOBILE6," Document No. M6.EVP.008, June 28, 1999, and
"Evaporative Emissions of Gross Liquid Leakers in MOBILE6,"
Report Number M6.EVP.009, June 30, 1999. Both of these are draft
reports.
Overall, we think that the reports are good, and they present
some new data analysis techniques that are attractive. Since, in
the past, we have had to do similar data analyses and modeling
for evaporative emissions from vehicle test data, we can
appreciate many of the difficulties and data limitations you are
subject to. We hope the comments below help you with this
effort.
Document No. M6.EVP.008 (June 28, 1999)
We have the following questions, comments, and recommendations on
this draft report. For each item we give the page number and
paragraph that the comment refers to, if it is a specific
comment.
Overall this report was clearly written and the general
methodology seems alright. We do not have any recommendations of
any alternate datasets. It seems to us that the more serious
problem with the report is that we are not convinced that the
MOBILES predictions adequately describe the new CRC E-35 and E-41
data. In the comments below, we make suggestions which would
help clarify this comparison to the reader.
1. Page 1, Section 1.0 - We agree with the general methodology
used for data collection in past studies. That is, we agree
that the running loss emissions do not need to be tested at
combinations of temperatures and volatilities that are
either both low or both high.
EPA, of course, agrees with its own methodology.
-------
-10-
2. Page 4, Paragraph 1 - The report suggests a number of
possible explanations for the differences between
corresponding values in Table 1 and Table 2. A short
paragraph listing and very briefly discussing alternative
explanations would give credence to the conclusion that
gross liquid leakers is the most likely explanation.
It was not EPA's intention to develop several possible
explanations as to why the CRC running loss results were
substantially higher than the predictions from MOBILES.
Rather, EPA's goal was to find (and test) a reasonable
hypothesis. This report concludes that the sum of the
running loss emissions attributable to "gross liquid
leakers" (GLLs) and the MOBILES estimates of running loss
emissions from the non-GLLs is a close approximation of the
actual results found in the CRC sample.
3. Page 4, Paragraph 1 - A comparison of the Table 2 averaged
results from the CRC projects with the Table 2 MOBILES
predictions by age is a weak comparison. In general,
whenever averages are used in data analysis, information is
lost. We suggest that instead, a tougher, and therefore
more revealing, comparison should be made by comparing the
individual running loss minus resting loss values from the
CRC studies with the MOBILES predictions by vehicle age.
This could be conveniently done in a plot of running loss
versus vehicle age with the CRC data points on the plot and
the MOBILES curve on the plot. If it is possible to make
such a plot with the CRC data, the result avoids the loss of
information produced by taking averages. Also, if the plot
were made in this way, it would not be necessary to delete
suspected gross liquid leaker running loss values from the
plot. These points would merely be points with high running
loss values and could be highlighted as those which are
suspected of being gross liquid leakers.
The CRC data will be provided (in a spreadsheet) with this
report. So, the users may create their own scatter plots if
they desire. The plots in this report have been revised to
include the (90 percent) confidence intervals at each point
rather than the full scatter plot of all the data.
4. Page 5, Table 3 - The values for mean age in Table 3 are
exactly the same as the values in Table 1. Presumably, if
six vehicles have been removed to produce Table 3 the mean
ages will be different.
The reviewer is correct about that error. However, based on
some of the comments from this reviewer, EPA decided to
change the approach / emphasis (not the actual analysis or
-------
-11-
conclusions) to improve clarity. One of the results of that
change was the dropping of that table from this revision.
5. Page 5, bottom half -
[This material is now on the bottom half of page 6.]
At this point in the report, the discussion centers around
the significance of the differences between the MOBILES
prediction curve and the CRC data values. There are two
problems with this analysis.
First, statistical tests of significance are usually made in
a space where the variance is relatively homogeneous and
normally distributed. The fact that the standard deviations
of the measured values in Tables 1 and 3 are one and a half
to three times the means indicates to us that the running
loss values have skewed distributions. Since automotive
emissions values typically are skewed approximately in a
log-normal fashion, we suggest that instead the test of
significance be based in log space. For us, this would most
convincingly be put forward by plotting the running loss
emissions on a log scale in the plot suggested by Comment 3.
The MOBILES curve should pass somewhere through the center
of the 200 CRC measurements.
As noted in our response to the third comment, the users may
create their own scatter plots (using a logarithm scale in
this case) to compare the results from the CRC sample with
the estimates from MOBILES and MOBILES.
Second, to verify that no significant difference exists
between the measured CRC values and the MOBILES predicted
value, some sort of formal statistical test should be
performed. We suggest that a paired t-test be used to
compare the average of the residuals (measured CRC value -
predicted MOBILES value) for each car with the standard
deviation of the mean of the residuals. These calculations
should be done in log space, where we presume the variance
is homogeneous and normal. If the mean residual is found to
be not significantly different from zero, then it can be
concluded that the measurements and the predictions are the
same. If the mean residual is significantly different from
zero, then either the MOBILES model needs to be changed or
an explanation needs to be provided that the significant
difference is small and is of small practical importance.
The approach suggested by the reviewer is valid; however, it
is more extensive than what EPA is attempting. In fact, a
statistical analysis of the results at the age of 20 years
may find the difference to be statistically significant, but
the relatively small number of in-use vehicles at that age
reduces the effect of that difference on the overall in-use
-------
-12-
fleet (composite) running loss value to be insignificant for
practical purposes.
6. Page 6, Figure 1 - It appears to us that this figure
indicates that MOBILES under predicts the CRC values by
about 40%. Since the standard deviations in linear space of
the running loss values from the CRC studies are greater
than the means of those running loss emissions, any model
which has a curve between 0 and the means will be within one
standard deviation of the means.
The reviewer is correct. We have, therefore, dropped this
figure (graph) from this revision (as noted in the response
to the fourth comment) to improve clarity. The new Figure 2
(in this revision) avoids these problems and still conveys
the desired information.
7. Page 5, Paragraph 2 -
[This material is now on page 6.]
The word "excellent" is glaring in light of the
relationships shown in Figure 1. Regardless of the outcome
of any further analysis on this data in this report, we
suggest selecting a less enthusiastic word.
As noted in the response to the preceding comment, that
figure has been dropped (replaced) to improve clarity.
However, the word "excellent" has been retained, but it now
applies to the fit in the (new) Figure 2 (for vehicles under
the age of 11). We believe that its use is appropriate.
8. Page 6, Section 5.0 -
[This section ("Conclusions") is now on page 7.]
Based on our comments above, the report does not convince us
that there is no significant or important difference between
the CRC running loss values and the MOBILES running loss
predictions. Nevertheless, it could very well be that the
conclusions stated in Section 5.0 are correct.
The differences between the MOBILES estimates of running
loss (plus resting loss) emissions and the results obtained
by CRC in its recent testing programs are significant. We
believe that (for practical purposes) those differences are
explained (and eliminated) by including the emissions from
the "gross liquid leakers" as illustrated in Figure 2 on
page 6 (of this revision) which was added to improve
clarity.
-------
-13-
Appendix B
Response to Comments from Stakeholders
No comments were submitted in response to EPA's posting a
draft of this report on the MOBILE6 website.
------- |