EPA-AA-EOD-80-15
EPA Technical Report
Analysis of Humidity Effects on
Fuel Economy in Response to a
GM Request for CAFE Adjustments
by
Don Paulsell
August, 1980
Engineering Staff
Engineering Operations Division
Office of Mobile Source Air Pollution Control
Environmental Protection Agency
Ann Arbor, Michigan 48105
-------
Analysis of Humidity Effects on Fuel Economy in
Response to a GM Request for CAFE Adjustments
Introduction
GM submits that EPA1 s change in test humidity in 1977 (48 to 75
grains/lb) caused an estimated fuel economy penalty of .29 mpg on the
EPA tests. This value was then adjusted by the fraction of tests EPA
conducted for GM's CAFE. These data are shown in attachment E-II of
their exhibits, and result in the requested adjustments of .16, .13,
and .07 mpg to the GM CAFE values for 78, 79, and 80 MY.
Discussion
GM's penalty factor is based on a value they cited from the DOT/EPA
Panel Report No. 6 (1/10/75) and from some of their own data on 19
vehicles from model years 75-79. The value cited from Report No. 6
(.069% decrease in MPG per one grain/lb humidity increase) was based
on one study referenced in a letter from Ford Motor Co. to EPA in
1974. This letter in turn referenced an internal Ford program report
entitled "Request for Barometric Pressure and Specific humidity
Adjustment Factors for Emissions and Fuel Economy Results, dated
7/30/74". Specific details could not be obtained regarding the test
program, number of tests, confidence levels, or data source.
The same EPA/DOT report section also says, "Others have reported simi-
lar effects although some doubt that such an effect exists." This
statement reflects the fact that the theoretical effect of increased
humidity would be to enrichen the fuel/air mixture, thus reducing fuel
economy. However, the actual occurrence of this theoretical effect is
dependent on the calibration of the particular fuel system and the
form of emission control used. Many late model control systems are
insensitive to humidity differences because of feedback sensors that
control the fuel/air ratio at the optimum value. In deriving a humi-
dity factor from test data, one must carefully design the experiment
to minimize the effects of other variables on fuel economy. Several
studies and correlation programs over the past five years have shown
both positive and negative effects, as well as a wide range of values
for the sensitivity of fuel economy to humidity. The information ob-
tained during these analyses are summarized in the Appendix to this
report.
A specific study done by Juneja et. al. of GM in 1977 (SAE 770136)
showed both positive and negative differences on five cars and the
overall average was only half of what GM is now claiming as the proper
adjustment.
GM estimates average MPG differences from test results where the humi-
dity "ranged" from 30 to 90 grains/lb. Eight of the 19 tests reported
were on 1975 MY vehicles. For the model years involved in the peti-
tion, only two data points are presented for each year. Several
questions immediately come to mind regarding these data. How many
-------
tests were conducted on each vehicle and what are the confidence
intervals for the data reported? Were the tests done just at 30 and
90 grains/lb and if-, so, does this represent the same sensitivity that
might be obtained, between 50 ajid 75 grains/lb? Why dia GM only run
two highway fuel economy data points, when the HWFE test accounts for
45% of the overall MPG value? How can two test vehicles per year
properly represent the fleet used to determine the CAFE for 78, 79,
and 80? Were these data collected specifically for this report, or
were they retrieved from other sources? Hence, one can see that the
data GM collected was not presented in sufficient detail to make an
analysis of whether it is representative, appropriate, or statisti-
cally significant.
The entire subject of ambient correction factors to emissions or fuel
economy results has been discussed previously between EPA and specific
manufacturers. EPA1 s position has been to reject the concept for three
reasons - vehicles do not operate in a world of constant ambient con-
ditions, universal factors can not be equitably applied because they
are constantly changing and are not precisely quantifiable, and
finally, vehicles should be capable of meeting emissions and fuel
economy standards throughout a normal range of ambient conditions.
GM's claim that the change in test humidity levels induced a change to
their CAFE has been recognized by EPA. The humidity levels have been
reset to 50 grains/lb. and are being controlled more precisely than in
1975. Nevertheless, EPA's review of GM's submission and other litera-
ture and data on this subject indicate that much more information and
analysis are needed to assess whether a correction factor could be
adequately determined and properly applied.
One source of data which was considered and analyzed was the EPA
"Paired Data" test results file. This file contains the test results
and ambient conditions for both the EPA and manufacturer's test on the
same vehicle. These data were analyzed for MY 78, 79, 80, and 81 for
both city and highway tests and were stratified into six groups (AMC,
CHRY, FORD, GM, OTHERS, and ALL). Plots and regressions of fuel
economy differences as a function of humidity differences for 6600
test pairs were obtained. The results indicated that the correlation
between fuel economy and humidity differences is very weak. This is
apparent from the high amount of scatter on the plots. The regres-
sions, even though they have no statistical significance, show that it
is possible to get both positive and negative effects on fuel economy
from increases in humidity.
The paired data results incorporate almost all the variable differ-
ences one could encounter between two tests - labs, conditions, equip-
ment, drivers, etc. Although it does not represent a well controlled
experiment, nevertheless, it does represent a large data set from the
actual certification and fuel economy tests which would be expected to
reflect a directional and significant adverse impact on fuel economy
values as a result of EPA's 1977 humidity increase. Based on the data
analysis, EPA cannot state that their is no effect of humidity on the
fuel economy of individual vehicles. At the same time, the analysis
shows that a CAFE adjustment in which EPA could have any confidence
cannot be quantified from the available data.
-------
The entire concept of accepting manufacturer's data in lieu of EPA
confirmatory test^s- under the abbreviated certification program is a
change froth the-'program of 1975. One assumes that the manufacturer
can generate data that are essentially equivalent to EPA's official
values. Hence, the concept of correction factors would have to also
address the laboratory correlation aspects of testing at two
facilities.
For example, data from MY81 tests (at EPA's reduced humidity levels)
on GM vehicles show that GM humidity levels average about 7 grains/lb.
higher than those at EPA. However, GM's fuel economy values have been
and are still about 1.5% higher than EPA's results. Starting in the
1979 model year, EPA accepted data from GM's lab in lieu of confirma-
tory testing (which accounts for about 50% of the data used to
generate CAFE values). GM failed to account for this "bonus" in its
calculation of adjustment factors related to EPA's humidity change.
Therefore, for the 79 and 80 model years, GM may have received a net
CAFE bonus as a result of EPA's acceptance of GM laboratory data.
Conclusions/Recommendations
The data presented in support of GM's proposed adjustment to their
CAFE for humidity cannot be assessed for validity and significance.
The confidence one can have in the universal nature of a correction
factor is generally not sufficient to make predictive or corrective
adjustments. In other words, some data would be grossly overcorrected
and other results would be undercorrected.
The subject of ambient correction factors, their standardization,
validity, feasibility, and magnitude should not be based on limited
data of questionable representativeness. A change to EPA's practices
in this area has significant implications for both emissions and fuel
economy results for all manufacturers. A comprehensive analysis of
all effects, both positive and negative, should be part of any such
study. Even if an acceptable test program could be done, application
of the data to other model years and control systems may not be valid.
-------
APPENDIX
Fuel Economy Differences versus1" Humidity Differences
Table A - Regression Data for 1979 MY80
Table B - Regression Data for 1978 MY79
Table C - Regression Data for 1977 MY 78
Table D - Average Differences on Paired Data 1979
Table E - Average Differences on Paired Data 1978
Table F - Average Differences on Paired Data 1977
Table G - Average Slopes of Regressions 77, 78, 79
Figure 1 - FTP % &MPG versus A Hum 1979
Figure 2 - KWFE %A MPG versus A Hum 1979
-------
ic:it NI.:Y i IIMI
TA6LE. A -
;N
*P£idbErtT
MfclABLEi
2
FTP 3
4
£^ i\ i G^ 5
6
7
8
ILL MPG-9
10
11
12
13
14
15
HUJFE. '6
17
^MPG- 18
19
20
21
*AMKSr "
23
24
25
26
27
28
29
30
31
'TP
(z.)
(3)
ITEM
N
m
b
l^
h^
b
^
M
/v\
b
f^
m
b
j^
y =
=
^ —
3
iv\ L
Aw
^An/=
4 /O
/)AIC
32.
-.3ej
-.60
.OZ
-8.$!
-3.^5
.08
15
-2.6
-.75
.2?
-32.1
-5,15
.35
kKM +
(5 =(M
&• - ? ^
5 2o [• 3o
^fj^\/
\~rf&f
-79
/./xfX>
//*?
;.?o
-.23
> /6
;5.j4
-/.2?
.21
^
2.2^
.26
./2L
/7.6
2-36
JS
^Kii
0
' yo
^>/^
1 1/ *y
/.3I
-.01
.10
7.sy
-.o/
,O«|
/^z
/./z
.3V-
.06
I6f
A76
.0^
"
*(l*f&
5o-6e>O I9 >)^_
ornees
Z88
.6/6
~. 2/
.05
/.oz.
-55
.0 ^
2V5
-/7*
— . /z
,11
-/^.3
"". /7
.21
~£PA)
?¥
-------
ZTT~' [3-FO!rN( ,'Y L INI NLl ('jljLrlCJ
TABcE. B -
15.18
FTP,
AMPG-
7
8
9
10
12
13
1 TP
C3>
y=
N
*b
N
N
b
N
X +b
PATA F o£ FOEu ECOAJOAIX
EP£_
lioJ
CV
X * Zi KM * MO
AJo
PAT ft
N)O
DATA
No
DATA
K)0
- .2.6
,13
-3.^0
.001
13
06
• 0V-
-A67
,3o
103
A73
,0V-
237
, /I
.12.
23S
,65
/S*5
,05
155
Z03
-3.93
,22.
,23
63*?
-,/^f
,17
-/A 6
-53
S/3
S/S
-------
~ I i I- 1C .11 Nl :Y I INJI I Ji )
TflBLE. C -
FTP
2
3
4
5
6
7
8
9
10
r11
12
13
14
15
16
17
18
19
2O
21
22
23
24
25
26
27
28
29
30
31
TP
C3)
ITEM
N
b
b
M
b
TEST
NO
NO
DATA
**
•067
-•38?
.ou
20
-fctf
357
.0$^
37V
-U7
-------
TEST-
/o
PILL.
-2,
^LSCL
.=t«3jL
10
-7
11
12
13
14
—..M
ISO
15
16
17
18
19
20
Mr
,07
-395"
PELL
21
22
23
24
25
26
27
28
29
3O
31
ftTk W
D£t
DEC
-.07T
JS/i-6
3*0
-3
-7
/ry
-------
TA8UL
OP
TEST
/
-------
1*977
-------
TABLE.
-------
TP:2 C|TV TEST"
N= 677 OUT OF 763 116.0El%_rE VS. 12<*.DEU_KH
OEL%_FE
4.0000
7.0000
S.OOOO
3.0000
1.0000
-1.0000
-3.0000
-S.OOOO
-7.0000
-9.0000
yMFfc-EPAV.^o
* « * * *
•
' 0 O
oo oo o
* o o
0 >
ww 2 x O O
» O O O O O O g
0 0 00 0 0?0000 00
* o ?oo o fo ooo*
0 « 00 OpOO 000
* o o oo 2 oo jo 2»
« 00 0 0 OOpO Ol>
o 002 « 232 * 32**** » »
o'o 2000 « «» ?|>«» »*
« « « o » oo oo Tl> oo
» O O O OOO 2**** O O O
4} 4} tJ o ^ o "^ ** f^ o J* -^ '^ ^
O ?OO ?^VO x' OO /'l'*^ OO
o •» ^. o 2 »»K 3
*>« «3 2 2 »* 3 C2 «»
o ooo oo o ' o
» « * * 22 2 o oo
» « oo 002 J; 2
o o o
?*0 »*
» o 2
+ O O » 4>O I
0 O ' 0 fr 0
o o oo oo o
• o o o i o
o oof
* o
v J ji A A (^ /** t ^A 1 1
1^7^ FTP P/An
o
^
0
> 00
0 00 0
o 2 *
2 o o oo o
0
oo o o
ooo o o ooo
> o oo o o
'/> O » « O
> 2 « ?2 » «»* « «
i 0 O O O O
12 o o oo o
oo o o oo
>3 0 00 . 00 f 0
> o o o op o
>oo 300000 oo o
>o o o oo o
oo 2 ** * °
oo, 300 ooo
o ? o o o n
ooo 0300 .
It 1> tt *
o o
oo o o
o o o o o
« o
PATA
-.18000
-.10000
-.20000 -
.60000 -1
.14000
uu
Uil
}
-------
SCATTER PLOT <3> Tp:3
N= 612 OUT OF 7^3 '116.0EL%_FE VS. 12^.DEL KH
DEL*_FE ; _
9.0000 +1 Writ. -g.rt-' i^ ,^^<7 o
-1.0000 *
-3.0000 «
-5.0000
-7.0000
-9.0000 »
7.0000 +
5.0000 *
3.0000 *
1.0000 * «
00 O
o o
O 000 O
O O 00 O
oo o o
o o oo oo I o
o oo el o
2 » 2 » I <
o o oo p ooo o o
» 2 2
^ o
oooo « o ooooo «
ooooo *oo o o o
2 oo oo 2* 22
» o o o o 2 3
o oooo po ooo oo o
o oo 002 * *i *
o <>? oolpoo
o o o ooo o ol 3 o
ooo o oop o
oo ooo o 02
o o o
oo o »2 * *
2 o o2 «»
o oo o
o oooo
02* * oo
o o o o
o oto «
o o
ooo
I
o o o
o o ' o
o o
o« o
o . o
o o o
o, o o
02 *
oooo
o o
ooo o o
o
t> o
o o
^> o io ooo
o op 2
o o o o o o
ooo o
oo » * o
o o ft t> o o
oo o »oo 22 ?
20 o o
OO V 7
3 oo o oo o o
o o o
00*00 o o
o oo
2 *» • *
o o oo o
o 2
o
o
-.18000
-.10000 -.20000
- -""000 -1
•1 .60000 -1
.?nnnn -i
.14000
------- |